The experimental protocols followed EU Directive 2010/63/EU guidelines for animal experiments and the care and use of laboratory animals, and were approved by the Ethical Committee (CREMEAS) under agreement number APAFIS#3809-2016012713151858.

The 18 hamsters tested in this experiment were adults (2-year old) born and bred in the laboratory breeding centre of the Institut Pluridisciplinaire Hubert Curien in Strasbourg, France. A balanced sex ratio of nine males and nine females was studied. To align with the solitary behaviour of this species, all individuals were housed individually in cages (dimensions 38.5 × 121.5 × 18 cm), which were enriched with nesting materials (cellulose and hay). Animals were maintained at a constant ambient temperature of 18 °C and a relative humidity of 55%. Animals of this centre are bred to be released in the wild for repopulation purposes and are reared under normal photoperiod conditions. Therefore, during the daytime photoperiod, fluorescent lamps emitting 944 lux at mid-height, were synchronised with the sunrise and sunset and with the natural progressive light intensities at this latitude. Kibbles (pellets 105; Safe, Augy, France, composed of 19.3% protein, 54.9% carbohydrates, 5.1% lipids, 4.2% cellulose, 5.0% minerals and 11.5% water) and water were available ad libitum. Hamsters were studied for 8 weeks after emerging from hibernation at the beginning of April 2016. Animals were manipulated using a box to move them from their cages to the experimental test arena in order to reduce stress and limit their contact with humans.

This experiment investigated whether ALAN modifies the foraging patterns of hamsters. To ensure that analogous suburban lighting conditions were successfully reproduced, a lux meter was used prior to testing to measure the light intensity a hamster would experience if 1) directly below the streetlight and 2) in the peripheral zone (at a distance of 15 m) of a streetlight in a suburban area. The distance between two street lamps in a Strasbourg periphery is an average of 35 m. The resulting intensities of 40 lux and 4 lux, respectively, were then reproduced in experimental conditions using light bulbs (Luxeon, type Rebel PC Amber-LXM2-PL010000, dominant wavelength 591.0 nm) that imitate the high-pressure sodium bulbs used in the current lighting system of Strasbourg, Alsace, France.

We built a test arena (dimensions 150 × 80 × 50 cm) composed of two compartments connected via a 10 cm x 10 cm opening, and placed it in a dark room (Figure 1A). An identical source of artificial light (an amber LED emitting at 0, 4 or 40 lux) was placed 80 cm above the floor of each compartment. One of the compartments was randomly chosen as the ALAN environment, or “illuminated compartment” (IC). The other compartment, the “dark compartment” (DC), remained unlit (i.e. 0 lux). At the opening between the compartments, on a perimeter of 5 cm², we measured 0.2 lux; this region was considered a “secure” area of the arena (Wolfe and Summerlin 1989, Kotler et al. 1991, Clarke 1983) because in the dark, hamsters are less visible to predators. At the beginning of each trial, the animal was removed from its cage and placed in the arena’s entrance opening located midway between the LEDs (i.e. between the two compartments).

The hamsters were tested in two different conditions (Figure 1B). A light intensity of 4 or 40 lux was applied in the IC in both conditions, but the food source differed. In the high-quality food (HQF) condition, two live mealworms (1.2 cm, Truffaut) were placed below the functioning LED in the IC. This food is considered to be highly palatable for this species (Yves Handrich, personal communication). A low palatability food (two kibbles that they were accustomed to consume) was placed in the DC. In the low-quality food (LQF) condition, two kibbles were placed in each of the compartments. Eight trials of 15 min each at 4 or 40 lux (Figure 1B) were conducted for each individual in HQF and LQF conditions in randomly assigned order to limit any experimental design bias related to the location of the illuminated compartment (left or right) and trial schedule (morning or afternoon). This trial duration was chosen to let individuals exploring all the arena and eating food without possibility of habituation. In addition, because we aimed to study the natural behaviour of hamsters encountering light in their environment, no acclimation period was considered.

Each 15-min trial was filmed using an infrared camera placed above the arena. Data were recorded by video monitoring. During the viewing, observers were blinded to the light intensity applied but not to the food quality condition because mealworms and kibbles are easy to distinguish. The time spent by individuals in IC and DC was calculated for each test. The proportion of time of foraging and grooming behaviour was recorded as well as the compartment in which the behaviour was performed was recorded. “Foraging” refers to any behaviour related to the active seeking or manipulation of food. Both mealworms and kibbles require several bites by hamsters in order to be entirely consumed. The term “grooming” refers to cleaning or scratching. Grooming behaviours were recorded because they could reflect individual stress level (Smolinsky et al. 2009). At the beginning of the testing period, none of the hamsters had previous experience of our experimental procedure.

Experiments were conducted between 9:00 am and 5:00 pm. For safety reasons, no testing was permitted by our administration during nights at the lab, so tests were conducted during the day. As this corresponds to the period when the hamsters would usually be asleep, we first ensured that individuals were sufficiently active. For 4 weeks, we filmed hamsters of a breeding room during the whole day and recorded their frequency of activities: mean hourly frequency of exploration was 2.45 ± 0.275; mean hourly frequency of foraging was 1.18 ± 0.202; mean hourly frequency of maintenance was 0.65 ± 0.085; with some individuals being quite active with 15 occurrences of exploration per hour, 12 occurrences of foraging per hour. Moreover, their frenzy observed in the presence of mealworms indicated they were sufficiently active during our trials. In addition, wild hamsters have been reported to display surface activity during the day in urban environments despite that they are nocturnal animals (Schmelzer and Millesi 2008).

All statistical analyses were performed using R.3.1.2 (R core Team 2015). To analyse the behavioural responses of hamsters exposed to ALAN, we ran generalised linear mixed-effect models (GLMM) using the R package ‘lme4’. Individual identity was included as a random effect to consider pseudo-replication (i.e., the same hamster was used in 4 trials : 2 in the HQF (1 per light intensity - 4 lux and 40 lux) and 2 in the LQF condition (1 per light intensity - 4 lux and 40 lux), Hurlbert 1984) in each model. Best-fitting models were selected via an information-theoretic approach using the R package ‘multcomp’ and its ‘dredge’ function. This package uses the corrected Akaike Information Criterion (AICc), the delta AIC (ΔAIC) and the Akaike weight (Burnham and Anderson 2002). When ΔAIC > 2, the decision-making rule applied was to select the model with the lowest AIC value. When ΔAIC < 2, we chose the model with the lowest number of parameters by applying the principle of parsimony (Suppl. material 1: Tables S1, S2).

The first analysis concerned the time spent in each light condition. We ran two sets of models to explain our variable of interest by separating the two light intensity conditions (4 or 40 lux) in either HQF or LQF conditions. The response variable was expressed as the time spent by individuals in each compartment of the arena divided by the total proportion of time of a trial. We investigated the ability of hamsters to discriminate between our two compartments according to the light intensity thanks to paired Student’s t-tests. We compared the time spent by each individual to chance level (50%: 450 out of 900 recorded seconds) using a binomial test.

The second analysis was the proportion of time spent engaged in behaviours (grooming or foraging). We ran two sets of models to explain our variables of interest in HQF and LQF conditions. The response variables were expressed as: (1) the time spent grooming in the illuminated compartment divided by the total time spent grooming, (2) the time spent foraging in the illuminated compartment of the arena divided by the total time spent foraging. The resulting proportions were considered a binomial distribution. ‘Light intensity’ (4 vs 40 lux) and ‘sex’ (male vs female) were fixed factors. Wald tests were performed to evaluate the effect of each fixed factor on these response variables.

The final analysis sought to determine the effect of food palatability on the time spent foraging in the illuminated compartment of the arena. We again considered the time spent foraging in the illuminated compartment divided by the total time spent foraging. The resulting proportion was considered a binomial distribution. ‘Light intensity’ (4 vs 40 lux), ‘food quality’ (HQF vs LQF) and ‘sex’ (male vs female) were fixed factors. Wald tests were again performed to compare the proportion of time of foraging in each condition and the effect of the light intensity on this variable.

To exclude a possible learning effect, only the first trial of each light intensity per food-quality condition for each hamster was considered in the analysis (N = 18 hamsters for each experimental condition). Analyses thus used data from four trials (LQF and HQF at 4 lux; LQF and HQF at 40 lux). We used Student’s t-test to ensure that the time of day (morning or afternoon) and laterality (illuminated compartment on the left or the right of the arena) had no effect on the proportion of time of foraging and grooming and excluded these variables from our analysis (grooming proportion of time: Student’s t-Test, P_{Time of day} = 0.765; P_Laterality = 0.158; Foraging proportion of time: Student’s t-test, P_{Time of day} = 0.114; P_Laterality = 0.275). We did not observe an effect of the ‘trial order’, considered as a fixed variable, in any model, so we decided to exclude this variable from our analyses.

In both HQF and LQF conditions, we found that individuals spent significantly more time in the DC than the IC when the IC was illuminated at 40 lux (Table 1). More precisely, we highlighted that individuals spent significantly less time in the IC compared to the chance level (Student’s t-test, t = -2.62, P-value = 0.018; t = -2.14, P-value = 0.044 respectively for HQF and LQF). We found no significant difference in the proportion of time spent in the DC and the IC when the IC was illuminated at 4 lux (Table 1). More precisely, we observed that individuals spent equal time in the IC compared to 50% (Student’s t-test, t = -0.350, P-value = 0.731; t = -1.50, P-value = 0.149 respectively for HQF and LQF).

The best-fitting model for the proportion of time hamsters spent grooming in the HQF and LQF conditions included only the light intensity (Suppl. material 1: Table S1 and S2, respectively). More precisely, individuals groomed 4.34 and 4.08% longer in the IC when it was illuminated at 4 lux rather than 40 lux for the HQF and LQF treatments, respectively (Figure 2 and Table 2).

The best-fitting model for the proportion of time spent foraging in the HQF condition only included the light intensity (Suppl. material 1: Table S1). Individuals foraged 8.26% longer in the IC at 4 lux rather than at 40 lux (Figure 3 and Table 2). For the LQF condition, the best-fitting model also only included light intensity (Suppl. material 1: Table S2), but unlike in the HQF condition, individuals foraged only marginally (2.07%) longer in the IC at 4 lux than at 40 lux (Table 2).

Considering the proportion of time of foraging in the IC, the best-fitting model solely included the food quality. Individuals foraged 12.5% longer in the IC when we placed a mealworm (HQF condition) rather than a kibble (LQF condition) below the light, whatever the light intensity (Wald’s test, z = 6.78, P-value < 0.001; Figure 4 and Table 1). Then, hamsters increased their visibility to predators (i. e. frenzied and risky behaviours) in presence of highly attractive food.