This study was conducted in the facilities of CRILAR (Centro Regional de Investigaciones Científicas y de Transferencia Tecnológica), located in Anillaco, La Rioja, Argentina (28°48′S; 66° 56′ W; altitude: 1,350 m). The study area belongs to the northern region of the Monte Desert. At the summer solstice, the duration of the photophase (the “L” portion in an LD cycle) in this area is approximately 14 h, being 3 h and 40 min longer than the photophase during the winter solstice (approx. 10 h and 20 min) (Time and Date AS, 2021). The population of tuco-tucos studied (Ctenomys aff. knighti) occurs naturally in this area.

Animals were trapped using tubular traps made of rigid plastic (PVC), placed in the entrance of active burrows. The traps did not injure or offer any threat to animals’ bodily integrity and, to minimize any discomfort and stress, each trap was checked at least every 2 h. Only adult animals (>100 g) were used in the experiments.

In the laboratory, tuco-tucos from experiment 2 were directly transferred to the experimental condition, and those in experiments 1 and 3 were previously maintained for at least 7 days in a room with minimum noise and natural photoperiod provided by a glass window. Relative humidity ranged from 30 to 60% and room temperature was maintained at 24 ± 2°C, which is within the thermoneutral zone for this species (Tachinardi et al., 2017). Data loggers (HOBO U10/003, Onset Computer Corporation, Bourne, MA) recorded room temperature and relative humidity every 15 min. Animals were individually housed in a cage with wood shavings for bedding and equipped with an activity wheel (23 cm in diameter, 10 cm wide, 1 cm between the bars) connected to a data acquisition system which recorded wheel revolutions at 5-min intervals (ArChron Data Acquisition System—Simonetta System, Universidad Nacional de Quilmes, Buenos Aires, Argentina). Food (carrots, sweet potatoes, native plants, sunflower seed and commercial rabbit pellets) was provided ad libitum, with daily replacement at random times.

During laboratory experiments, animal cages were maintained inside opaque insulation boxes, equipped with dedicated systems for ventilation and lighting. Incandescent red-light bulbs provided continuous dim red light (1–5 lux) to facilitate animal care. A fluorescent bulb (200–250 lux at cage level) was turned on and off by a timing device to control the light (L) and dark (D) regimens. Each isolation box held up to 4 acrylic cages (53 × 29 × 27 cm). Animals used in experiment 3 were also surgically implanted with temperature sensitive transponders (G2 E-Mitters, Mini-Mitter, Bend, OR) to allow for semi-continuous monitoring of core temperature (T_b; for surgical details, see Tachinardi et al., 2014). Data from the intra-abdominal transmitter were collected at 5-min intervals by a receiver (ER 4000, Mini-Mitter, Bend, OR) placed below the cage and processed using the software VitalView (Mini-Mitter, Bend, OR).

We used three semi-natural outdoor enclosures built in an area with native vegetation (enclosure 1: 10 × 5 × 1 m; enclosures 2 and 3: 12 × 6× 1.5 m). Enclosures were surrounded aboveground by wire mesh fencing and 1 m deep underground by concrete blocks to prevent tuco-tucos from escaping (Supplementary Figure 1). A nylon mesh also covered each enclosure to prevent aerial predation. Since native vegetation was enough for foraging, no extra food was provided.

Animals were released in the semi-natural enclosures with biologgers that recorded their activity and light exposure. From January to March (summer season) of 2015, 2016, and 2017, 19 freshly caught tuco-tucos (145.5 ± 37 g; 8 males; 11 females) were released individually inside each enclosure for different deployment durations (n = 15 short term recordings of 6–24 days; n = 4 long term recordings of 100–152 days). Deployment duration in the field enclosures was adjusted along the experiment, considering the trade-off between long duration recordings and increasing chance of death, predation or escape of animals with time. Both light and activity biologgers were mounted on a collar, built with cable ties inserted through silicon tubing, and deployed on tuco-tucos. Light loggers (15 × 6 × 6 mm; weight 0.65 g; model W65, Migrate Technology, United Kingdom) detected bouts of surface emergence during daylight within civil twilight limits, with no sensitivity to moonlight. Light intensity in the range of 1–19,000 lux (resolution of 249 discrete levels) was recorded every 5 min. In addition, accelerometers (23 × 12 × 10 mm; weight 2 g; model Axy-3, Techno Smart, Italy) were used in 6 individuals to record daily rhythm of general activity. Acceleration in three spatial axes (XYZ) was recorded in the range of −4 to +4 G-forces (8 bits resolution) every second. Activity was extracted from raw data by calculating the Vectorial Dynamic Body Acceleration (VeDBA) as in Qasem et al. (2012).

We named field-activity duration the interval between activity onset and offset measured in the field. Field-activity duration was measured in the accelerometry data from 3 animals recaptured in summer in the current work and another 3 animals in autumn/winter from Jannetti et al. (2019). The fourth summer animal with activity data did not show stable activity rhythms during the field condition, thus its data could not be used in this analysis. For each of the 6 accelerometry recordings, a subset of 10 days was extracted, based on the robustness of activity/rest rhythm. Onset and offset of activity phase for each day were determined through El Temps software eye-fitting tools (Díez-Noguera, 2020). Field-activity duration was calculated for each day and it was averaged by animal. Comparison between averaged values from summer and winter season was performed with one-sided Student’s T-test (de Winter, 2013). This analysis as well as the next ones described were performed in R software (R Core Team, 2020), unless specified otherwise. The first 24 h of both sensors’ data were excluded from all analyses.

Soil temperature was recorded by a buried temperature logger (HOBO^®, Onset Computer Corporation, Bourne, MA; accuracy ± 0.5°C from 0 to 50°C) at 20 cm deep (T_soil). Data from external temperature at the surface (Jannetti et al., 2019) were obtained using a temperature logger inserted inside a taxidermied tuco-tuco exposed to the sun, to account for radiation and convection effects. This measurement is called operative temperature (T_e) (Chappell and Bartholomew, 1981). Due to technical difficulties, T_e data could not be collected for all field recordings, thus, we used an estimation of T_e from a semi-continuous satellite dataset to cover the entire experiment duration. This was possible due to numerous factors that diminished the microenvironmental variations present in our T_e measures: the region studied has sparse vegetation and low retention of humidity; T_e data were summarized in 3-h averages; and measures were taken in open areas, where tuco-tucos usually emerge to the surface. Average earth surface “skin” temperature (T_skin) measured by satellite (Jin and Dickinson, 2010) was obtained from GLDAS_NOAH025_3H v2.1 dataset (Beaudoing and Rodell, 2020). The dataset is available in the Giovanni online data system, developed and maintained by the NASA GES DISC (Acker and Leptoukh, 2007). T_skin detects radiation emission from the earth surface, discounting the effect of atmosphere. This dataset collected data averaged every 3 h for each point of a grid with 0.25 degrees resolution (approximately 760 km²). T_e was estimated through a linear regression model as a function of T_skin and season (categorical variable with levels “summer,” “autumn,” “winter,” and “spring”). The resulting R² of 87% of the regression model was considered sufficient to replace T_e by the predicted values of the model (Supplementary Figure 2). The new temperature variable without gaps was named surface temperature (T_sur).

Statistical models can disentangle the relative contributions of candidate environmental factors that modulate biological variables in the field (Kowal et al., 1976; Bennett, 1987). Generalized linear mixed models were used to test the role of environmental temperatures in modulating tuco-tucos surface emergences during the photophase, in summer and winter separately (glmmTMB function of “glmmTMB” package; Brooks et al., 2017). Counts of surface episodes every 3 h were considered as the response variable. Surface episodes were defined as events in which light-loggers detected Illuminances higher than 2 lux, based on the minimum sensitivity of the loggers and on previous observations that light loggers do not detect light when tuco-tucos are inside tunnels. The following independent variables were included in both models: T_soil (soil temperature at 20 cm below ground); T_sur (surface temperature described above); Hour [5 categories representing 3 h blocks: A (06:00–08:55 AM); B (09:00–11:55 AM); C (12:00–02:55 PM); D (03:00–05:55 PM); E (06:00–08:55 PM)]. For winter data, Hour categories A and E were excluded, since more than 50% of these blocks corresponded to nighttime, when light loggers do not detect emergences. The final model was chosen from a biologically plausible initial set, based on lowest Akaike Information Criterion (AIC) values (Zuur et al., 2009). The variables included had variance inflation factors (VIF) (Zuur et al., 2009) lower than 5. A total of 20 extreme values of T_soil and T_sur (more than 2 standard deviations or less than −2) were underrepresented in the datasets and were excluded to avoid bias, resulting in a sample size of 973 observations from 12 animals in the current summer data and 227 observations from 9 animals in the winter data published previously (Jannetti et al., 2019).

For summer data, surface emergence during the photophase had significant temporal dependence (Potvin et al., 1990), i.e., outside the 95% confidence interval of the autocorrelation function (acf function from “forecast” package) (Hyndman et al., 2020). Since autocorrelation violates the assumption of independence of the model, it was accounted for by adding an auto-regressive structure of order 1 (AR-1, Zuur et al., 2009), modeled as a function of Hour and allowing for this effect to be different according to the animals’ ID number (ID). A random intercept according to ID was not included in the summer model due to its negligible variance (6.841e-08), as in Pasch et al. (2013).

Although our datasets had a much finer time resolution, the choice to use five blocks (factors) in the “Hour” variable took into account that a higher number of factors would increase the number of regression parameters to be estimated and this would diminish the significance of the model, unless we increased the sample size (Zuur et al., 2009). At the same time, using fewer blocks, e.g., day vs. night only, would not give us enough time resolution to probe the varying effects of T_sur and T_soil along the day. Alternatively, we could have analyzed each time block separately with an independent model, eliminating the need for an “Hour” variable. However, that would have worsened the estimation of the other variables’ effects, since the sample size to each analysis would have been five times smaller.

Freshly caught tuco-tucos were brought to the laboratory in summer and winter to evaluate aftereffects of natural entrainment. Captures were held in 2016 and 2017, between January 12th and February 1st for summer and between July 8th and 26th during winter, which implies that, in both seasons, the associated solstice had already occurred. After each capture, animals were taken to the laboratory (located at a walking distance), where they were weighed and sexed and then immediately released into constant conditions to enable measurement of aftereffects. The average mass of the 35 analyzed animals was 148 ± 55 g, including 16 females, 18 males, and one of undetermined sex. Running-wheel activity was measured under constant darkness (dim red lights—5 lux).

Data were plotted in actograms in the software El Temps (Díez-Noguera, 2020). In each actogram, activity onset and offset times were eye-fitted from the first 10 days in DD and lab-activity duration (α) was calculated as the difference between them. Significant differences were evaluated using Student’s T-Test.

To evaluate splitting in activity and temperature rhythms, nine tuco-tucos (6 females and 3 males) were subjected to the following conditions: first, LD 12:12 (lights on at 07:00) for 13 days; then, LL (200–300 lux) for either (i) 92 days (individuals #69, #99, #100, # 101, # 102, # 104, and # 106) or (ii) 145 days (individuals # 46 and # 52). The sample size was chosen based on previous reports of splitting occurrence in at least half of both male and female tuco-tucos exposed to prolonged LL (Valentinuzzi et al., 2009). Wheel-running and body temperature of all individuals were monitored at 5-min intervals. Data were plotted in actograms for visual analysis and chi-square periodograms (Sokolove and Bushell, 1978) were performed to evaluate periodicities in the data. Splitting events were recognized by visual inspection of the actograms when there was more than one activity phase per day and by the periodogram when periods close to 12 h were detected. Analyses were performed using the software El Temps (Díez-Noguera, 2020).

The two-oscillator model of the circadian clock proposed by Pittendrigh and Daan (1976c) suggests that the clock is composed of a morning (M) oscillator and an evening (E) oscillator. We simulated this model for the circadian clock of tuco-tucos, with Pavlidis-Pittendrigh oscillators. The equations were used in our previous studies of the entrainment of M and E oscillators to regular photoperiods (Flôres and Oda, 2020) and the splitting phenomenon in hamsters in constant light (Oda and Friesen, 2002). State variables and parameters of M and E are indicated in the equations below. The terms for each oscillator are identified by subscribed letters.

Morning oscillator (M):

Evening oscillator (E):

Briefly, R and S state variables describe the phase of the oscillator at each time point. R is prevented from assuming negative values (R > 0). a, b, c and d are fixed parameters that compose an oscillator configuration, with intrinsic period, amplitude and phase-response. We assigned the following configuration to the oscillators: a = 0.85, b = 0.3, c = 0.8, d = 0.5. L is the light variable set to 0 to represent darkness and changed to 1.1 arbitrary units for 1 h to make a light pulse. The non-linear term K smooths the numerical integration [K = k₁ / (1 + k₂R²), k₁ = 1, k₂ = 100]. Finally, C_EM indicates the coupling strength of the E oscillator onto the M oscillator and C_ME defines the opposite coupling strength. The coupling parameters were assigned a symmetrical value: C_EM = C_ME = 0.03. These model parameters were chosen based on our previous work with the equations (Flôres and Oda, 2020). They generate a model with a free-running period close to 24 h and that is responsive to skeleton photoperiods. Computer simulations were performed in the CircadianDynamix extension of the Neurodynamix II software (Friesen and Friesen, 1994), using numerical integration at 1,000 steps per 24 h, with the Euler method.

The M-E model was exposed to daily light-pulse regimens that mimic the light exposure patterns of tuco-tucos in the field. In each simulation, the oscillator system was left in constant darkness for 20 days, followed by the light pulse regimens for 50 days, and constant darkness again for another 20+ days. To mimic the irregular light exposure of tuco-tucos, daily light pulses were applied at random times, uniformly distributed in pre-defined time-intervals within the photophase. Random pulse times were generated in the software R (R Core Team, 2020), as described in the Supplementary Text. Three light pulse regimens of increasing complexity were defined, to probe different features of the tuco-tucos light exposure, as described below.

Model I—one single pulse per day that can occur at any random time throughout the photophase. It is the simplest light exposure model, with minimal information on the timing of the photophase.

Model II—two pulses per day, one constrained to the first half of the photophase and the other pulse to the second half. The extra pulse adds more timing information about the photophase.

Model III—two pulses per day each constrained to a 4 h-interval phase-locked to one of the simulated twilights. This is still simplified, but it adds yet more timing information on the photophase, by replicating a feature observed in tuco-tucos: twilight exposure in summer, and noon exposure in winter.

In each of the models, the photophase onset and offset times were then systematically varied to simulate short and long-day photoperiods, with photophase durations varying from 8 h (LD 8:16) to 16 h (LD 16:8). Each light pulse was constrained to input only the M or E oscillator, depending on whether the pulse occurred in the first or in the second half of the photophase, respectively. This strategy successfully replicated experimental results of nocturnal rodents exposed to “skeleton” photoperiods (Flôres and Oda, 2020).

The phase relationship between E and M (ΔΦ_EM) was calculated on each day during the light pulse regimens and also in constant darkness after the pulses, to verify aftereffects of entrainment to the different photoperiods. We first determined the reference phases for each oscillator, as the maximum of the state variable S within each cycle. To obtain the ΔΦ_EM on a given day, we measured the absolute difference between the time of the E reference phase and the time of the M reference phase. Mean and standard deviation of ΔΦ_EM were calculated in the last 20 days under each light pulse regimen.

From the 19 animals deployed in the field enclosures in summer, 13 were successfully recaptured (153 ± 39 g; 7 males; 6 females). Four of the recaptured tuco-tucos (169.7 ± 46 g; 2 males; 2 females) had collars containing accelerometers. Data from 9 animals in winter and 1 animal in autumn were taken from Jannetti et al. (2019) and used for comparisons between seasons.

Light-logger data indicated when the subterranean tuco-tucos were on surface during daylight and, consequently, the timing of their exposure to light. Daily surface emergences and light exposure occurred near the twilights during mid and late summer (Figure 1 and Supplementary Figure 3), contrasting with the pattern in winter (Flôres et al., 2016; Jannetti et al., 2019), when both were concentrated in the middle of the day (Figures 1, 2A). The two animals with longer deployment duration (Figure 1 #177 and #193) show the gradual transition between these two contrasting patterns during autumn.

Accelerometer data, on the other hand, informed the timing of general activity, either on surface or underground, based on its sensitivity to body movement. As seen in Figure 1, general activity is organized in activity bouts that occur both during the day and night, with highest levels during the day. This occurs for both summer and winter data. Furthermore, when accelerometer data are compared to light-logger data, surface emergence roughly coincides with 47% of the highest levels of general activity in autumn/winter (Jannetti et al., 2019). Using the same calculations, surface emergences occur in 20% of highest activity of tuco-tucos in summer. Daily field activity duration was significantly greater for summer animals (n = 3, current dataset) than for late autumn or early winter animals (n = 3, from Jannetti et al., 2019) (2.13 h difference between means; Student’s T-test t = −4.8, p = 0.02, df = 2.2).

We next used statistical models to verify the role of soil (T_soil) and surface (T_sur) temperatures in stimulating or inhibiting the surface emergence during the day of tuco-tucos in summer and winter. We fitted two Negative Binomial generalized linear mixed models, for summer and winter data separately. They were selected from putative models with different combinations of environmental measures (Table 1). In the final summer model, T_sur had a significant contribution to explain surface emergences, showing a negative correlation to the counts of emergence episodes during the photophase (estimated coefficient of −0.7 ± 0.1, p < 0.001), while T_soil had no significant effect. In the winter model, however, both T_sur and T_soil had significant contributions. T_sur showed a positive correlation with emergence episodes (0.5 ± 0.2, p = 0.003), and T_soil, a negative correlation according to Hour (time of day) (no significant effect at 09:00–11:55 AM; −0.5 ± 0.2 at 12:00–02:55 PM, p = 0.01; −0.4 ± 0.2 at 03:00–05:55 PM, p = 0.03).

These findings in the winter data agree with the analysis performed in Jannetti et al. (2019) in the sense that T_soil had a significant negative correlation with emergence episodes, although interaction between T_soil and Hour was not considered before. The inclusion of this interaction here did not improve the model greatly: compared to the winter model without T_soil-Hour interaction, the model with the interaction decreased less than 2 AIC units (Table 1; Bolker, 2007) and the difference between the models was only significant at the 5% level of significance (X²-test, p = 0.02). However, the T_soil-Hour interaction revealed that surface emergences do not correlate with T_soil in the morning, while the correlation is significantly negative in the afternoon. These results indicate that T_soil does not modulate the onsets of tuco-tucos surface emergences in the morning. On the other hand, they do not exclude the possibility of modulation of the offset of emergences in winter.

Contrary to the current winter results, our previous analysis of the autumn/winter data (Jannetti et al., 2019) had not shown an effect of surface temperature on the tuco-tucos surface emergences. Part of this divergence may be due to different techniques used to measure surface temperature (temperature logger vs. satellite measurements). Additionally, the divergent results may be due to the inclusion, in the previous work, of autumn and winter animals in the same model. Notably, our current analysis reveals a seasonal variation in the role of T_sur, with opposite effects in summer and winter. In this sense, pooling together autumn and winter data in the previous analysis might have caused an underestimation of the effects of surface temperature on tuco-tucos’ surface emergence (Jannetti et al., 2019).

Integrating all variables’ effects, surface emergence episodes in summer are less dependent on exogenous temperature conditions than in winter, since the summer crepuscular pattern can be deduced without the T_sur effect, as can be seen in the scenario of constant temperature in Figure 2B. Despite a significant T_sur effect detected in our analysis, there is a rhythmic component in the summer surface emergences (details in the methods), not fully explained by T_soil or T_sur effects. These results suggest that endogenous factors may have a more important role to generate this emergence pattern rather than other environmental variables not considered here.

Our data support the seasonal role of exogenous temperature in modulating surface emergences in tuco-tucos. However, the surface activity is only one behavioral component of the daily activity-rest rhythm, which also includes what the animals do underground throughout day and night. We observed a significant difference in the daily field-activity duration of the general activity, between summer and winter from the accelerometer data (Figure 1). In other words, not only surface activity but the gross activity (surface+underground) was shown to have different duration between the seasons. This seasonal change could be generated by masking effects of environmental cycles, but it could also be due to changes in the endogenous regulation by the circadian clock. Moreover, the analysis of surface emergences alone suggests an endogenous contribution to the patterns observed, mainly in summer. Thus, in the next experiment, we tested the role of endogenous circadian timing in the regulation of seasonal activity.

New sets of freshly caught animals were brought to the laboratory, to evaluate the endogenous contribution to the observed seasonal variations in activity patterns. We used a standard experiment in chronobiology which consists of releasing synchronized organisms into constant laboratory conditions, to eliminate exogenously generated rhythmicities. In the first days under this condition, before their rhythms start to free-run, they display “aftereffects” of previous entrainment (Pittendrigh and Daan, 1976a; Tackenberg et al., 2020). The aftereffects indicate rhythmic parameters that have been encoded in the endogenous circadian clock.

Animals captured during summer and winter were immediately released into DD conditions to evaluate aftereffects in the lab-activity duration (α). Representative actograms are presented in Figure 3, with the mean α of 5.4 h (n = 17) for summer and 7.0 h (n = 18) for winter animals (Student’s T-test t = −8.8907, df = 32.086, p-value = 3.621e-10). As an important remark, tuco-tucos are nocturnal in the lab and diurnal in the field (Tachinardi et al., 2017). For this reason, field-activity duration is inversely proportional to lab-activity duration α, as seen comparing Figures 1, 3, in summer and winter.

The results confirmed seasonal α differences when animals were free from any masking effect, in laboratory DD conditions. These α aftereffects indicate a seasonal change in the endogenous control of activity, reminiscent of season-dependent entrainment of the circadian clock by different photoperiods. Persistent aftereffects in α in DD are explained in terms of a slowly relaxing coupling between the component oscillators within the SCN. Thus, the following experiment tests the multi-oscillatory composition of the circadian clock of tuco-tucos.

We next evaluated the splitting phenomenon (Pittendrigh and Daan, 1976c) in tuco-tucos under LL to probe the two-oscillator composition of their circadian clock. Splitting occurred in seven out of the nine animals (77, 7%), displaying great variability of splitting patterns (Figure 4 and Supplementary Figure 4). The splitting happened gradually in three individuals (#46, #100, and #104) and abruptly in four (#52, #99, #101, #102). Preceding splitting, the free-running period of the rhythms (τ) shortened in two animals (#52, #100), lengthened in two (#102 and # 104) and remained stable in three (#46, #99, and #101).

Rhythmic patterns of wheel-running and body temperature were similar in all but one of the animals that experienced splitting (Supplementary Figure 4). The exception was individual # 52 in which wheel running became arrhythmic 10 days after the splitting occurred. During the days when wheel running was arrhythmic, a third component, with a τ of 25.2 h, could be detected in the body temperature rhythm and became undetectable after wheel running rhythmicity was reestablished.

The finding of multiple components in activity and temperature rhythms under LL is an evidence that the underlying circadian pacemaker of tuco-tucos is composed of multiple oscillators. Next, we use mathematical modeling to test whether a minimal two-oscillator model could account for the seasonal adjustments in activity α, when exposed to light regimens that mimicked the light exposure of tuco-tucos in summer and winter.

In mammals, the morning (M) and evening (E) oscillators (Pittendrigh and Daan, 1976c) are thought to be composing parts of the SCN. It is predicted that, as the photoperiod changes, the M oscillator tracks dawn and the E oscillator tracks dusk, resulting in adjustments of the phase relationship between them (ΔΦ_EM) (Figure 5). In turn, adjustments in ΔΦ_EM are reflected in activity α, since the SCN control the circadian rest-activity rhythm.

Our hypothesis is that the tuco-tucos seasonal light exposure acts on this photoperiod-encoding mechanism, generating the observed endogenous adjustments in the tuco-tucos daily activity α at different seasons. The hypothesis is supported by our current findings that (i) light exposure patterns in tuco-tucos are different under different photoperiods; (ii) seasonal aftereffects in α are seen in animals brought from the field directly into laboratory constant conditions, which suggests seasonal adjustments of the circadian clock; (iii) splitting of tuco-tucos activity/rest rhythms into two components, under constant light in the laboratory, indicates a multi-oscillatory circadian clock.

To test the hypothesis, two limit-cycle oscillators representing the M-E model were computer simulated (Figure 6A). The oscillators were exposed to simplified light cycles that mimicked the observed natural light exposure of tuco-tucos in winter and summer, with increasing complexity from Model I to Model III (Figures 6B–D, upper panels) (details in section “Materials and Methods”). Changes in ΔΦ_EM were used as a proxy for α and as an indication of the efficacy of these light signals in informing day length.

The dynamics of M and E under the different light exposure models are reported for 2 extreme photoperiods in Figures 6B–D (lower panels). Results for intermediate photoperiods can be found in Supplementary Figure 5. Despite the simplicity of Model I, with a single light pulse per day, it already presented a modest adjustment in ΔΦ_EM, in short days vs. long days (Figure 6B, lower panels). The average ΔΦ_EM was 1.6 h in the short “winter” days (LD 8:16), and 2.7 h in the long “summer” days (LD 16:8), with intermediate values for the other photoperiods (Figure 7A). Upon termination of the pulses, in constant conditions (DD), ΔΦ_EM gradually decreased to the steady state value (Figure 7B). The initial values in DD were reminiscent of the previous entrainment to the different photoperiods, which suggests aftereffect in ΔΦ_EM. The greater the initial value, the longer it took for the model to reach its steady state.

Tuco-tucos are exposed to more than one single light pulse per day in the field, thus, in Model II, we applied two pulses each day. The effect of photoperiod (Figure 6C, lower panels) was more pronounced than Model I, with ΔΦ_EM ranging from 2.4 h under short winter days to 5.2 h under long summer days (Figure 7A). When the pulses were turned off, in DD, there was once again a gradual adjustment of ΔΦ_EM, tending toward the steady-state null value (Figure 7B). In contrast to Model I, however, there were larger initial differences in ΔΦ_EM between the photoperiods, which resulted in a longer duration of summer aftereffects (LD 16:8) in Model II.

In the field data, summer light exposure is concentrated in the twilights, while winter light exposure is concentrated around noon (Figure 1). This feature could contribute even further to the photoperiodic regulation of the circadian system. To test this hypothesis, we devised Model III, in which the pulses are concentrated in 4 h windows locked-on to the twilights. In Figure 6D (lower panels) we can see a greater difference between summer and winter synchronization in Model III, compared to the other two models. ΔΦ_EM ranged from 2.4 h in the summer to 12.7 h in the winter (Figure 7A). These large differences were also reflected in DD, resulting in a much longer duration of the aftereffects in ΔΦ_EM (Figure 7B).

The acute effect of temperature on daily activity patterns has long been demonstrated for several diurnal rodent populations. Many rodent populations in nature present a bimodal activity distribution in summer, in contrast to a unimodal pattern in winter (Halle and Stenseth, 2000) and this seasonal change is attributed to avoidance of high midday heat loads in summer. Accordingly, summer activity is bimodal in Piute ground squirrels (Spermophilus mollis) and degus (Octogon degus) living in open habitats, as they retreat to shade spots or burrow entrances during midday. In contrast, those living in shrubby habitats, with access to the extensive shadows, still display activity in the summer midday (Sharpe and Van Horne, 1999; Kenagy et al., 2002; Bacigalupe et al., 2003). Similarly, summer activity is not bimodal in squirrel species with adaptive morpho-physiological and behavioral strategies that reduce heat load and increase heat dissipation (Chappell and Bartholomew, 1981; Fick et al., 2009). Together, these studies support the causal relationship between summer bimodal activity and the thermoregulatory pressure of high midday heat loads (Kenagy et al., 2002).

In contrast to the epigeous species that use burrows for temporary retreats from heat load (Hainsworth, 1995), subterranean tuco-tucos spend most of their time in the tunnels and just emerge to the surface in brief episodes, never wondering around (Tomotani et al., 2012). Thus, in principle, soil temperature could be more determinant than surface temperature in modifying activity patterns of subterranean rodents. Indeed, soil temperature was observed to modulate activity patterns in other subterranean rodent species (Rado et al., 1993; Vlasatá et al., 2017). Nevertheless, hot external temperatures have also been shown to curtail surface activity of subterranean and fossorial species in summer, generating a bimodal pattern of aboveground incidence (Rezende et al., 2003; Hinze and Pillay, 2006). We then hypothesized that both soil and surface temperatures could have a role in shaping emergence patterns of tuco-tucos during summer in a desert habitat.

We had previously shown differences in the time course of surface and soil temperature cycles in our study area, in northern Monte desert, due to the delay of daily heat flow in the soil (Jannetti et al., 2019). As a consequence, the highest soil temperatures occur at night when surface temperatures are the lowest, while, surprisingly, the lowest soil temperatures occur at noon when surface temperatures are the highest. In the previous autumn/winter records, surface emergence episodes occurred mostly at midday when not only it was warmest aboveground but also coolest below ground (Jannetti et al., 2019). The current results confirmed that, in winter, emergence episodes statistically correlate to lower T_soil and higher T_sur. In summer, however, no effect of T_soil was observed on emergence episodes. On the other hand, T_sur had a significant inhibitory role, although the crepuscular pattern of time on surface could be deduced without the T_sur effect. Thus, to some extent, the model confirmed the masking role of surface temperature, which partly explains avoidance of midday hours and concentration in twilight times, as pointed out for other species (Kenagy et al., 2002; Bacigalupe et al., 2003; Rezende et al., 2003; Hinze and Pillay, 2006; Vlasatá et al., 2017). Nevertheless, the magnitude of the temperature effect was potentially lower than that encountered for some epigeous species (Long et al., 2005). This suggests that other factors, such as variable foraging demands and social interactions (Amaya et al., 2021), may also have an important role modulating timing of surface activity in tuco-tucos. Interestingly, while epigeous species retreat to shelters as a reaction to extreme surface temperatures (Chappell and Bartholomew, 1981; Long et al., 2005), subterranean tuco-tucos refrain from exiting to the surface if aboveground temperatures are too high in summer, being a subtly distinct behavioral response.

Taken together, our results point to a significant role of the daily ambient temperature cycle in masking surface activity of tuco-tucos. However, our model projects more subtle temperature effects in summer than in winter. The lack of association of summer emergences with T_soil and the fact that crepuscular patterns can be deduced disregarding T_sur effects suggest that other factors, e.g., the endogenous control, should contribute to driving the observed temporal pattern of emergences.

The surface activity of a subterranean rodent is just one component of the general activity which also includes behaviors displayed underground, out of our sight. Accelerometers can detect body movements above or underground and they have revealed that, in both summer and winter, tuco-tucos display bouts of general activity throughout day and night (Figure 1). However, there is a predominant active phase during the day—an interval when activity bouts are concentrated—which allows definition of activity onsets and offsets. These allowed comparison of field-activity duration between summer and winter, which was significantly different. In other words, not only surface activity but the gross activity (surface+underground) was shown to have different onset and offset times between the seasons. The endogenous nature of this phenomenon was shown by releasing freshly captured tuco-tucos in winter and summer into constant laboratory conditions (Figure 3). They displayed clear aftereffects of natural entrainment, with summer and winter animals showing significantly different α. This is in accordance with previous reports in laboratory experiments with model organisms, using artificial LD cycles with different photoperiods (Pittendrigh and Daan, 1976a; Tackenberg et al., 2020). To our knowledge, this is the first time α aftereffects are demonstrated in organisms previously free-living in the field, as a result of natural entrainment in different seasons.

This result indicates that part of the seasonal changes in activity patterns are endogenous, likely due to different patterns of SCN entrainment by different photoperiods. The endogenous regulation means that we would likely observe seasonal changes in activity patterns even in the absence of the strong thermal constrains of the desert. Accordingly, seasonal variation of activity onset and offset times are also seen in species that face less dramatic temperature variations between summer and winter, like grassland rodents (Paise and Vieira, 2006; Pita et al., 2011). Photoperiod-dependent entrainment of the SCN is the very first step of photoperiod processing within the seasonal neuroendocrine physiology in mammals (see below). Thus, changes in endogenous activity α are not only a proxy for the SCN entrainment but may also indicate a preliminary photoperiodic response within the body.

In summary, our results support that tuco-tucos display seasonal differences in daily activity patterns, due both to endogenous and exogenous factors. The endogenous component is best shown by change in activity onset and offset times between summer and winter, indicating that tuco-tucos start activity time earlier and finish it later in summer, compared to winter. Within this endogenously regulated α interval, the animals emerge to the surface at random times, but the probability of exiting the tunnels increases with a season-dependent combination of subterranean and surface temperatures, which are the exogenous factors that shape their time on surface.

The seasonal variations in tuco-tucos’ time on surface generate different light exposure patterns. Our next step was to verify if the contrasting parameters of light exposure between summer and winter could account for photoperiod measurement in tuco-tucos, via mathematical modeling. The α aftereffects of natural photoperiod (Figure 3) and splitting of activity/temperature rhythms into two bouts under LL (Figure 4) set the stage for a two-oscillator model of the circadian clock for photoperiod encoding in tuco-tucos. In addition, our light-logger data (Figure 1) provided the basic parameters used to simulate tuco-tucos’ natural daily light exposure patterns in summer and winter.

Several mechanisms have been proposed for photoperiodic time measurement and the circadian clock plays a central role in one such proposition, namely the “internal coincidence model.” It proposes that different photoperiods are transduced in the form of internal reorganization among components of the circadian system (Pittendrigh, 1972; Tackenberg and McMahon, 2018). In mammals, this model is supported by photoperiod-induced changes in the daily electrical activity of the SCN (VanderLeest et al., 2007), which is likely mediated by the phase relationships among component oscillators within the nucleus (Evans et al., 2013; Yoshikawa et al., 2017; Olde Engberink et al., 2020).

In its simplest form, the “internal coincidence” within the circadian clock can be modeled as the two oscillators E (evening) and M (morning), proposed by Pittendrigh and Daan (1976c). The E-M model was developed to explain circadian activity rhythms in nocturnal rodents, including the splitting of activity rhythms under constant light conditions (Pittendrigh and Daan, 1976c), which we also see in tuco-tucos (Figure 4; Valentinuzzi et al., 2009). In this model, as the day length varies throughout the seasons, there is an adjustment of the phase relationship between E and M (ΔΦ_EM) (Figure 5). In our previous study, we used mathematical modeling to simulate this two-oscillator clock and how it transduces different photoperiods into changes in ΔΦ_EM, provided that “dawn” and “dusk” signals input separately on M and E oscillators, respectively (Flôres and Oda, 2020). The model successfully replicated the correlations between photoperiod, ΔΦ_EM and activity α. The same model has also replicated the splitting phenomenon in constant conditions (Oda and Friesen, 2002).

The former E-M simulations (Flôres and Oda, 2020) were performed under “skeleton photoperiods,” a minimal representation of photoperiod in artificial lighting experiments (Pittendrigh and Daan, 1976b; Olde Engberink et al., 2020; Tackenberg et al., 2020). It consists in two light pulses per day, applied at the times corresponding to lights-on and off in complete photoperiod regimens (Figure 5). In the present study, the E-M model was exposed to light inputs that mimic the light exposure of tuco-tucos in the field. In contrast to skeleton photoperiods, in the new light regimens the pulse times were distributed randomly within the photophase, not at a fixed time. In principle, these regimens inspired by tuco-tucos’ light exposure should carry even less information about the timing and duration of the photophase, posing a greater challenge to photoperiod encoding.

Three light exposure models were tested, with increasing complexity, and we verified their capacity to modify ΔΦ_EM as a function of photoperiod. Surprisingly, photoperiodic information was already conveyed by the simplest light exposure model (Model I). Expectedly, the light inputs became more effective in informing photoperiod as we added further complexity in Models II and III. The additional features in these models more closely resemble the light/dark pattern experienced by tuco-tucos in the field. Nonetheless, the models are still gross approximations to the rich natural LD cycles, which also include changes in light intensity and spectral composition. The effectiveness of the simplified light exposure models, even in the absence of these natural features, corroborates that tuco-tucos get much more temporal information from their light exposure patterns than the theoretical minimum needed to decode photoperiods.

From a functional perspective, the ability to decode photoperiod is relevant not only to the adjustment of seasonal activity, but also as an anticipatory cue to time seasonal physiology in general. In mammals, the photoperiod-induced changes in the SCN trigger downstream seasonal physiological responses (Goldman, 2001). One crucial output of the SCN in this photoperiodic signaling cascade is the rhythm of melatonin secretion in the pineal gland. In nocturnal mammals, photoperiod modifies equally the durations of activity α and of nocturnal melatonin release (Illnerová and Vanìèek, 1982; Elliott and Tamarkin, 1994), as a result of photoperiod-dependent entrainment of the SCN. Sequentially, the duration of nocturnal melatonin informs photoperiod to downstream neuroendocrine systems that control seasonal reproduction (Dardente et al., 2019; Nakane and Yoshimura, 2019). In the tuco-tucos (Ctenomys spp.), most species studied so far display a seasonal reproductive pattern (Fanjul et al., 2021). Moreover, laboratory studies of female reproductive status under artificial photoperiods provided some evidence of photoperiodic responsiveness in C. talarum (Fanjul and Zenuto, 2008). Identifying these seasonal reproductive events, as well as other seasonal physiological changes, may provide clear output phase markers for future experiments, which would confirm neuroendocrine responses to changes in photoperiod.

Putting it all together, tuco-tucos display seasonal differences in time on surface, which results in seasonal differences in light exposure, transduced into different photoperiods. Conversely, the photoperiodic signals feed-back on the internal organization of the LD-entrained SCN, contributing to the different seasonal activity patterns. Understanding the complementary regulation of seasonal physiology by endogenous clocks and temperature in natural settings is urgent, in view of disruptions to phenology across the globe, associated with global warming (Walker et al., 2019). This knowledge is needed to estimate vulnerability of a species to the growing misalignment between photoperiod and seasonal environmental conditions caused by climate change (Walker et al., 2019; Meyer et al., 2021).