We used 17 male Sprague–Dawley rats (Charles River, Germany) weighing on arrival 176–200 g, which corresponds to the age of 6–8 weeks and is the standard age in this type of experiment. The sex of animals has been chosen to enable comparison of the results with previous data from ours and other laboratories, which, in the overwhelming majority, have been generated using male rats (for review see (Rygula et al., 2018)). Animals were group-housed (four animals per cage) in an enriched environment (plastic pipes 25 cm long) under controlled temperature (21 ± 1°C) and humidity (40–50%) and a 12 h light/dark cycle (lights on at 7: 00 AM). Throughout the experiment, rats were mildly food restricted to 85% of their free-feeding weight (according to the normal growth curve recommended by the laboratory rodent supplier - Charles River Research Models and Services Catalogue) by providing 15 g of food pellets/rat/day (standard laboratory chow). The food restriction used in our study is a standard and broadly applied procedure in experiments using operant conditioning techniques allowing to maintain motivation and performance of the experimental animals (Rygula et al., 2013; Cieslik et al., 2022; Surowka et al., 2022). Water was always available ad libitum. All behavioural procedures were performed during the light phase of the light/dark cycle.

The PRL tests were conducted in operant conditioning chambers (Med Associates; St Albans, Vermont, USA) enclosed within a sound-attenuating box. Each chamber was equipped with a fan, house light, speaker, a food dispenser set to deliver a sucrose pellet (Dustless Precision Pellets, 45 mg; Bio-Serv, New Jersey, USA), and two retractable levers located at the sides of the feeder.

After the initial instrumental training described in detail elsewhere (Rygula et al., 2018) and upon reaching the initial training criterion of less than 7.5% omissions on each lever (i.e., less than 15% total omissions but equally distributed between the 2 levers) for 3 consecutive training days, the rats were trained in the PRL paradigm. In brief, each PRL training session lasted until the completion of 200 trials, and each trial lasted for a maximum of 22 s. The start of a trial was signalled by the house light, which remained on until the end of the trial. Two seconds after the trial had started, both levers were presented, and one of them was randomly assigned as the “correct” lever, which delivered a reward (one sucrose pellet) 80% of the times it was pressed. A press on the other lever, the “incorrect” lever, would result in a rewarding outcome only 20% of the times it was pressed. No response in 10 s triggered the intertrial interval (ITI) and was counted as an omission. During the ITI, both levers remained retracted, and the house light was turned off. The same ITI directly followed an unrewarded outcome, i.e., no reward on 20% of the “correct” and 80% of the “incorrect” lever presses. After every 8 consecutive “correct” lever presses (regardless of the outcome), the criterion for the reversal of the outcome probabilities was reached. The previously “correct” lever now became “incorrect” and vice versa. This pattern was followed until the end of the session. The PRL training phase was repeated daily until the individual animals achieved sufficient performance levels. The criteria to be met were a minimum of 3 reversals completed during 3 consecutive training sessions, with less than 15% omissions per session.

To measure rats’ sensitivity to NF (measured as the ability of animals to ignore infrequent and misleading lack of reward), their decisions were monitored on a trial-by-trial basis. Unrewarded outcomes on the “correct” lever, after which an animal decided to switch levers (probabilistic lose-shifts), were scored and expressed as a ratio of all unrewarded outcomes on that lever. Additional measured parameters included the proportion of all rewarded outcomes followed by a decision to stay with the lever that delivered them (win-stay behaviours), and the number of reversals completed during the test, which relies on the ability to both suppress previously rewarded action and engage in previously unrewarded actions, and was used as a measure of the animal’s performance (Bari et al., 2010).

After achieving a stable performance in the PRL test (a minimum of 3 reversals and less than 15% omissions in three consecutive sessions), animals were subsequently tested in 10 consecutive PRL tests over 10 days. Based on this “NF sensitivity screening,” the rats were divided using the median split into the NF-insensitive and NF-sensitive groups. The division was made based on the average ratio of lever changes following misleading unrewarded outcomes (probabilistic lose-shifts) made by the animals across all 10 screening tests. Because the results of our previous studies have clearly indicated that such a dichotomous categorization based on median split is well suited to investigate NF sensitivity as a stable and enduring cognitive trait in rats (Rygula and Popik, 2016; Noworyta-Sokolowska et al., 2019; Surowka et al., 2020; Noworyta and Rygula, 2021), this method of data analysis has been extended to the present research. The number of screening days following meeting the performance criterion was also based on the results of our previous experiments (Rygula and Popik, 2016; Noworyta-Sokolowska et al., 2019; Surowka et al., 2020; Noworyta and Rygula, 2021).

After decapitation, the brains were quickly removed, frozen on dry ice and stored at-80°C until processed. The tissue of the dorsal and ventral hippocampi was manually isolated using sterile tweezers by a person experienced in this type of procedure, and according to The Rat Brain Atlas (Paxinos and Watson, 1998). The structures were collected between coordinates from the bregma in mm-dHipp (CA1): AP ~ − 2.6 to −5.2 mm, ML ~ 0 ± 5 mm, DV ~ 3–4.5 mm; vHipp (CA3): AP ~ − 5.2 to −6.04 mm, ML ~4 ± 6 mm, DV ~4.5–9 mm (Paxinos and Watson, 1998).

The RNA Mini Kit (A&A Biotechnology, Gdańsk, Poland) was used for RNA extraction. The quantity of the RNA was checked with a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies Inc., Wilmington, DE, USA).

The High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Waltham, MA, USA, Life Technologies, Waltham, MA, USA) was used for reverse transcription into cDNA. RT–qPCR was performed by using Quant Studio 3 (Thermo Fisher Scientific, Life Technologies, Waltham, MA, USA) and TaqMan Gene Expression Assays (Applied Biosystems, San Francisco, CA, USA) for Htr1a (Rn00561409_s1), Htr2a (Rn00568473_m1), Htr2c (Rn00562748_m1), and Htr7 (Rn0056048_m1). The PCR conditions were described previously by Gawlinski et al. (2021). The relative level of mRNA was assessed using the comparative CT method (2^−ΔΔCt) and normalized to the level of hypoxanthine phosphoribosyltransferase 1 (Hprt1), a housekeeping control (Rn01527840_m1).

Total RNA (20 ng) and miRNA-specific stem–loop RT primers (Applied Biosystems, San Francisco, CA, USA) were used for the reverse transcription reactions of miRNA. The cDNAs were then synthesized with the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems, San Francisco, CA, USA) according to the manufacturer’s protocol. RT–qPCR was performed with TaqMan MicroRNA assays (Applied Biosystems, San Francisco, CA, USA) to analyse the expression of the following mature miRNAs: miR-16-5p (assay ID: 000391) and miR-15b-5p (assay ID: 000390). The relative level of miRNA was assessed using the comparative CT method (2^−ΔΔCt) and normalized to the level of the U6 small nuclear RNA (U6 snRNA). One sample from this analysis was excluded due to technical problems.

In the next step, the levels of proteins encoded by the genes that were differentially expressed in rats classified as NF-sensitive and NF-insensitive were measured using ELISA kits (Bioassay Technology Laboratory, Shanghai, China). Quantities of the 5-HT_2A (#CAT E1825Ra) and 5-HT₇ (#CAT E3324Ra) receptors were measured according to the manufacturer’s protocol. Duplicates of each sample and series of standards were transferred to ELISA plates. The absorbance was measured at a wavelength of λ = 450 nm using a Multiskan Spectrum spectrophotometer (Thermo LabSystems, Philadelphia, PA, USA). The concentration of proteins was calculated from a standard curve and expressed as ng/mg of protein. For total protein measurement, a bicinchoninic acid assay (BCA) protein assay kit (Serva, Heidelberg, Germany) was used.

The data were analysed using GraphPad Prism (version 8.0.1). The distribution of the experimental data was tested using the Kolmogorov–Smirnov test. The screening data were analyzed using one-way (for the whole cohort) or two-way (for the animals classified as sensitive and insensitive to NF) ANOVAs with repeated measures and the within-subject factor of test day (10 levels: test day 1 … test day 10) and between-subject factor of NF sensitivity (2 levels: sensitive and insensitive). For pairwise comparisons, the values were adjusted using the Sidak correction (Howell, 1997). In molecular studies, statistical analyses were performed using a t-Student test. Additionally a Pearson correlation coefficients were computed to assess the linear relationship between the mRNA levels of all investigated genes, and between the levels of miRNAs and the sensitivity to NF (proportion of lose-shift behavior). The tests of significance were performed at α = 0.05.

For the animals classified as NF insensitive, the average proportion of lose-shift responses following misleading NF ranged from 0.385 to 0.505, with an average of 0.454 ± 0.042. For those classified as NF sensitive, the average proportion of probabilistic lose-shift responses ranged from 0.523 to 0.740, with an average of 0.606 ± 0.013. The sensitivity to NF in both subgroups was stable across 10 consecutive screening days {nonsignificant screening Day x NF sensitivity interaction [F(9, 135) = 0.853, p = 0.569, Figure 1A]}. There were no significant inter-trait differences neither in the proportion of win-stay behaviours [F(1, 15) = 0.034, p = 0.977, Figure 1B] nor in the number of reversals [F(1, 15) = 0.353, p = 0.561, Figure 1C].

Because each RT–qPCR reaction was performed separately, and because analysis of the correlation between the relative mRNA levels of all analyzed genes revealed no statistically significant correlations between them (see Supplementary Table S1), the intergroup differences between the NF sensitive and NF insensitive animals were analyzed using separate t-tests. In the vHipp of rats classified as NF sensitive, the mRNA level of Htr2a was statistically significantly higher than in the vHipp of rats classified as NF insensitive (t = 2.886, df = 15, p = 0.011, Figure 2B). There was also a positive correlation between the sensitivity to NF (proportion of lose-shift behaviour) and the mRNA levels of Htr2a [r(15) = 0.670, p = 0.003]. There were no statistically significant intergroup differences in the mRNA levels of Htr1a (t = 1.551, df = 15, p = 0.142, Figure 2A), Htr2c (t = 1.969, df = 14, p = 0.069, Figure 2C), and Htr7 (t = 0.438, df = 15, p = 0.668, Figure 2D) in this region. The correlation between sensitivity to NF and the mRNA levels of Htr2c [r(14) = −0.048, p = 0.857], and Htr7 [r(15) = −0.048, p = 0.857] was not significant. Interestingly, the mRNA levels of Htr1a, turned out to be also positively correlated with sensitivity to NF [r(15) = 0.499, p = 0.041].

In the dHipp, the mRNA levels of Htr1a (t = 1.155, df = 15, p = 0.266, Figure 2E), Htr2a (t = 1.015, df = 15, p = 0.326, Figure 2F), and Htr2c (t = 0.413, df = 15, p = 0.685, Figure 2G) did not significantly differ between the rats classified as NF sensitive and NF insensitive. The level of Htr7 mRNA in the dHipp was significantly lower in rats classified as NF sensitive compared to their NF insensitive conspecifics (t = 4.685, df = 15, p < 0.001, Figure 2H). There was also a negative correlation between the sensitivity to NF (proportion of lose-shift behaviour) and the mRNA levels of Htr7 [r(15) = −0.529, p = 0.029]. The correlation between sensitivity to NF and the mRNA levels of Htr1a [r(15) = −0.200, p = 0.443], Htr2a [r(15) = −0.165, p = 0.528], and Ht2c [r(15) = −0.048, p = 0.857] was not significant.

After observing significant differences in mRNA levels, these differences were validated at the protein level. Statistically significantly higher levels of the 5-HT_2A receptors were detected in the vHipp (t = 3.045, df = 14, p = 0.009) of rats classified as NF sensitive compared to their insensitive conspecifics (Figure 2I). Surprisingly, no significant intertrait differences in the 5-HT₇ receptor levels (t = 0.941, df = 15, p = 0.362) were observed in the dHipp (Figure 2J).

In the next step, the potential epigenetic mechanism of the observed differences in the expression of mRNA encoding the Htr2a gene in vHipp was evaluated by detecting miRNAs (miR-16-5p and miR-15b-5p). The miRNA analyses revealed statistically significant lower levels of miR-16-5p (t = 2.498, df = 14 p = 0.0256, Figure 2K) and miR-15b-5p (t = 12.68, df = 14, p < 0.0001, Figure 2L) in the vHipp of rats classified as NF sensitive compared to their NF-insensitive conspecifics. There was also a negative correlation between the sensitivity to NF (proportion of lose-shift behaviour) and the levels of miR-15b-5p [r(14) = −0.725, p = 0.002]. The correlation between sensitivity to NF and the levels of miR-16-5p was not significant [r(14) = −0.390, p = 0.136].