Expression of HO1 and PER2 can predict the incidence of delirium in trauma patients with concomitant brain injury

Intensive care unit (ICU)-acquired delirium is associated with adverse outcome in trauma patients with concomitant traumatic brain injury (TBI), but diagnosis remains challenging. Quantifying circadian disruption by analyzing expression of the circadian gene period circadian regulator 2 (PER2) and heme oxygenase 1 (HO1), which determines heme turnover, may prove to be potential diagnostic tools. Expression of PER2 and HO1 was quantified using qPCR from blood samples 1 day and 7 days after trauma. Association analysis was performed comparing mRNA expression levels with parameters of trauma (ISS—injury severity score), delirium, acute kidney injury (AKI) and length of ICU stay. 48 polytraumatized patients were included (equal distribution of TBI versus non-TBI) corrected for ISS, age and gender using a matched pairs approach. Expression levels of PER2 and HO1 were independent of age (PER2: P = 0.935; HO1: P = 0.988), while expression levels were significantly correlated with trauma severity (PER2: P = 0.009; HO1: P < 0.001) and longer ICU length of stay (PER2: P = 0.018; HO1: P < 0.001). High expression levels increased the odds of delirium occurrence (PER2: OR = 4.32 [1.14–13.87]; HO1: OR = 4.50 [1.23–14.42]). Patients with TBI showed a trend towards elevated PER2 (OR = 3.00 [0.84–9.33], P = 0.125), but not towards delirium occurrence (P = 0.556). TBI patients were less likely to develop AKI compared to non-TBI (P = 0.022). Expression levels of PER2 and HO1 correlate with the incidence of delirium in an age-independent manner and may potentially improve diagnostic algorithms when used as delirium biomarkers. Trial registration: German Clinical Trials Register (Trial-ID DRKS00008981; Universal Trial Number U1111-1172-6077; Jan. 18, 2018).

www.nature.com/scientificreports/ The cytoprotective enzyme heme oxygenase 1 (HO1) serves as the rate-limiting step in heme metabolism and is strongly upregulated following various cellular stressors 19 . HO1 catalyzes the degradation of heme into iron, biliverdin and CO. Biliverdin is then further reduced to bilirubin by oxidation of NAD(P)H to NAH(P) + . HO1 expression may serve as a potential predictive blood biomarker since it is closely linked to circadian regulation as it: (a) catabolizes the heme moiety contained in the circadian regulatory proteins NPAS-2 and RevErb-α, and (b) produces carbon monoxide (CO) as a bioactive product. Endogenous CO in turn modulates the transcriptional activity of NPAS-2, RevErb-α and CLOCK that control expression of PER2 20 . Moreover, antioxidant qualities of HO1 (and CO) are well documented. Especially in the case of AKI HO1 induction has been identified as a rapid cytoprotective response in multiple animal models of AKI [21][22][23][24][25][26] .
Here, we hypothesized that PER2 and HO1 expression could serve as potential serum biomarkers to diagnose and predict delirium and/or AKI. We posited that alterations in PER2 and HO1 mRNA expression would indicate circadian disruption. Being at high risk to develop both delirium and AKI 27,28 , we tested this hypothesis in polytraumatized patients. As circadian rhythm is controlled by the hypothalamus, a trauma patient collective afforded us the opportunity to simultaneously investigate the effect of traumatic brain injury (TBI) on circadian disruption and clinical outcome. We studied a sub-group of patients with TBI and measured PER2 and HO1 mRNA expression and asked if expression correlated with subjects developing delirium and/or AKI.

Results
Patient characteristics and recruitment. A total of 72 patients were screened of which three patients were excluded due to pre-existing kidney disease. Of the remaining 69, a total of 48 polytrauma patients were included in the study (Fig. 1b). Using a matched pairs approach, two 24-patient groups with TBI and non-TBI patients were formed. These groups did not significantly vary regarding age at trauma, gender, trauma ISS, preexisting cognitive or cardiovascular disease, diabetes mellitus, time spent in the ICU and maximum leucocyte counts during the first week after trauma (Table 1).

HO1 expression correlates with bilirubin concentration. To verify that expression levels mirror
actual enzymatic activity, we tested HO1 expression and bilirubin concentration for correlation, as bilirubin is the final product of heme degradation. Indeed, HO1 expression levels on the first day correlated with concurrent bilirubin concentration (r = 0.315, P = 0.029, r 2 = 0.378, P < 0.001, Fig. 2a) on day one and after one week (a) Schematic illustration of hypothesized relationship between input (parameters of trauma) and output (negative outcome) parameters with mediating variables (circadian disruption and heme metabolism measured via PER2/HO1 expression). (b) Selection of polytrauma patients that were analyzed for circadian mRNA expression. Patients were sorted in matched pairs (criteria: gender, age, trauma severity (ISS injury severity score)) to form patient groups of TBI and non-TBI for later between-group comparison. TBI traumatic brain injury, ISS injury severity score, PER2 period circadian regulator 2, HO1 heme oxygenase 1, AKI acute kidney injury, ICU intensive care unit, LOS length of stay.
Biomarkers correlate with occurrence of delirium. A significant correlation was discovered between the expression of both PER2 and HO1 on day one after trauma and the maximum nuDesc (PER2: r = 0.319, P = 0.027, Fig. 4a; HO1: r = 0.409, P = 0.004, Fig. 4b). Correlation of delirium and PER2 expression remained significant 1 week post trauma (r = 0.366, P = 0.024, r 2 = 0.083, P = 0.048, Fig. 4c). In accordance with age as a major risk factor for the development of delirium 29 , maximum nuDesc in the first week after trauma correlated significantly with a patient's age at the time of trauma (r = 0.370, P = 0.010, r 2 = 0.105, P = 0.024, Fig. 4d). No correlation could be found between a patient's age at the time of trauma and PER2 and HO-1 expression (PER2: r = 0.012, P = 0.935, r 2 = 0.005, P = 0.634; HO1: r = 0.002, P = 0.988, r 2 = 0.029, P = 0.247, Fig. 4e), suggesting an age-independent correlation between these gene expression patterns and delirium. Trauma severity measured by ISS did not correlate with maximum nuDesc after trauma (r = 0.070, P = 0.638, r 2 = 0.001, P = 0.825, data not shown). Mean values of PER2 and HO1 mRNA expression on day one after trauma (P = 0.013) and after one week (P = 0.016 and P = 0.004 respectively) were significantly elevated in delirious patients (Table 2). When dividing patients into categories of high and low mRNA expression in relation to ISS scoring, delirium (nuDesc > 1) was significantly more frequent in patients within the group of high PER2 (P = 0.032, Fig. 4f) and HO1 (P = 0.019,  www.nature.com/scientificreports/   Fig. 4g).
Bilirubin concentration neither correlated with maximum nuDesc (r = 0.065, P = 0.663, r 2 = 0.011, P = 0.486, data not shown) nor were patients with high bilirubin levels more likely to develop delirium compared to those with low levels (P = 0.770, data not shown). Table 2, no significant differences in mRNA expressions were apparent for either PER2 or HO1 between groups of TBI and non-TBI patients. However, there was a tendency towards elevated values of PER2 on day one after trauma in TBI patients with a relative risk of 2.00 (95% CI 0.94-4.52) and an OR of 3.00 (95% CI 0.84-9.33, P = 0.125). A similar non-significant tendency was apparent when comparing the distribution of high and low PER2 mRNA expression on day one after trauma comparing TBI and non-TBI (P = 0.135, Fig. 5a). Distribution of high and low HO1 mRNA expression and bilirubin concentration showed no difference between TBI and non-TBI patients (HO1: P = 0.770; bilirubin: P = 0.561). Similarly, analysis of delirium occurrence in TBI and non-TBI patients revealed no difference in delirium when comparing TBI with non-TBI groups (P = 0.556).

TBI influences PER2 expression but not HO1 expression or occurrence of delirium. As shown in
TBI patients show smaller likelihood of developing AKI. As a second clinical readout, the occurrence of AKI was investigated in relation to the biomarkers defined above and in relation to the presence or absence of TBI. Compared with non-TBI trauma patients, TBI patients showed a significantly lower risk of developing AKI (P = 0.022, Fig. 5b). However, higher prevalence of AKI did not significantly coincide with either high PER2 (P > 0.999, Fig. 5c) or HO1 mRNA expression (P = 0.219). Patients with high bilirubin concentrations on day one non-significantly tended towards developing AKI (P = 0.073, data not shown).  (Fig. 5d). However, each of these associations remained non-significant (TBI: P = 0.609, high PER2: P > 0.999, high HO1: P > 0.999, high bilirubin: P = 0.103, delirium: P = 0.286, AKI: P = 0.071).

Discussion
In this study, we measured systemic PER2 and HO1 mRNA expression on day one after trauma. We conclude that both HO1 and PER2 may serve as novel blood biomarkers for trauma severity, length of ICU-stay and, most interestingly, the risk of developing trauma-associated delirium.
There is an unmet need for objective laboratory parameters to improve diagnosis and treatment algorithms for delirium. Many have been tested but have so far proven incapable of having diagnostic value in a clinical setting 11 . Delirious patients commonly demonstrate compromised sleep-wake-cycles, and molecular circadian disruption is considered to play a major role in delirium pathogenesis 2 . Nonetheless, circadian gene expression has not been investigated as potential biomarkers for delirium. Our data suggests that PER2 and HO1 mRNA expressions may prospectively assess delirium and trauma severity. Further we measured concentrations of bilirubin, the final product of heme degradation by HO1. By demonstrating a correlation between HO1 expression and bilirubin concentration we can confirm the notion that mRNA expression mirrors actual enzymatic activity. Table 2. Median HO1 and PER2 mRNA expression between groups of TBI/non-TBI and delirium/nondelirium over time. Sample acquisition on first day and after 1 week. P-values of between-group comparisons of PER2 and HO1 mRNA expression (median and IQR) calculated via Mann Whitney test between groups of patients with and without delirium as well as between TBI and non-TBI groups. PER2 period circadian regulator 2, HO1 heme oxygenase 1, TBI traumatic brain injury, IQR inter-quartile ratio. www.nature.com/scientificreports/ Expression of PER2 was not only significantly elevated on day one, but remained elevated one week post trauma. These findings suggest not only predictive value, as is the case with HO1, but also provided diagnostic value for occurrence and severity of trauma-associated delirium. Although the patient's age at trauma is a known major risk factor for developing delirium 29 it could not explain the predictive aspects of PER2 and HO1, suggesting age-independence and hence potentially additive diagnostic value. Trauma severity did not contribute to identifying patients at risk for trauma-associated delirium.
Although circadian disruption is thought to contribute to kidney disease 3,4 , no predictive or diagnostic capacity of PER2 or HO1 expression could be found for AKI. However, this may be due to the limited amount of AKI cases in our patient cohort. Work by others strongly suggests that HO1 plays a major role in stress response to acute kidney injury [21][22][23][24][25][26]30 . Indeed, patients with more GT-repeats in the HO1-gene promoter region have been shown to be less likely to develop AKI 30 . Therefore, further studies should investigate the predictive and diagnostic capacity of HO1 and PER2 for those with AKI.
Similarly, 6-month mortality could not be attributed to high HO1 and PER2 expression as well as TBI. With a total of only four deceased patients, our cohort of n = 48 was underpowered for this endpoint and similar to AKI, larger cohort studies might prove to yield more conclusive results. On a descriptive statistical level limited by our patient cohort, our data suggest that TBI and high PER2 expression may be favorable for survival, while high HO1 expression, high bilirubin, delirium and AKI carried a greater risk of death. The last two of which are already considered to contribute to increased mortality 5,6 .
Intracranial bleeding has been shown to cause disruption in circadian gene expression in mice 20 . It was investigated whether TBI has an effect on the likelihood of developing delirium and AKI. TBI did not show any significant impact on occurrence of trauma-associated delirium or systemic mRNA expression levels of PER2 and HO1. Thus, our findings cannot support our hypothesis of TBI on top of polytrauma causing more severe molecular circadian disruption and thereby contributing to the pathogenesis of delirium and AKI. It is possible, however, that our definition of TBI as a radiographical diagnosis might have been too broad. Our categorization may not have allowed for sufficient effect sizes and significant differences between TBI and non-TBI groups. In addition, systemic changes in mRNA expression were measured in peripheral blood leucocytes, which might not be representative of the regional changes in PER2 and HO1 expression in brain tissue following TBI. TBI . TBI traumatic brain injury, PER2 period circadian regulator 2, AKI acute kidney injury, HO1 heme oxygenase 1, RR relative risk, 95% CI 95% confidence interval, ISS injury severity score, 95% CI 95% confidence interval. www.nature.com/scientificreports/ patients tended to have elevated PER2 expression on day one after trauma when compared with non-TBI patients. Further studies evaluating regional PER2 and HO1 expression following TBI, e.g. in the cerebrospinal fluid might help fill this gap. Interestingly, PER2 expression in peripheral blood cells as a surrogate for disturbance of circadian rhythmicity has recently been studied in both polytrauma and general ICU patient collectives 31,32 . In both cases, low PER2 expression was associated with systemic inflammation and thereby may mirror progression towards multi-organ dysfunction syndrome 33 . Strikingly, in our patient cohort the likelihood of AKI was significantly reduced in TBI patients compared with the non-TBI control group while non-significantly tending towards elevated PER2 expression. Taking into account work by others concerning nephro-protective qualities of PER2 34 , along with our findings, it is tempting to speculate that TBI induces PER2 expression, which on the one hand indicates central circadian disruption that contributes to the development of delirium while on the other hand exerting anti-inflammatory organ protection peripherally by preserving kidney function. Fittingly, on a descriptive statistical level and limited to our patient cohort, TBI and PER2 induction showed a trend towards a beneficial impact on survival.
Several limitations of this study have to be noted in relation to the findings discussed above. First, our patient sample size was limited and restricts generalizability of our findings. Second, the ICU as a "polluted" environment in terms of external circadian entrainment factors poses challenges for making conclusions about true endogenous circadian signaling. Third, we performed static analysis of circadian gene expression at two time points that might not reflect true circadian oscillation of these genes.

Conclusions
In summary, systemic mRNA expression of PER2 and HO1 is induced after trauma that correlates with trauma severity. Furthermore, the extent of expression correlates with the incidence of delirium in an age-independent manner. TBI in our cohort did not contribute to delirium development. However, it was associated with reduced incidence of AKI, suggesting a link between TBI and secondary organ injury. Utilizing expression of PER2 and HO1 as biomarkers could potentially improve diagnosis algorithms for delirium by adding an objective laboratory parameter.

Methods
Study. This observational biomarker study focused on TBI in the setting of polytrauma and regulation of PER2 and HO1 expression in leucocytes along with ICU-relevant clinical outcome data (Fig. 1a). In this singlecenter cohort study, patients were studied under a protocol that was approved by the Institutional Ethics Committee of the University of Freiburg (Protocol No. 293/15). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration. Data reporting is done in accordance with the STROBE guidelines. www.nature.com/scientificreports/ Clinical assessment. Delirium was assessed by trained ICU staff every 8 h using the nursing delirium screening scale (nuDesc) 35 . In patients with deep sedation for therapeutic reasons, delirium screening was resumed as soon as sedation was terminated. Maximum nuDesc scores during the hospital stay were used to correlate with biomarker expression. To perform between-group comparisons patients were classified into groups of "delirium" and "non-delirium" according to their maximum nuDesc. However, the maximum nuDesc is a static evaluation of delirious symptoms and may not be specific to delirium alone when applied at a single point of time. Therefore, the classification was validated based on a dynamic criterion of delirium that is mentioned in the DSM-5 (fifth edition of the Diagnostic and Statistical Manual of Mental Disorders): fluctuation of delirious symptoms. Symptom fluctuation was assessed through SD (standard deviation) of all daily nuDesc values during patient's hospital stay. This SD was then defined as the independent variable in a between-group comparison of "delirium" and "non-delirium" via Mann-Whitney-U test. Acute kidney injury was assessed using the first KDIGO (kidney disease improving global outcomes) criterion (0.3 mg/dl serum creatinine increase within 48 h (Stage 1)). The injury severity score (ISS) was obtained from the emergency ward admission sheet that is routinely assessed in polytrauma patients by the admitting staff.
Biomarkers. Blood  Gene expression analysis was performed relatively using ΔΔCt-method. Rpl13a served as reference gene as it provides stable expression profiles in various tissues and cell types including peripheral blood and across various experimental conditions [36][37][38][39] . Lacking pre-trauma sampling, cDNA samples of five healthy individuals without a history of trauma or known preconditions were recruited in order to obtain normal baseline expression patterns. These pooled technical baseline values were than used to calculated relative changes in mRNA expression in trauma patients.
Furthermore, maximum plasma concentrations of bilirubin (as the final product of heme degradation) from blood gas analyses were obtained from the digital patient file matching the corresponding time of blood withdrawal for gene expression analysis. Plasma bilirubin quantification was done by direct spectrophotometry using a blood gas analyzer (ABL 800 Flex, Radiometer).
Statistics. Data analysis was done using Prism 8 software (GraphPad Software Inc.). Based on previous data 40 , an a priori power analysis (contingency tables PER2/HO1 high/low vs. delirium/non-delirium; effect size Cohen's w 0.6; α = 0.05; power 95%, df 2) indicated a total sample size of n = 43 to be sufficient to characterize biomarker expression levels. Patients for whom data on outcomes were missing were excluded from the study. When performing measurements of central tendency in ordinal-scaled data (PER2 expression, HO1 expression, bilirubin concentration, maximum leucocyte count of first week) the median value is presented along with the corresponding IQR (inter-quartile ratio). In instances of cardinal-scaled data (age, ISS (injury severity score), time on ICU) the mean value is presented with SD. Between-group differences in metrical data (age, ISS, time on ICU, PER2 expression, HO1 expression, bilirubin concentration, maximum leucocyte count of first week) were analyzed via Mann-Whitney-U test. Between-group comparisons in categorical data were performed using Fisher's exact test (yes/no: pre-existing neurological disease, pre-existing cardiovascular disease, diabetes mellitus, TBI, delirium, AKI, 6-month mortality; high/low: PER2 expression, HO1 expression, bilirubin concentration). Tests for correlation were performed between metrical data sets (ISS, PER2 expression, HO1 expression, bilirubin concentration, time on ICU, age, maximum nuDesc) using Spearman correlation (r, P) as well as linear regression (r 2 , P). Categories of high and low expression levels of PER2 and HO1 were formed from linear regression of ISS and mRNA expression levels of PER2 or HO1 as well as maximum bilirubin concentration on day one after trauma. A linear regression formula of the format Y = a × x + b with the corresponding curve was calculated. Then, data points with mRNA expression/bilirubin concentration located above the regression curve, mathematically corresponding to values higher than expected when using the regression formula for the individual data point, were defined to be of high expression/concentration. Data points with mRNA expression/bilirubin concentration located below the regression curve, mathematically corresponding to values lower than expected using the regression formula, were defined to be of low expression/concentration. Descriptive statistics of relative risk (RR) and odds ratio (OR) were assessed via Koopman asymptotic score (RR) and Baptista-Pike (OR). RR and OR are presented along with their 95% confidence interval (95% CI). With all statistical analyses, a P-value www.nature.com/scientificreports/ of less than 0.05 was considered to be significant. A P-value of less than 0.15 was considered to be indicative of a possible association, and is referred to as "non-significant tendency".

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.