Acute caffeine differently affects risk-taking and the expression of BDNF and of adenosine and opioid receptors in rats with high or low anxiety-like behavior

Anxiety disorders are common psychiatric conditions with a partially elucidated neurobiology. Caffeine, an unspecific adenosine receptor antagonist, is a common psychostimulant with anxiogenic effects in sensitive individuals. High doses of caffeine produce anxiety-like behavior in rats but it is not known if this is specific for rats with high baseline anxiety-like behavior. Thus, the aim of this study was to investigate general behavior, risk-taking, and anxiety-like behavior, as well as mRNA expression (adenosine A2A and A1, dopamine D2, and, μ , κ , δ opioid, receptors, BDNF, c-fos, IGF-1) in amygdala, caudate putamen, frontal cortex, hippocampus, hypothalamus, after an acute dose of caffeine. Untreated rats were screened using the elevated plus maze (EPM), giving each rat a score on anxiety-like behavior based on their time spent in the open arms, and categorized into a high or low anxiety-like behavior group accordingly. Three weeks after categorization, the rats were treated with 50 mg/kg caffeine and their behavior profile was studied in the multivariate concentric square field (MCSF) test, and one week later in the EPM. qPCR was performed on selected genes and corticosterone plasma levels were measured using ELISA. The results demonstrated that the high anxiety-like behavior rats treated with caffeine spent less time in risk areas of the MCSF and resituated towards the sheltered areas, a behavior accompanied by lower mRNA expression of adenosine A2A receptors in caudate putamen and increased BDNF expression in hippocampus. These results support the hypothesis that caffeine affects individuals differently depending on their baseline anxiety-like behavior, possibly involving adenosine receptors. This highlights the importance of adenosine receptors as a possible drug target for anxiety disorders, although further research is needed to fully elucidate the neurobiological mechanisms of caffeine on anxiety disorders.


Introduction
Anxiety disorders, such as panic disorder, are common and debilitating psychiatric conditions affecting up to one-third of the adult population across their lifetime (Kessler et al., 2005;Penninx et al., 2021).First-line treatments include cognitive behavioral therapy and prescription of pharmacotherapeutics such as selective serotonin reuptake inhibitors (SSRI) and serotonin norepinephrine reuptake inhibitors (SNRI) (Baldwin et al., 2011;Domenici et al., 2019), only a small majority responds to treatment and relapse is common (Baldwin et al., 2011).Furthermore, benzodiazepines are sometimes used to dampen anxiety, but their use is hampered by unwanted side effects and the short and long-term effects are under question (Martin et al., 2007).Hence, there is a growing need for new pharmacotherapies.Several different neurotransmitter systems have been evaluated with regards to anxiety, including adenosine receptor signaling (Domenici et al., 2019;Klevebrant and Frick, 2022;van Calker et al., 2019).Adenosine receptors are involved in several physiological functions, and have been suggested as interesting drug targets in a variety of diseases, such as inflammatory diseases, cancer, and neurodegenerative diseases, in addition the adenosine A2A receptor (A2AR) antagonist istradefylline was recently approved as an add-on treatment towards Parkinson's disease (Ijzerman et al., 2022).The neuromodulators of the brain are involved in synaptic plasticity and neurotransmitter regulation, including dopamine release (Ijzerman et al., 2022;van Calker et al., 2019).The adenosine receptor antagonist caffeine (1,3,7-trimethylxantine) stimulates the central nervous system (Fredholm et al., 1999) and has long been implicated in anxiety.High doses of caffeine induce anxiety and anxiety-like behavior in both humans, prominently females and animals (Guillen-Ruiz et al., 2021;Paz-Graniel et al., 2022).Certain individuals are more sensitive to caffeine-induced anxiety, for example, patients with panic disorder are more prone to experience caffeine-induced panic attacks and anxiety (Klevebrant and Frick, 2022).The exact mechanism underlying these effects is not elucidated, it is known that caffeine binds with strong affinity to the adenosine A1 receptors (A1R) and A2AR, and with low affinity to A2B and A3 receptors (Borea et al., 2017).The effect of caffeine on the A2AR works in an antagonistic manner regulating synaptic plasticity such as long term potentiation (LTP), and on the A1R caffeine's antagonistic effects regulates basal synaptic transmission (Lopes et al., 2019).The adenosine receptors expression includes several regions of the brain, for example amygdala, caudate putamen, frontal cortex, hippocampus, and hypothalamus (Svenningsson et al., 1997;Villanueva-Garcia et al., 2020).These brain regions are among other things associated with anxiety and several regions are shown to be affected by caffeine treatment.For example, rats consuming 0.3 g/l caffeine in drinking water for 3 weeks had altered LTP in hippocampus, amygdala, and frontal cortex compared to control (Lopes et al., 2023).In bulbectomized mice the A1R expression in frontal cortex increased when consuming 0.3 and 1.0 g/l caffeine 12 h/day for seven weeks, while the A2AR expression increased in the hippocampus and caudate putamen (Machado et al., 2020).Adult rats exposed to restrained stress for 14 days had an upregulation of A2AR receptor binding in prefrontal cortex and hippocampus (Dias et al., 2021), and lower levels of A2AR are documented in the paraventricular nucleus of hypothalamus in outbreed high anxiety rats (Niedzielak et al., 2020).In addition, several other signaling systems are connected to adenosine receptor signaling, and anxiety-like behavior, including dopamine D2 receptors (Ferre et al., 1997;Fuxe et al., 2005), and opioid receptors (Bailey et al., 2002;Hang et al., 2015;Van't Veer and Carlezon Jr., 2013).Caffeine also have a documented effect on neural activity, possibly via c-fos (O'Neill et al., 2016;Savchenko and Boughter Jr., 2011), brain derived neurotrophic factor (BDNF) (Ardais et al., 2014), and insulin-like growth factor 1 (IGF-1) pathways (Du et al., 2018).
With the growing problem of anxiety disorders in mind, it is paramount to improve the treatments.For this, experimental models are important to further elucidate the underlying neurobiology and identification of potential treatment targets.One of these targets may be adenosine receptors.In humans and animals alike, anxiety levels and behavior differ between individuals.In rodents, several studies have used the elevated plus maze (EPM), the most common test used for quantifying anxiety-like behavior in rodents (Pellow et al., 1985), as a method to identify animals with contrasting anxiety-like behavior (Borta et al., 2006;Ho et al., 2002;Lundberg et al., 2019;Naslund et al., 2015;Pawlak et al., 2003;Vautrin et al., 2005).The test is based on the animal's natural behavior to avoid open areas and the proclivity towards more sheltered spaces (Walf and Frye, 2007).Here, the EPM task was used to categorize the animals and to study behavioral effects of caffeine treatment in individuals with varying degrees of anxiety-like behavior.In addition, the multivariate concentric square field (MCSF) test, a multivariate test for behavioral profiling in rats (Meyerson et al., 2006), was used to study the caffeine-induced effects on general behavior, risk taking, and shelter seeking behavior in rats expressing high or low baseline anxiety-like behavior.Caffeine-induced changes in risk-taking has rarely been studied, previous studies have described risk-taking to increase in humans consuming energy drinks containing caffeine (3.57ml/kg (1.2 mg caffeine) energy drink at four sessions) (Peacock et al., 2013) and adolescent consumption of caffeine is correlated with certain high-risk situations (1 or 2 mg/kg caffeine at three sessions) (Temple et al., 2017).However, adults given 220 mg caffeine when intoxicated reduced their desire to continue drinking (Heinz et al., 2013), and a recent study by Liley et al. found no impact of caffeine on risky decision making in rats (10, 17, 30 mg/kg, administered every other day for 8 days) (Liley et al., 2022), but they did not consider individual differences in baseline anxiety-like behavior.
Altogether, evidence suggests that caffeine affects individuals differently depending on their anxiety state, but caffeine's precise role and the possible involvement of adenosine receptors need to be further evaluated.Thus, the aim of the present study was to, in rats categorized as having a high or low anxiety-like behavior, investigate risk-taking, shelter seeking, and anxiety-like behavior, with and without caffeine treatment.The aim also included to study changes in mRNA expression of selected genes adenosine A2A and A1 receptors, dopamine D2 receptor, c-fos, opioid μ, κ, λ receptors, BDNF, IGF-1, in amygdala, caudate putamen, frontal cortex, hippocampus, hypothalamus, regions associated with anxiety-like behavior and risk-taking.

Animals
Male Wistar rats (n = 72) were obtained from Envigo (Horst, Netherlands), 13 weeks old, 338-432 g at the beginning of the study, and were put under a reversed light regimen (lights off at 07:00, lights on 19:00).The rats were placed in groups of three per cage in type VI cages, bottom area of 1820 cm 2 , with raised cage lid, a wooden house, and food and water provided ad libitum.Temperature was controlled to 20 ± 1 • C with 40 ± 10 % humidity.Rats were allowed to acclimatize 12 days prior to be exposed to behavioral testing, during this time the animals were handled multiple times by the test executioner.The rats were weighed two days a week, with the addition of weighing prior to injection.This study was approved by the local Uppsala animal ethics committee (Dnr 5.8.18-08450-2020).All animal handling were in accordance with Swedish rules and guidelines regarding animal experiments (Animal Welfare Act SFS1998:56) and the European directive (86/609/EEC).

Behavioral testing
All behavioral testing was performed 8:30-14:30 Monday to Sunday.The rats were transported to the behavior room using a small cage with a towel over for light protection.On experimental day 1 the rats first underwent the EPM screening test and were subsequently divided into three groups based on their open arm time in the EPM with equal distribution (n = 24) in each group (Fig. 1).The two extreme groups were retained for further analyses, with low open arm time (LOAT) translating to high anxiety-like behavior, and high open arm time (HOAT) indicating low anxiety-like behavior (Borta et al., 2006;Ho et al., 2002;Naslund et al., 2015).Each group was then divided further into one caffeine-treated and one control group giving in total four groups with twelve rats in each group.On experimental day 21, the rats were subcutaneously injected with 50 mg/kg caffeine (Caffeine sodium benzoate 1:1; Sigma Aldrich, Darmstadt, Germany), dissolved in saline, or vehicle, 50 mg/kg sodium benzoate (Sigma Aldrich) dissolved in saline, 30 min prior to the MCSF test.The dose based on earlier studies (O'Neill et al., 2016;Paz-Graniel et al., 2022) was selected to obtain a high acute dose of caffeine.On experimental day 28, one week after the MCSF test, the rats were subjected to a second injection of caffeine sodium benzoate or vehicle and tested in the EPM test.Two hours after the second injection, animals were euthanized by decapitation.Brain regions (amygdala, caudate putamen, frontal cortex, hippocampus, and hypothalamus) were dissected and immediately placed on dry ice.The brain regions were selected due to their expression of adenosine receptors and their suggested involvement in anxiety disorders.The brain regions were kept at − 80 • C until further analysis.Trunk blood was collected in heparincoated tubes, centrifuged a 4 • C, 3000 rpm, for 10 min and plasma was collected and kept at − 80 • C until qPCR and ELISA were performed.All doses were injected in a volume of 0.5 ml/kg.

Elevated plus maze (EPM)
The EPM had four arms, two open arms (40 × 10 cm, length × width) and two closed arms (40 × 10 × 40 cm, length × width × height) in a plus shape with the closed arms facing each other and likewise for the open arms, with a 10 × 10 cm center connecting them.The EPM was elevated 50 cm from the floor and illuminated with 30 lx.The maze was monitored with a video camera from above.Prior to each session an outof-test rat was placed in the EPM arena to evade any first-in-line effects.All rats included in the experiment were allowed to explore the EPM arena for 5 min, and the EPM was cleaned with 10 % EtOH and let dry between each trial session.All trials were recorded and the tracking software Ethovision XT version 15 (Noldus Inc., Wageningen, Netherlands) was used to track parameters in the EPM i.e., open and closed arm time (s), frequency to each arm, distance moved (cm), and velocity (cm/s).Manual scoring of a person blinded to the rat's treatment was performed for head dips, rearing, grooming, and SAP (stretch attend posture).

Multivariate concentric square field (MCSF)
The multivariate concentric square field is a multivariate test designed to test several behaviors simultaneously, such as general activity, explorative activity, risk assessment, risk taking and shelter seeking behavior (Fig. 2).It was developed by Meyerson et al. in 2006 and used as described (Meyerson et al., 2006).It consists of an outer arena 100 × 100 cm; and an open center area 70 × 70 cm with surrounding corridors on three of the sides, referred to as the south, north, and west corridor, and a highly illuminated elevated bridge on the fourth side.The south corridor is linked to a dark corner room in one end, the other two corridors (west and north) meet in the hurdle at the next corner, and the north corridor is further linked to the slope leading up to the bridge in the other end.For a more detailed presentation see (Meyerson et al., 2006).All rats were tested in the MCSF individually on experimental day 21, 30 min post injection of 50 mg/kg caffeine or vehicle.Prior to each session an out-of-test rat were exposed to the MCSF arena to evade any first-in-line effects.Thereafter each rat included in the experiment was placed in the center face towards the wall without openings and allowed to explore for 20 min.Between each trial session the arena was cleaned with 10 % EtOH and let dry.
A camera placed above the test arena recorded each trial, and automatic tracking with Ethovision XT version 15 was used to obtain the following parameters: duration for each zone (Duration, s); latency to first time in each zone (Latency to first, s); frequency for each zone (Frequency); distance moved in the arena (cm); velocity in the arena (cm/s), and latency to leave the center (s).Manual scoring of rearing, grooming and SAP was performed by a person blinded to the rat's treatment.Nose pokes in the hurdle were recorded by a photocell, fecal boli and urination were counted after each animal.Calculated parameters were duration per visit (Duration/Frequency, s), total activity (sum of all zone frequencies), and Duration, Frequency, and Duration/Frequency for all corridors combined (total corridors).The risk shelter index is the ratio of time spent in risk versus shelter areas of the MCSF arena.The bridge is considered a risk area due to the uncomfortable floor (steel mesh), its elevation, and the bright light (600 lx).The dark corner room is presented as a shelter area due to the similarity to the rat's natural habitat, such as small and dark confinements (Meyerson et al., 2006).The behavioral parameters of most interest for anxiety-like behavior were those earlier defined as shelter seeking (duration, frequency and duration/frequency for dark corner room), risk taking (frequency, duration, and duration/frequency for bridge and central circle), and the relation between these (Alsio et al., 2009).Meyerson et al. (2013) developed a method (trend analysis) to categorize and sort the large number of descriptive parameters that the MCSF provides, the parameters are correlated to functional categories i.e., general activity, explorative activity, risk assessment, risk taking, and shelter seeking.For each descriptive parameter the individual rats (independent on treatment group) are ranked against each other.The ranking procedure enables a normalization for each parameter and they can be grouped together.Thereafter, Fig. 1.Timeline of the study.Arrival on day − 14, experimental day 1 the rats ran the Elevated plus maze (EPM), based on this the rats were divided in to groups based on high or low open arm time, on day 21 the rats were injected with a subcutaneous dose of caffeine 50 mg/kg, and then ran through the Multivariate concentric square field (MCSF), on experimental day 28 the rats were again injected with caffeine and ran in the EPM for a second time, 2 h after last injection the rats were euthanized.the individual rank values for the descriptive parameters of a functional category are summed, resulting in group-wise comparison based on the animal's relative position to the entire tested population.The parameters chosen for the different categories are: for General Activity: total activity, frequency of all corridors, frequency center, duration/frequency of all corridors (inversely ranked); for Explorative Activity: duration center (inversely ranked), hurdle, and for all the corridors combined (inversely ranked), rearing; for Risk Assessment: frequency for slope and bridge entrance, duration for slope and bridge entrance, duration/frequency slope and bridge entrance.;for Risk Taking: frequency, duration, and duration/frequency for bridge and central circle; and for Shelter Seeking: duration, frequency and duration/frequency for dark corner room.

RNA extraction
Qiagen RNeasy Mini kit (Qiagen, Hilden, Germany) was used according to instruction to extract RNA from frozen brain tissue.Brain sections were lysed in QIAzol Lysis reagent and chloroform was added in order to create a homogen solution followed by centrifugation to obtain two phases where the upper phase transferred to a Rnase free spin column to clean and elute the sample.The RNA concentration was measured using a NanoDrop® ND-1000 Spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE, USA).

cDNA synthesis
cDNA was obtained from the RNA using iScript cDNA synthesis kit (Bio-Rad, Hercules, California) containing 250 ng RNA, 5 × iScript reaction mix, iScript reverse transcriptase, and water, with a total volume of 20 μl.Positive and negative controls (without reverse transcriptase) were also conducted.

Quantitative polymerase chain reaction (qPCR) SYBR Green
The gene expression of the adenosine A1(A1R, Adora1) and adenosine A2A receptor (A2AR, Adora2a), dopamine D2 receptor (D2R, Drd2), proto-oncogene c-fos, brain-derived neurotrophic factor (Bdnf), insulin growth factor 1 (IGF-1), and the opioid receptors μ (Oprm1), δ (Oprd1), and κ (Oprk1), were evaluated using qPCR.These genes were selected due to their suggested involvement in anxiety-like behavior, and caffeine signaling.The samples were divided on two 96 w/p (Brand) due to the number of samples.Duplicates were made for all samples including controls.Positive controls were added in order enable comparison over the plates for each gene.A qPCR master mix was made with a volume of 23 μl per reaction containing iQ SyBr Green Supermix (Bio-Rad), 29 μM forward primer, 20 μM reverse primer, and Rnase free water with the addition of 2 μl (2.5 ng/μL) cDNA sample for each well.
Primer sequences (Table 1) were obtained from Primer-BLAST (NCBI, Maryland, USA) and validated in silico using UCSC genome browser.The CFX96 real-time PCR detections system, version 3.1 (Bio-Rad) was used to obtain the gene expression with the following cycle: 95 • C for 3 min, followed by 40 cycles of 95 • C for 15 s, 60 • C for 20s, 72 • C for 10 s, a melt curve was incorporated after each run, in order to ensure specific amplification.The software LinRegPCR was used for calculations of amplification efficiency (mean value and standard deviation for each gene) and the software qBASEplus software, version 3.2 (Biogazelle, Gent, Belgium), was used to analyze Cq-values obtained from CFX imaging software, in addition to three reference genes: ribosomal protein 19 (Rpl19), actin beta (Actb), and ribosomal protein large (Rplp0).The reference genes were selected based on evaluation from GeNorm (a part of qBASEplus), to obtain normalized gene expression levels.

ELISA
An Abcam corticosterone kit (ab108821) was used to examine corticosterone concentrations in plasma.Briefly, the kit components were equilibrated to room temperature and buffers were prepared.Standards were prepared and samples were diluted 1:100, all standards, controls, and samples were assayed in duplicates.The mean absorbance was detected at 450 nm and 570 nm using the FLUOstar Omega Spectrophotometer (BMG Labtech, Germany).

Statistical analysis
Data from the behavior tests are expressed as median ± interquartile range.The Shapiro-Wilk test for normality was performed on all data.For non-normally distributed data, Kruskal Wallis test with Dunn's multiple comparisons test was used to compare groups, comparisons were made for the following pairs: HOAT treated and control; LOAT, treated and control; HOAT treated vs. LOAT treated; and HOAT control vs. LOAT control, to resemble column and row factor in a two-way ANOVA.A significance level of p = 0.05 was determined for all statistical tests.The data was managed using Microsoft Excel version 16.48 (Microsoft, Redmond, Washington, USA) and all statistical analysis were performed in GraphPad Prism version 9 software (GraphPad Software Inc., USA).
The EPM data are presented as % of total activity (sum of all frequency) and % of total time 5 min, as well as the distance moved in cm and the velocity in cm/s.The frequencies for the MCSF parameters are presented as number of times entered to each zone, and the duration is the number of seconds spent in that zone.
Data from the qPCR gene expression is presented as mean ± standard deviation, values were excluded where an error had occurred during the preparation of the samples thus presenting incorrect values.Two-way ANOVA followed by Šídák's multiple comparisons test were performed, comparing row, column, or both factors, with anxiety level as row factor and treatment as column factor.The data from the ELISA assay were analyzed using two-way ANOVA followed by Šídák's multiple comparisons test, with the same comparisons as mentioned above.A significance level of p = 0.05 was determined for all statistical tests.The data was managed using Microsoft Excel version 16.48 (Microsoft) and all statistical analysis were performed in GraphPad Prism version 9 software (GraphPad Software Inc.).

Rats categorized into groups of high or low anxiety-like behavior based on their time spent in the open arms
The rats were categorized into three groups, based on their time spent in the open arms in the EPM (Section 2.2.1), designated as having a high (LOAT group), intermediate, or low (HOAT group) anxiety-like behavior (Fig. 3).The open arm time ranged from 1 % to 11 % for the LOAT group and from 21 % to 67 % in the HOAT group.The intermediate group, with an open arm time from 11 % to 21 % was excluded from further analyses.The time spent in the open arms differed significantly (F(2, 69) = 7.255, p < 0.0001) between the groups with multiple comparison specifying the difference between the HOAT and LOAT groups to p < 0.0001.

Caffeine decreases risk-taking behavior in rats with a high anxietylike behavior
On experimental day 21, the rats were treated with either 50 mg/kg caffeine or control substance, and tested in the MCSF.The descriptive parameters analyzed in the MCSF are presented in Supplementary Table 1.Using trend analysis, a significant difference in the functional category risk taking (Kruskal-Wallis test, H(3) = 12.25, p = 0.0066) was found, and following post hoc test identified a difference between HOAT and LOAT caffeine groups (Dunn's multiple comparisons test, p = 0.0256), and LOAT caffeine and control groups (Dunn's multiple comparisons test, p = 0.0084) (Fig. 4A) where the LOAT caffeine group evidenced decreased risk taking in comparison to HOAT caffeine and LOAT control.The functional category general activity presented a significant difference (Kruskal-Wallis test, H(3) = 13.77,p = 0.0032), identified between LOAT caffeine and control group (post hoc, p = 0.0245) (see Fig. 4B), where LOAT caffeine animals had a lower general activity than its control.The remaining functional categories presented no significant differences (Fig. 4C-E).

Effects of caffeine on behavior in the EPM
28 days after the first EPM trial the animals were injected with 50 mg/kg caffeine or control substance and exposed to a second EPM trial for 5 min.Kruskal-Wallis test (see Fig. 6A) yielded a significant difference (H(3) = 9.258, p = 0.0261) between the groups in their duration of the open arms, but post hoc-tests showed no differences between specific groups.The time the rats spent in the center differed significantly (H(3) = 14.97, p = 0.0018), with the caffeine-treated HOAT group spending longer time in the center than the caffeine-treated LOAT group (Fig. 6B) (Dunn's multiple comparisons test, p = 0.0420).The distance moved showed a significant difference, Kruskal-Wallis test (H(3) = 23.43,p < 0.0001) (Fig. 6C), Dunn's multiple comparisons test stated a difference between HOAT caffeine and control (p = 0.0035), and LOAT caffeine and control (p = 0.0021) in the EPM where the caffeine treated groups moved a longer distance than their controls.There was no significant difference between the groups within the parameter closed arm time (H (3) = 3.937, p = 0.2683) (Fig. 6D).Open arm frequency (Fig. 6E) and closed arm frequency (data not shown) did not show any significant group differences.The velocity in the EPM was significantly different between the groups (Kruskal-Wallis test, H(3) = 23.43,p < 0.0001) (Fig. 6F), Dunn's multiple comparisons test stated a difference between HOAT caffeine and control (p = 0.0035) and LOAT caffeine to control (p = 0.0021) where the caffeine treated rats had a higher velocity than its corresponding control.Thus, the animals receiving caffeine moved more and faster in comparison to their controls.

Effects on mRNA expression after caffeine administration in rats with high and low anxiety-like behavior
The mRNA expression of A1R, A2AR, D2R, c-fos, BDNF, IGF-1, Oprm1, Oprd1, and Oprk1 was assessed in the brain regions amygdala, caudate putamen, frontal cortex, hippocampus, and hypothalamus, and the results from the two-way ANOVA are presented in Table 2.In the amygdala, a significant difference of c-fos gene expression was found for both anxiety level and treatment group, however the post hoc test could not specify the difference in either treatment or anxiety level.In the frontal cortex, the genes BDNF and c-fos presented a significant difference (treatment factor respectively both factors), however, the difference could not be specified in either of the genes using the post hoc test.In the hippocampus, significant differences were detected for A2AR (anxiety level, F(1, 43) = 5.009, p = 0.0304), c-fos (treatment, F(1, 44) = 14.53, p = 0.0004), Oprm1 (treatment F(1, 44) = 8.200, p = 0.0064) and BDNF (anxiety level F(1, 44) = 11.03,p = 0.0018).The post hoc test detected a higher gene expression of c-fos in the LOAT caffeine rats compared to LOAT controls (p = 0.0061), for BDNF the HOAT and LOAT caffeine treated groups differed (p = 0.0190), and for Oprm1 the HOAT groups differed between caffeine and control (p = 0.042).The significant difference of A2AR could not be defined in the post hoc test.Regarding the remaining genes in the amygdala, frontal cortex, and hippocampus no significant differences were found.In the caudate putamen, significant differences were seen for the A2AR mRNA expression F(1, 44) = 7.328, p = 0.0096, a difference was found between HOAT and LOAT controls (p = 0.0186), where the mRNA expression of A2AR was higher in the HOAT rats.For the Oprk1 gene in the caudate putamen, there were a significant difference between the groups F(1, 44) = 9.855, p = 0.0030, and also here the post hoc test showed that the HOAT control had a higher expression than LOAT controls (p = 0.0026), where the mRNA expression of the Oprk1 was elevated in the HOAT controls.There was also a significant difference for the BDNF mRNA expression F (1, 44) = 9.027, p = 0.0044, and the post hoc test revealed a difference between the caffeine-treated rats, where the caffeine-treated HOAT rats had an increased BDNF expression compared to the caffeine-treated LOAT rats (p = 0.0054).No significant difference could be seen for the other genes in caudate putamen.Hypothalamus did not present any significant differences except for the Oprm1 gene; however, post hoc test could not define the significant difference.The Oprd1 gene expression was below the detection limit in all five examined brain regions.

Caffeine increases corticosterone plasma concentration in rats with high and low anxiety-like behavior
The corticosterone concentration was measured in plasma and the result is presented in Fig. 7.The two-way ANOVA showed a significant difference of corticosterone plasma concentration for both treatment (F (1, 42) = 19.29,p < 0.0001) and anxiety level (F(1, 42) = 4.308, p = 0.0441), Šídák's multiple comparisons test showed a significantly higher corticosterone concentration in HOAT caffeine treated group (p = 0.0016) than the HOAT controls and the same for the LOAT groups (p = 0.0257).No further significant differences could be seen for the anxiety level factor.No difference in corticosterone concentration was observed between the HOAT and LOAT control groups.

Discussion
The main findings of this study demonstrate that both behavior and mRNA expression were affected by an acute dose of caffeine in rats with a high or low anxiety-like behavior.Acute caffeine significantly affected risk-taking behavior of the rats with a high anxiety-like behavior at baseline (LOAT), but not of the rats with low anxiety-like behavior (HOAT).This is in line with findings from humans showing that patients with the anxiety disorder panic disorder are more sensitive than healthy controls to caffeine-induced anxiety and panic attacks.Frequency groomings, F. Frequency SAP.Significant differences between LOAT Caffeine treated animals and their control were seen in risk shelter index, frequency to enter the bridge, time spent on the bridge, and in number of rearings.Significant difference between Caffeine treated HOAT and LOAT differences could be identified in duration bridge and number of groomings.HOAT control versus caffeine differed in number of groomings.Data presented as median for each group.Kruskal-Wallis test were performed on all data sets followed by Dunn's multiple comparisons test, p < 0.05 were considered statistically significant.*p ≤ 0.05, **p ≤ 0.001.Selected comparisons for Dunn's multiple comparisons test were HOAT Caffeine -HOAT Control, HOAT Caffeine -LOAT Caffeine, LOAT Caffeine -LOAT Control, HOAT Control -LOAT Control.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) The MCSF test, assessing behaviors such as general behavior, risktaking, and shelter seeking, has never before been used to examine effects of caffeine treatment, resulting in our study being one of the first in this aspect.The bridge in the MCSF test is highly illuminated and is considered to be a risk area to enter (Meyerson et al., 2006), our results show that the caffeine treated LOAT rats decreased their time spent on the bridge, and also had a lower number of entries to the bridge.Thus, caffeine had a significant effect on the LOAT rats but not on the HOAT rats.Using trend analysis, sum of ranks for the composite risk-taking behavior was significantly decreased in the caffeine LOAT rats, further emphasizing these results.Caffeine and risk-taking behavior in humans are assessed in several studies, where contradictory results are presented.Caffeine seems to reduce risk behavior when the subject is sleep deprived for over 75 h (4 × 200 mg caffeine/night of sleep deprivation, 3 nights in total) (Killgore et al., 2011), but in young adults the consumption of caffeinated energy drinks had significant correlation, according to a self-assessment online questionnaire, to alcohol consumption, drunk driving, and extreme sports (Spierer et al., 2014), as well as nicotine consumption (Martin et al., 2008).Very few have investigated the impact of caffeine on risk-taking in rats, however the effect of adenosine A2A receptor antagonist, SCH58261 (0.01 mg/kg/ day), was found to decrease risk-taking behavior in male Wistar rats with methylphenidate induced mania (5 mg/kg/day for 3 weeks) (Shalaby et al., 2022).Furthermore, in a recent study, the effect of components in green tea, including caffeine (10, 17, 30 mg/kg, administered every other day for 8 days), was shown not to affect risky decision making in male Long-Evans rats (Liley et al., 2022), however, their study did not specifically address risk-taking or anxiety-like behavior.In addition, there was a significant difference of the risk shelter index in the LOAT caffeine treated rats, but the index for the HOAT rats was not affected by caffeine.The risk shelter index is calculated based on the time in the dark corner room versus the time on the bridge (Meyerson et al., 2006), giving a measurement on if the animal is more prone to enter the risk or the sheltered areas of the MCSF.Although the MCSF not yet is validated for anxiety-like behavior, the risk shelter index is a parameter that is suggested to be linked to anxietylike behavior (Belovicova et al., 2017;Meyerson et al., 2013).The relationship between a reduced risk-taking behavior and anxiety has been confirmed in several studies in humans evaluating risk behavior when performing a gambling task (Giorgetta et al., 2012), selfassessment of risk avoidance and social anxiety symptoms, and a behavioral computer task for risk avoidance (Lorian and Grisham, 2010), behavioral decision-making task between risk avoidance and social anxiety (Maner et al., 2007) and summarized in (Bishop and  Gagne, 2018).Thus, our results demonstrate that individuals with a high anxiety-like behavior are more affected by acute caffeine exposure, decreasing their risk-taking behavior, and possibly also increasing their anxiety-like behavior.This is in line with findings in humans that patients with anxiety disorders (e.g., panic disorder) are more sensitive than healthy individuals to the anxiogenic effects of caffeine (Klevebrant and Frick, 2022).This is also supported by studies in animals, where in a study in WAG/G and F-334 rats the authors used individual housing to increase the anxiety-like behavior.Caffeine treatment (0.1 % in drinking water ad libitum, 3 weeks) only affected those rats being individually housed (Sudakov et al., 2001) but no anxiogenic response could be detected in either male Wistar rats with chronic caffeine treatment ad libitum and stress-induced increased anxiety (0.3 g/l, 17 days) (Dias et al., 2021), or in C57BL/6, A2AR-knock out, and helpless mice (1 g/l, 3 weeks; 0.3 g/l, 12 h/day, 4 weeks) (Kaster et al., 2015;Machado et al., 2017).These studies used different stressor methods (single housing, 4 h restraint stress for 14 days, chronic mild stress for 3 weeks, breeding towards certain behaviors, respectively) which could have influenced the conflicting results.It is important also to mention that these studies had a chronic caffeine treatment and in this present study the effects of an acute caffeine treatment were investigated.The time spent in the center of the EPM have been positively correlated to efficiency and better response time in reinforcement tests, suggesting that the time in the center of the EPM would be linked to impulsivity of the animal (Rico et al., 2016).In contrast to our results, a study where male Long-Evans rats characterized with high and low anxiety-like behavior in the EPM and then exposed to the differential reinforcement of low rate of responding (DRL) operant box, showed that the rats with high anxiety-like behavior had an increased impulsivity (decreased reward and efficiency in the DRL operant task), and lower A2AR levels in paraventricular nucleus (Niedzielak et al., 2020).Acute (2 mg/kg or 6 mg/kg, single exposure) caffeine treatment decrease the impulsivity of spontaneous hypertensive rats, while chronic treatment (2 mg/kg two times/day for 21 days) seems to stimulate an impulsive response in combination with cannabinoid receptor agonists (Leffa et al., 2019).Lister hooded rats with outbred high impulsivity were entering the open arms of the EPM with shorter latency (8 vs. 24 s, respectively) than rats with outbred low impulsivity, although, no correlation between impulsivity and anxiety-like behavior (time spent on open arms with latency to enter open arm in EPM) were observed (Molander et al., 2011).In addition to this, the time spent in the center of the EPM is suggested to be related to decision making both in mice (DBA/2 mice) (Rodgers and Johnson, 1995), and rats.Anxious Dark Agouti rats spent more time in the center of the EPM and had less motor activity overall however more sniffing activity together with less time spent in the open arms compared to Wistar rats.The authors discussed that the more anxious strain might take longer time investigating the open arm from the center of the EPM, compared to the Wistar rats that investigated by entering the open arm (Casarrubea et al., 2015;Casarrubea et al., 2013).This is in difference to our study where low anxiety HOAT rats spent more time than the other groups both in the open arms and the center.The increased time in the center could indicate the animal's decisiveness whether the open arms were assessed a risk, speculating that it often was assessed as not a risk for the low anxiety-like group, while the high anxiety-like animals, directly without considering, assessed the open arms a risk.This assessment may be linked to the impulsiveness mentioned earlier (Rico et al., 2016) where the high anxiety-like rat's lack of impulsivity perhaps made them disregard the open arms altogether, or it was a result of the acute caffeine treatment, causing a decreased impulsivity (in line with Leffa et al., 2019) and the caffeine treated rats would actively select the safer, closed arm.
In the elevated plus maze, the Kruskal-Wallis test showed a significant difference between the groups regarding time spent on the open arms, however the post hoc test did not detect any differences.The medians of the group presents the LOAT controls with the lowest % of time spent on the open arms, whereas the caffeine treated HOAT rats had the highest %, giving a tendency of spending more time on the open arms in the EPM, i.e., suggesting that caffeine in these rats had an anxiolytic effect.As mentioned in the introduction, earlier studies have shown both anxiolytic and anxiogenic effects from caffeine treatment, this discrepancy may be caused by several factors such as dose variations, treatment length, or as suggested by our results, underlying individual differences of anxiety-like behavior.Our rats received an acute treatment of 50 mg/kg subcutaneously at two isolated events before being exposed to the behavioral tests, which according to previous studies supposedly would create an anxiogenic response (single dose of 50 mg/kg resulting in an anxiogenic response) (Lawther et al., 2018).In general, lower caffeine doses seem to give an anxiolytic effect and higher doses to present an anxiogenic effect, although acute and chronic treatments have displayed different outcome in previous studies.Acute treatment of caffeine in animals, 50 to 100 mg/kg, single dose prior to open field, EPM, social interaction test, and novelty suppressed feeding (Bhattacharya et al., 1997) and 50 or 100 mg/kg IP single dose prior to EPM and light/dark box (El Yacoubi et al., 2000) have presented anxiogenic effects, as well as discomfort and depression (100 mg/kg single dose prior to forced swim test, entered helplessness earlier than control) (Gan et al., 2009).In humans a high continuous dose of caffeine (>200 mg/day) increased the risk of general anxiety in women (Paz-Graniel et al., 2022), and (>400 mg/day) induced panic attacks in individuals with panic disorder and increased anxiety (Klevebrant and Frick, 2022).However, high acute doses, 25 and 50 mg/kg, have also resulted in anxiolytic behavior, increased entries to open arm, increased rearing, and no difference to control when exposed to light induced stress (Hughes et al., 2014).Lower acute doses (10 mg/kg, single dose) prior to open field, EPM, social interaction test, and novelty suppressed feeding (Bhattacharya et al., 1997) and 10 mg/kg (single dose) (Van Schaik et al., 2021), are mainly presented as not affecting the anxiety level in rats.In contrast to this, 10, 20 and 40 mg/kg caffeine, single dose, gave an anxiogenic response in EPM, and light/dark box, in female rats during metestrus-diestrus, but the effects of 10 mg/kg caffeine was attenuated during proestrus-estrus, presenting higher sensitivity to caffeine during some of the phases of the cycle (Guillen-Ruiz et al., 2021).In contrast to the high doses often used for acute treatment, the doses used for chronic treatment are often lower (0.3 g/l in drinking water, 8 days or 2 months), but also a high chronic caffeine treatment (50 mg/kg, twice a day for 7 days) is evaluated, both have presented an anxiogenic response in mice (El Yacoubi et al., 2000).In rats, an increased anxiety-like behavior was observed when given 0.1 % of caffeine in drinking water for one day, but no effect was found if caffeine was administered for three weeks (Sudakov et al., 2001).However, Dias et al. (2021), reported no anxiogenic or anxiolytic response in rats given 0.3 g/l caffeine in drinking water ad libitum for 2 weeks.In contrast to the decreased activity seen in the MSCF test, the caffeine treated rats locomotor activities were increased in the EPM, a common indicator of caffeine stimulation (Svenningsson et al., 1995(Svenningsson et al., , 1999)).This discrepancy between the tests may be caused by several factors, but one important difference is that the rats spent longer time, 20 min, in the MCSF, compared to 5 min in the EPM.In addition, it is worth to mention that the rats were tested in the EPM twice.When the second EPM test was performed, all rats decreased their open arm time, and the rats in the control groups did not show the same clear separation into the high and low open arm time groups in the retest.The interval between the test trials of 28 days is in earlier studies suggested to reduce the one trial tolerance problem, comparing 28 days between test with 14 days between tests, and comparing 28 days with a seven day period, respectively (Schneider et al., 2011;Zhou et al., 2015).However, even if there were four weeks between the occasions the animals may still have a memory of the test.In addition, the animals were exposed to the MCSF one week before the second EPM test, which also may have affected the behavior in the second EPM trial.
The antagonistic actions of caffeine on the A2AR and A1R are well known (Fredholm et al., 1999), but, with regards to anxiety disorders the mechanism is not yet fully elucidated.Our results show that the A2AR mRNA expression in caudate putamen differs in rats with a high or low anxiety-like behavior, where the expression was significantly higher in the HOAT control group compared to the LOAT control group.This further highlights the involvement of adenosine A2A receptors in anxiety-like behavior.The A2AR, highly expressed in the caudate putamen, are important for striatal function.They are localized both preand postsynaptically, and can form heteromers with dopamine D2 receptors, cannabinoid CB1 receptors, as well as with adenosine A1R, reviewed in (Ferre et al., 2011), suggesting an important role in modulating G-protein coupled receptor signaling in striatum.Recently reviewed in (Chen et al., 2023) the A2AR expression in striatum is linked to inhibitory control in cognitive processes, especially goal-directed behavior, meaning that the different levels of A2AR expression in the animals could result in a variation of cognitive flexibility (Chen et al., 2023).Cognitive flexibility is known to be affected by anxiety disorders, but previously studied primarily in the prefrontal cortex (Park and Moghaddam, 2017).Changes in cognitive flexibility may as well be involved in the effects on decision making discussed earlier.The acute caffeine treatment did not affect A2AR expression in the caudate putamen in this study, whereas long-term caffeine treatment, (0.3 or 1.0 g/l ad libitum in drinking water, in mice increased the protein expression of A2AR in the striatum (Machado et al., 2020).In humans, polymorphism in the adenosine receptor genes and especially in the A2AR gene, ADORA2A, is linked to dissimilar tolerance towards caffeine and anxiety, with higher risk towards caffeine-induced anxiety.For example, oral 150 mg caffeine gave self-reported anxiety linked to 1976C > T and 2592C > T polymorphisms in A2AR gene (Alsene et al., 2003), oral 150 mg caffeine at four separate occasions, demonstrated an association between caffeine induced anxiety and polymorphisms in rs5751876, rs2298383, and rs4822492, of the A2AR gene (Childs et al., 2008), and the polymorphism rs187442 (A1R) and rs5751876 (A2AR) were found to have connections to regulation arousal further linking them to mental disorders such as anxiety (Hohoff et al., 2020).The precise function of A2AR in anxiety-like behavior is not elucidated, but increased activation of A2AR has been linked to an increased anxiety-like behavior.Rats with over expression of the A2AR in forebrain had an increased depressive like behavior, increased locomotor activity and altered explorative behavior from control (Coelho et al., 2014).Fear related behavior has been associated with regulation of A1 receptor density in hippocampus, striatum, and amygdala in C57B1/6 mice.Two days after fear conditioning an increase in hippocampus and dorsal striatum was observed, followed by an decreased density eight days after fear conditioning, in combination with an increased A2AR density in amygdala two and eight days after fear conditioning (Simoes et al., 2016).In addition, LTP was inhibited when A2AR was blocked.A recent study in mice demonstrated that disruption of the A2AR in the ventral hippocampus in C57BL/6 mice blocks caffeine's effect (1.0 g/l in drinking water, 3 weeks) and further interrupts the anxiogenic effects (no difference between treatment and control in open field and EPM test) in comparison to wild-type mice where an effect from caffeine could be seen (Xu et al., 2022).In caudate putamen, the mRNA expression of the κ opioid receptor was increased in the HOAT control rats compared to the LOAT controls.κ opioid receptors are strongly associated with anxiety, for example it has been suggested that κ opioid receptor signaling and control of dopamine function in the caudate putamen may underlie the pathophysiology of anxiety disorders, reviewed in (Van't Veer and Carlezon Jr., 2013).As with the A2AR, increased activation of κ opioid receptors has been linked to anxiety-like behavior, however, biphasic effects of κ opioid receptor agonists have been demonstrated, and it has been speculated that κ opioid receptors in specific brain regions may have opposite effects in the regulation of anxiety-like behavior summarized in (Hang et al., 2015), although further studies are needed on this topic.It should also be noted that mRNA expressions not always are directly correlated to protein expressions.
Single exposure for 5 h of 0.5, 2.0 and 5.0 mM caffeine has been shown to increase BDNF gene expression in cortical neurons from mouse (Connolly and Kingsbury, 2010), and long term caffeine treatment, 0.1, 0.3, 1.0 mg/l in drinking water 12 h/day for 24 days, in adolescent rats was shown to decrease BDNF and its precursor pro-BDNF protein density in hippocampus, and increase BDNF and pro-BDNF in cortex (Ardais et al., 2014).However, in the present study, no significant differences between caffeine-treated rats and controls were found in frontal cortex or hippocampus in the post hoc tests.Besides its role in cell growth, and synaptic modulation connected to cognitive functions, BDNF has been implicated to be involved in the pathogenesis of anxiety disorders, male mice whose mothers were on a tryptophan deficiency diet (0.00 %) and corticosterone treatment (80 μg/ml) presented an increased hippocampal BDNF expression (Zoratto et al., 2013), the BDNF involvement in anxiety is also reviewed in (Angoa-Perez et al., 2017).A2AR have been suggested to be involved in fear conditioning, knocking out the gene in several regions of the brain (prefrontal cortex, hippocampus and striatum) caused a decreased fear conditioning, while when removed only in striatum, the opposite result was observed (Wei et al., 2014).When A2AR were knocked out in hippocampus, BDNF decreased, however, not in the other regions.The BDNF expression in frontal cortex demonstrated a difference in treatment in this study, as well as in A2AR, which could indicate caffeine's antagonistic effect on A2AR could be involved in the regulation of the BDNF levels in the frontal cortex.In the caudate putamen, caffeine increased BDNF expression in the caffeine-treated HOAT rats but not the LOAT rats, showing that caffeine affects mRNA expression in rats with high or low anxiety-like behavior differently.Our results suggest an involvement of striatal BDNF in caffeine's effects in association with anxiety-like behavior, although further studies are needed.A reduction of BDNF concentration in striatum was observed in the male mice mentioned earlier with increased anxiety related behavior (Zoratto et al., 2013), and a genetic variant causing a decreased secretion of BDNF from neurons was associated with, when the mouse was stressed, anxiety-like behavior (tested in contextual and cue dependent fear conditioning task) that could not be normalized with antidepressant fluoxetine (Chen et al., 2006).Furthermore, caffeine increased expression of c-fos in the hippocampus.This is in line with previous studies demonstrating caffeine-induced increase of the transcription factor c-fos protein expression, for example a single acute dose of 100 mg/kg caffeine in male mice increased c-fos protein expression in hippocampus, as well as in hypothalamus and amygdala (Savchenko and Boughter Jr., 2011), and short term treatment for 5 h of 0.5, 1.0, 2.0, 5.0, and 10 mM of caffeine increased c-fos expression in cortical neuronal cells (Connolly and Kingsbury, 2010).In fact, c-fos expression has been suggested as a marker of neuronal activity after caffeine administration.In a recent study, the effects on caffeine treatment (0.3 g/l in drinking water for 24 h or 2 weeks) on metabolomic, epigenetic and proteomic changes in mice hippocampi were examined (Paiva et al., 2022) suggesting multiple pathways for how gene expression in the brain, i.e. hippocampus, may be altered from the stimulus of caffeine.In addition to emphasizing that acute and chronic caffeine treatment have different effects, this study showed that transcriptional effects of caffeine in the hippocampus were amplified during activation of neuronal networks, possible related to a priming effect of caffeine on neuronal activity.C-fos expression in hippocampus is also proven to have a role in stress behavior.Chronic restraint stress test performed one time per day for 21 days displayed increased anxiety-like behavior in female WT mice compared to females with a hippocampal mutation giving a 95-98 % decreased expression of the c-fos gene (McQuade et al., 2006).However, after chronic exposure of caffeine (13 to 18 days, 0.3 g/l) in mice no change in hippocampal plasticity or altered behavior was observed, although transmission via the A1R increased in efficiency in the hippocampus together with metabolic changes (Lopes et al., 2023).This is partially in contrast to our results, however, since the focus on this paper is evaluating acute caffeine treatment compared to the chronic treatment in Lopes et al., this could be the cause of the divergence in results.Further studies of the mechanism of action regarding caffeine and anxiety-like behavior are needed, including to elucidate if the effects are the result of specific A2AR antagonism, and the possible involvement of other signaling systems such as BDNF, c-fos, and κ opioid receptors.
The A2AR are suggested to be involved in stress response, for example mice exposed to the stressor maternal separation (postnatal day 2-14, 180 min/day) showed increased stress in adult life (14 weeks), however, when the A2AR were blocked using the A2AR antagonist KW6002 (3 mg/kg/day) the stressed behaviors and elevated corticosterone levels were reverted (Batalha et al., 2013).Our results show that the acute caffeine treated rats have a higher concentration of corticosterone in their plasma, thus indicating a higher stress response.This comparison to chronic caffeine exposure (1 g/l in drinking water, 3 weeks prior and during 3 weeks of chronic unpredictable stress) where caffeine inhibited an increased stress response (Kaster et al., 2015).Chronic caffeine consumption (minimum of one caffeinated coffee per day) in humans were seen to give lower connectivity in the majority of networks in the brain decreasing connectivity with increased caffeine consumption.The chronic caffeine group in this case also showed higher stress, and higher anxiety in males (Magalhaes et al., 2021).The contradictory response of acute and chronic caffeine treatment can be due to that the A2AR respond differently when the stimulus is continuous and a tolerance is established, or perhaps there are other signaling pathways at work during acute treatment with caffeine contributing to the acute response.
In summary, the present study confirms that male Wistar rats display individual variation in anxiety-like behavior as deduced from time spent in the open arms of the EPM.We demonstrate that these rats differ in their general behavior and gene expression in the brain when under the influence of an acute dose of caffeine.Rats with higher anxiety levels were less prone to enter and spend time in risk areas after treatment with caffeine.Furthermore, mRNA expression of adenosine A2A receptors, and κ opioid receptors, differed between control rats with high or low anxiety-like behavior.In conclusion, these results strengthen the hypothesis that caffeine affects individuals differently depending on their anxiety-like behavioral state, and that caffeine targets important mechanisms, probably involving adenosine receptors, underlying these effects.Hence, our results are in line with findings in humans that patients with anxiety disorders (e.g., panic disorder) are more sensitive than healthy individuals to the anxiogenic effects of caffeine.
Further research is needed to fully elucidate the neurobiological mechanisms underlying the effects of caffeine on anxiety, and the importance of adenosine receptors as a possible drug target for anxiety disorders.

Fig. 4 .
Fig. 4. Trend analysis from ranking the individuals against each other for each parameter followed by the sum of selected parameters for each functional group.Caffeine treated rats (grey and turquoise) and control rats (black and teal).HOAThigh open arm time = low anxiety-like behavior, LOATlow open arm time = high anxiety-like behavior.Treatment with 50 mg/kg caffeine (100 mg/kg caffeine sodium benzoate) and 50 mg/kg sodium benzoate for control.Selected parameters for A. Risk taking were frequency, duration, and duration/frequency for bridge and central circle, and for B. General activity were total activity, frequency tot corr, frequency center, duration/frequency tot corr, for C. Risk assessment were for slope and entry zone 1, frequency, duration, duration/frequency, and rearing, for D. Explorative activity, duration total corridors, center duration and hurdle duration, for E. Shelter seeking, frequency, duration and duration per frequency for DCR.Significant differences between LOAT (high anxiety-like behavior) Caffeine treated animals and their control were seen in both Risk taking and General activity.Significant difference between HOAT (Low anxiety-like behavior) and LOAT caffeine treated groups were found in Risk taking.Data presented as median for each group.Kruskal-Wallis test were performed on all data sets followed by Dunn's multiple comparisons test, p < 0.05 were considered statistically significant.*p ≤ 0.05, **p ≤ 0.001.Selected comparisons for Dunn's multiple comparisons test were HOAT Caffeine -HOAT Control, HOAT Caffeine -LOAT Caffeine, LOAT Caffeine -LOAT Control, HOAT Control -LOAT Control.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .
Fig. 5. Behavior parameters from the Multivariate concentric square field test Caffeine treated rats (gray and turquoise) and control rats (black and teal).HOAThigh open arm time = low anxiety-like behavior, LOATlow open arm time = high anxiety-like behavior Treatment with 50 mg/kg caffeine (100 mg/kg caffeine sodium benzoate) and 50 mg/kg sodium benzoate for control.A. Duration risk/shelter index, B. Duration bridge, C. Frequency bridge, D. Frequency Rearing, E.Frequency groomings, F. Frequency SAP.Significant differences between LOAT Caffeine treated animals and their control were seen in risk shelter index, frequency to enter the bridge, time spent on the bridge, and in number of rearings.Significant difference between Caffeine treated HOAT and LOAT differences could be identified in duration bridge and number of groomings.HOAT control versus caffeine differed in number of groomings.Data presented as median for each group.Kruskal-Wallis test were performed on all data sets followed by Dunn's multiple comparisons test, p < 0.05 were considered statistically significant.*p ≤ 0.05, **p ≤ 0.001.Selected comparisons for Dunn's multiple comparisons test were HOAT Caffeine -HOAT Control, HOAT Caffeine -LOAT Caffeine, LOAT Caffeine -LOAT Control, HOAT Control -LOAT Control.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6 .
Fig. 6.Behavioral parameters of the Elevated plus maze test.Caffeine treated rats (gray and turquoise) and control rats (black and teal).HOAThigh open arm time = low anxiety-like behavior, LOATlow open arm time = high anxiety-like behavior Treatment with 50 mg/kg caffeine (100 mg/kg caffeine sodium benzoate) and 50 mg/kg sodium benzoate for control.A. Open arm time (%) B. Time in center (%), C. Distance moved (cm/s), D. Closed arm time (%), E. Open arm frequency (OAF) (%) (total entries per group/total entries for all rats), F. Velocity (cm/s).Data presented as median for each group.Kruskal-Wallis test were performed on all data sets followed by Dunn's multiple comparisons test, p < 0.05 were considered statistically significant.*p ≤ 0.05, **p ≤ 0.001, ***p ≤ 0.0001.Selected comparisons for Dunn's multiple comparisons test were HOAT Caffeine -HOAT Control, HOAT Caffeine -LOAT Caffeine, LOAT Caffeine -LOAT Control, HOAT Control -LOAT Control.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7 .
Fig. 7. Corticosterone levels (ng/ml) in plasma for caffeine (gray and turquoise) treated and control (black and teal) rats.HOAThigh open arm time = low anxiety-like behavior, LOATlow open arm time = high anxiety-like behavior.Treatment with 50 mg/kg caffeine (100 mg/kg caffeine sodium benzoate) and 50 mg/kg sodium benzoate for control Two-way ANOVA (p < 0.0001, treatment factor) followed by Šídák's multiple comparisons test stated a significant difference between the HOAT Caffeine to its control(p = 0.0016) and the same for the LOAT groups (p = 0.0257).No significant differences could be seen within the other comparisons.p < 0.05 were considered statistically significant.Data is presented as mean for each group.*p ≤ 0.05, **p ≤ 0.001, ***p ≤ 0.0001.

Table 2
Gene expression levels in brain sections following caffeine treatment.
(continued on next page)