Regular Positive Human Contacts Do Not Improve Pigs’ Response to a Lipopolysaccharide Immune Challenge

The relationship between humans and domestic animals has important implications for animal welfare [1]. The valence of the interaction for the animal, i.e., positive or negative, has most often been characterized using a behavioural approach in human-animal interaction (HAI) studies [2]. It has been reported that positive interactions can improve health and well-being [3‒5], but the biological pathways leading to these effects remain poorly understood. Furthermore, studies have generally focused on changes during interactions or shortly after (e.g., a few days to a few weeks), rather than longer-term, potentially protective, changes. Improved immune competence has been demonstrated in several species using environmental enrichment as a positive treatment. For example, living in enriched environment has neuroprotective and therapeutic effects on mice submitted to lipopolysaccharide (LPS) challenge [6] and increases the immune system of pigs [7, 8]. It has been proposed that HAIs could act as an environmental enrichment for zoo-housed animals [9]. A study included the provision of positive human contacts as part of an environmental enrichment treatment [10]. However, the contacts in these studies were quite brief, i.e., 1 min of daily brushing of the pigs, and provided for a short period post-weaning, i.e., 2 weeks, and the authors did not find a difference in the pigs’ response to the LPS challenge although they observed a lower cortisol level after a frustration challenge. This triggers the question of whether positive contact provided for a longer period of time could improve the immune system of pigs.

Response to experimentally induced immune challenges is a test of immune competence that has been used in pigs (e.g., [11‒13]) and other animals, including humans. A commonly used immune challenge is the administration of the bacterial endotoxin LPS, as the most abundant pathogen-associated molecular pattern in the cell wall of Gram-negative bacteria (e.g., Escherichia coli), which provokes an immune response in the animal without being pathogenic. Therefore, LPS administration induces sickness behaviour without being harmful: pigs usually show a peak of the sickness symptoms (e.g., somnolence, shivering, and vomiting) around 2–3 h post-injection and recover within 6 h–14 h post-injection [13, 14]. For ethical concerns, we selected one of the lowest doses of LPS administration known to be effective in the scientific literature [15‒17].

Studies vary in the measures used to assess the response of animals to an immune challenge. The pleiotropic cytokine interleukin-6 (IL-6), which has both pro- and anti-inflammatory properties [18], and the anti-inflammatory cytokine interleukin-10 (IL-10) are widely used as markers of the innate immune system’s response. Cortisol is a steroid hormone with anti-inflammatory properties, secretion of which is activated through the stimulation of the hypothalamo-pituitary-adrenal axis by cytokines in response to inflammation [19]. Psychosocial stress (e.g., marital conflict [20] or negative feedback [21] in humans, social isolation or regrouping in pigs [12]) affects inflammation-associated cytokines produced by macrophages in innate immune responses such as IL-6, IL-1β, IL-10, and tumour necrosis factor alpha (TNF-α). For instance, the TNF-α response was reduced in response to an LPS challenge in piglets that were repeatedly (daily over the first 2 weeks post-partum) isolated from their mother and littermates for 2 h alone [14, 22] or as a group [14]. This reduced cytokine response was also observed in mice who underwent maternal separation as pups [23]. However, results on the interleukins are less clear: Tuchscherer et al. [22] found diminished IL-10 response in the stressed group, but not Brückmann et al. [14], and both found no effect of the (total) social isolation on the IL-6 response. Conversely, there is also some evidence of protective health effects (e.g., faster wound healing) induced by positive interactions and social bonds [5]. Immunoglobulins are also immune markers that can reflect health status; in particular, secretory immunoglobulin A (IgA, associated with mucosal immunity) has been shown to increase with positive events (e.g., appropriate social pairing, voluntary exercise in rats) [24]. However, some studies did not find differences in secretory IgA related to positive events (e.g., play behaviour in calves [25]) and others even found elevated levels of IgA in stressful situations (e.g., restraint in pigs [26]). Behavioural observations are used in all studies to assess the magnitude of the response to the immune challenge. Increased behavioural responses, characterized by higher duration, occurrence, or magnitude of sickness behaviours, are usually associated with higher pro-inflammatory cytokine responses. However, some studies found increased behavioural responses in animals submitted to stressful events (e.g., maternal separation in pigs [14] and mice [23]), while decreased responses are found in animals submitted to positive events (e.g., environmental enrichment or ventral tegmental area stimulation in rats [27]). Beyond the acute response of the immune system to the infection, the return to baseline or healing after the infection is also indicative of greater resilience [28]. Speed of recovery from injury or infection seems to be influenced by emotional states in humans. For instance, positive emotional style was linked to faster recovery from rhinovirus or influenza exposure [29], and positive affect was linked to faster wound healing [30].

The present study aims at filling the gap of knowledge on the effects of providing regular positive human-animal contacts on pigs’ immune competence, by assessing their response to an immune challenge. Beyond the importance of the results for the improvement of animal welfare, this paradigm could also deepen our knowledge on the effects of social interaction on immune competence. We predicted that pigs that received positive contacts would develop a more positive relationship with the human (evidenced by a shorter latency to approach the human and greater acceptance of being approached and touched by the human), which will provide them with protective effects to an acute immune challenge (i.e., higher inflammation-associated cytokines [TNF-α, IL-6, IL-10] and higher cortisol concentrations). We also expected the pigs in the positive contacts group to have an altered behavioural response (i.e., show less sickness behaviours) across the time of the challenge, and to return faster to baseline in terms of rectal temperature and behaviour. Finally, we expected to see higher levels of IgA and brain-derived neurotrophic factor (BDNF) at T0 in pigs, which received positive human contacts, as markers of increased welfare and health.

In total, 26 pigs (Sus scrofa domesticus; Swiss Large White × Pietrain breed) were used in this study. Two batches of 12 female pigs were purchased from the Vetmeduni farm “Hof Medau” and housed in the same facility until the end of the study. Two additional female pigs (15 weeks old) were used to pilot the potency of the LPS solution and refine the dose to be used in the main study. The pigs were 4 weeks old at the start of the experiment and 16 weeks old at the time of LPS challenge.

At the age of 5 and 15 weeks, pigs were transported in a vehicle purposely fitted for the transport of pigs to the Swine Clinic at the Vetmeduni campus (50-min drive). They were housed in the same groups in those facilities for up to a week for a related experiment that they participated on to be scanned for magnetic resonance imaging on days prior to the LPS challenge [31].

For the 12-week duration of the experiment, pigs were housed in groups of 3 in enriched pens (200 × 300 cm, 5.9 m² total floor space) with straw bedding on the lying area (200 × 133 cm, 0.88 m²/pig) and addition of a jute rope and a dog-toy orange plastic ball. At 5 and 15 weeks old, the pigs were housed in the animal facilities of the Swine Clinic, in pens (2.25 × 1.80 m², 3.96 m² total floor space, non-slatted) also enriched with jute ropes and deep straw bedding.

All pens were equipped with a feeder and a drinker, and standard weaner diet was provided ad libitum in the home pen throughout the experiment. Spot cleaning of the pen (i.e., removal of faeces and soiled straw) and top-up of straw was done once to twice a week to provide a clean and comfortable environment to the pigs. Windows allow natural lighting of the room and the pens. Care of the pigs was kept as per the farm routine procedures, with no other human interaction apart from feeding, cleaning (by the same person delivering the treatments), and daily visual health inspection (by farm staff or the person delivering the treatments).

Littermates were split equally across treatments and pens (1 pig per litter per pen) to avoid genetic and ontogenetic differences, thereby creating equal litter representation in each of 2 pens of each treatment (i.e., pair-matched pen composition design in terms of litters).

Imposition of the treatments was randomized among conditions and groups between days. There was a 5-min break between pens.

Pigs in the positive contacts group received positive contacts from a (female) familiar human for 10 min, twice a day, three times weekly from 5 to 16 weeks of age in their home pen, a schedule sufficient to improve growth and induce behavioural changes [32]. During these interaction sessions, the human entered the pen while talking softly to the pigs and sat in the lying area. She then encouraged pigs to approach and interact with her by using her voice and moving her hands (e.g., rub or soft tap on the floor and legs). If the pigs approached her, she delivered positive physical contacts (e.g., stroking, scratching, belly rubbing) while keeping on talking with a soft voice. The positive contacts were first directed to the head (avoiding the ears) and chest of the pigs, and then stroking moved further towards the back and abdomen. The positive contacts were attempted to be spread equally across all pigs that approached and remained within an arm distance. If access to the human was prevented by another pig, then the human moved (stood up and walked if necessary) to allow access to the denied pig(s). Pigs that did not approach were still encouraged to do so.

The experimenter brought a dog tug-toy (plastic yellow ring) inside the pen but placed it away from the pigs’ reach, as it was only offered to them if they bit the human to redirect unpleasant behaviour in a positive manner. If the pigs continued to bite, they were suddenly (but softly) tapped on the snout with 2–3 fingers, or gently pushed away, so that the surprise effect would prevent them from repeating the behaviour for human safety purposes.

The control groups were exposed to human presence using the same schedule, with the familiar human standing outside the home pen, not moving besides shifting weight, and ignoring the pigs, i.e., avoided eye contact by looking straight at the wall in front of her, and not talking.

The human-animal relationship was assessed by a standard approach test and a standard avoidance test, carried out by the person providing the positive contacts, at the start (i.e., when all pigs were naïve to human experience) and at the end of the experiment (i.e., before the LPS challenge), in order to assess the quality of the human-animal relationship. The tests were carried out in the pigs’ home pens at the Medau research VetFarm (start) and at the Swine Clinic facilities (at the end). During the approach test, the human entered the home pen, went to the opposite corner, and stood still until the pigs touch her (i.e., snout in contact with boot or overall). The latency for each pig to touch the familiar human was recorded. The approach test ended when all pigs touched the interacting human or after 5 min. In the human avoidance test, the human progressively moved towards a focal pig, starting around 3 m away from it, in steps of 0.5 m, until the pig moved to “avoid” the experimenter (defined as moving its two front feet). The distance at which the human could approach the pig before it moved was estimated in increments of 0.5 m. If the pig did not move and could be touched on the forehead, the distance recorded was 0 m; if the pig could not be touched, the distance recorded was 0.5 m. A 5-min break was taken between rounds of human avoidance test (within one pen), and between the human approach test and the human avoidance test.

At 16 weeks of age, the pigs were subjected to an immune challenge about 24 h after the last human contact treatment session. Because of time constraints, one group of each treatment (i.e., two groups of three pigs) was challenged on 1 day, and the other two groups were challenged on the following day. Bacterial endotoxin “LPS” from E. coli was first solubilized in commercial sterile physiological saline, aliquoted as stock solution of 1 mg/mL, and stored at −20°C until used. On the day of the challenge, the stock solutions were thawed and diluted 1:5 in saline, making doses of 200 µg/mL. The volume of solution administered was then calculated to obtain the dose of 2 μg/kg body weight (weighed the day before the challenge). The solution was administered intravenously in the ear vein by a practicing veterinarian, in order to induce an immune response and sickness behaviour [14]. Pigs were restrained using a snout snare for the injection and the blood samplings (30 s–2 min).

Blood samples (5 mL) were collected by jugular venipuncture at 0 h (just before the LPS administration), 1 h and 3 h after LPS administration to measure plasma IgA, TNF-α, IL-6, IL-10, and cortisol. A single aliquot from the baseline sampling was assigned to measure BNDF levels in blood serum, as environmental enrichment can increase circulating BDNF levels [33, 34] and BDNF has been linked to higher resilience (e.g., in response to social defeat stress in mice [35]). The BDNF concentrations were determined as previously described [33]. Blood samples were stored in a −80°C freezer until transported to the FBN laboratory facilities (Dummerstorf, Germany), where they were analysed using commercially available ELISA kits (IL-10, IL-6, and TNF-α: R&D Systems, Minneapolis, MN, USA; cortisol: DRG Instruments, Marburg, Germany; IgA: Bethyl Laboratories Inc., USA). Samples were run in duplicates. The cytokine kits have been validated as described in Brückmann et al. [14]. The cortisol kit has been validated for porcine plasma as previously described [36]. Rectal temperature was measured using a digital thermometer at the same timepoints as blood sampling to assess the fever response and recorded at the closest digit (precision: 0.1°C). Behaviours indicative of sickness (Table 1) were recorded (absence/presence) by performing live scan sampling every 5 min for 6 h by one observer to monitor the evolution of the pigs’ health condition.

A coding system anonymized the samples and datasets, which allowed to blind all personnel involved in the laboratory and statistical analyses to the treatments. Latency to approach the human, and the avoidance of humans score, before and after the 9 weeks of treatment was analysed using a Wilcoxon signed rank test [37].

All data collected during the LPS challenge were analysed using mixed models with the fixed effects of treatment condition, time of sampling, the interaction of treatment × time, and replicate, with the random effects of pig, litter, pen, and group. Diarrhoea was only observed once in one pig and therefore was not analysed statistically. Vomiting was only observed in the first hour (12 occurrences across 10 pigs); thus, the statistical investigation only looked at the effect of treatment during the first hour post-injection. Even if IgA concentration data was available for other timepoints, the baseline measure was the more relevant one for our study. BDNF concentration was only analysed in baseline samples. Consequently, the statistical models for vomiting, IgA, and BDNF excluded the effects of time and the interaction of treatment × time.

A full-null model comparison (based on a likelihood ratio test [38]) tested the effects of treatment and its interaction with time [39] and aimed at avoiding cryptic multiple testing. The null model lacked the two fixed effects, but the random effects were retained. We tested the effect of individual fixed effects by means of the Satterthwaite approximation [40] using the function lmer of the package lmerTest (version 3.1-3 [41]) and a model fitted with restricted maximum likelihood.

Prior to fitting the model, we inspected all quantitative predictors and the response for whether their distributions were roughly symmetrically distributed. To overcome the left-skewed distribution of the cytokines data and maximize models’ fitness, IL-6 variable was square-root transformed, and IL-10, cortisol, and TNF-α variables were log-transformed. Behavioural recordings were averaged per hour (reducing from 54 values to 6 values per pig), changing the distribution of the data from discrete (absence/presence of the behaviour for each scant) to continuous (i.e., proportion of scans). The distribution was left-skewed for most variables, but given that data transformation (log, square root) did not improve their distribution, or the fitness of the statistical models, these variables were left untransformed.

After fitting the model, we checked whether the assumptions of normally distributed and homogeneous residuals were fulfilled by visual inspection of a QQ-plot [42] of residuals and residuals plotted against fitted values [43]. These indicated no deviations from these assumptions. Collinearity, determined for a model lacking the interaction, appeared to be no issue (maximum Variance Inflation Factor: 1.04 [43]).

Model stability was assessed on the level of the estimated coefficients and standard deviations by excluding the levels of the grouping factors one at a time, using a function written by Rodger Mundry. This revealed the models to be of acceptable stability for the cytokines and temperature variables and showed small problems for the behaviour variables.

We fitted the model in R (version 4.1.2 [44]) using the function lmer of the package lme4 (version 1.1-27.1; [45]). We determined Variance Inflation Factors using the function VIF of the package car (version 3.0-11 [46]).

The sample for each model encompassed 48, 144, or 288 values taken from 24 individuals out of two batches on 1 (BDNF, IgA, vomiting), 3 (TNF-α, IL-6, IL-10, cortisol), or 6 (temperature, behaviours except vomiting) timepoints, respectively. Significance was considered when the p value was inferior to 0.05.

The application of the positive contacts successfully induced a difference in the response of pigs to the human between the treatment groups. Before the start of the treatment, both positive contacts’ pigs and control pigs took longer than 4 min to approach the human (267.4 ± 12.58 s vs. 286.8 ± 9.31 s, respectively; Wilcoxon test = 60.5, p = 0.3); but after the 9 weeks of treatment, positive contacts’ pigs approached the human sooner than control pigs (4.4 ± 0.95 s vs. 20.8 ± 4.84 s, respectively; Wilcoxon test = 24.5, p = 0.007). Similarly, both positive contacts pigs and control pigs had an avoidance distance of 1.3 m (1.3 ± 0.13 m vs. 1.3 ± 0.18, respectively, Wilcoxon test = 73.5, p = 0.95) in the beginning; but after the 9 weeks of treatment, positive contacts pigs could all be approached and touched, while control pigs could not be touched or avoided the human (0.0 ± 0.00 m vs. 0.3 ± 0.07 m, respectively, Wilcoxon test = 42, p = 0.016).

The concentrations of BDNF did not significantly differ between treatment groups (positive contacts: 24.1 ± 12.6 pg/mL, control: 21.7 ± 11.1 pg/mL, F(1, 10.349) = 0.0515, p = 0.82; data not shown).

IgA concentration before the LPS challenge did not significantly differ between treatment groups (positive contacts: 0.8 ± 0.07 ng/mL, control: 0.7 ± 0.06 ng/mL, F(1, 12.1) = 0.75, p = 0.48; data not shown).

Pigs from the two treatments did not differ in the proportion of scans displaying sickness behaviours (somnolent, cramping, shivering, panting, vomiting), and there was also no significant difference according to the interaction between treatment and time (Table 2).

There was an effect of time on somnolence (F(5, 4.29) = 24.56, p = 0.003) and cramping (F(5, 5.8) = 7.04, p = 0.018), as these behaviours increased in the first 3 h post-injection. Compared to before LPS injection, the proportion of scans spent being somnolent was higher 3 h (−0.6 ± 0.06, t(4.64) = −10.44, p = 0.001) and 4 h (−0.6 ± 0.11, t_7.88 = −5.16, p = 0.008) post-administration and tended to be higher 2 h post-administration (−0.5 ± 0.11, t_3.33 = −4.59, p = 0.08). The proportion of scans spent cramping was higher at 3 h post-administration (−0.4 ± 0.09, t(5.31) = −4.39, p = 0.04), compared to 1 h post-administration, and tended to be higher 2 h post-administration (−0.3 ± 0.07, t_5.09 = −3.85, p = 0.07), T4 (−0.4 ± 0.09, t_6.53 = −3.74, p = 0.06) and 5 h post-administration (−0.3 ± 0.1, t_9.03 = −3.32, p = 0.07) compared to 1 h post-administration.

There was no significant difference according to the treatment or the interaction between treatment and time, on any of the cytokines level analysed (Fig. 1). There was a significant time effect on all variables (TNF-α: F(2, 5.11) = 261.77, p < 0.001; IL-6: F(2, 13.79) = 21.53, p < 0.001; IL-10: F(2, 8.6) = 439.87, p < 0.001; cortisol: F(2, 6.06) = 24.98, p = 0.001).

The concentration in TNF-α was significantly higher 1 h (−51,207 ± 12,183 pg/mL, t_7.86 = −4.2, p = 0.01) and 3 h post-administration (−4,391 ± 847 pg/mL, t_6.75 = −5.18, p = 0.005), compared to baseline. It was also higher 1 h compared to 3 h post-administration (46,816 ± 11,801 pg/mL, t_6.75 = 3.97, p = 0.02). IL-6 increased over time (baseline vs. 1 h: −75 ± 25 pg/mL, t_30.45 = −3.02, p = 0.02; baseline vs. 3 h: −1,352 ± 384 pg/mL, t_7.04 = −3.52, p = 0.03; 1 h vs. 3 h: −1,277 ± 384 pg/mL, t_7.04 = −3.32, p = 0.04). The concentration of IL-10 was higher 1 h (−1,244 ± 113 pg/mL, t_3.58 = −10.97, p = 0.002) and 3 h post-administration (−253 ± 22 pg/mL, t_3.58 = −11.3, p = 0.002), compared to baseline. IL-10 was also higher 1 h compared to 3 h post-administration (991 ± 103 pg/mL, t_5.32 = 9.63, p = 0.0004). The cortisol concentration tended to be higher 3 h post-administration compared to baseline (−120 ± 33 ng/mL, t_3.98 = −3.68, p = 0.06).

There was no significant difference between treatments or the interaction of time by treatment on rectal temperature. However, there was a significant difference according to time, as the rectal temperature increased over the two first hours (T0 vs. all: p < 0.005, T1 vs. all: p < 0.05) (Fig. 2).

This study investigated the effects of positive contacts with a human, provided regularly over several weeks, on the response of pigs to an immune challenge. We hypothesized that the pigs which received positive contacts would have a greater response (i.e., higher levels of cytokines and cortisol, and fewer sickness behaviours) and a faster recovery (i.e., faster return to the baseline rectal temperature and reduction of sickness behaviours) to the immune challenge. Although the human contact treatments created a difference in the pigs’ response to humans, with positive contacts pigs being more receptive to humans (shorter latency to approach, lower reaction to being approached and touched) than control pigs, our results show no statistically significant difference in their behavioural or physiological response to the immune challenge. The immune challenge may have been too severe and overwrote the effects of the provision of positive human contacts, or the protective effects of positive interventions are more subtle and therefore more difficult to detect than the detrimental effects of stressful events. We discuss below other possible explanations for the lack of effect of positive human contacts on the immune response.

We observed values of rectal temperature, cytokines, and cortisol similar or higher than those observed in comparable studies, using low doses of LPS with pigs of similar age or weight (e.g., [16, 17, 47, 48]). A greater secretion of cytokines was expected in animals that show a better response to the challenge, which suggests that our pigs all had a good immune competence. This could be due to the quite good welfare conditions in which pigs were housed (i.e., having a low stocking density of 2 m² per pig, no social mixing after weaning and initial group composition, ad libitum feeding and enough feeding space, access to toys and fresh straw), which have been proven to enhance the pigs’ immune system and resilience [8, 49].

Consequently, it is possible that in such good welfare conditions, pigs already had a healthy and functional immune system that could not be further improved by providing positive contacts with a human. Baseline IgA concentration did not differ significantly between the pigs, thus suggesting that providing regular positive contacts to the pigs did not positively affect this welfare and immune indicator. BDNF levels were measured before the LPS challenge, as a proximate assessment of enrichment effect of the positive contacts, as higher BDNF levels are usually observed in animals living in enriched environments [33, 50]. The fact that we did not find a difference between our treatment groups further supports that the provision of positive contacts may not have represented additional environmental enrichment for the pigs. Mere visual contact with the humans, for the pigs in the control treatment, may have been sufficient to reduce fear of humans, as has been shown in previous studies [5, 51].

However, the regular and active human contacts in the positive contact group had further effect, as evidence by the shorter latency to approach the human at the end of the treatment in the pigs who experienced regular positive human contacts. The large majority of studies that investigated the effect of social interactions on the immune competence compared (humans or non-human animals) subjects submitted to psychosocial stressors (e.g., conflicts, social isolation) to subjects not submitted to psychosocial stressor (controls), or to subjects which had social support. Overall, the results from these studies show a diminished response in the stressed subjects compared to control subjects, but the effects of social support are less straightforward (e.g., [5, 14]). Because the effect of human-animal relationship has seldom been investigated in the context of immune competence, we cannot compare our work to other studies in order to find methodological differences. The study of Luo et al. [10] included short daily positive contacts with a human (1 min of brushing) as a form of environmental enrichment, but they did not observe differences in the pigs’ response to the immune challenge (also using a LPS dose of 2 µg/kg but with younger pigs). Like our study, it is arguable that the schedule/duration of contacts may have been insufficient to enhance the immune response of pigs. Increasing the number or the duration of contacts sessions would be possible in an experimental setting, but it would be unrealistic for livestock farming, and therefore will have little chance to be relevant for large-scale pig farming and influence change in husbandry practices.

Finally, because of ethical, practical, and economical reasons, our sample size was rather small compared to other studies, and we observed a large variability of response between the pigs. This likely hindered the statistical significance of our results. For instance, as in the study of Parois et al. [48], who tested 46 pigs from each of the two treatments, we observed a similar peak of rectal temperature (41°C, at T3) and a difference of 0.5°C between the treatment groups, but this was statistically non-significant in our study. However, in the absence of data from previous studies on the effects of positive HAIs on the response to an LPS challenge, we were unable to better estimate the minimum sample size. Limiting our experiment to only females was meant to reduce the variability of the response to the LPS challenge, but it might have hindered some differences that could have been detected in male pigs (e.g., [14]). A posteriori power calculations revealed that a minimum of 44 pigs per treatment group should have been used to allow detection of differences at the statistical threshold of 95% certainty (online suppl. material; for all online suppl. material, see https://doi.org/10.1159/000544748).

In conclusion, despite successfully improving the human-animal relationship, providing regular positive contacts to the pigs did not improve their response to an LPS immune challenge under the conditions of this study. Despite the absence of effects of the positive contacts between humans and pigs on the immune response, the design of this study could be applied to study the implications of social interactions on immune competence in other species and, e.g., to study the effects of the pet animal ownership and animal-assisted therapy for human and animal health.

We would like to thank the staff at the Medau farm and at the Swine Clinic of the University of Veterinary Medicine Vienna for taking care of the pigs and providing help in the conduction of this experiment.

This experiment was conducted in accordance with the national legislation and the European Directive 2010/63/EU. It was approved by the BMBWF (Bundesministerium für Bildung, Wissenschaft und Forschung) under the Approval No. 2022-0.118.102.

Dr. Ulrike Gimsa was a member of the journal’s Editorial Board at the time of submission. All other authors have no conflicts of interest to declare.

This study and its publication as Open Access were funded by the Austrian Science Fund (Fonds zur Förderung der Wissenschaftlichen Forschung, FWF; Project No. P33669-B).

Conceptualization and methodology: O.S., J.-L.R., and U.G.; formal analysis: O.S., J.-L.R., U.G., B.S., and H.V.; funding acquisition: J.-L.R.; investigation, project administration, visualization, and writing – original draft preparation: O.S.; resources: J.-L.R., U.G., and C.K.; and writing – review and editing manuscript: O.S., J.-L.R., U.G., B.S., H.V., and C.K.

All data generated or analysed during this study are included in this article and its online supplementary material files. All datasets (raw data) and R files used for the statistical analyses can be found on Open Science Framework repository: https://osf.io/e534k/?view_only=c746112599b147b6a3195f73507873cf. Further enquiries can be directed to the corresponding authors.