OPEN
`
`received: 06 September 2015
`accepted: 22 January 2016
`Published: 29 March 2016
`
`The impact of acute stress on
`hormones and cytokines, and how
`their recovery is affected by music-
`evoked positive mood
`
`Stefan Koelsch1,2, Albrecht Boehlig3,4, Maximilian Hohenadel3, Ines Nitsche3, Katrin Bauer3 &
`Ulrich Sack3
`
`Stress and recovery from stress significantly affect interactions between the central nervous system,
`endocrine pathways, and the immune system. However, the influence of acute stress on circulating
`immune-endocrine mediators in humans is not well known. Using a double-blind, randomized study
`design, we administered a CO2 stress test to n = 143 participants to identify the effects of acute stress,
`and recovery from stress, on serum levels of several mediators with immune function (IL-6, TNF-α,
`leptin, and somatostatin), as well as on noradrenaline, and two hypothalamic–pituitary–adrenal axis
`hormones (ACTH and cortisol). Moreover, during a 1 h-recovery period, we repeatedly measured these
`serum parameters, and administered an auditory mood-induction protocol with positive music and a
`neutral control stimulus. The acute stress elicited increases in noradrenaline, ACTH, cortisol, IL-6, and
`leptin levels. Noradrenaline and ACTH exhibited the fastest and strongest stress responses, followed by
`cortisol, IL-6 and leptin. The music intervention was associated with more positive mood, and stronger
`cortisol responses to the acute stressor in the music group. Our data show that acute (CO2) stress affects
`endocrine, immune and metabolic functions in humans, and they show that mood plays a causal role in
`the modulation of responses to acute stress.
`
`The interactions between the central nervous system, the endocrine system, and the immune system are sub-
`jects of intense investigation in the field of biomedical science1. Particularly, effects of stress (and recovery from
`stress) on immune function have received a vast amount of attention due to their impact on human health and
`well-being. However, there is lack of information regarding the influence of acute (transient) stress on circulating
`immune-endocrine mediators such as hormones and cytokines in healthy individuals. Investigating such influ-
`ence is challenging, due to the fact that effects of stress are difficult to differentiate from confounding factors such
`as exercise, current health status, psychological states (e.g., depressed mood and anxiety), and inter-individual
`variability in responsiveness to stress2,3. Moreover, such alterations of endocrine and immune parameters are
`usually moderate and mostly stay within the reference values of healthy persons, although they can be precisely
`detected4. To obtain sufficient statistical power in the face of these methodological difficulties, our study employed
`a CO2 stress test, high sensitivity laboratory tests, and a relatively large population of participants (n = 143) to
`identify the effects of acute stress on hormonal and immunological parameters5.
`The CO2 stress test used in the present study (inhalation of 35% CO2) is a test originally employed as an exper-
`imental model of panic6. The CO2 test incites noticeable vegetative reactions (such as changes in heart rate and
`blood pressure)6, an increase of serum noradrenaline and salivary alpha-amylase7, as well as an increase in serum
`cortisol6. The CO2 test has therefore been taken to represent an acute physiological stressor suited to model the
`human stress response6. Thus, purely physiological effects of stress can be investigated, independent of physical
`stressors (e.g., exercise) or psychological stressors (e.g., mental stress). Whether CO2 stress also affects immune
`
`1Max Planck Institute for Human Cognitive and Brain Science, Stephanstr. 1a, 04103 Leipzig, Germany. 2Department
`of Biological and Medical Psychology, University in Bergen, Jonas Liesvei 91, 5009 Bergen, Norway. 3Institute of
`Clinical Immunology, Medical Faculty, University of Leipzig, Johannisallee 30, 04103 Leipzig, Germany. 4Clinic for
`gastroenterology and rheumatology, Medical Faculty, University of Leipzig, Liebigstr. 20, 04103 Leipzig, Germany.
`Correspondence and requests for materials should be addressed to S.K. (email: koelsch@cbs.mpg.de) or U.S. (email:
`ulrich.sack@medizin.uni-leipzig.de)
`
`1
`
`www.nature.com/scientificreports
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`SPRUCE - EXHIBIT 2004
`Neurocrine Biosciences, Inc. v. Spruce Biosciences, Inc. - PGR2022-00025
`
`

`

`Figure 1. Illustration of the experimental design. In the beginning, participants (n = 143) rated their current
`mood using the Profile of Mood States (POMS, see third row). Then, the first blood sample was drawn (time point 1,
`bottom row), and about 10 min later the CO2 stress-test was administered (top row). Following the stress test, another
`blood sample was drawn (time point 2), and the acoustic stimulus (positive music or neutral control stimulus) was
`delivered (second row). Then, every 15 min, four further blood samples were obtained. After the last blood sample
`was obtained (at time point 6), participants rated again their mood (see third row).
`
`parameters is unknown, although knowledge about effects of physiological stress on immune function is impor-
`tant for the understanding of neuro-endocrino-immunological function.
`To investigate this, we measured effects of CO2 stress on several mediators with immune function, in addition
`to two hypothalamic–pituitary–adrenal (HPA) axis hormones (adrenocorticotropic hormone [ACTH] and cor-
`tisol), and on serum noradrenaline (released into the blood stream by cells of the adrenal medulla in response to
`stimulation by sympathetic preganglionic neurons). The mediators with immune function included two cytokines
`(interleukin-6 [IL-6] and tumor necrosis factor [TNF]-α ), leptin (which also acts as pro-inflammatory peptide)8,
`and somatostatin (which down-modulates a number of immune functions, such as lymphocyte proliferation,
`immunoglobulin production and the release of pro-inflammatory cytokines)9. Generally, it is assumed that
`stress-related HPA and sympathetic nervous system activity inhibits the functions of inflammatory cells, in par-
`ticular cytokines (such as IL-6 and TNF-α )10. Whether this mechanism is also activated during acute stress has
`remained elusive in previous research. This lack of knowledge is surprising, given that acute stress in everyday
`life might have an impact on immune system activity (in both positive and negative ways)11, and thus on an indi-
`vidual’s health.
`Moreover, to investigate latencies of peak responses, and to shed light on the recovery of parameters affected
`by the stress test, we obtained serum values of these parameters not only directly before and after the CO2 stress
`test, but also at several time points following the stress test (see also Fig. 1). This was intended to provide useful
`information about the most relevant time points for sample collection in future studies.
`Another aim was to investigate the influence of mood as a psychological factor on endocrine and immune
`reactions to stress and recovery from stress. As a mood-induction procedure we used positive music in one study
`group, and an acoustical control stimulus in a control group. Music as a stimulus was chosen because music has
`been shown to be a potent mood modulator12,13, and because music-evoked engagement of neurochemical sys-
`tems underlying stress and immunity has received increasing interest during the past years (see, e.g, refs 14–17
`and below).
`Several laboratory studies with healthy participants reported cortisol reductions in response to music (without
`a stressor), compared to a control group18–21. In general, it seems that a decrease in cortisol is more likely during
`tranquilizing music, a reduction in negative mood, and in music therapy settings, whereas an increase in cortisol
`levels is more likely when music is perceived as vitalizing, or when individuals are physically active14,22. In addi-
`tion, several studies measured salivary cortisol levels following a stressor23–25. Two studies observed lower cortisol
`levels in a music listening group (compared to silence group)24,26, and one study reported higher cortisol levels
`post-stressor in a music group (compared to a group listening to the sound of rippling water)25. Thus, the effects
`of mood (evoked by music) on cortisol responses to stress are not well understood, at least not in the context of
`laboratory studies. In clinical settings, listening to music before, during, and after medical interventions has been
`reported to correlate with lower cortisol levels, associated with reductions of anxiety (for reviews see refs 15,27).
`With regard to immune-parameters, there is scarcity of music studies investigating effects of music on
`cytokines (for reviews see refs 28,15,16), and to our knowledge only two studies with a control-group design
`investigated effects of music on Immunoglobulin A29,14. Our study aims at shedding more light on effects of music
`on immune functions, and interactions between cortisol and immune parameters. Insights into psychological
`mechanisms associated with immune function are particularly relevant with regard to mood disorders, such
`as depression, and with regard to somatic diseases with affective components, such as chronic diseases of the
`immune system30–32.
`We addressed the following four core hypotheses in our study: (1) Acute CO2-induced stress leads to an
`increase of serum ACTH, cortisol, and noradrenaline; (2) even short-term (acute) stress induces inflammatory
`activity involving IL-633, TNF-α 34, and somatostatin35; (3) beyond these HPA, sympathetic, and inflammatory
`parameters, we expected an increase of serum leptin in response to acute stress (due to its role as mediator of
`
`2
`
`www.nature.com/scientificreports/
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`

`

`Group
`music
`control
`
`n
`71
`72
`
`Males/Females
`36/35
`36/36
`
`Age (years) Height (cm) Weight (kg)
`24.8 (2.6)
`174.2 (9.9)
`67.4 (12.0)
`24.9 (2.9)
`175.6 (9.5)
`69.7 (10.8)
`
`BMI
`22.4 (2.4)
`22.2 (2.6)
`
`Table 1. Descriptive statistics of the study population. The sample consisted of 143 participants who were
`randomly assigned to either a music group or a control group. Following the CO2 stress, the music group
`listened to positive music, and the control group was presented with an auditory control stimulus. Groups were
`balanced with regard to group size (n), male/female ratio, age, body height, body weight, and body mass index
`(means, with standard deviation in parentheses). BMI: body mass index.
`
`stress-induced obesity)36; (4) finally, we tested whether music-evoked positive mood has effects on both stress and
`immune parameters during recovery from acute CO2 stress.
`
`Methods
`Summary of study design and procedures.
`In our double-blind, randomized study, acute stress was
`induced in n = 143 participants by CO2 insufflations. Serum levels of cortisol, ACTH, interleukin-6 (IL-6),
`noradrenalin (NA), leptin, somatostatin (SIH), and TNF-α were measured at six time points: directly before and
`after the CO2 challenge, and then every 15 minutes over the course of one hour (see Fig. 1). After the CO2 stressor,
`the mood of participants was systematically influenced by presenting them either with a positive music stimulus
`(music group) or with a neutral auditory control stimulus (control group).
`
`Participants. 143 healthy volunteers participated in the study, all participants were students from the
`University of Leipzig (for descriptive statistics of the study population see Table 1). Using a computer algorithm,
`participants were randomly assigned either to the music group (n = 71, mean age = 24.8 years, range 20–33 years,
`35 females and 36 males, mean body mass index 22.4) or to the control group (n = 72, mean age = 24.9 years,
`range 20–32 years, 36 females and 36 males, mean body mass index 22.2). Moreover, the computer algorithm
`assigned participants randomly to one of three measurement times (11:00 a.m., 2:00 p.m. and 5:00 p.m.). The com-
`puter algorithm ensured that all study groups were balanced with regard to group size, male/female ratio, mean
`age, mean height, mean weight, and body mass index. By balancing the study groups across three different times
`of day (instead of measuring at only one fixed time), we can guarantee that the effects observed not only hold for
`one particular time of day (e.g., morning or afternoon), but can be generalized to most time points during the
`day. Moreover, by balancing subjects carefully across the different times of day, we can exclude that our data are
`biased by the circadian rhythmicity of the measured parameters (and although this was not the focus of this study,
`we will report data informing about the circadian rhythmicity of the measured parameters in the Supplementary
`Information). Note that, as the results will show, the circadian rhythm of all of the investigated parameters did not
`interact with the stress test or with the recovery from stress.
`Exclusion criteria were pregnancy, uncontrolled hypertension (blood pressure above 160 over 90 mmHg),
`chronic lung diseases (including asthma), history of psychiatric or neurologic disorders (including panic disorder,
`migraine or seizure disorder), or immunological disorders. None of the subjects took any medication (including
`anti-inflammatories or steroids), and none of the subjects had needle phobia. Written informed consent was
`obtained from all participants. The study was approved by the ethics committee of the University of Leipzig
`(#166/2006), and conducted according to the Declaration of Helsinki.
`
`Stress induction. Subjects took a single full vital capacity breath of a gas mixture of 35% CO2 and 65%
`O2. Breaths were controlled and recorded using a spirometer (ZAN 100 Flowhandy, ZAN Messgeräte GmbH,
`Oberthulba, Germany). The single full vital capacity breath was practised beforehand, and observed by both the
`participant and the experimenter on the spirometer screen.
`
`Procedure. Participants were not informed as to whether they were in the music or the control group. After
`giving written informed consent, participants filled out questionnaires on personal data, and the short (35-item)
`German version of the Profile Of Mood States (POMS)37 which consists of the four scales “Depression/Anxiety”,
`“Fatigue”, “Vigor” and “Irritability”. Then, blood pressure was measured, and an intravenous cannula was inserted
`in an antecubital vein of the left arm. Immediately after the insertion of the cannula (within the first seconds after
`insertion), the first blood sample was drawn (time point 1). About 10 min later, the CO2 stress test was adminis-
`tered (see also Fig. 1). Directly after the CO2 challenge, participants lay down in supine position, closed their eyes,
`and were presented via headphones either with the musical stimulus or with the control stimulus (stimuli are
`described below), and about 90 seconds after the CO2 challenge, a second blood sample was obtained (time point 2).
`Subsequently (while participants were listening to either the music or the control stimulus), four further blood
`samples were taken over the course of one hour. Thus, blood samples were drawn at six different time points (TP):
`Before the CO2 challenge (TP 1), 90 seconds after the CO2 challenge (TP 2), and then every 15 minutes (TP 3–6,
`see also Fig. 1). Note that the parameters measured at TP 1 might have been influenced by some participants’
`anxiety associated with the insertion of the cannula. Thus, the serum values of TP 1 do not represent a normal
`baseline value. Moreover, the effects of acute stress on the parameters measured at TP 2 (taken directly after the
`CO2 stress test) might have been due to both the CO2 stress test and in part the venipuncture38. These issues will
`be discussed further in more detail below.
`
`3
`
`www.nature.com/scientificreports/
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`

`

`Experimenters were blinded with regard to which of the two stimuli the participant was exposed to. Likewise,
`participants were only informed that an auditory stimulus would be presented to them, and they were asked to
`tap the beat of the stimulus with their right index finger, allowing us to guarantee that subjects of both groups
`paid similar attention to the stimulus, and engaged in a similar way with the task (a procedure that has been
`employed successfully in previous studies, e.g. refs 39,40). After each auditory stimulus, there was a silent inter-
`val of 40 seconds in which participants opened their eyes and provided ratings on how they felt during the last
`piece regarding valence and emotional arousal. Ratings were obtained by using 9-point scales. As the results will
`show, the control stimulus was perceived as neither pleasant nor unpleasant. Thus, our control condition might
`also provide a valuable method for the investigation of recovery from stress in general, perhaps being even more
`adequate than simply letting participants relax in silence (see also Discussion). Directly after the very last auditory
`stimulus, participants filled out the POMS mood-questionnaire again (to assess effects of the experiment on the
`mood of participants). Finally, participants were debriefed.
`
`Biochemical analyses. Blood samples were cooled and centrifuged immediately after the respective sample
`was taken. Plasma and serum samples were stored at −80° C until analysis. The following parameters were ana-
`lyzed using test kits according to the manufacturers’ guidelines (manufacturers and detection limits are listed in
`parentheses): cortisol (IBL, Hamburg, Germany; 2.4 ng/ml); ACTH (Biomerica, Newport beach, CA; 0.46 pg/ml);
`Interleukin-6 (IL-6; R&D Systems, Minneapolis, MN; 0.11 pg/ml); Noradrenaline (NA; Labor Diagnostika
`Nord, Nordhorn, Germany; 0.2 ng/ml), Leptin (R&D Systems, Minneapolis, MN; 7.8 pg/ml); Somatostatin
`(SIH; Peninsula Laboratories LLC, San Carlos; 10 pg/ml); and Tumor Necrosis Factor (TNF)-α (R&D Systems,
`Minneapolis, MN; 0.1 pg/ml). Extreme values (i.e., values above or below 2 SD of the mean) of plasma and serum
`parameters were regarded as artefacts and excluded from the statistical analysis (~4% of all data).
`
`Music and control stimuli. Stimuli were taken from a previous study12. For the music group, 18 pieces of
`instrumental music (without lyrics) from various styles and epochs were used as stimuli (Classical, Jazz, Irish
`folk, South American, Reggae). Range of beats per minute was 106–132. These pieces had been shown to evoke
`feelings of pleasure and happiness12. The control group was presented with 18 computer-generated pieces that
`were melodic sequences of random tones of the chromatic scale, presented in isochronous intervals. All twelve
`pitch classes of the chromatic scale occurred with equal probability in these pieces (that is, pieces were not tonal,
`and had no tonal centre). The control stimuli had the same pitch range, mean pitch, tempo and duration as the
`musical comparison stimuli. Moreover, the timbres of the control stimuli were similar to those of the respective
`musical stimulus (e.g., for a musical piece with flutes, the corresponding control stimulus was created with a
`flutes-timbre). Control stimuli were synthesized with Reason 3.0 (Propellerhead Software, Stockholm, Sweden).
`In the previous study that also used these stimuli (ref. 12), participants rated the felt emotional valence of these
`stimuli as neutral, that is, they neither liked nor disliked the control stimulus. Moreover, in that study ratings of
`felt physiological arousal evoked by music and control stimuli did not differ12. Duration of each set of stimuli
`(music and control stimuli) was ~41 minutes.
`
`Data Analysis. Effects of stress test, group (music/control), time of day, and sex, as well as interac-
`tions between variables were calculated using MANOVAs with time-point (six time points, see also Fig. 1) as
`within-subjects factor, and stimulus group (music, control), sex (male, female), and time of day (11:00 a.m.,
`2:00 p.m., 5:00 p.m.) as between-subjects factors. Bonferroni-corrected significance threshold for main effects
`was p = 0.005. Statistical analysis was performed using PASW Statistics 18 (IBM Deutschland GmbH, Ehningen,
`Germany). Parametric tests were computed because none of the parameters deviated significantly from a normal
`distribution (and all parameters showed homogeneity of variance).
`Results
`Effects of the stress test. The stress test elicited significant changes in serum levels of NA, ACTH, cortisol
`and IL-6 (Fig. 2). This was reflected in main effects of time point (p ≤ 0.001 in the MANOVAs computed for each
`of these serum parameters, see Methods; for details see Table 2), with effect sizes (η2) of .39 (NA), .26 (ACTH), .67
`2
`(cortisol), and .16 (IL-6). The effect of the stressor on leptin approached statistical significance (p < 0.05, η = .0 1
`
`for the main effect of time point; see also Table 2). No effect of the stress test was observed for SIH, nor for TNF-α .
`IL-6 levels were lowest at time point 3 (around 20 min after the stress test, see Fig. 2), and highest at time
`point 6 (i.e., at the end of the recovery period, around 1 h after the stress test; a paired t-test indicated that IL-6
`levels were significantly higher at time point 6 than at time point 1, p < 0.005). Note that NA and ACTH showed
`maximum serum levels at time point (TP) 2, and an immediate decrease after TP 2, thus exhibiting a fast and
`phasic response to the stress test. Paired t-tests indicated significant differences between TP 1 and TP 2, as well
`as between TP 2 and TP 3, for both NA and ACTH (p < 0.0001 in each test). The spline interpolation of NA and
`ACTH data across time points suggests a peak latency of less than 7 min (note that this spline interpolation can
`only approximate the true time course of parameter levels). By contrast, cortisol levels were maximal at TP 3 (i.e.,
`more than 15 min after the stress test), thus showing a much slower response, and a gradual decrease until the last
`TP (this is to be expected because both ACTH and NA are released following neuronal signalling, while cortisol
`release is mediated humorally). Paired t-tests showed that cortisol differences were not significant between TP
`1 and TP 2 (p = 0.4), but between TP 2 and TP 3 (p < 0.05), and between TP 3 and TP 4 (p < 0.0001). At TP 6,
`Cortisol, NA, and ACTH levels were below levels of TP 1. Paired t-tests showed that cortisol, NA, and ACTH
`levels differed between TP 1 and TP 6 (cortisol: p < 0.0001, NA: p < 0.0005, ACTH: p < 0.05), probably owing to
`the stress and anxiety that participants felt prior to the venipuncture and the stress test.
`
`4
`
`www.nature.com/scientificreports/
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`

`

`Figure 2. Serum levels of parameters that showed sensitivity to the CO2 stress test (NA, ACTH, cortisol,
`IL-6 and leptin), pooled across both groups, and all three times of day (n = 143 subjects). Mean serum levels
`are indicated by the circles, separately for each of the six time points (error bars represent SEM). Dotted lines
`are spline interpolations of mean values, thus approximating the time course of parameter levels between the
`measured time points. The bottom right panel shows z-standardized values so that responsiveness to the stress
`test can be compared between parameters. ACTH: adrenocorticotropic hormone; IL-6: interleukin-6; NA:
`noradrenaline.
`
`The bottom right panel of Fig. 2 shows z-standardized values of parameters that exhibited significant sensi-
`tivity to the CO2 stress test (so that responsiveness to the stress test can be compared between parameters). These
`data indicate that NA had the strongest sensitivity in response to the stress test, followed by ACTH and cortisol.
`
`Interactions with music and mood. When comparing music and control groups directly with each other,
`the music stimulus had a significant influence on the recovery of cortisol levels, as indicated by an interaction
`2
`between time point x stimulus group for cortisol (p < 0.02, η = .0 1
`; see Fig. 3a; for details see left column of
`Table 3). This interaction was due to cortisol levels being higher in the music group during TPs 3–6 (two-samples
`
`5
`
`www.nature.com/scientificreports/
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`

`

`Cortisol
`NA
`ACTH
`IL-6
`Leptin
`SIH
`TNF-α
`
`Time point (TP)
`F = 77.35 (3.015); p < 0.001
`F = 35.85 (3.130); p < 0.001
`F = 11.62 (4.145); p < 0.001
`F = 5.57 (3.151); p = 0.001
`F = 2.74 (4.749); p = 0.021
`F = 0.27 (3.652); p = 0.88
`F = 0.78 (4.12); p = 0.54
`
`Table 2. Statistical values of the main effects of time point (TP), as indicated by the MANOVAs (F-and
`p-values, with degrees of freedom in parentheses). Main effects of time point reflect that serum levels differed
`between the six measured time points, thus indicating the effects of the CO2 stressor (recall that TP 1 preceded,
`and TP 2-6 followed the CO2 stress test). Significant effects (p < 0.005, corrected for multiple comparisons)
`are marked in bold. ACTH: adrenocorticotropic hormone; IL-6: interleukin-6; NA: noradrenaline; SIH:
`somatostatin; TNF: tumor necrosis factor.
`
`Figure 3. Serum levels of cortisol, separately for music group and control group (a), and for participants
`in whom mood increased vs. decreased over the course of the experiment (b) (as indexed by the difference
`between pre- and post-values of the mood questionnaire). Mean serum levels are indicated by the circles,
`separately for each of the six time points (error bars represent SEM). Dotted lines are spline interpolations of
`mean values, thus approximating the time course of parameter levels between the measured time points.
`
`t-tests on the standardized cortisol values showed significant differences between the music and the control group
`for TP 3 and TP 4, p < 0.05, a marginally significant difference at TP 5, p < 0.07, and a clear difference at TP 6,
`p < 0.005). To substantiate these findings we computed a discriminant function analysis (DFA, probability for
`entry was p = 0.05, for removal p = 0.10) with group (music, control) as group variable, and standardized values
`
`6
`
`www.nature.com/scientificreports/
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`

`

`Cortisol
`NA
`ACTH
`IL-6
`Leptin
`SIH
`TNF-α
`
`TP x Stimulus group (music, control)
`F = 3.77 (3.015); p = 0.011
`F = 0.24 (3.130); p = 0.877
`F = 1.34 (4.145); p = 0.252
`F = 0.45 (3.151); p = 0.725
`F = 1.62 (4.749); p = 0.156
`F = 1.81 (3.652); p = 0.132
`F = 1.46 (4.12); p = 0.211
`
`TP x POMS group (increase, decrease)
`F = 4.91 (2.986); p = 0.002
`F = 0.92 (3.093); p = 0.436
`F = 1.3 (4.063); p = 0.268
`F = 0.79 (2.957); p = 0.498
`F = 1.09 (4.838); p = 0.365
`F = 0.23 (3.606); p = 0.905
`F = 2.74 (4.081); p = 0.028
`
`Table 3. Statistical values of interactions between TP and group, as indicated by the MANOVAs reported
`in the main text (F- and p-values, with degrees of freedom in parentheses), pooled across all n = 143
`participants. Significant interactions (p < 0.005, corrected for multiple comparisons) are marked in bold.
`ACTH: adrenocorticotropic hormone; IL-6: interleukin-6; NA: noradrenaline; SIH: somatostatin; TNF: tumor
`necrosis factor; TP: Time point.
`
`of cortisol, NA, and ACTH as dependent variables. 61% of cross-validated cases were correctly classified (Wilks’
`, p < 0.005) based on two variables, which were (standardized) cortisol levels at TP 5 and TP 6.
`λ = 0.9,  = .10 4
`Music was causal for a more positive mood in the music group: Participants with mood increase from pre- to
`post-measurements (as measured with the POMS) were clearly over-represented in the music group ( = .10 4
`,
`p = 0.001). In the music group, mood increased in 41, and decreased in 27 participants. In the control group, on
`the other hand, mood decreased in 47, and increased in 23 participants. Thus, not all participants in the music
`group experienced a positive effect of the music on their mood.
`To assess effects of mood on the recovery of the stress test in more detail, participants were divided into
`those in whom mood increased or decreased over the course of the experiment (as indexed by the difference
`between pre- and post-values of the POMS). This group membership had a significant impact on the effects of
`the CO2-stressor on cortisol levels, as reflected in a significant interaction of time point x POMS group (mood
`increase, mood decrease) for cortisol (F = 4.91; p = 0.002; see Fig. 3b; for details see right column of Table 3). This
`interaction was due to group differences in cortisol at TPs 2–6, during which (standardized) cortisol levels were
`significantly higher in the group with mood increase compared to the group with mood decrease (two-samples
`t-tests on the standardized cortisol values showed significant differences between the music and the control group
`for TP 2: p < 0.05, TP 3 & TP 4: p < 0.005, TP 5: p < 0.01, and TP 6: p = 0.001).
`Valence and arousal ratings. Average valence ratings (provided on a 9-point scale ranging from −4 to +4)
`were M = − 0.33 (SD = 1.68) in the control group (and not significantly different from zero according to a
`two-tailed one-sample t-test testing against zero, p > 0.1) and M = 2.06 (SD = 1.13) in the music group, the dif-
`ference between groups being significant (T(124) = 10, p < 0.0001). Thus, the control stimulus was perceived as
`neutral, whereas the music stimulus was rated as positive. Average arousal ratings (provided on a 9-point scale
`ranging from 0 to 8) did not differ between control group (M = 2.88, SD = 1.29) and music group (M = 2.64,
`SD = 1.29).
`Circadian rhythm and sex. Data on the circadian rhythm of serum parameters and of differences between
`males and females are reported and discussed in the Supplementary Information. Note that neither circadian
`rhythm (i.e., time of day), nor sex, interacted with the factor time-point in any of the MANOVAs (for details see
`Supplementary Information)
`Discussion
`The CO2-stressor elicited significant changes in serum levels of NA, ACTH, cortisol and IL-6, and the effect of
`the stressor on leptin approached statistical significance (no effect of the stress test was observed for SIH, nor for
`TNF-α ). The effect of the stress test on NA replicates previous findings7, and the effect on cortisol is consonant
`with previous studies investigating effects of CO2-stress on serum cortisol levels6. In addition, our results provide
`the first evidence for effects of a CO2-stressor on ACTH levels. Thus, our results corroborate the notion that
`CO2-stress incites a stress response involving both the HPA axis (ACTH and cortisol) and sympathetic nerv-
`ous system activity (NA). As a note of caution, it is likely that, in addition to the CO2-stressor, the insertion of
`the intravenous cannula (and anxiety associated with the venipuncture) contributed to the stress-related effects
`(this issue is also discussed further below).
`More importantly, our results indicate that a cytokine (IL-6) was also affected by the acute stressor. This sug-
`gests that acute physiological stress affects IL-6 levels, independent of physical stress (e.g., exercise) or psycho-
`logical stress (e.g., mental or social stress), extending meta-analytic data on effects of acute psychological stress
`on IL-6 serum levels (provided by Steptoe et al.)41. IL-6 levels were lowest at time point 3 (around 20 min after
`the stress test), and highest at time point 6 (i.e., at the end of the recovery period, more than 1 h after the stressor;
`see also Fig. 2). Again, this is consistent with the meta-analysis by Steptoe et al. (ref. 41), in which effects of acute
`psychological stress on IL-6 serum levels were stronger in studies sampling 30–120 min post-stress (compared
`to studies sampling immediately after a stressor). The time course of IL-6 responses has not been established
`previously. Our data indicate that the IL-6 peak response does not occur before 60 min after the acute physio-
`logical stressor. The findings of effects of acute stress on IL-6 levels contradict data showing that stress-related
`HPA and sympathetic activity simply inhibits the function of pro-inflammatory cytokines such as IL-610. Thus,
`
`12
`
`22
`
`7
`
`www.nature.com/scientificreports/
`
`Scientific RepoRts | 6:23008 | DOI: 10.1038/srep23008
`
`

`

`while immune function is depressed by chronic stress (as shown in previous studies), immune function may be
`enhanced by acute stress, in line with re