Blood Levels of S-100 Calcium-Binding Protein B, High-Sensitivity C-Reactive Protein, and Interleukin-6 for Changes in Depressive Symptom Severity after Coronary Artery Bypass Grafting: Prospective Cohort Nested within a Randomized, Controlled Trial

Background Cross-sectional and retrospective studies have associated major depressive disorder with glial activation and injury as well as blood–brain barrier disruption, but these associations have not been assessed prospectively. Here, we aimed to determine the relationship between changes in depressive symptom severity and in blood levels of S-100 calcium-binding protein B (S-100B), high-sensitivity C-reactive protein, and interleukin-6 following an inflammatory challenge. Methods Fifty unselected participants were recruited from a randomized, controlled trial comparing coronary artery bypass grafting procedures performed with versus without cardiopulmonary bypass for the risk of neurocognitive decline. Depressive symptom severity was measured at baseline, discharge, and six-month follow-up using the Beck Depression Inventory II (BDI-II). The primary outcome of the present biomarker study was acute change in depressive symptom severity, defined as the intra-subject difference between baseline and discharge BDI-II scores. Blood biomarker levels were determined at baseline and 2 days postoperative. Results Changes in S-100B levels correlated positively with acute changes in depressive symptom severity (Spearman ρ, 0.62; P = 0.0004) and accounted for about one-fourth of their observed variance (R2, 0.23; P = 0.0105). This association remained statistically significant after adjusting for baseline S-100B levels, age, weight, body-mass index, or β-blocker use, but not baseline BDI-II scores (P = 0.064). There was no statistically significant association between the primary outcome and baseline S-100B levels, baseline high-sensitivity C-reactive protein or interleukin-6 levels, or changes in high-sensitivity C-reactive protein or interleukin-6 levels. Among most participants, levels of all three biomarkers were normal at baseline and markedly elevated at 2 days postoperative. Conclusions Acute changes in depressive symptom severity were specifically associated with incremental changes in S-100B blood levels, largely independent of covariates associated with either. These findings support the hypothesis that glial activation and injury and blood–brain barrier disruption can be mechanistically linked to acute exacerbation of depressive symptoms in some individuals.


Design, setting, and participants
This is a prospective cohort study nested within a randomized, controlled trial. An unselected group of 50 participants was recruited from among 201 patients undergoing coronary artery bypass grafting in the context of a randomized, controlled trial comparing conventional vs off-pump procedures for the risk of neurocognitive decline (Scarecrow trial, 2001-2004) [98]. Inclusion criteria were ages 40 to 80 years, clinical indication for urgent or elective coronary artery bypass grafting, referral to the Cardiothoracic Surgery service at Eastern Maine Medical Center (Bangor, Maine, United States), and capacity to provide written informed consent. Exclusion criteria were concomitant surgical procedures, severe calcification of the ascending aorta or deep intra-myocardial left anterior descending coronary artery, recent administration of an inotropic agent in excess of 3 mg/kg/min, requiring a cardiac-assist device for hemodynamic instability, and not providing written informed consent. Details of the surgical and anesthesia protocols used in the trial have been described previously [98]. All participants provided written informed consent. The Institutional Review Board of Eastern Maine Medical Center approved this study.

Biochemical measurements
Blood samples were collected by standard venipuncture at baseline and approximately 2 days postoperative. Upon collection, samples were stored at 280uC and then transferred on dry ice to the Laboratory for Clinical and Biochemical Research (Colchester, Vermont, United States) where S-100B, hs-CRP, and interleukin-6 blood levels were determined in either plasma or serum. Enzyme-linked immunosorbent assays were used to determine S-100B (#364701D, DiaSorin, Bromma, Sweden) and interleukin-6 blood levels (#HS600B, R&D Systems, Minnesota, United States). An automated BNII nephelometer was used to perform an immunonephelometric assay to determine hs-CRP blood levels (Siemens Healthcare Diagnostics, Illinois, United States). All laboratory personnel were blinded to participant characteristics and clinical information.
Lower limits of detection were 0.02 mg/l for S-100B, 0.16 mg/l for hs-CRP, and 0.16 pg/ml for interleukin-6. Inter-assay coefficients of variance at baseline and 2 days postoperative were 6.01% and 3.31% for S-100B, 2.93% and 1.82% for hs-CRP, and 7.95% and 8.71% for interleukin-6. In determining S-100B and interleukin-6 blood levels, each sample was run twice, the arithmetic mean of which was taken as the final value. In determining hs-CRP blood levels, each sample was run once, as the assays were automated. The approximate upper limits of normal blood levels for each biomarker are 0.15 mg/l for S-100B, 10.0 mg/l for hs-CRP, and 5.0 pg/ml for interleukin-6. hs-CRP levels less than or equal to 1.0 mg/l are indicative of low risk for ischemic heart disease, 1.0 to 3.0 mg/l moderate risk, and 3.0 to 10.0 mg/l high risk. hs-CRP levels greater than or equal to 10.0 mg/l are indicative of acute systemic inflammation of a noncardiac origin (e.g., as occurs with infection or cardiac surgery). The change (i.e., intra-subject difference) between blood levels of each biomarker at baseline and 2 days postoperative was computed for each participant. Instances where 2-day postoperative blood levels were greater than baseline levels (for a given biomarker and participant), were considered a positive change (i.e., perioperative increases). Likewise, when 2-day postoperative blood levels were less than baseline levels, this was considered a negative change (i.e., perioperative decreases).

Main outcomes
Depressive symptom severity was assessed at baseline, discharge, and six-month month follow-up using the Beck Depression Inventory II (BDI-II). Scores ranging from 0 to 13 indicate subclinical or minimal depressive symptom severity, 14 to 19 mild, 20 to 28 moderate, and 29 to 63 severe [99]. Participants, family members, treating clinicians, BDI-II test examiners, and all other trial staff in contact with participants were blinded to the results of blood biomarker levels. The primary outcome of the present biomarker study, acute change in depressive symptom severity, was defined as the intra-subject difference between baseline and discharge BDI-II scores. The secondary outcome, delayed change in depressive symptom severity, was defined as the intra-subject difference between baseline and six-month follow-up BDI-II scores.

Statistical analysis
Analyses were done using Stata version 12.1 for Mac (StataCorp, LP. College Station, Texas, United States). Participant characteristics were summarized as the arithmetic mean 6 standard deviation for continuous data and frequency and percentage for categorical data. These descriptive statistics were then compared between the biomarker and non-biomarker cohorts that comprised the Scarecrow trial with the use of two- tailed Student t tests for continuous data and Cochran-Mantel-Haenszel x 2 tests for categorical data. To identify covariates of blood biomarker measures at baseline and their differences between baseline and postoperative assessments and/or the main outcomes, Spearman rank correlation coefficients (Spearman r) were calculated for all possible two-term combinations, applying the pairwise deletion method for comparisons involving missing data. Participant characteristics that yielded statistically significant   Missing data for: n = 3; n = 3. || Acute D refers to the primary outcome, acute change in depressive symptom severity, defined as the intra-subject difference between BDI-II scores at baseline and discharge.

"
Delayed D refers to the secondary outcome, delayed change in depressive symptom severity, defined as the intra-subject difference between BDI-II scores at baseline and six-month follow-up. doi:10.1371/journal.pone.0111110.t001 Table 2. Spearman rank correlation coefficients for comparisons between participant characteristics and main outcomes as well as blood biomarker levels. { Acute D refers to the primary outcome, acute change in depressive symptom severity, defined as the intra-subject difference between BDI-II scores at baseline and discharge.

BDI-II
Delayed D refers to the secondary outcome, delayed change in depressive symptom severity, defined as the intra-subject difference between BDI-II scores at baseline and six-month follow-up. 1 Total D refers to the intra-subject difference in blood biomarker levels between baseline and 2 days postoperative. || Within 7 days before the index procedure.
Delayed D refers to the secondary outcome, delayed change in depressive symptom severity, defined as the intra-subject difference between BDI-II scores at baseline and six-month follow-up.
correlations were entered into the relevant multiple linear regression models (see below).
Spearman r values were calculated and scatterplots generated for all possible two-term combinations between and among blood biomarker measures at baseline and postoperative assessments, and the difference between them and BDI-II scores at baseline, discharge, six-month follow-up, the difference between baseline and discharge BDI-II scores (primary outcome), and the difference between baseline and six-month follow-up BDI-II scores (secondary outcome). For statistically significant correlations between blood biomarker measures and the main outcomes, we did sensitivity analysis by recalculating Spearman r values using Bonferroni-adjusted significance levels and again using Sidakadjusted significance levels. Last, the main outcomes were regressed on blood biomarker measures at baseline and the difference between blood biomarker levels at baseline and postoperative assessments using ordinary-least squares simple and multiple linear regressions. Before entering terms into the models, variable transformations were first applied to all variables that did not have a normal distribution. The transformation used for a given variable was based on the lowest x 2 value yielded by a ladder of powers analysis [100].

Biomarker cohort
The pairwise distributions of BDI-II scores and blood biomarker levels are displayed in Figure 1. Participant characteristics and BDI-II scores were generally similar among those comprising the biomarker cohort relative to the remaining participants in the Scarecrow trial (Table 1). One exception was discharge BDI-II scores, which were lower on average among those in the biomarker cohort (Table 1).

Covariate assessment
Spearman r values for comparisons between participant characteristics and the main outcomes and blood biomarker measures are displayed in Table 2. None of the participant characteristics showed a statistically significant correlation with either outcome (Table 2). Intra-subject differences between baseline and postoperative S-100B levels correlated positively with age and negatively with weight and body-mass index, and were significantly lower on average among those taking b-blockers (Table 2). Perioperative increases in hs-CRP levels correlated negatively with age. Perioperative increases in hs-CRP and interleukin-6 levels were significantly greater on average among participants randomized to (intention-to-treat) and having undergone (as-treated) off-pump procedures, as compared with those randomized to and having undergone conventional procedures ( Table 2). Spearman r values and scatterplots for all possible twoterm combinations between and among BDI-II scores and blood biomarker measures are displayed in Table 3 and Figure 2.

Acute change in depressive symptom severity
Intra-subject differences between baseline and postoperative S-100B levels exhibited a strong positive correlation with intra-subject differences between baseline and discharge BDI-II scores (Spearman r, 0.62; P = 0.0004; Table 3; Figure 3). Neither the strength nor the statistical significance of this correlation was affected by Bonferroni-or Sidak-adjusted significance levels (for both adjustments: Spearman r, 0.62; P = 0.0004). There were no other statistically significant correlations between blood biomarker measures and the primary outcome (Table 3). Significant negative correlations were observed between baseline BDI-II scores and changes in S-100B levels and between baseline BDI-II scores and acute changes in depressive symptom severity ( Table 3).
The results of simple and multiple linear regressions of the primary outcome on blood biomarker measures are displayed in Table 4. A simple linear regression model showed that each 0.10mg/l increase in intra-subject differences between baseline and postoperative S-100B levels was associated with a 1.  Table 4; Figure 3). This association remained statistically significant after individual adjustments for baseline S-100B levels, age, weight, body-mass index, and b-blocker use, but not baseline BDI-II scores (R = 9.18, R 2 = 0.40, t = 1.94, P = 0.064) adjusting for which improved the explanatory power of the model overall (b = 0.32, R 2 adjusted = 0.35, F (2,25) = 8.29, P model = 0.0017; Table 4). There were no statistically significant associations between the primary outcome and the remaining blood biomarker measures (Table 3; Table 4).

Delayed change in depressive symptom severity
Spearman r values and scatterplots for all possible two-term combinations between the secondary outcome and blood biomarker measures are displayed in Table 3 and Figure 2. There were no statistically significant correlations between blood biomarker levels and the secondary outcome ( Table 3). The results of simple and multiple linear regressions of the secondary outcome on blood biomarker levels are displayed in Table 5. None of these models, unadjusted or adjusted, yielded a statistically significant relationship between blood biomarker levels and delayed change in depressive symptom severity (Table 5).

Discussion
This is the first study, to our knowledge, to have prospectively assessed the relationships between blood biomarkers of glial  activation and injury, blood-brain barrier disruption, and depressive symptom severity in the context of the systemic inflammatory challenge paradigm of coronary artery bypass grafting. We found that perioperative increases in the glial activation and injury and blood-brain barrier-disruption biomarker, S-100B, exhibited a strong positive correlation with acute changes in depressive symptom severity, largely independent of covariates associated with either. Given the magnitude of the test statistic for this correlation (P = 0.0004), the statistical significance of this finding cannot readily be attributed to the fact that we did multiple pre-specified comparisons that included assessing the relationship between depressive symptoms and two other biomarkers besides S-100B (a/comparisons = [0.05]/[3] = 0.001. 0.0004). We also found that baseline BDI-II scores correlated negatively with both perioperative increases in S-100B levels as well as acute changes in depressive symptom severity, and that adjusting for baseline BDI-II scores diminished the strength and statistical significance of their association (the term for perioperative increases in S-100B levels still approached statistical significance at P = 0.064). Adjusting for baseline BDI-II scores also increased the proportion of variance in the primary outcome explained by the model (from 23% to 35%). Exploratory analyses indicated that the association between perioperative increases in S-100B levels and acute changes in depressive severity remained statistically significant after adjusting for baseline BDI-II scores among participants with BDI-II scores greater than or equal to 5 at baseline (Spearman r, 0.66; P = 0.0007). This was not the case, however, among those with BDI-II scores less than 5 at baseline (Spearman r, 0.15; P = 0.15).
Two earlier studies assessed relationships between depression and inflammation in the context of coronary artery bypass grafting [38,39]. In the first of these, Yang and colleagues found that mean levels of hs-CRP at baseline were significantly higher among those with clinical depression at six-month follow-up (defined as a Patient Health Questionnaire (PHQ-9) score $9) compared with those who were not [38]. By comparison, the correlation between baseline blood levels of hs-CRP blood levels and six-month BDI-II scores in the present study was not statistically significant. A possible explanation for this apparent discrepancy is that sixmonth follow-up depressive symptom severity was defined as a dichotomous outcome in Yang and colleagues' study, whereas it was defined as a continuous outcome in the present study. Supporting this interpretation, results of an exploratory analysis we performed showed that baseline hs-CRP blood levels were significantly higher among participants who were clinically depressed at six-month follow-up (defined as a BDI-II score $ 14), as compared with participants who were not (26.9641.1 mg/l (n = 4) vs 5.666.3 mg/l (n = 21); mean difference, 21.3 mg/l; 95% confidence interval, 3.3 to 39.3 mg/l; P = 0.022). The same between-group comparison for interleukin-6 blood levels at baseline approached statistical significance (16.0615.4 pg/ml (n = 4) vs 7.067.9 pg/ml (n = 21); mean difference, 8.5 pg/ml; 95% confidence interval, 1.4 to 19.4 pg/ml; P = 0.09). In the second study, Poole and colleagues found that elevated hs-CRP levels at postoperative day four mediated the association between depression and an increased length of stay [39]. Differences in the aims and designs of this study compared with the present one limit the ability to directly compare and contrast their findings.
The present study has limitations. First, some studies have shown that cardiopulmonary bypass can produce a transient elevation in blood S-100B levels, potentially confounding the Table 4. Cont.  observed correlation between perioperative increases in S-100B levels and acute changes in depressive symptom severity [101]. However, subsequent studies, which performed serial assays of blood S-100B levels among patients undergoing coronary artery bypass grafting found that increases in blood S-100B levels due to cardiopulmonary bypass are no longer apparent after 24 hours [102,103]. Accordingly, because we collected postoperative blood samples at about 48 hours, it is unlikely that perioperative increases in S-100B levels in our study were related to cardiopulmonary bypass. Supporting this interpretation, perioperative increases in S-100B levels among participants in our study did not differ significantly between patients who had cardiopulmonary bypass and those who did not. Second, whether the observed perioperative increases in blood S-100B levels are reliably attributable to glial activation and injury and blood-brain barrier disruption rather than non-cardiac extracerebral sources remains subject to interpretation. This concern arises primarily in the context of two animal model studies having shown a disconnect between S-100B levels in blood and those in CSF or brain parenchyma [68], and one recent cross-sectional study that found a negative correlation between CSF levels of S-100B and depressive symptom severity among patients with non-inflammatory neurological disorders [62]. Arguing in the opposite direction, however, other studies have shown that extracerebral sources do not influence blood S-100B levels [104]. In addition, a recent study of patients undergoing coronary artery bypass grafting found that S-100B levels in both blood and CSF increased together with those of the quintessential astroglial marker glial fibrillary acidic protein as well as CSF-to-serum albumin ratios [81]. Last, since data on antidepressant and statin use were not collected for the original Scarecrow trial, the effect of these drugs on the observed correlation between changes in S-100B levels and acute changes in depressive symptom severity cannot be excluded.
In conclusion, perioperative increases in S-100B levels exhibited a strongly positive correlation with acute changes in depressive symptom severity in the systemic inflammatory challenge paradigm of cardiac surgery. There were no significant associations between either acute or delayed changes in depressive symptom severity and baseline S-100B levels, baseline hs-CRP or interleukin-6 levels, or perioperative changes in hs-CRP or interleukin-6 levels. Among most participants, levels of all three biomarkers were normal at baseline and markedly elevated at 2 days postoperative. Taken together, these findings are consistent with the hypothesis that depression can be mechanistically linked to glial activation and injury as well as blood-brain barrier disruption in the context of systemic inflammation challenge paradigms, to include cardiac surgery. Large prospective cohort studies that can inform the reproducibility of these findings are warranted.

Acknowledgments
We thank all participants and staff involved in the Scarecrow trial.