Plasma mitochondrial DNA and metabolomic alterations in severe critical illness
Cell-free plasma mitochondrial DNA (mtDNA) levels are associated with endothelial dysfunction and differential outcomes in critical illness. A substantial alteration in metabolic homeostasis is commonly observed in severe critical illness. We hypothesized that metabolic profiles significantly differ between critically ill patients relative to their level of plasma mtDNA.
We performed a metabolomic study with biorepository plasma samples collected from 73 adults with systemic inflammatory response syndrome or sepsis at a single academic medical center. Patients were treated in a 20-bed medical ICU between 2008 and 2010. To identify key metabolites and metabolic pathways related to plasma NADH dehydrogenase 1 (ND1) mtDNA levels in critical illness, we first generated metabolomic data using gas and liquid chromatography-mass spectroscopy. We performed fold change analysis and volcano plot visualization based on false discovery rate-adjusted p values to evaluate the distribution of individual metabolite concentrations relative to ND1 mtDNA levels. We followed this by performing orthogonal partial least squares discriminant analysis to identify individual metabolites that discriminated ND1 mtDNA groups. We then interrogated the entire metabolomic profile using pathway overrepresentation analysis to identify groups of metabolite pathways that were different relative to ND1 mtDNA levels.
Metabolomic profiles significantly differed in critically ill patients with ND1 mtDNA levels ≥ 3200 copies/μl plasma relative to those with an ND1 mtDNA level < 3200 copies/μl plasma. Several analytical strategies showed that patients with ND1 mtDNA levels ≥ 3200 copies/μl plasma had significant decreases in glycerophosphocholines and increases in short-chain acylcarnitines.
Differential metabolic profiles during critical illness are associated with cell-free plasma ND1 mtDNA levels that are indicative of cell damage. Elevated plasma ND1 mtDNA levels are associated with decreases in glycerophosphocholines and increases in short-chain acylcarnitines that reflect phospholipid metabolism dysregulation and decreased mitochondrial function, respectively.
KeywordsMitochondrial DNA Metabolite Metabolomics Homeostasis Critical illness Acylcarnitine Glycerophosphocholine
- APACHE II
Acute Physiology and Chronic Health Evaluation II
Cross-validation analysis of variance
Damage-associated molecular patterns
False discovery rate
Human Metabolome Database
Intensive care unit
Metabolomic pathway analysis
NADH dehydrogenase 1
Orthogonal partial least squares discriminant analysis
Registry of Critical Illness
Significance analysis of microarrays
Systemic inflammatory response syndrome
Variable importance in the projection
Mitochondrial function is a major determinant of outcome in critical illness. Circulating mitochondrial damage-associated molecular patterns (DAMPs), such as cell-free mitochondrial DNA (mtDNA), contain unmethylated CpG and formylated peptides that activate immune responses through Toll-like receptor 9 and formyl peptide receptors, respectively [1, 2, 3, 4].
Plasma mtDNA is measurable in critically ill patients, with increasing levels associated with sepsis, sepsis disease severity, and mortality [5, 6, 7]. The primary factors in the extracellular release of mtDNA are cell stress and necrosis . Experimental data published in abstract form showed an increase in extracellular mtDNA by induction of endothelial cell necroptosis following transfusion . Mitochondria-related DAMPs from damaged, dying, or dead cells appear to be important for the early systemic endothelial response to sepsis . mtDNA is shown to increase endothelial cell permeability, either directly or through interactions with endothelial cells and polymorphonuclear leukocytes . These findings suggest that plasma mtDNA levels could reflect the level of injury and may also reflect the level of dysfunction or damage that mitochondria undergo in response to physiologic stress .
Because metabolic homeostasis is often disrupted in critical illness, substantial alterations of several intrinsic pathways can be expected in septic patients . To date, a number of metabolomic studies have been published in experimental sepsis models , pediatric sepsis , and adult critically ill patients [14, 15, 16, 17]. Circulating metabolic signatures showing alteration in fatty acids, lipids, and tryptophan pathways are prominent in cohorts of septic patients [14, 15, 16, 17].
Existing data support that mtDNA is related to the activation of inflammation and organ dysfunction . However, there is limited understanding of the metabolic alterations associated with elevated mtDNA levels in critical illness. Therefore, we analyzed metabolite profiles with regard to NADH dehydrogenase 1 (ND1) mtDNA levels in a prospective study of adult patients with systemic inflammatory response syndrome (SIRS) and sepsis . The ND1 protein is a subunit of NADH dehydrogenase found in the inner membrane of mitochondria . We hypothesized that the metabolomic profile of critically ill patients near intensive care unit (ICU) admission differs in patients with elevated ND1 mtDNA levels and that this difference can illuminate important biologic pathways related to the response to mitochondrial DAMPs.
Study design and patients
The Registry of Critical Illness (RoCI) is a registry of adult medical ICU patients based at the Brigham and Women’s Hospital (Boston, MA, USA), created to record patient data and store samples for plasma, RNA/DNA analysis, and protein isolation. The protocol for patient recruitment has been previously described at length . Between September 2008 and May 2010, 90 medical ICU patients had metabolic profiling performed; of these, 29 patients satisfied SIRS criteria, 30 patients satisfied criteria for sepsis, and 31 patients satisfied criteria for sepsis and acute respiratory distress syndrome . Cases were not selected with regard to risk of death or any known metabolic feature. We conducted a subanalysis involving 73 RoCI patients who had been selected for metabolic profiling  and in whom cell-free plasma ND1 mtDNA levels were determined in a prior study of mtDNA .
Exposure of interest and comorbidities
The exposure of interest was cell-free circulating plasma ND1 mtDNA assessed by measuring copy number of the ND1 gene using qRT-PCR . ND1 mtDNA level was assessed as a binary variable (ND1 mtDNA ≥ 3200 copies/μl plasma vs. ND1 mtDNA level < 3200 copies/μl plasma). The cut point of ND1 mtDNA level of 3200 copies/μl plasma was determined in our prior study to maximize the AUC for the prediction of 28-day mortality . The preparation and quantification of plasma ND1 mtDNA is outlined in Additional file 1. Demographic and physiologic data were collected from the clinical record as described previously . In addition to data collected by the RoCI, supplemental data on all patients were compiled through a hospital-based computerized data registry  as outlined in Additional file 1.
Metabolomic profiling identified 411 metabolites for the complete RoCI cohort (N = 90 plasma samples within 72 h of ICU admission) using Metabolon, Inc. (Morrisville, NC, USA) . Gas and liquid chromatography mass spectroscopy (GC-MS, LC-MS) were performed as described previously [22, 23]. We removed metabolites with the lowest IQR of variability in the RoCI data, leaving 308 metabolites. This strategy is commonly used to reduce baseline noise by removing constant or very weak variables [24, 25]. All metabolite concentrations were log2-transformed to normalize the data that were used for all of the models and all of the metabolite data analyses. Details on metabolomic sample processing have been described at length previously and are outlined in Additional file 1 .
We used MetaboAnalyst 4.0 software (www.metaboanalyst.ca) to identify key metabolism alterations related to ND1 mtDNA level . Univariate tests, including fold change analysis and volcano plot visualization based on false discovery rate (FDR)-adjusted p values, were performed to evaluate the distribution of individual metabolite concentrations in individuals with elevated ND1 mtDNA levels (≥ 3200 copies/μl plasma) relative to those with ND1 mtDNA levels < 3200 copies/μl plasma . Cross-sectional correlations were calculated using Pearson’s product-moment correlation (r) between metabolites and ND1 mtDNA levels. For data visualization purposes, a bipartite graph was generated of metabolites that were significantly changed (increased or decreased) with elevated ND1 mtDNA level (≥ 3200 copies/μl plasma). Significant features were further identified by significance analysis of microarrays (SAM) , in which the FDR was determined by running multiple tests on high-dimensional data that distinguish between patients with ND1 mtDNA ≥ 3200 copies/μl plasma relative to those with ND1 mtDNA < 3200 copies/μl plasma, with a q value (upper limit of FDR) < 0.01 considered to be significant.
We performed logistic regression with ND1 mtDNA ≥ 3200 copies/μl plasma as the exposure and 28-day mortality as the outcome, after adjustment for Acute Physiology and Chronic Health Evaluation II (APACHE II) score and sepsis. Linear regression was performed with ND1 mtDNA ≥ 3200 copies/μl plasma as the exposure and acylcarnitine metabolites as the outcome after adjustment for age, sex, race, and APACHE II score. Linear regression was also performed with ND1 mtDNA copies/μl plasma as the exposure and individual metabolites as the outcome adjusted for age, sex, race, and APACHE II score. STATA 14.1/MP software (StataCorp, College Station, TX, USA) was used for all regression analyses.
Orthogonal partial least squares discriminant analysis (OPLS-DA), a supervised method, was used to select variables representing the greatest contribution to classification of the ND1 mtDNA groups . The quality of the multivariate model developed was described by R2 and Q2, which corresponded to the model’s goodness of fit and predictive performance, respectively. Permutation testing was performed to validate the OPLS-DA model [29, 30]. Sevenfold cross-validation analysis of variance (CV-ANOVA) was applied to determine OPLS-DA model significance . Variables that contributed the most to ND1 mtDNA group recognition were identified with SIMCA (Umetrics, Umeå, Sweden) using variable importance in the projection (VIP) scores. VIP scores > 1 are considered to be important for the explanatory/predictive ability of an OPLS-DA model . We used an S-plot to visualize the variable influence between ND1 mtDNA groups in the OPLS-DA model by combining the contribution/covariance and reliability/correlation loading profiles . A correlation coefficient of ± 0.410 was adopted as a cutoff value to select the variables that are most correlated with the OPLS-DA discriminant scores.
For pathway overrepresentation analysis of case-control metabolite data, metabolomic pathway analysis (MetPA)  was used. MetPA was used to evaluate a list of the 308 metabolites and their log-normalized concentration data in the 73 samples by comparing patients with ND1 mtDNA ≥ 3200 copies/μl plasma (cases) with those with ND1 mtDNA < 3200 copies/μl plasma (controls). Metabolite set enrichment analysis was performed by mapping the metabolite data onto the Human Metabolome Database (HMDB) . Metabolites were evaluated for pathway enrichment using the “Homo sapiens” library with the default parameters (“Global Test” and “Relative Betweenness Centrality”) specified as the algorithms for pathway enrichment and topological analysis, respectively. The resulting metabolic networks were represented as directed graphs, and centrality measures of a metabolite within a given network were then applied to estimate the relative importance of that metabolite in the network. Fisher’s exact test p values were adjusted for multiple testing using the Holm-Bonferroni method .
ND1 mtDNA copies/μl plasma
No. of patients
Age, years, mean (SD)
Male sex, n (%)
White race, n (%)
APACHE II score, mean (SD)
Sepsis, n (%)
28-Day mortality, n (%)
Cross-validation and permutation
Permutation (n = 200)
(0.00, − 0.264)
Metabolomics pathway analysis
Total no. of metabolites
No. of overlapping metabolites
Unadjusted p value
Pathway impact score
Starch and sucrose metabolism
In the present study, our goal was to determine if metabolite signatures in critically ill patients would be distinct relative to cell-free plasma ND1 mtDNA levels. Using high-resolution metabolomics, we demonstrated substantial differences in glycerophospholipid and acylcarnitine family member metabolism based on the level of ND1 mtDNA liberated in the plasma. Specifically, patients with high levels of plasma ND1 mtDNA, indicative of cellular damage, have very low levels of multiple glycerophosphocholine esters and increased levels of several short-chain acylcarnitines.
In cohorts of septic patients, alterations in circulating kynurenines, fatty acids, lysophosphatidylcholines, and/or carnitine esters [14, 15, 16, 17] indicate a substantial disturbance in energy and lipid homeostasis that occurs with increasing severity of illness. Large decreases in glycerophosphocholines are demonstrated in patients with experimental infection with Bacillus anthracis spores , bacteremia , and sepsis  and appear to correlate with sepsis mortality . Glycerophosphocholines are water-soluble compounds formed in the breakdown of phosphatidylcholine via phospholipase A1 and phospholipase A2 activities, and they are degraded by glycerophosphodiester phosphodiesterases . Glycerophosphocholines are essential components of biological membranes that modulate membrane trafficking and control cell viability . Glycerophosphocholines function in glycerophospholipid, prostaglandin, and leukotriene metabolism; are important in energy storage, signal transduction, and membrane physiology; provide mitochondrial support; and are neutrophil-activating factors .
The observed substantial decreases in glycerophosphocholines during sepsis may be related to increased glycerophosphocholine hydrolysis . Circulating phospholipase A2 activity is found in sepsis . Additionally, endothelial cells secrete the phospholipase endothelial lipase (EL) involved in phospholipid homeostasis [41, 42]. EL is produced by macrophages in addition to the endothelium in response to plasma inflammation markers . In human experimental models of low-dose endotoxemia, significant augmentation of plasma EL concentrations has been shown . The combination of circulating phospholipase A2 and EL activity may be responsible for the low glycerophosphocholine metabolites observed in our study.
Lipidomic alterations are prominent in sepsis and critically ill patients [45, 46, 47, 48]. We have shown that carnitine esters, important for immune response to pathogens , are the most pronounced metabolites that differed between sepsis nonsurvivors and survivors . Alterations of acylcarnitines are found in studies of severe sepsis/septic shock , the prediction of death in sepsis , and an integrative omics study in primates that was validated also in human patient cohorts .
In critical illness, metabolic pathways are altered to preferentially catabolize fatty acids and amino acids. Substantive literature demonstrates that an early indicator of critical illness outcomes is mitochondrial biogenesis [51, 52, 53, 54]. Elevated short-chain acylcarnitines found in plasma are due to incomplete mitochondrial fatty acid β-oxidation downstream of carnitine palmitoyltransferase I and are suggestive of impaired mitochondrial function [55, 56, 57]. The increase in plasma short-chain acylcarnitines with elevated ND1 mtDNA in our study may reflect less efficient fatty acid β-oxidation, potentially thorough worsening of mitochondrial bioenergetics.
Accelerated tryptophan catabolism along the kynurenine pathway occurs with sepsis. The enzyme responsible for kynurenine production is upregulated by bacterial products and is critically involved in CD4+ and CD8+ effector T-cell suppression as well as in generation and activation of regulatory T cells [58, 59]. We and others have found that modulation of kynurenine is associated with 28-day mortality in sepsis [14, 15]. Increased production of kynurenine has been proposed to contribute to hypotension in sepsis  and has been associated with dysregulated immune response and impaired microvascular reactivity .
Strengths of the present study include using cell-free plasma for ND1 mtDNA measurement. Because platelets secrete their mitochondria following activation during inflammation and sepsis, they may serve as a source of extracellular mtDNA . Using cell-free plasma allowed us to draw the inference that the source of ND1 mtDNA is more likely the endothelium. Further, we employed several types of statistical procedures, and data visualization processes were used to identify differential metabolites, including Student’s t test , Pearson correlation, volcano plot, bipartite graph, SAM , OPLS-DA, and MetPA .
The present study is not without potential limitations. Metabolites were measured early in the ICU course of severe critical illness, from a relatively small number of patients, at a single time point, and from a single biofluid (plasma). Our assumption that plasma is an integrative biofluid may not account for tissue- or organ-specific metabolism. Our observational study included patients who were critically ill for various reasons, creating a heterogeneous study sample with high severity of illness. Further, selection bias may be present because we analyzed only a subset of patients of the RoCI cohort who had ND1 mtDNA determined. We are unable to account for the impact of race on metabolic profiles because our cohort was mostly white. Because our study was performed on a convenience sample and not replicated in other cohorts, our results may not be generalizable to all critically ill patients. Our bioinformatics approaches, while robust, are not without risk of introducing sources of bias. Single–time point metabolomics provides important information but does not capture the dynamic changes over time . We were not able to determine the stability of metabolites over storage time . Although OPLS-DA is well-suited for metabolomic data with much larger numbers of predictors than observations and multicollinearity, it is prone to overfitting; however, permutation testing indicated a low likelihood of seeing results this strong by chance (p ≤ 0.05) [29, 30]. Like in our study, mapping the metabolite data onto the HMDB  does not always result in HMDB number assignment to each metabolite. Finally, we cannot fully account for potential confounding, reverse causation, and the lack of a randomly distributed exposure .
In summary, elevated levels of cell-free plasma ND1 mtDNA are associated with differential metabolic profiles in early severe critical illness. Glycerophospholipids, which are important in phospholipid metabolism and mitochondrial support, are significantly depressed in patients with high plasma ND1 mtDNA. Short-chain acylcarnitines indicative of mitochondrial dysfunction are increased with high plasma ND1 mtDNA. These data, although observational, provide an important window into the metabolite signatures of cell stress and necrosis in response to sepsis.
This article is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD. The authors thank Shawn Murphy and Henry Chueh and the Partners Health Care Research Patient Data Registry group for facilitating use of their database.
KN is supported by National Institutes of Health (NIH)/National Center for Advancing Translational Sciences grant KL2-TR-002385. AJR is supported by NIH grant K23 HL125663. MJM is supported by NIH grant R01 HL139634. LEF is supported by NIH grant R01 HL114839. RMB is supported by NIH grants R01 HL091957, R01 HL112747, and P01 108801. AMKC is supported by NIH grants P01 HL108801, R01 HL079904, and R01 HL112747. KBC is supported by the ASPEN Rhoads Research Foundation and NIH grant R01 GM115774.
Availability of data and materials
The dataset supporting the conclusions of this article is not available.
KN participated in the design of the study, assembled input data, and helped to draft and revise the manuscript. PIJ helped to draft and revise the manuscript. AJR assembled input data, participated in the design of the study, and helped to draft the manuscript. MJM assembled input data, participated in the design of the study, and helped to revise the manuscript. RMB recruited cohort patients, assembled input data, and helped to draft the manuscript. LEF recruited cohort patients, assembled input data, and helped to draft the manuscript. JH helped to draft and revise the manuscript. AMKC recruited cohort patients, assembled input data, and helped to draft the manuscript. KBC conceived of the study, participated in its design and coordination, assembled input data, wrote code, ran the model, analyzed output data, and drafted and revised the manuscript. All authors read and approved the final manuscript.
Ethics approval and consent to participate
Approval for the study was granted by the Partners Human Research Committee Institutional Review Board. Written informed consent was obtained from all subjects or their legal surrogates.
Consent for publication
Dr. Choi is a co-founder of Proterris, Inc and served on the TEVA Advisory Board meeting as a consultant on July 12-13, 2018. The remaining authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 8.Mangalmurti N, Qing D, Hotz M, Siegel DL, Sondheimer N, Mangalmurti NS. Mitochondrial DNA released following necroptosis accumulates on RBCs [abstract]. Am J Respir Crit Care Med. 2016;193:A4309.Google Scholar
- 21.Murphy SN, Chueh HC. A security architecture for query tools used to access large biomedical databases. Proc AMIA Symp. 2002:552–6.Google Scholar
- 34.Holm S. A simple sequentially rejective multiple test procedure. Scand J Stat. 1979;6:65–70.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.