Mitochondrial DNA Haplogroups and Delirium during Sepsis

Abstract

Objectives:

Studies suggest that mitochondrial dysfunction underlies some forms of sepsis-induced organ failure. We sought to test the hypothesis that variations in mtDNA haplogroup affect susceptibility to sepsis-associated delirium, a common manifestation of acute brain dysfunction during sepsis.

Design:

Retrospective cohort study.

Setting:

Medical and surgical ICUs at a large tertiary care center.

Patients:

Caucasian and African-American adults with sepsis.

Measurements and Main Results:

We determined each patient’s mtDNA haplogroup using single nucleotide polymorphisms genotyping data in a DNA databank and extracted outcomes from linked electronic medical records. We then used zero inflated negative binomial regression to analyze age-adjusted associations between mtDNA haplogroups and duration of delirium, identified using the Confusion Assessment Method for the ICU. Eight hundred ten patients accounted for 958 sepsis admissions, with 802 (84%) by Caucasians and 156 (16%) by African Americans. In total, 795 (83%) patient-admissions involved one or more days of delirium. The 7% of Caucasians belonging to mtDNA haplogroup clade IWX experienced more delirium than the 49% in haplogroup H, the most common Caucasian haplogroup (age-adjusted rate ratio [RR] for delirium 1.36, 95% confidence interval [CI] 1.13 to 1.64, p = 0.001). Alternatively, among African Americans the 24% in haplogroup L2 experienced less delirium than those in haplogroup L3, the most common African haplogroup (adjusted RR for delirium 0.60, 95% CI 0.38 to 0.94, p = 0.03).

Conclusions:

Variations in mtDNA are associated with development of and protection from delirium in Caucasians and African Americans during sepsis. Future studies are now needed to determine whether mtDNA and mitochondrial dysfunction contribute to the pathogenesis of delirium during sepsis so that targeted treatments can be developed.

Over 30 million patients worldwide develop sepsis each year (1). Improved patient management has reduced mortality (2), but many adverse effects of sepsis remain unaddressed. Delirium, for example, is a manifestation of acute brain dysfunction that affects up to 80% of patients with sepsis (3). In a multicenter study of mechanically ventilated ICU patients, 75% of whom had sepsis, delirium was associated with delays in extubation and ICU discharge as well as increased mortality (4). In addition, many sepsis survivors struggle with long-term cognitive impairment (5). Recent studies have shown that longer episodes of delirium in the ICU predict long-term cognitive impairment (6, 7), suggesting that therapies directed at sepsis-associated delirium may improve long-term cognitive outcomes in this population. To develop effective therapies, however, the pathophysiology of sepsis-associated delirium must be better understood.

Animal and human studies have shown that the inflammatory response of sepsis imposes stress on cells and tissues characterized by increased production of reactive oxygen species (ROS), decreased oxygen delivery, impaired oxygen extraction, increased lactate production, and dysregulation of proliferation/apoptosis homeostasis (8). These phenomena converge in the mitochondria, which regulate ROS production, oxygen and nutrient metabolism, and apoptosis. Mitochondrial dysfunction, e.g., impaired mitochondrial metabolism, failure of mitochondrial biogenesis, etc., has been associated with sepsis-induced organ failure in numerous studies (9), but few have examined effects on the brain, which is particularly vulnerable to metabolic and oxidative insults. Given evidence linking mitochondrial dysfunction with neurodegeneration (10), one would expect factors that negatively impact mitochondrial function during sepsis to increase risk for delirium.

We therefore conducted the first study to examine whether mitochondrial DNA (mtDNA) haplogroups (11)—ancestry-related variations in mtDNA (Figure 1) that are associated with differential mitochondrial function (12–14)—are associated with sepsis-associated delirium during critical illness. Since European mtDNA haplogroups H, J and T were reported to predict improved survival among Caucasians with sepsis (15–17), we hypothesized that haplogroups H, J and T would be associated with reduced delirium duration among Caucasians. Among African Americans, we conducted an exploratory analysis since no prior studies of African mtDNA haplogroups and outcomes in sepsis were available to inform a hypothesis.

Figure 1.

Partial phylogenetic tree demonstrating approximate relationships between mitochondrial DNA haplogroups examined in the current investigation. Haplogroups, which are groups of similar haplotypes that share a common ancestor, are defined by specific single nucleotide polymorphisms that have emerged in the genome over millennia. Because mitochondrial haplogroups have a well-defined phylogeny (11), they can be used to understand variations in mitochondrial genetics within populations.

MATERIALS AND METHODS

Study Design and Population

We conducted this retrospective cohort study at Vanderbilt University Medical Center using BioVU (18), Vanderbilt’s DNA databank composed of DNA and the Synthetic Derivative, a de-identified electronic medical record (EMR)-derived database (18). BioVU consists of Vanderbilt patients who provide informed consent, which allows investigators to use DNA extracted from residual blood collected for routine clinical tests and to analyze de-identified clinical data. The institutional review board approved the research protocol, which was classified as non-human subjects research since the data were completely de-identified.

We queried the Synthetic Derivative and BioVU to identify all adult (≥18 years old) Caucasians or African-Americans with existing genotyping data who were treated for sepsis and assessed for delirium using the Confusion Assessment Method for the ICU (CAM-ICU) (19). We identified sepsis using ICD-9 codes (995.92 for severe sepsis and 785.52 for septic shock) shown to have high specificity for sepsis (20). We included all observations for patients with multiple eligible hospital stays and accounted for correlation between patients.

Exposure

We determined each participant’s mtDNA haplogroup—the exposure variable—using existing single nucleotide polymorphisms (SNP) genotyping data deposited in BioVU, as described in the Supplemental Digital Content. Since mitochondrial haplogroups are closely related to continental ancestry, we stratified analyses according to race as recorded in the EMR. Administratively assigned race has been shown to be as good a good proxy for genetic ancestry as self-reported race (21). For Caucasians, we analyzed haplogroups H, Uk, J, T, the IWX clade (combining I, W, and X, three rarer haplogroups that descend from a common node in the mitochondrial phylogenetic tree) (22), and a small number of “Other” haplogroups, which we grouped together. For African Americans, we analyzed the major African haplogroups, L1, L2, and L3, and “Other” (11). We limited our analyses to these a priori-specified haplogroups.

Outcome

Duration of delirium was defined as the number of days with a positive CAM-ICU assessment from the first CAM-ICU or RASS assessment (considered a surrogate for ICU admission since these assessments are only performed in the ICU) until ICU discharge, death, or ICU day 14. Since few patients remain delirious > 14 days, we limited analyses to a 14-day period to avoid undue influence from outliers. As described in the Supplemental Digital Content, clinical practice bedside nurses use the CAM-ICU to assess patients for delirium (19). The sensitivity and specificity of bedside-nurse-implemented CAM-ICUs recorded in the Vanderbilt EMR are high (both 81%) when compared with CAM-ICU assessments performed by highly trained, experienced research nurses (23).

Statistical Analysis

To determine whether mtDNA haplogroups were associated with delirium duration, we used zero-inflated negative binomial regression (24), which models overdispersed count variables with excessive zeros. A priori, we selected delirium duration as the outcome rather than ever/never delirium prevalence for three reasons: delirium during sepsis is so common that incidence provides little insight; delirium duration predicts cognitive function up to one year after critical illness (6, 7); and analyses of a continuous variable are better powered than analyses of a dichotomous variable.

We stratified the analysis by race since major mtDNA haplogroups differ between races, and we included age as a covariate in the count portion of the models, allowing age to be nonlinear via restricted cubic splines. Results are expressed as rate ratios (RR) with the largest haplogroup within each race (H among Caucasians and L3 among African-Americans) serving as the reference group. In this context, a RR estimates relative differences in delirium duration based on mtDNA haplogroup; e.g., a RR of 1.5 indicates a 50% increase in the duration of delirium relative to the reference group.

We conducted a secondary analysis that included only data from one hospitalization per patient and selected the first hospitalization for the 106 patients who were admitted more than once. As described in the Supplemental Digital Content, we also conducted several post hoc analyses to examine possible mechanisms underlying observed associations between mtDNA haplogroups and delirium. We used R version 3.1.2 for all statistical analyses (25) and considered a two-sided alpha of 0.05 to indicate statistical significance. Additional details regarding statistical analyses are provided in the Supplemental Digital Content.

RESULTS

In the BioVU database, 958 sepsis hospital admissions (representing 810 patients) met eligibility criteria. We included these 958 patient-admissions in the study cohort; 802 (84%) admissions were by Caucasian patients, and 156 (16%) were by African Americans (Table 1). In total, 795 (83%) patient-admissions involved one or more days of delirium. Among the 84% of Caucasians affected, delirium lasted a median of 6 days, whereas it lasted a median of 4 days among the 77% of African Americans affected. Coma was also common in both cohorts, but the duration was shorter than that of delirium. Both delirium and coma were more common during the early portion of the ICU stay and tended to resolve over time (Figure 2); only 159 (17%) patient-admissions involved delirium lasting longer than 14 days.

Table 1.

Clinical and Genetic Characteristics for 958 Patient-Admissions^a

Characteristic	Caucasian N = 802	African American N = 156
Age, median [IQR], y	62 [52–71]	59 [48–68]
Female, N (%)	371 (46%)	75 (48%)
Hospital length of stay, median [IQR], days	13 [7–26]	8 [5–17]
Days assessed with CAM-ICU/RASSb	9 [5–18]	6 [3–12]
Deliriumc
Ever, N (%)	675 (84%)	120 (77%)
Duration, median [IQR], daysd	6 [3–13]	4 [2–9]
Comac
Ever, N (%)	274 (34%)	35 (22%)
Duration, median [IQR], dayse	1 [1–3]	1 [1–2]
Caucasian mitochondrial DNA haplogroups
H	392 (49%)	-
IWX	54 (7%)	-
J	66 (8%)	-
T	101 (13%)	-
Uk	164 (20%)	-
Other	25 (3%)	-
African mitochondrial DNA haplogroups
L1	-	24 (15%)
L2	-	38 (24%)
L3	-	49 (31%)
Other	-	45 (29%)

CAM-ICU Confusion Assessment Method for the ICU, ICU intensive care unit, IQR interquartile range, RASS Richmond Agitation-Sedation Scale.

^aEight hundred ten patients accounted for 958 hospital admissions.

^bPer usual care, all ICU patients were assessed with the CAM-ICU and RASS at least once per nursing shift (i.e., once every 8 hours) until discharge from the ICU.

^cDuring the 14-day study period beginning on the day of first CAM-ICU and/or RASS assessment.

^dAmong patient-admissions when delirium occurred during the study period.

^eAmong patient-admissions when coma occurred during the study period.

Figure 2.

Mental status according to day after ICU admission. Delirium was identified on days the CAM-ICU was positive; coma was identified when RASS was −4 or −5. Days when both coma and delirium were present (at different times) were considered delirium days. Missing CAM-ICU/RASS assessments on ICU days were imputed. The first assessment was considered ICU day 1, CAM-ICU use was limited to ICUs. The proportion of assessed patients declined as patients were discharged or died; the denominator remains fixed at 958, i.e., the size of the entire study cohort.

The frequencies of Caucasian and African mtDNA haplogroups are reported in Table 1. As expected based on mtDNA haplogroup frequencies in the general population (22), haplogroup H was the most common Caucasian haplogroup, representing nearly half of the Caucasian cohort. L3 was the most common African haplogroup, representing one third of the African American cohort.

Among both Caucasians and African Americans, mtDNA haplogroups were significantly associated with delirium. Caucasian mtDNA haplogroup clade IWX was associated with increased probability of delirium compared with haplogroup H, the most common haplogroup (age-adjusted RR for delirium 1.36, 95% confidence interval [CI] 1.13 to 1.64, p = 0.001; Figure 3). Among African Americans, haplogroup L2 was associated with reduced probability of delirium compared with haplogroup L3 (age-adjusted RR for delirium 0.60, 95% CI 0.38 to 0.94, p = 0.03; Figure 4). Both bootstrap validation and a secondary analysis using data from only the first hospital stay for each patient yielded similar results (Table 1 in Supplemental Digital Content) in both Caucasians and African Americans.

Figure 3.

Caucasian mitochondrial DNA haplogroups and delirium during sepsis. After adjusting for age, patients in haplogroup clade IWX were estimated to spend 36% (95% confidence interval, 13%−64%; p = 0.001) more days delirious on average than those in the reference group, haplogroup H.

Figure 4.

African mitochondrial DNA haplogroups and delirium during sepsis. After adjusting for age, patients in haplogroup L2 were estimated to spend 40% (95% confidence interval, 6%−72%; p = 0.03) fewer days delirious on average than those in the reference group, haplogroup L3.

The observed associations between Caucasian mtDNA haplogroup clade IWX and delirium as well as between African haplogroup L2 and delirium were unchanged in the model adjusting for liver dysfunction and the model adjusting for benzodiazepine exposure (Table 2 in Supplemental Digital Content). Additionally, we found no significant association between Caucasian mtDNA haplogroup and 14-day survival (p = 0.35; Figure 1 in Supplemental Digital Content). African mtDNA haplogroup was associated with 14-day survival with patients in haplogroup L1 at increased risk of death compared with those in L2, L3, and Other (p = 0.03; Figure 2 in Supplemental Digital Content).

DISCUSSION

Our investigation, the first to examine mitochondrial genetics as a risk factor for delirium, found that specific mtDNA haplogroups were associated with delirium during sepsis among both Caucasians and African Americans. Since mtDNA haplogroups can impact mitochondrial function (12–14), our results support the hypothesis that mitochondrial dysfunction during sepsis is a key determinant of sepsis-associated delirium. Future research is needed to confirm our findings and to specifically examine measures of mitochondrial dysfunction as risk factors for delirium during sepsis.

Sepsis, which occurs when immune responses to infection lead to multiple organ failure, is characterized pathophysiologically by impaired cellular respiration despite adequate tissue oxygen (8), with dysfunction of mitochondria—the site of cellular respiration—contributing to tissue damage, organ failure, and adverse clinical outcomes during sepsis (8, 9, 26–28). Multiple studies have shown that sepsis can cause neuroinflammation (29, 30), but animal models have also revealed that mitochondrial dysfunction occurs in the brain during sepsis. d’Avila and colleagues (27) demonstrated uncoupling of oxidative phosphorylation in the brains of septic mice, whereas Comim and coworkers (28) found septic rats had decreased mitochondrial respiratory chain activity in multiple brain structures. Since the effects of mitochondrial dysfunction during sepsis occur in multiple organs, our findings could be explained by mitochondrial dysfunction in the liver rather than the brain. Altered hepatic function and/or drug metabolism during sepsis could increase plasma concentrations of deliriogenic medications. Yet, we found no evidence that the associations between mtDNA haplogroups and delirium were mediated by liver failure or benzodiazepine exposure, suggesting that mitochondrial dysfunction in the brain, rather than liver, underlies delirium.

Distinct from nuclear/chromosomal DNA, mtDNA does not undergo recombination and is exclusively maternally inherited, leading to distinct ancestry-related variations known as haplogroups (Figure 1). mtDNA encodes 13 of the approximately 90 oxidative phosphorylation proteins that control mitochondrial function (12, 13, 31–34), and mtDNA haplogroups have been associated with neurodegenerative disorders, such as Alzheimer’s disease (35–37) and Parkinson’s disease (38, 39). In sepsis, mtDNA haplogroups have been associated with mortality among Asian (40) and European (15–17, 41) populations, findings that may have resulted from haplogroup-related changes in oxidative phosphorylation or altered expression of inflammation, complement, and apoptosis genes, each of which have been linked to haplogroup-defining mtDNA variants (14). Among Caucasians, Baudouin and coworkers (15) found haplogroup H to be associated with improved survival in sepsis, whereas Lorente and colleagues (16, 17) found the clade JT was associated with better survival. In contrast with our cohort, which included nearly 1,000 patient-admissions and both Caucasians and African Americans, the cohorts analyzed by Baudouin et al. (15) and Lorente et al. (16, 17) included only Caucasians and had limited sample sizes that were too small to examine the Caucasian IWX clade with sufficient power. Alternatively, Jimenez-Sousa et al. (41) recently published a study of 507 Caucasians with postoperative sepsis, in which the IWX clade was a risk factor for death.

Though they did not study delirium, Yang and colleagues (42) found that Asian mtDNA haplogroup R was a predictor of recovery from sepsis-associated encephalopathy diagnosed using the Glasgow coma scale, a result that is consistent with our finding regarding Caucasian haplogroup clade IWX given that the SNPs defining the R branch of the human mtDNA phylogenetic tree (T12705C and T16223C) also distinguish the IWX clade from non-IWX patients in a European population. As described in the Supplemental Digital Content, the functional significance of IWX-defining SNPs is known whereas that of the SNPs defining African haplogroup L2 is not yet known and should be the subject of future investigations. It is possible that the associations we report are due to lower frequency variants that define branches within these groups. Full mtDNA sequencing data from a larger cohort would be needed to assess this possibility.

Based on our findings, future studies should test the hypotheses that, relative to patients in other Caucasian mtDNA haplogroups, those in clade IWX have reduced cerebral metabolic performance during sepsis, and those in African haplogroup L2 have increased cerebral metabolic performance relative to those in other African mtDNA haplogroups. If future studies confirm that mtDNA variation and mitochondrial dysfunction have pathophysiologic roles in sepsis-associated delirium, clinical trials testing whether mitochondrial-targeted therapeutics (e.g., mitochondrial antioxidants) can prevent and/or treat delirium during sepsis would be warranted. In addition, haplogroup-guided personalized sepsis management may have promise given that antibiotic choices, temperature management strategies, electrolyte and metabolite management, nutrition strategies, sedative choices, and other aspects of critical care may affect and/or be influenced by mitochondrial function. For example, patients who are at highest mtDNA-related risk for sepsis-associated delirium could be more prone to develop delirium due to benzodiazepines, which decrease cerebral metabolic rate by 20%−30% (43) and have been associated with increased delirium risk during critical illness in numerous observational (44–46) and randomized studies (47, 48). In addition, mtDNA haplogroups might guide glucose management and nutrition during sepsis in the future, given that global cerebral metabolic rate decreases dynamically with glucose ingestion (49) and common medications can impede adaptive changes in carbon substrate trafficking that support the injured brain. The growing list of factors that influence cerebral metabolism and cognitive function during acute illness suggests that new knowledge regarding the associations we report between mtDNA haplogroups and sepsis-association delirium will likely guide future sepsis treatment trials and may eventually direct personalized interventions.

Our investigation has several important strengths and limitations. Strengths include a large sample size, which was 3–5 times larger than those of previous studies of mtDNA haplogroups in sepsis (15–17); examination of both Caucasian and African mtDNA haplogroups, with the latter never before studied in this context; and diagnosis of delirium using a validated assessment tool, which was shown to be sensitive and specific not only when used by research staff but also when implemented by clinical staff in the institution where this cohort was treated (23). Limitations include investigation of a single cohort at one institution, which may limit generalizability and requires confirmation in a separate cohort; an observational study design, which precludes us from proving causal relationships; and a predominantly Caucasian study population, such that other haplogroups (e.g., Asian) could not be examined and power to analyze African mtDNA haplogroups was low. Additionally, since participants were genotyped using multiple chips, we were unable to adjust for genetic principal components, which may be helpful when analyzing African mtDNA haplogroups since the European admixture in African Americans makes them genetically more complex. Our results regarding African haplogroup L2 should be viewed as hypothesis generating, and our sample size was not large enough to support analyses examining potential interactions between mtDNA and nuclear DNA genetic variants.

Given the retrospective study design, we were not able to ascertain preexisting cognitive impairment, which may confound the associations in question, or specific subtypes of delirium, such as the rapidly reversible sedative-related form of delirium reported in one study to affect 12% of ICU patients (50). Though all patients in the current study had sepsis, some may have had delirium due to factors other than sepsis, a possibility that would likely bias our findings towards the null hypothesis. Future studies may benefit from more specific delirium phenotyping and from analyses examining the influences of host factors (e.g., chronic diseases, source of infection, etc.). We also did not have granular severity of illness data and could not adjust for this potential confounder, but the mtDNA haplogroups that were associated with increased delirium were not associated with increased mortality, suggesting that general illness severity was not primarily responsible for our findings. Importantly, although bootstrap validation suggested our models will perform similarly in a new cohort of similar patients, the findings of this investigation need to be replicated in a separate cohort.

CONCLUSIONS

In this large sepsis cohort, we found that Caucasians belonging to mtDNA haplogroup clade IWX (7% of Caucasians) had significantly longer periods of delirium than did other Caucasians independent of age, whereas African Americans belonging to mtDNA haplogroup L2 (24% of African Americans) had significantly shorter periods of delirium compared with other African Americans. These results suggest that variations in mtDNA, which may be associated with mitochondrial dysfunction, play a role in the pathogenesis of delirium during sepsis.