Introduction
Septic shock is a leading cause of morbidity and mortality in the pediatric intensive care unit (PICU) [
1],[
2]. With improved therapies to reverse shock, progressive multi-organ failure and secondary infection from acquired immunoparalysis are now the main antecedents to sepsis-associated death [
3],[
4]. Increasing evidence supports a role for mitochondrial bioenergetic dysfunction in the pathobiology of organ injury and immune dysregulation in sepsis [
5]-[
7].
Circulating blood cells from critically ill patients with septic shock exhibit decreased oxidative respiration, electron chain complex activity, mitochondrial turnover, and mitochondrial membrane potential [
8]-[
12]. Blood is an easily accessible tissue that can be used to directly measure mitochondrial dysfunction in immune cells and may reflect a systemic process affecting other vital organs. Mitochondrial dysfunction in blood cells has been associated with severity of illness, organ dysfunction, mortality, and immunoparalysis in human sepsis [
8]-[
12], including children [
13]. Differential expression of mitochondrial genes in blood cells has been reported for several diseases in which bioenergetic failure is a postulated mechanism [
14]-[
16], and injection of endotoxin has been shown to cause widespread suppression of genes encoding for mitochondrial ATP production and protein synthesis within human leukocytes [
17]. However, there are no data about the blood cell mitochondrial transcriptome in pediatric sepsis. Identification of mitochondrial genomic changes within blood cells could provide clinically relevant biomarkers, offer insight into biological mechanisms, and inform therapeutic targets related to mitochondrial bioenergetic dysfunction for children with sepsis.
While mitochondria contain their own circular genome, the majority of the mitochondrial proteins comprising subunits of the electron transport system (ETS) are encoded by nuclear genes, including 38/45 for complex I, 4/4 for complex II, 10/11 for complex III, 10/13 for complex IV, and 17/19 for complex V (ATP synthase). In addition, all of the 79 known mitochondrial ribosomal proteins (MRPs) are encoded by the nuclear genome [
18],[
19]. These ETS and ribosomal proteins are synthesized within the cytoplasm and then imported into the mitochondria.
Over the last decade, we have generated an extensive genome-wide expression database of children with septic shock drawn from multiple centers in the U.S. [
20]. The database has enabled the discovery of gene expression-based subclasses of pediatric septic shock [
21]-[
23], stratification biomarkers [
24]-[
31], diagnostic biomarkers [
32]-[
35], and novel therapeutic targets [
36]-[
42]. Here, we mined the database to test the hypothesis that expression of whole blood-derived nuclear-encoded mitochondrial genes will be differentially regulated between pediatric patients with septic shock and nonseptic controls within the first 24 hours of presentation to the PICU. We further hypothesized that nuclear-encoded mitochondrial genes would be differentially regulated across genotypic and phenotypic distinct subclasses of pediatric septic shock. We tested these hypotheses using a focused analytical approach in which we restricted the working gene list to 296 nuclear-encoded mitochondrial genes, as previously reported by Lunnon
et al. [
14].
Discussion
In this focused analysis of a comprehensive genomic expression database, we found that nuclear-encoded mitochondrial genes are differentially regulated early in pediatric septic shock compared to healthy controls. We also compared expression of the nuclear-encoded mitochondrial genes across previously defined and validated subclasses of pediatric septic shock with distinct phenotypic characteristics. Although these subclasses were defined primarily by differential expression of genes corresponding to adaptive immunity, glucocorticoid receptor signaling, and PPARβ signaling, we found that the nuclear-encoded mitochondrial genes were also differentially regulated across these subclasses, with a greater degree of repression in the subclass of patients with the most organ dysfunction and highest mortality.
Fifty-one nuclear genes encoding subunits of the mitochondrial ETS complexes were differentially regulated in blood samples from children with septic shock, with a greater degree of downregulation overall. These findings parallel the decrease in leukocyte gene expression for subunits of the ETS complexes I to V that occurs four to six hours after endotoxin injection in healthy human volunteers [
17]. The mitochondrial ETS involves over 100 proteins derived from the nuclear and mitochondrial genomes assembled into five complexes. Although it is difficult to predict how the sum of the changes we observed might alter total function of the respiratory chain, our data raise the notion that genomic influences could affect mitochondrial oxidative respiration and therefore cellular bioenergetic and organ function in sepsis. In particular, the prominence of NADH dehydrogenase activity in the top gene networks is consistent with decreased complex I gene expression [
56] and activity noted in skeletal muscle in human adult sepsis [
57],[
58]. Prior studies using blood samples from patients with sepsis have also demonstrated altered respiratory chain activity in peripheral blood mononuclear cells and platelets [
9],[
11],[
13],[
59],[
60].
A similar genome-wide analysis using skeletal muscle from 17 adult patients with sepsis-induced multi-organ failure found that 82 mitochondrial genes were differentially regulated compared to healthy controls (74 upregulated, 8 downregulated) [
61]. Decreased ETS enzyme activity was also observed and attributed to a loss of mitochondrial content. The authors concluded that deficits in ETS activity could not be explained by decreased mitochondrial gene expression because of the overall absence of repression of mitochondrial genes. However, these patients were studied later in their septic course when a decrease in mitochondrial content seems to be most prominent, as opposed to our study in which patients were studied in the acute phase when decreased ETS function predominates. More recently, Carre
et al. performed muscle biopsies early after onset (one to two days) of critical septic illness in adult patients and observed that an overall decrease in ETS gene expression was associated with decreased protein content and activity of ETS complexes I and IV [
56]. Concurrent measurements of mitochondrial gene expression and function are needed to determine if the changes we observed in mitochondrial gene expression are sufficient to alter mitochondrial function in blood cells (and other tissues) in pediatric sepsis. We note, however, that serum lactate concentrations were highest in group A. While this suggests a potential association between downregulation of mitochondrial genes and mitochondrial dysfunction, these data should be interpreted cautiously because they are incomplete and represent an indirect, nonspecific measure of mitochondrial function.
Genes encoding the MRPs were also predominately downregulated in pediatric septic shock. This finding should not be construed as specific to mitochondria, as global downregulation of ribosomal gene transcription may occur nonspecifically during extreme biological conditions such as septic shock [
17],[
62]. Similar downregulation of MRPs has been previously described in blood and brain tissue from patients with Alzheimer's disease [
14]. Like bacteria and eukaryotic cell cytoplasm, mitochondria contain their own ribosomes. The MRPs are encoded in nuclear genes, synthesized in the cytoplasm, and then imported into mitochondria. MRPs assemble with mitochondrial-transcribed rRNAs to form two ribosomes that are responsible for translating the 37 mitochondrial-encoded genes, including 13 protein subunits critical to the function of ETS complexes I, III, IV, and V (ATP synthase). Thus, MRPs play a critical role in mitochondrial protein synthesis and bioenergetic function [
18],[
19]. Known mutations in MRPs are associated with lactic acidosis, organ dysfunction, and early death [
18], and several MRPs (MRPS29, MRPS30) have been implicated in apoptosis [
19]. Moreover, changes in mitochondrial turnover (including biogenesis, mitophagy, and fusion/fission) have been associated with clinical outcomes in sepsis [
5],[
63]. The significant repression of MRPs in our study points to changes in the mitochondrial protein synthesis machinery as one potential mechanism that could lead to mitochondrial dysfunction and diminished mitochondrial content in sepsis.
The differential regulation of mitochondrial genes across subclasses of pediatric septic shock supports a link between mitochondrial gene expression and clinical outcomes, including organ failure and mortality. These subclasses were previously identified based on hierarchical clustering, with patients in group A exhibiting the greatest repression of genes corresponding to key signaling pathways of the adaptive immune system, glucocorticoid receptor signaling, and PPARα signaling [
21]. In the current study, patients in group A also exhibited a greater repression of nuclear-encoded mitochondrial genes than groups B and C, especially in ETS complexes II and III. However, we caution that since we observed minimal differential regulation of mitochondrial genes when directly comparing survivors and nonsurvivors, or patients with and without a complicated course, we cannot rule out a `coupling effect' in which changes in mitochondrial gene expression are enhanced by other biologic pathways that differ between genomically defined subclasses. Mitochondria are involved in a variety of cell signaling pathways underlying the immune response, including cytokine release, inflammasome formation, and formation of reactive oxygen species [
64]. It will be important to establish the extent to which mitochondrial gene expression may truly affect phenotypic differences in septic shock through its role in the immune system and other cell signaling pathways versus more direct effects on cellular bioenergetics.
We note the limitations of our study. First, this was a post hoc, focused analysis using a limited set of 296 nuclear-encoded mitochondrial genes. To reduce the likelihood of false-positive results, we used a relatively stringent FDR of 1% and conducted control analyses to determine expected rates of differential gene expression based on either all available genes on the array, or 10 iterations of analyses based on randomly selected genes. In both cases, we found that the observed rate of differential gene expression in mitochondrial genes was greater than the expected random rate. Nonetheless, our focused analytical approach does not allow us to conclude that the pediatric septic shock transcriptome is specifically enriched for genes corresponding to mitochondrial function. We can only conclude that if the analytical approach is limited to nuclear-encoded mitochondrial genes, we find differential regulation of these genes in children with septic shock, and across subgroups of patients with septic shock.
Second, sufficient data were available from only one time point. Although blood sampling within 24 hours of initial presentation to the PICU with septic shock was likely to capture maximum clinical acuity, we were unable to test temporal changes in the mitochondrial transcriptome with evolution of the septic course. Third, the data are based on whole blood-derived RNA, which carries the potential for confounding by differential white blood cell counts. Although platelet counts did not differ between the three septic shock subclasses, group A had a lower total leukocyte count with a greater percentage of lymphocytes and fewer neutrophils. Lymphocytes have a relatively lower mitochondrial content than neutrophils, though how this effects nuclear-encoded mitochondrial gene expression is not clear [
65],[
66], and have been shown to have slightly less gene upregulation in sepsis [
67]. However, we have previously shown that whole blood-derived RNA can yield biologically meaningful data, gene expression profiles have revealed similar themes in leukocyte subsets and whole blood [
62], and our current data are consistent with mitochondrial gene expression profiles from previous laboratory- and clinical-based studies [
17],[
56],[
68]-[
70]. Fourth, since concomitant measures of mitochondrial function were not available we cannot determine how the observed changes might alter the total function of respiratory chain and ATP production. Although the fold change in gene expression was modest in most cases, our findings were similar to magnitude of changes in mitochondrial gene expression profiles observed in prior studies [
56],[
68],[
71]. However, it also possible that primary mitochondrial bioenergetic dysfunction itself leads to changes in nuclear gene expression [
72]. Finally, because we used a nuclear gene array platform, changes in mitochondrial-encoded genes were not included in this study. These genes are critical to the function of the ETS and mitochondrial protein synthesis and should be considered in future studies.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SLW and HRW conceived and developed the study, obtained funding for the study, conducted the analyses, and wrote the manuscript. NZC, GLA, NJT, RJF, NA, KM, PAC, TPS, MTB, and JF enrolled subjects at the participating institutions, provided clinical data and biological samples, and edited the manuscript. SB, EB, K Howard, and EF maintained the clinical database and coordinated all inter-institutional research activity. K Harmon maintained the biological repository and processed all biological samples. All authors read and approved the manuscript.