Association of acute psychosocial stress with oxidative stress: Evidence from serum analysis

Abstract

Growing evidence implicates an association between psychosocial stress and oxidative stress (OxSt) although there are not yet reliable biomarkers to study this association. We used a Trier Social Stress Test (TSST) and compared the response of a healthy control group (HC; N=10) against the response of a schizophrenia group (SCZ; N=10) that is expected to have higher levels of OxSt. Because our previous study showed inconsistent changes in conventional molecular markers for stress responses in the neuroendocrine and immune systems, we analyzed the same serum samples using a separate reducing capacity assay that provides a more global measurement of OxSt. This assay uses the moderately strong oxidizing agent iridium (Ir) to probe a sample's reducing capacity. Specifically, we characterized OxSt by this Ir-reducing capacity assay (Ir-RCA) using two measurement modalities (optical and electrochemical) and we tuned this assay by imposing an input voltage sequence that generates multiple output metrics for data-driven analysis. We defined five OxSt metrics (one optical and four electrochemical metrics) and showed: (i) internal consistency among each metric in the measurements of all 40 samples (baseline and post TSST for N=20); (ii) all five metrics were consistent with expectations of higher levels of OxSt for the SCZ group (three individual metrics showed statistically significant differences); and (iii) all five metrics showed higher levels of OxSt Post-TSST (one metric showed statistically significant difference). Using multivariant analysis, we showed that combinations of OxSt metrics could discern statistically significant increases in OxSt for both the SCZ and HC groups 90 min after the imposed acute psychosocial stress.

Graphical abstract

Highlights

• •Ir-reducing capacity assay (Ir-RCA) provides a robust global measure of oxidative stress in serum.
• •The multiple oxidative stress (OxSt) output metrics of this Ir-RCA are useful for data-driven analysis.
• •The combination of OxSt metrics can discern significant increases in OxStwithin 90 mins of an imposed psychosocial stress.

The primary conclusions from this study are: the Ir-RCA provides robust global measurements of OxSt that can detect changes over comparatively short time scales (i.e., hours); and the results from this study support growing evidence for an association between psychosocial and oxidative stress.

1. Introduction

Fig. 1a illustrates that the biological response to external stresses (e.g., psychosocial stresses) involves a complex interconnection among the nervous, endocrine, and immune systems [1,2]. In response to stress, the activated hypothalamic-pituitary-adrenal (HPA) axis in the neuroendocrine system leads to secretion of the stress hormone cortisol that serves to interconnect the CNS with the immune system [[3], [4], [5], [6], [7]]. Also, activation of the immune system can result in the generation of cytokines that affect the CNS (e.g., neuroinflammation) and other systems [5,[8], [9], [10], [11], [12]]. Thus, it appears that cortisol and cytokines are key components for bidirectional communication between the CNS and immune system.

Fig. 1.

Psychosocial stress and oxidative stress. (a) The integrated biological stress response system includes the central nervous, endocrine, and immune systems. (b) Cortisol and pro-inflammatory cytokines were used as molecular biomarkers in a prior study involving acute psychosocial stress (results are adapted with permission under a Creative Commons Attribution License from Ref. [18]). The same serum samples were used for the work reported here. (c) Summary of the Trier Social Stress Test procedure.

Previously, we designed a pilot study to investigate the response to acute psychosocial stresses by the endocrine system (as measured by cortisol) and the immune system (as measured by the pro-inflammatory biomarkers, TNF-α, and IL-6). Because the stress-response system of Fig. 1a is implicated in the pathology of schizophrenia [[13], [14], [15], [16], [17]], we also focused on a comparison between persons diagnosed with schizophrenia (SCZ; N=10) and healthy controls (HC; N=10) [18]. In that study, we employed the Trier Social Stress Test (TSST) protocol that is one of the most widely used laboratory stress tests to examine the effects of acute stress on psychological and physiological functioning in humans [[19], [20], [21], [22], [23], [24], [25]]. The results from that previously-published study are replotted in Fig. 1b and show significant changes in cortisol levels during that short term study. Specifically, the HC group showed a transient increase in cortisol levels following the imposed stress, while the SCZ group showed a markedly different response pattern. The results with the two pro-inflammatory biomarkers were more difficult to interpret. At all measurement times, these two pro-inflammatory biomarkers were lower for the SCZ vs HC group which is inconsistent with evidence linking inflammation and schizophrenia [[26], [27], [28]]. In addition, both groups showed trends of gradual increases in IL-6 but no changes in TNF-α: recent studies suggest psychological stress activates the immune system [1,15,29,30] although inconsistent responses of inflammatory biomarkers across studies have been reported [9,26,31,32]. These discrepancies in inflammation biomarkers have been explained by: i) the immune response susceptibility to various confounding factors [9,33]; ii) inter-individual differences [26,34]; and iii) inconsistencies in sample collection, storage and measurement protocols [26,35].

Fig. 1a also suggests that psychosocial stress is linked to oxidative stress (OxSt) although the mechanistic details remain unclear. While OxSt is believed to result from the excessive generation of reactive oxygen species (ROS), there is growing evidence that ROS also perform important redox signaling functions [[36], [37], [38], [39]] and are embedded within a broader network of redox reactions (i.e., the redox interactome [36,40]) important for maintaining homeostasis. Further, this redox network appears to serve as an interface: at the molecular level with other biological networks (e.g., transcription and protein interaction networks) [41,42]; and at a systems level between exposure to various internal and external stressors (the exposome) [43] and the biological stress-responses [7,[44], [45], [46]]. Thus, there is increasing interest in understanding the relationships among ROS, redox signaling, OxSt and the integrated stress response systems [7,36,40,47,48].

There is no generally-accepted measurement for OxSt but various molecular biomarkers have been considered [[49], [50], [51], [52], [53], [54]]. Such molecular biomarkers include the reactive oxidants (e.g., ROS and RNS) [50,54,55], the endogenously generated protective antioxidants (e.g., GSH and ascorbic acid) [56,57], proteins for antioxidant protection (e.g., defense enzymes) or the ROS-induced damage (e.g., lipid peroxidation and protein carbonylation) [[58], [59], [60], [61], [62], [63], [64]]. An alternative approach to assess OxSt is to obtain a global measurement of antioxidant status that integrates the contributions from various molecular species [[65], [66], [67], [68], [69]].

We are developing a new global measurement for the OxSt using an Ir-reducing capacity assay (Ir-RCA) that probes a serum sample using the moderately strong oxidant K₂IrCl₆(IV) (designated Ir^OX) [70,71] to determine the electron-donating activities of the various serum components (e.g., glutathione, ascorbate and proteins). In our first report [70], we showed that the Ir^OX could be reduced upon incubation with diluted serum sample, which could be detected by electrochemical and optical measurements. In addition, the Ir-RCA results were compared with a commercial Cu-reducing capacity assay (Cu-RCA). Ir-RCA could discern higher levels of OxSt in the serum from a SCZ group (N=10) vs a HC group (N=5) while a commercial Cu-RCA could not discern differences in OxSt between the SCZ and HC groups. In our second report [71] using only an optical measurement, we validated that the Ir-RCA can discern higher OxSt in the SCZ group (N=73) vs HC group (N=45).

Here, we modified the Ir-RCA and investigated its ability to discern changes in OxSt in response to acute psychosocial stress. Specifically, we modified the assay to detect reducing capacity in serum using independent optical and electrochemical measurements, and we imposed a sequence of electrical voltage inputs to enrich the output signal to generate multiple metrics suitable for data-driven analysis. We applied this assay method to the same serum samples collected from our previous study (i.e., in Fig. 1b [18]). Consistent with our previous studies [70,71], we observed that this Ir-RCA can discern higher levels of OxSt in the SCZ group vs the HC group. More importantly, our measurements showed that increases in OxSt for both groups were observable within 2 h of the imposed psychosocial stress. By incorporating an enhanced data analytics approach, our results demonstrate the value of an Ir-RCA in providing a global indicator of OxSt. Moreover, our data reinforces the link between psychosocial stress and oxidative stress.

2. Materials and methods

2.1. Chemicals

The following were purchased from Sigma-Aldrich: K₂IrCl₆ (IV), and phosphate buffered saline (PBS, pH 7.4). The water (>18 MΩ) used in this study was obtained from a Super Q water system (Millipore). A stock solution of 10 mM K₂IrCl₆ (Ir^OX) was prepared in PBS (pH 7.4) and its aliquot was stored in −80 °C freezer.

2.2. Study participants

The recruitment of participants was described in the previous report [18] and more details are provided in the Supplementary Material.

2.3. Design of the study

This study consists of two visits as described in Fig. 1c.

2.3.1. Consent and screening in the first visit

In the first visit, all participants signed informed consent and completed a screening to evaluate eligibility. Details of screening process were previously described [18] and are provided in the Supplementary Material. During the first visit, the participants completed various clinical assessments to evaluate psychiatric symptoms including the Structured Clinical Interview for the DSM-IV Diagnoses (SCID) [72], Brief Psychiatric Rating Scale (BPRS) (18-item) [73], Hamilton Depression Scale (HAMD) [74], Scale for the Assessment of Negative Symptoms (SANS) [75], and the Clinical Global Impression Scale (CGIS) [76]. Participants also completed the Repeatable Battery for the Assessments of Neuropsychiatric Status (RBANS) [77].

2.3.2. Trier social stress test (TSST) in the second visit

Upon arrival of the second visit, participants had an intravenous catheter inserted and then the baseline blood was drawn (i.e., 15 min before TSST). After the participants watched the introducing video about the role-playing and prepared for their speech for 15 min, they returned to the room with the “panel of judges” to deliver their speech and complete the mental math (e.g., TSST). Blood was drawn just prior to the TSST (t=0), and 15, 30, 60, 90 min post stressor as described in Fig. 1c. In this study, we assayed the serum samples collected at 15 min before TSST (t-15 min: Baseline) and 90 min after TSST (t+90 min: post-TSST). More details for TSST are provided in Supplementary Material.

2.4. Biochemical analysis (cortisol, cytokines, C-reactive protein)

The analysis of serum cortisol and cytokines (IL-6 and TNF-α) was described in the previous study [18] and Supplementary Material. Here, we used the previously published results and we measured serum C-reactive protein (CRP) using a commercial Human C-Reactive Protein ELISA Kit (MilliporeSigma, MO, USA), following manufacturer's recommended protocol.

2.5. Ir-reducing capacity assay (Ir-RCA)

Before the measurement, we thawed each aliquot of serum and a 10 mM Ir^OX stock solution that had all been frozen at −80 °C. Sample solution (3000 μL) was prepared by diluting serum (1000-fold) and Ir^OX stock solution (20-fold) with 0.1 M PBS. After 1-h incubation at room temperature, two 200 μL aliquots of each sample were transferred to each of 2 wells of a 96 well microplate for duplicate optical absorbance measurements at 490 nm using a standard absorbance microplate reader (iMark, Bio-Rad Laboratories, Inc.). At the same time, 2600 μL of the samples were analyzed electrochemically (Fig. S1 provides more information of an electrochemical system); and an imposed voltage sequence that included a 2 min wait (i.e., OFF) period followed by a sequence of 3 separate voltage inputs (between each input the samples were stirred and a 2 min OFF period was used). Fig. 2b explains each imposed voltage input and its electrochemical output in detail.

Fig. 2.

Optical and electrochemical metrics of the Ir-reducing capacity assay. (a) A redox-mediator (K₂IrCl₆, Ir^OX) is used to detect oxidative stress by probing the reducing capacities of a diluted serum sample. During a 1 h incubation, Ir^OX is converted to Ir^Red by accepting electrons from the sample's reductants. Independent measures of the residual Ir^OX are obtained optically (Abs at 490 nm) and electrochemically (reductive charge; Q.RED), while measures of the generated Ir^Red are obtained electrochemically (oxidative charge; Q.OX). (b) For the electrochemical measurement, three sequential electrical input voltages are imposed: (i) E1 is a 2-min imposed oxidation voltage (+0.7 V vs Ag/AgCl) that yields one metric (the cumulative oxidative charge during these 2 min; Q.OX); (ii) E2 is a 2-min imposed reduction voltage (+0.1 V vs Ag/AgCl) that yields a second metric (the cumulative reductive charge during these 2 min; Q.RED); and (iii) E3 is a 10-min imposed cyclic voltage (CV) that generates 2 additional metrics (Q.CV1 and Q.CV3) that reflect the dynamic balance between the reduction of the residual Ir^OX and oxidation of the generated Ir^Red. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Two independent analyses of the 40 serum samples were performed over a 60-day period. At each analysis, each sample was measured in triplicate using both optical and electrochemical methods, and the triplicate measurements were averaged for the further data analysis.

2.6. Statistical analysis

Statistical analyses were performed with R version 3.4.3. Group differences in demographic information, clinical assessments, and Ir-reducing capacity measurements were assessed using Kruskal-Wallis test for continuous variables and chi-squared test for categorical variables. The intraclass correlation coefficient (ICC) is calculated using the ‘irr’ package [78] based on an absolute-agreement, one-way random effect analysis of variance (ANOVA) model [79,80]. Spearman's correlation coefficients were calculated to examine the relationship between the 5 OxSt metrics and the 5 clinical assessments. A logistic regression model [[81], [82], [83]] (‘nnet’ package) [84], was used to assess the combination effect of multiple OxSt metrics on i) the discerning ability of SCZ group from HC group and ii) discriminating ability of post-stressed status relative to baseline. We used Akaike Information Criterion (AIC) to estimate the quality of each model. The discriminating ability was evaluated using ROC (Receiver Operating Characteristic) metric with AUC (Area Under Curve) (i.e., c-statistic)) and p-values [[85], [86], [87], [88], [89], [90], [91]].

3. Results

Table 1 summarizes the characteristics of the HC group (N=10) and the persons with SCZ (N=10) at baseline study entry [18]. Table 1 shows that there is no significant difference between the two groups for age, gender, race, current smoking status or perceived stress. However, the psychiatric symptoms assessments (BPRS, SANS, HAMD, CGIS) shows the significantly higher scores and the RBANS shows significantly lower scores in the SCZ group compared with HC group as expected. With respect to physiological measurements, there is no significant difference in pulse or systolic blood pressure, while the SCZ group had a significantly higher diastolic blood pressure.

Table 1.

Demographic information for healthy controls and schizophrenia group (*; p<0.05).

Variable	Healthy Controls (HC, N =10)	Schizophrenia Group (SCZ, N=10)	Test Statistics
Age/years	35.4 ± 12.3	40.9 ± 14.7	χ2= 0.46, p = 0.49
Sex/Male	10 (50%)	10 (50%)	χ2= 0.20, p = 1
Race			χ2= 0.28, p = 1
African American	7 (70%)	7 (70%)
Caucasian	2 (20%)	3 (30%)
Mixed	1 (10%)
Smoker	3 (30%)	4 (40%)	χ2= 0.95, p = 0.62
Clinical Assessment (Symptoms)
*BPRS Total Score	19.6 ± 1.8	33.2 ± 5.3	χ2= 12.59, p = 3.9 × 10−4
*SANS Total Score	1.0 ± 2.8	25.7 ± 9.2	χ2= 15.26, p = 9.36 × 10−5
*HAMD	0.67 ± 1.4	8.8 ± 5.8	χ2= 11.61, p = 6.57 × 10−4
*CGIS	1.0 ± 0.0	4.5 ± 0.8	χ2= 16.59, p = 4.63 × 10−5
*RBANS Total Score	86.4 ± 15.1	71.0 ± 16.0	χ2= 4.66, p = 0.031
Perceived Stress Scale	11.2 ± 9.7	14.5 ± 5.2	χ2= 0.57, p = 0.45
Physiological Measurement
Systolic Blood Pressure (mmHg)	122.7 ± 21.7	117.7 ± 13.1	χ2= 0.036, p = 0.85
*Diastolic Blood Pressure (mmHg)	67.2 ± 10.8	74.9 ± 5.2	χ2= 4.18, p = 0.041
Pulse (bpm)	73 ± 14.4	83.9 ± 12.9	χ2= 1.56, p = 0.21

3.1. Description of Ir-RCA and OxSt metrics

Fig. 2a illustrates the redox probing method (i.e., Ir-RCA) used in this study. Ir^OX is incubated with diluted serum samples and accepts electrons from serum components converting the yellow-colored Ir^OX into its colorless reduced form (Ir^Red). This decrease in Ir^OX and increase of Ir^Red are measured both optically and electrochemically after 1-h incubation. The optical absorbance (Abs) at 490 nm measures the yellow-color associated with remaining Ir^OX in the sample after the 1-h incubation: samples from persons with higher levels of OxSt (i.e., less serum reducing capacity) are expected to have more Ir^OX and thus higher Abs values.

Fig. 2b shows that we used three independent electrochemical measurements of Ir^OX and Ir^Rred. Specifically, an electrode was inserted into the sample and three different input voltage sequences were imposed while measuring the output current responses. Importantly, electrochemistry measures a small region of the sample (we estimate ∼0.15% of the sample volume) [92] and thus allows multiple measurements to be performed while minimally altering the sample (note between each of the three voltage input sequences we stirred the sample to refresh the measurement region and then used a 2 min OFF time).

The first voltage input (designated E1) was a 2-min step change to an oxidative value (+0.7 V vs Ag/AgCl reference electrode). The number of electrons transferred during this oxidative pulse (Q.OX; μC) provides a measure of how much Ir^Red was formed during the 1-h incubation: Fig. 2b illustrates that samples from persons with higher levels of OxSt are expected to form less Ir^Red and show a smaller Q.Ox response (by convention, oxidative Q values have a negative value and thus a smaller response corresponds to a less negative Q.Ox).

The second voltage input (E2) was a 2-min step change to a moderately reductive value sufficient to reduce Ir^Ox (+0.1 V vs Ag/AgCl). The number of electrons transferred during this reductive pulse (Q.RED; μC) provides a measure of how much Ir^OX remained in the sample after the 1-h incubation: Fig. 2b illustrates that samples from persons with higher levels of OxSt are expected to have higher Ir^OX and thus larger Q.RED values.

The final voltage input (E3) was a 3-cycle (10 min) oscillating cyclic voltage (CV) between the oxidative value (+0.75 V) and a reductive value (+0.25 V). Fig. 2b shows an oscillating output Q is observed with a trend toward larger (i.e., reductive) values and we characterized this output response by two metrics: the first metric is the minimum charge after the first cycle (Q.CV1) and the second is the final charge after the third cycle (Q.CV3). This sequence provides an independent measure of the Ir^OX remaining in the sample after the 1-h incubation: Fig. 2b illustrates that serum from persons with higher levels of OxSt are expected to have higher Ir^OX levels and thus higher Q.CV1 and Q.CV3 values. Here, we defined five output metrics of Ir-RCA including Abs, Q.OX, Q.RED, Q.CV1 and Q.CV3 as OxSt metrics.

3.2. Reliability of Ir-RCA

To investigate the reliability of Ir-RCA, we analyzed the 40 samples (20 samples at baseline and 20 samples post TSST) at two different times that spanned 60 days. At each analysis, a single aliquot for each of the 40 frozen serum samples (-80°C) was thawed, and measured in triplicate to generate our 5 OxSt metrics (Abs, Q.Ox, Q.RED, Q.CV1 and Q.CV3). The triplicate measurements of each analysis were averaged to 1^st mean and 2^nd mean. Fig. 3 represents the correlation between two mean values of each sample for each metric.

Fig. 3.

Reliability of Ir-RCA. The cross plots of the individual metrics show good agreement between two independent analyses of 40 samples (20: Baseline, 20: Post TSST) that were performed at two different times spaced 60-days apart. Each point represents the mean value and the error bars are for triplicate measurements for each sample in the 1^st and 2^nd analysis. The intraclass correlation coefficient (ICC) is calculated to assess the agreement of 6 replicate measurements for each of the 5 metrics.

To quantify the reliability of our assay, we calculated the intraclass correlation coefficient (ICC) [79,93,94] which assesses the absolute agreement between different analyses. Each calculated ICC for 5 metric measurements was relatively high indicating a good agreement for the 6 replicate measurements. Importantly, the high reliability of this assay indicates that the assay is measuring a stable feature in the serum (i.e., it is not measuring unstable reactive species) and it is not sensitive to the presence of air (i.e., no precautions were made to exclude oxygen when drawing the blood, processing the serum or assaying the samples).

3.3. Correlation between OxSt metrics and clinical assessments

Fig. 4 presents the correlation heat map between 5 OxSt metrics and 5 clinical assessments at the baseline. Each colored cell of the heat map indicates a statistically significant correlation between variables. First, the lower left region of this heat map shows the strong correlations between the 5 OxSt metrics, which supports the internal consistency of the optical and electrochemical metrics of OxSt. Second, the upper right region of the heat map of Fig. 4 shows strong correlations between the 5 different clinical assessments (Fig. S2 provides correlation plots). The symptom severity-related assessments (BPRS, CGIS, SANS) show positive correlations with each other, but a negative correlation with RBANS scales that assess cognition [32,[95], [96], [97]].

Fig. 4.

Correlations of the metrics from the Ir-RCA with clinical assessments. Heat map of Spearman correlation coefficients of the 5 OxSt metrics and 5 clinical assessments (BPRS, RBANS, CGIS, SANS, HAMD) at baseline show: excellent internal consistency among the 5 OxSt metrics (lower left); strong internal consistency among the clinical assessments (upper right); and some correlation between OxSt metrics and clinical assessments (lower right). Q.OX metric shows correlation with BPRS and RBANS and Q.CV1 shows a correlation with BPRS and SANS: these correlations indicate that symptom severity is associated with OxSt.

Third, the lower right region of the heat map of Fig. 4 shows the correlations between the 5 OxSt metrics and the 5 different clinical assessments. Importantly, Fig. S2a shows the response of each metric is consistent with expectations that higher OxSt is associated with more severe symptoms [[98], [99], [100]] and lower cognition [95,97]. As indicated in the Table of Fig. 4, four of these correlations are statistically significant. The BPRS assessment of psychiatric symptoms shows a negative correlation with the Q.OX metric (r = −0.45, p = 0.047) and a positive correlation with the Q.CV1 metric (r = +0.53, p = 0.017). Also, the SANS assessment of negative symptoms shows a positive correlation with the Q.CV1 metric (r = +0.51, p = 0.02). The RBANS assessment of cognition shows a positive correlation with the Q.OX metric (r = +0.45, p = 0.048). Thus, the OxSt metrics are consistent with clinical expectations and, in some cases, statistically significant correlations were observed between individual metrics and clinical assessments despite the small sample size.

When we checked the correlation between the 5 OxSt metrics and other biochemical markers including cortisol and inflammation markers (CRP, IL-6, TNF-α) in Fig. S2a, only one metric, Q.OX, shows a positive correlation (r = +0.46, p = 0.005) with TNF-α. The relationship is not consistent with the other studies that show an elevation of TNF-α levels with higher level of OxSt biomarkers [101,102].

3.4. Discriminating abilities of individual OxSt metrics

In addition to identifying correlations to baseline clinical assessments, we also examined the abilities of the individual assay metrics to discern differences between the two groups (SCZ and HC) and the two states (pre- and post-TSST). Fig. 5a shows the ability of Ir-mediated OxSt assay to discriminate the SCZ group from the HCs at baseline and post-TSST using the 5 OxSt metrics. The first important observation in Fig. 5a is that all 5 metrics consistently show higher OxSt for the SCZ (vs HC) group both at baseline and post-TSST. The differences between SCZ and HC groups for three of these metrics was statistically significant (Q.OX at baseline (p=0.034) and post TSST (p=0.010), and Q.CV1 at baseline (p=0.034)). This observation agrees with our previous studies that show higher OxSt for persons diagnosed with SCZ vs healthy controls [70,71,103]. The second important observation in Fig. 5a is that all 5 metrics changed in the direction of increasing OxSt in post-TSST compared with baseline for both the SCZ and HC groups. Fig. 5b shows results when the groups were combined to consider only the effect of the TSST. Again, all five metrics showed the expected responses that OxSt was higher post-TSST compared to baseline and one metric showed a statistically significant difference (Q.OX (p=0.045)). These two observations - increasing OxSt for the SCZ group (vs. HC) and post-TSST (vs. pre-TSST) - are consistent with the expectations (Note: Fig. S3 provides additional plots to show the changes of OxSt metrics for individual samples pre/post TSST and Fig. S4 provides additional analysis of the difference of OxSt metrics between pre and post TSST).

Fig. 5.

Discriminating abilities of individual OxSt metrics. (a) Five OxSt metrics including one optical metric (Abs) and four electrochemical metrics (Q.OX, Q.RED, Q.CV1 and Q.CV3) consistently show the higher oxidative stress for the SCZ group vs HC group at baseline and post-TSST. Three metrics shows the statistically significant difference between the two groups (*; p<0.05, Q.OX at baseline, post-TSST, Q.CV1 at baseline). (b) Five OxSt metrics for the combined group show consistently increasing OxSt for post-TSST vs baseline. One metric (Q.OX) shows the statistically significant difference between baseline and post-TSST (*; p<0.05). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.5. Combining effect of multiple OxSt metrics

In Fig. 6, we investigated the additive effect of multiple metrics on discerning higher levels of oxidative stress using a logistic regression model [[81], [82], [83]]. First, Fig. 6a estimates the probabilities for discriminating the SCZ group from HC group using one or combinations of multiple OxSt metrics. The goodness-of-fit for each model is evaluated using the Akaike Information Criterion (AIC) [104] and the discriminating power is evaluated using ROC curve analysis [[86], [87], [88], [89], [90], [91]]. The table in Fig. 6a presents the calculated AIC and the AUC values (i.e., c-statistic) and the p-values from ROC curve for each combination of multiple metrics at the baseline and post-TSST (Fig. S5 provides the ROC analyzed results for each single metric). While the ROC analysis with a single metric (Abs) shows lower AUC with higher p-value (>0.05), the combination of OxSt metrics increased the AUC and the discriminating ability became statistically more significant (p-value < 0.05). This result indicates that the combination of multiple metrics increased the power for discriminating the SCZ and HC groups based on their levels of OxSt.

Fig. 6.

Receiver operating characteristic (ROC) analyses for assessing the combination effect of multiple metrics. (a) ROC analysis for discerning SCZ group from HC group at baseline and post-TSST. (b) ROC analysis for discerning post-TSST from baseline for HC group and SCZ group. ROC curves and Tables show that the combination of multiple OxSt metrics increased its discerning ability (SCZ vs. HC, Post-TSST vs. Baseline) with higher AUC (c-statistic) and lower p-value (*; p<0.05).

In Fig. 6b, a logistic model is used to estimate the probabilities to discern the post-TSST from baseline (pre-TSST) using one or combinations of multiple OxSt metrics. The table in Fig. 6b shows that the combination of OxSt metrics increases the AUC (c-statistic) and lowers the p-value. This result indicates that the combination of multiple OxSt metrics by logistic regression model can improve the ability to discern increases in OxSt post-TSST.

4. Discussion

There is growing evidence for an integrative stress response system [7,105,106] that includes the immune [11], nervous [2] and endocrine systems, and also that persons with schizophrenia may have dysfunctions in this stress response. In a previous pilot study using a short-time psychosocial stress [18], we observed that the response in cortisol levels (a marker of HPA axis response) for people with SCZ was markedly different from healthy controls (HC) as expected from other studies [6,7,105,107]. However, changes in the immune response biomarkers (TNF-α and IL-6) were less consistent with expectations. First, these measurements showed lower levels of TNF-α and IL-6 for the SCZ group compared to HC group (Fig. 1b), and these cytokine levels did not correlate to clinical measures of symptom severity (Fig. 4). These results are inconsistent with other studies and the neuroimmune hypothesis of schizophrenia [1,12,19,108,109]. Second, after psychosocial stress, IL-6 increased but TNF-α remained unchanged in both the SCZ and HC groups (Fig. 1b) which is also inconsistent across studies and expectations (based on a neuroimmune hypotheses). These inconsistencies suggest that pro-inflammatory cytokines may not be robust biomarkers [110].

There is also evidence that HPA activation and inflammation are linked to oxidative stress (OxSt) [19,26,111,112]. To examine this link, we used an Ir reducing capacity assay (Ir-RCA) to measure the same samples collected from this previous study [18]. In contrast to traditional molecular biomarkers or measures of oxidation [49,51,54,113,114], the Ir-RCA provides a more global measure that is detected through relatively simple optical and electrochemical methods. Here, we extended the Ir-RCA from a single optical [71] or electrochemical output metric [70,103] to multiple independently-measured output metrics useful for a data-driven analysis [41,[115], [116], [117]]. Each of the five output metrics showed that the SCZ group had higher OxSt compared with the HC group which is consistent with expectations [70,71,103] (3 of these metrics showed statistically-significant differences: Fig. 5). Further, each of the five output metrics showed that increased OxSt is associated with greater symptom severity and lower cognition which is also consistent with expectations [[118], [119], [120], [121], [122], [123]] (2 of these metrics showed statistically-significant correlation: Fig. 4). As expected, combining output metrics increased the discriminating power of the Ir-RCA (e.g., Fig. 6a).

Although many studies suggest that the chronic psychosocial stress is linked to OxSt and the pathologies of diverse diseases [[6], [7], [8],45,46,124,125], few studies focus on the association between acute psychosocial stress and OxSt [[126], [127], [128]]. Jansakova et al. [127] examined the acute effect of psychosocial stress on OxSt using saliva samples from children. They found increases in the total antioxidant capacity (as measured by ferric ion reducing antioxidant power) after exposure to acute psychosocial stress (e.g., TSST) during the stress day, and higher lipid peroxidation on stress day than control day but no significant difference prior to and post TSST. Wiegand et al. [126] found a higher expression of the microRNA, which is implicated in regulating the production of reactive species, in the saliva of healthy subjects after the psychosocial stress test (TSST). While both studies were performed in healthy groups, we investigated for the first time the acute effect of psychosocial stress on the OxSt in SCZ and HC groups. All five output metrics in our study showed an increase in OxSt after the TSST compared with the baseline (Fig. 5) and one metric (Q.OX) showed a statistically significant difference between baseline and post-TSST. Further, the results in Fig. 5 implicate an association between psychosocial stress and OxSt. The combination of OxSt metrics was well-fitted with the model to distinguish post-TSST from baseline (Fig. 6b). While our results indicate that the TSST test was associated with increased OxSt for both the SCZ and HC groups, our Ir-RCA could not discern differences in response between these two groups (using linear mixed effects model in Fig. S4).

There are several strengths and limitations of this study. One strength is that the Ir-RCA appears to access stable features of the serum samples. Specifically, we measured samples that had been stored at −80 °C for 3 years, and our replicate measurements that were performed 60 days apart showed strong interclass correlations (Fig. 3). Presumably, this assay is detecting stable differences in the oxidation state of proteins and other antioxidants that provide a stable historical record of the oxidative context that had been experienced by the serum at the time of sampling (e.g., trace levels of reactive species are expected to rapidly decay during sample processing while oxidized amino acid residues are expected to be stable) [71]. A second strength is the cost and simplicity of the Ir-RCA, as well as the ability to make independent measurements to facilitate a data-driven analysis. A third strength is that all 5 metrics measured in this study show responses consistent with expectations (greater oxidative stress for the SCZ group and for both groups post-TSST). One limitation of this study is that, since the optical and electrochemical metrics generated by the Ir-RCA reflect a summation of all oxidizable species (i.e., antioxidants) in the sample, it is generally not possible to relate the Ir-RCA measurements to individual chemical species or molecular mechanisms of OxSt. A second limitation of the Ir-RCA is that there are several assay variables that could be adjusted (e.g., mediator types/levels, electrochemical input sequences, and output response quantification) and optimizing this assay for clinical analysis is limited by the small volumes of sample available for methods development. A final limitation is the small sample size of this pilot study (N=10 schizophrenia and N=10 control). Thus, while this study supports the growing evidence for a link between psychosocial stress, inflammation, and oxidative stress, larger studies are needed to confirm these findings.

In summary, the results from this study indicate that acute psychosocial stresses can lead to detectable increases in oxidative stress over comparatively short times (<2 h).

Declaration of competing interest

All authors declare no conflicts of interest related to this work.

Contributors

EK, DLK, and GFP designed the study and wrote the first draft of the manuscript. JR built an electrochemical analyzer. EK performed the assay. MG collected serum samples. EK, ZZ and SC undertook the statistical analysis. All authors contributed to edit the manuscript and have approved the final manuscript.

Appendix A. Supplementary data

The following is the Supplementary data to this article: