We first conducted a Dose-finding experiment using meal schedule bS1vS2 to probe the effects of Ex on food intake in the two-session operant paradigm. Based on the results, we selected the optimal dose of Ex which reduced food intake without observable side effects and went on to investigate the hypothesised meal schedule dependent effects of Ex (see Fig. 1).

In the Dose-finding experiment the total daily energy intake significantly and dose-dependently decreased under Ex treatment compared to control conditions (F2.0,8.1 = 60.21, p = 1.3 × 10−5 using Vehicle control; F1.8,7.2 = 38.24, p = 0.0002 using Baseline control; see Fig. 2B). The Vehicle and Baseline control energy intake did not differ significantly from nutritional recommendations based on age and body weight, but the daily energy intake was significantly lower than Vehicle, Baseline, and also the recommended nutritional value for the 5 and 20 µg/kg doses of Ex, while it only marginally decreased for the 1 µg/kg dose (see Supplementary Table 1 for details).

Dose-dependent effects of Ex on food consumption (A) and on total daily energy intake (B). Effects of 1, 5 or 20 µg/kg Ex (x axis) on food consumption (y axis on A) in S1 (upper panel) and S2 (lower panel) and daily energy intake (y axis on B) of the two-session food intake paradigm compared to Vehicle control days (Vehicle, *: p < 0.05,**: p < 0.01) and no-treatment Baseline days (Baseline, #: p < 0.05, ##: p < 0.01). P-values are Dunnett-corrected for the 3 comparisons within each analysis. Colours indicate individual animals, points correspond to data from one session on A and one day on B. Black points and whiskers are condition averages and s.e.m. across animals, with the 6 Baseline and 3 Vehicle days shown separately. On the y axis of panel B inward pointing ticks show the daily energy intake recommendation for each animal (colour-coded) based on their current weight. On B, ® markers denote significant differences from nutritional recommendation values (®: p < 0.05, ®®: p < 0.01).

In S1, Ex significantly decreased food consumption in all three doses (against Vehicle: F1.6,6.5 = 26.36, p = 0.001; against Baseline: F1.8,7.2 = 22.82, p = 0.0009; see Fig. 2A, top; Supplementary Table 2, top), with close to zero consumption in the two higher dose levels (5 and 20 µg/kg). In S2 overall, Ex still caused strong, significant decrements in food consumption (against Vehicle: F1.7,6.8 = 22.86, p = 0.001; against Baseline: F1.9,7.7 = 20.01, p = 0.001; see Fig. 2A, bottom; Supplementary Table 2, bottom), however, the 1 µg/kg dose of Ex was not effective, and close-to-zero floor consumption levels were only reached for the highest dose of Ex (20 µg/kg).

After the 20 µg/kg dose of Ex, all the animals showed transient hypoactivity, coordination problems and gastrointestinal symptoms, while after the 5 µg/kg dose, side effects were observed only in one animal. For the 1 µg/kg dose, all animals showed normal behaviour and general activity.

We verified that food consumption did not differ significantly between the Baseline and the Vehicle control conditions, with no significant temporal trends in either (see ‘Control analyses of Vehicle and Baseline data’ section of the Supplementary Results).

Based on the results, we selected the 1 µg/kg dose of Ex which reduced food intake without observable side effects, and went on to investigate the hypothesised meal schedule dependent effects of Ex in all four possible meal schedule conditions (see Fig. 3).

The effects of 1 µg/kg Ex across meal schedule (panels from left to right) and treatment conditions (x axis) on food consumption (y axis, A) across sessions (lower and upper part of A), and on daily total energy intake (y axis, B). Pictograms represent meal schedule conditions. Colours indicate individual animals; coloured points correspond to data from one session on A and one day on B. Black points and whiskers are condition averages and s.e.m. across animals. Coloured tick marks on the edges of sub-panels on B show recommended energy intake of each animal calculated based on weight measured during the experiment.

In our previous study [34] validating and describing this paradigm, we found a palatability-driven anticipation effect: S1 food consumption in the two-session paradigm strongly decreased in the bS1vS2 condition in anticipation of the more palatable S2 meal type, compared to when the less palatable meal was offered in S2 (the bS1bS2 condition). Importantly, this palatability-driven anticipation effect was replicated in the present study (Vehicle – control condition in the present study, S2 meal type: F1,4 = 8.48, p = 0.044, Baseline – previous study data, S2 meal type: F1,4 = 13.80, p = 0.021; Control type × S2 meal type: F1,4 = 0.13, p = 0.73, see also Fig. 3). To facilitate comparison between the two studies, we include the data from the previous study as the no-treatment Baseline control condition (see also Methods) in addition to the placebo-treatment Vehicle control condition. In the Supplementary Results (‘Control analyses of Vehicle and Baseline data’) we show that the new Vehicle dataset is similar to the previously analysed ‘Baseline’ condition in the present study, except for a substantially weakened S1 meal type effect – an effect that is orthogonal to the focal anticipation effect described above.

Total daily energy intake significantly decreased under 1 µg/kg dose of Ex treatment relative to nutritional recommendations (d = −366 ± 84 kcal, t4 = −4.35, p = 0.012), Vehicle (d = −335 ± 62 kcal, t4 = −5.45, p = 0.0055) and Baseline controls (d = −309 ± 62 kcal, t4 = −4.99, p = 0.0075; see Fig. 3B and Supplementary Table 2). In contrast, energy intake in the Vehicle and Baseline conditions matched the nutritional recommendations (Vehicle relative to recommended: d = −31 ± 32 kcal, t4 = −0.97, p = 0.38; Baseline relative to recommended: d = −57 ± 30 kcal, t4 = −1.91, p = 0.13).

Under Ex treatment, food intake in S1 strongly decreased (main effect of Treatment against Vehicle: F1,4 = 29.28, p = 0.0056; against Baseline: F1,4 = 41.96, p = 0.003) to the same low level (from 119.0 ± 11.8 g in Vehicle to 29.9 ± 6.9 g in Ex) regardless of meal schedule condition (under Ex treatment, S1 meal type: F1,4 = 2.01, p = 0.23; S2 meal type: F1,4 = 1.14, p = 0.35; S1 meal type × S2 meal type: F1,4 = 0.35, p = 0.58), as also implied by the Treatment × S2 meal type interactions (against Vehicle control: F1,4 = 6.97, p = 0.058, against Baseline control: F1,4 = 9.72, p = 0.036). Thus, the palatability-driven anticipatory effect in S1 was completely erased by the 1 µg/kg dose of Ex.

In S2, the overall effect of Ex on food consumption was weak, also in accordance with the Dose-finding results (Treatment main effect against Vehicle: F1,4 = 2.37, p = 0.20, against Baseline: F1,4 = 1.52, p = 0.29). However, Ex had a slightly stronger effect on the consumption of the more palatable food (mean ± SE average Treatment effect: −15.4 ± 6.5 g against Vehicle and −16.2 ± 9.5 g against Baseline) compared to the effect for the less palatable food ( − 2.8 ± 5.9 g against Vehicle and −3.2 ± 6.7 g against Baseline) as supported by a Treatment × S2 meal type interaction (using Vehicle days as control: F1,4 = 12.37, p = 0.025; using Baseline days as control: F1,4 = 8.46, p = 0.044).

The characteristic pattern in S2 was the successive positive palatability contrast effect – a dramatic increase in food intake in the bS1vS2 meal schedule condition compared to all the remaining three meal schedule conditions [34]. Importantly, this pattern was fully preserved under Ex treatment (S1 meal type × S2 meal type under Ex: F1,4 = 8.94, p = 0.040), and was not moderated by Treatment (Treatment × S1 meal type × S2 meal type, Vehicle control: F1,4 = 2.50, p = 0.19; Treatment × S1 meal type × S2 meal type, Baseline control: F1,4 = 0.60, p = 0.48; S1 meal type × S2 meal type, Vehicle control: F1,4 = 4.94, p = 0.09; Baseline control: F1,4 = 9.53, p = 0.037).

Blood glucose levels (BGL) were measured during the four meal schedules both on the Vehicle and the Ex treatment days (see Fig. 4). BGL in general substantially dropped from pre-treatment to pre-S1 (F1,4 = 8.79, p = 0.041), change-score analysis – further analyses are baseline-adjusted models or raw values. (For more detailed statistics, see Supplementary Results.) During Vehicle treatment in S1 with very berry flavoured pellets offered, BGL increased (from pre-S1 to post-S1) by 0.76 ± 0.13 mmol/L, a significantly (Time × S1 meal type: t50.0 = 4.30, p = 8 × 10−5) larger increase compared to sessions when banana flavoured pellets were offered, where it stagnated ( + 0.06 ± 0.13 mmol/L). This effect was erased by Ex treatment (post-S1 minus pre-S1, very berry consumed, Ex treatment: +0.11 ± 0.18 mmol/L), which is not surprising given the similar low consumption rates across conditions under the effect of the drug.

Blood glucose levels (y axis) from the three sampling times (pre-treatment, pre-S1, post-S1, on x axis) in the four meal schedule conditions (panels left to right) under Vehicle and Ex treatment (green and orange, respectively).

Exenatide strongly decreased pre-S1 BGL specifically in the bS1vS2 meal schedule condition (d = −2.38 ± 0.25 mmol/L, t6.8 = − 9.7, p = 3 × 10−5, uncorrected contrast), and only to a smaller extent in the other conditions (d ≥ −0.73, t ≥ −3.5, p ≥ 0.006; pre-S1, Treatment: F1,4.2 = 41.50, p = 0.0025; Treatment × S1 meal type × S2 meal type: t50.4 = 3.50, p = 0.001). This effect substantially weakened in the post-S1 blood sample (Treatment × Time × S1 meal type × S2 meal type interaction in the main model: t50.0 = − 2.61, p = 0.012), though BGL was still significantly lower under Ex compared to Vehicle control treatment (post-S1, Treatment: F1,4.1 = 24.23, p = 0.0073).