Altered muscle transcriptome as molecular basis of long-term muscle weakness in survivors from critical illness

Abstract

Purpose: Critically ill patients requiring intensive care unit (ICU) admission suffer from muscle weakness that persists for years, compromising quality-of-life. The pathophysiology of this long-term weakness remains unclear. We hypothesized that former ICU-patients show a long-term abnormal RNA-expression profile, which may contribute to lower long-term strength and for which modifiable risk factors can be identified.

Methods: This pre-planned secondary analysis of the EPaNIC-trial compared muscle transcriptomes of 115 former ICU-patients 5 years after critical illness and 30 matched controls with RNA-sequencing, followed by pathway over-representation and differential co-expression analyses of the differentially expressed RNAs. We used multivariable linear regression analyses to identify which of the abnormal RNA-expressions associated with the long-term muscle strength of the patients and to identify potential risk factors for the abnormal RNA-expressions.

Results: In former patients, 234 down-regulated and 116 up-regulated RNAs were identified after adjustment for age, sex, and BMI. Pathway over-representation and further molecular and histological analyses indicated impaired mitochondrial energy metabolism, disturbed lipid metabolism, and increased collagen formation/fibrosis in former patients. Abnormal muscle RNA-expression in former patients correlated with lower long-term muscle strength. Several treatments given in-ICU and at 5-year follow-up associated with abnormal RNA-expression, most notably in-ICU early parenteral nutrition (early PN) and glucocorticoid use.

Conclusion: Abnormal RNA-expression profiles 5 years after critical illness suggest disrupted mitochondrial function, disturbed lipid metabolism, and fibrosis, associated with lower long-term muscle strength and partly attributable to possibly avoidable risk factors. These findings open perspectives for prevention and possibly treatment of long-term muscle weakness after critical illness.

Trial registration number and date: ClinicalTrials.gov-NCT00512122, July 31, 2007.

Keywords: Critical illness; Intensive care unit; Mitochondria; Muscle weakness; Post-intensive care syndrome; Transcriptome.

Fig. 1

Study design and flowchart of study participants. a Study design. Step 1: An RNA-sequencing analysis was performed on muscle of former ICU-patients, who were studied 5 years after critical illness, and on muscle of matched controls with no ICU history. Step 2: RNAs differentially expressed in former ICU-patients as compared with controls were identified, adjusting for age, sex, and BMI. Step 3: Pathways affected by the differentially expressed RNAs were identified via functional over-representation analyses starting from the KEGG, GOBP and Reactome databases. Step 4: Changes in how the differentially expressed RNAs work together in patients as compared with controls were identified with differential co-expression analysis. Step 5: Further histological and molecular analyses were performed to investigate whether major pathways dysregulated at gene expression level showed functional or structural correlates. Step 6: Abnormal RNA-expression 5 years after ICU-admission was associated with long-term muscle strength measures (strength of hand, wrist, elbow, shoulder, hip, knee, and ankle), adjusted for demographics and ICU baseline characteristics (sex, age, and BMI at 5-year follow-up, history of diabetes, history of malignancy, reason for ICU-admission, severity of illness upon ICU-admission, and sepsis upon ICU-admission). Step 7: Risk factors associated with abnormal RNA-expression in former ICU-patients were identified. Factors considered included demographics (sex, age, and BMI at 5-year follow-up), ICU baseline characteristics (reason for ICU-admission, severity of illness upon ICU-admission, sepsis upon ICU-admission, diabetes, and malignancy), in-ICU complications and treatments (ICU length of stay, duration of mechanical ventilation, duration of hemodynamic support, early versus late parenteral nutrition, duration of treatment with hypnotics, alpha2 agonists, glucocorticoids, benzodiazepines or opioids, and acquisition of a new infection in ICU), and medications taken at the time of follow-up (insulin, glucocorticoids, betablockers, antidepressants, antipsychotics, oral antidiabetics, and statins). b Flowchart of study participants. Of the 4640 patients who had been enrolled in the EPaNIC study, 674 participated in a follow-up assessment 5 years later, in comparison with 50 control individuals with comparable age, sex, and BMI as the former EPaNIC patients. The controls had never required an ICU-admission, did not suffer from conditions that could confound the morbidity endpoints, and were recruited via primary care givers and outpatient clinics. At that 5-year follow-up time point, 120 former ICU-patients and 31 of the controls gave a muscle biopsy. Apart from 5 samples with insufficient RNA, bad RNA-quality or low total read counts, 1 additional sample was excluded as it exhibited abnormally high variation in principal component (PC) analysis at PC1 (electronic supplementary material Fig. S1). Upset plots showing overlap of the different analyses in former ICU-patients and controls are shown in electronic supplementary material Fig. S2. * Patients with pre-ICU neuromuscular disorders, unable to walk without assistance prior to ICU or other disabilities present before follow-up potentially confounding morbidity endpoints, refusing participation or not contactable were excluded. DERNAs differentially expressed RNAs, GOBP Gene Ontology biological process, ICU intensive care unit, KEGG Kyoto Encyclopedia of Genes and Genomes

Fig. 2

Former ICU-patients show long-term differential RNA-expression patterns in muscle as compared with controls. a Volcano plot of RNA-expression in muscle of former ICU-patients as compared with controls. RNAs significantly down-regulated or up-regulated (FDR < 0.05) in former ICU-patients are marked with blue and red dots, respectively. The log₂ scale in the x-axis indicates the fold-change in former ICU-patients as compared with controls; for instance, a value of 1 indicates a twofold higher expression and a value of − 1 indicates a twofold decrease (meaning a reduction to 50%) in the former ICU-patients. Adjusted P values in the y-axis are shown on a -log₁₀ scale, the higher the − log₁₀ value, the lower the P value (e.g., a value of 2 corresponds to P = 0.01 and a value of 5 corresponds to P = 0.00001). b Over-represented pathways for the differentially expressed RNAs in muscle of former ICU-patients as compared with controls. The upper panel shows the top ten of over-represented GO biological processes that are down-regulated in former ICU-patients, the lower panel shows the KEGG and Reactome pathways that are up-regulated in former ICU-patients. Adjusted P values are shown on a − log₁₀ scale, the higher the -log₁₀ value, the lower the P-value. FDR false discovery rate, GO gene ontology, ICU intensive care unit, KEGG Kyoto Encyclopedia of Genes and Genomes

Fig. 3

Functional and structural correlates of abnormal RNA-expression in muscle of former ICU-patients. a Activity of mitochondrial respiratory-chain complex-I. b Activity of citrate-synthase. c Mitochondrial DNA (mtDNA) copy number normalized to the median of the controls. d Electron microscopy was performed for illustrative purposes to verify whether impaired mitochondrial activity coincides with abnormal mitochondrial morphology. C: Control, P: Patient. Former ICU-patients with impaired complex-I and citrate-synthase activity showed mitochondria with an irregular pattern of cristae (reduced number of visible cristae that are oriented toward the sides), mitochondria with vacuoles (arrowhead), as well as mitochondrial swirls (degenerating mitochondria, asterisk). e The distributions of the cross-sectional areas of all myofibers and of type-I and type-II myofibers are shown for former ICU-patients and controls. Cross-sectional area was categorized in blocks of 100 μm² for each subject. The graphs show smoothed curves of the proportion of myofibers in each category for subjects with at least 30 myofibers for the respective analyses. Illustrative photographs are shown with type-I myofibers stained in black and type-II myofibers stained in pink. Lining of myofibers for quantification of myofiber cross-sectional area is illustrated in electronic supplementary material Fig. S8. f Muscle sections were stained for collagen (blue staining). Samples with faint blue staining or only very rare presence of a thicker strand of blue staining were grouped as no fibrosis and samples with moderate-to-strong fibrosis were grouped as fibrosis. Illustrative photographs are shown for these classifications. Box plots in panels a-c depict medians with interquartile ranges (IQR), and whiskers are drawn to the furthest point within 1.5 times the IQR from the box. P values shown are adjusted for age, sex, and BMI in multivariable linear regression analysis. IQR interquartile range, mtDNA mitochondrial DNA

Fig. 4

Association of abnormal RNA-expression in muscle of former ICU-patients with lower muscle strength. This figure summarizes the results of the multivariable linear regression analyses that assessed associations between abnormal RNA-expressions in muscle of former ICU-patients and measures of muscle strength. The number of significant associations for each of the assessed strength measures is indicated next to each bar. A positive association for a down-regulated RNA or a negative association for an up-regulated RNA suggests that more abnormal RNA-expression in former ICU-patients associates with lower strength, pointing to harm, and vice versa. The x-axis shows the percentage of associations indicating harm (brown) or protection (green), calculated by dividing the number of harmful or protective associations by the total number of associations for each of the strength measures, split up for down-regulated and up-regulated RNAs. For example, we found the expression of 20 down-regulated RNAs and 6 up-regulated RNAs to be associated with lower dominant handgrip strength, thus all pointing to harm (100.0%)

Fig. 5

Risk factors for abnormal long-term RNA-expression in muscle of former ICU-patients. a The figure summarizes the results of the multivariable linear regression analyses that assessed associations between risk factors and abnormal RNA-expressions in muscle of former ICU-patients. The number of significant associations for each risk factor is indicated next to each bar. A negative association of a risk factor with a down-regulated RNA or a positive association of a risk factor with an up-regulated RNA indicates that (higher) exposure to the risk factor associates with more abnormal RNA-expression in former ICU-patients, pointing to harm, and vice versa. The x-axis shows the percentage of associations indicating harm (brown) or protection (green), calculated by dividing the number of harmful or protective associations by the total number of associations for each of the risk factors, split up for down-regulated and up-regulated RNAs. For example, we found age to be associated with the expression of 113 down-regulated RNAs and 45 up-regulated RNAs. For 111 (98.2%) of the down-regulated RNAs and all up-regulated RNAs (100.0%), higher age was associated with more abnormal expression. b Network analysis, with use of the protein-protein interactions database web-tool STRING, on the RNAs of which expression was affected by early PN and glucocorticoid treatment in the ICU. STRING includes direct (physical) and indirect (functional) associations derived from both experimental results and computational predictions, annotated with one or more confidence scores ranging from 0 (lowest confidence) to 1 (highest confidence). Only high confidence interactions (> 0.7) are shown