Insufficient Autophagy Contributes to Mitochondrial Dysfunction, Organ Failure, and Adverse Outcome in an Animal Model of Critical Illness*

www.ccmjournal.org January 2013 • Volume 41 • Number 1 Nonresolving multiple organ failure is a leading cause of morbidity and mortality in prolonged critically ill patients (1). The pathogenesis is multifactorial involving a complex interplay among inflammatory, hemodynamic, metabolic, endocrine, and immunological disturbances (2). Although inadequate oxygen delivery to vital organs is an important etiological factor, increasing evidence implicates mitochondrial dysfunction as a culprit in multiple Objective: Increasing evidence implicates mitochondrial dysfunction as an early, important event in the pathogenesis of critical illness-induced multiple organ failure. We previously demonstrated that prevention of hyperglycemia limits damage to mitochondria in vital organs, thereby reducing morbidity and mortality. We now hypothesize that inadequate activation of mitochondrial repair processes (clearance of damaged mitochondria by autophagy, mitochondrial fusion/fission, and biogenesis) may contribute to accumulation of mitochondrial damage, persistence of organ failure, and adverse outcome of critical illness. Design: Prospective, randomized studies in a critically ill rabbit model. Setting: University laboratory. Subjects: Three-month-old male rabbits. Interventions: We studied whether vital organ mitochondrial repair pathways are differentially affected in surviving and nonsurviving hyperglycemic critically ill animals in relation to mitochondrial and organ damage. Next, we investigated the impact of preventing hyperglycemia over time and of administering rapamycin as an autophagy activator. Measurements and Main Results: In both liver and kidney of hyperglycemic critically ill rabbits, we observed signs of insufficient autophagy, including accumulation of p62 and a concomitant decrease in the microtubule-associated protein lightchain-3-II/microtubule-associated protein light-chain-3-I ratio. The phenotype of insufficient autophagy was more pronounced in nonsurviving than in surviving animals. Molecular markers of insufficient autophagy correlated with impaired mitochondrial function and more severe organ damage. In contrast, key players in mitochondrial fusion/fission or biogenesis were not significantly different regarding survival status. Therefore, we focused on autophagy to study the impact of preventing hyperglycemia. Both after 3 and 7 days of illness, autophagy was better preserved in normoglycemic than in hyperglycemic rabbits, which correlated with improved mitochondrial function and less organ damage. Stimulation of autophagy in kidney with rapamycin correlated with protection of renal function. Conclusions: Our findings put forward insufficient autophagy as a potentially important contributor to mitochondrial and organ damage in critical illness and open perspectives for therapies that activate autophagy during critical illness. (Crit Care Med 2013; 41:182–194)

www.ccmjournal.org January 2013 • Volume 41 • Number 1 N onresolving multiple organ failure is a leading cause of morbidity and mortality in prolonged critically ill patients (1). The pathogenesis is multifactorial involving a complex interplay among inflammatory, hemo-dynamic, metabolic, endocrine, and immunological disturbances (2). Although inadequate oxygen delivery to vital organs is an important etiological factor, increasing evidence implicates mitochondrial dysfunction as a culprit in multiple Objective: Increasing evidence implicates mitochondrial dysfunction as an early, important event in the pathogenesis of critical illness-induced multiple organ failure. We previously demonstrated that prevention of hyperglycemia limits damage to mitochondria in vital organs, thereby reducing morbidity and mortality. We now hypothesize that inadequate activation of mitochondrial repair processes (clearance of damaged mitochondria by autophagy, mitochondrial fusion/fission, and biogenesis) may contribute to accumulation of mitochondrial damage, persistence of organ failure, and adverse outcome of critical illness. Design: Prospective, randomized studies in a critically ill rabbit model. Setting: University laboratory. Subjects: Three-month-old male rabbits. Interventions: We studied whether vital organ mitochondrial repair pathways are differentially affected in surviving and nonsurviving hyperglycemic critically ill animals in relation to mitochondrial and organ damage. Next, we investigated the impact of preventing hyperglycemia over time and of administering rapamycin as an autophagy activator. Measurements and Main Results: In both liver and kidney of hyperglycemic critically ill rabbits, we observed signs of insuf-ficient autophagy, including accumulation of p62 and a concomitant decrease in the microtubule-associated protein lightchain-3-II/microtubule-associated protein light-chain-3-I ratio. The phenotype of insufficient autophagy was more pronounced in nonsurviving than in surviving animals. Molecular markers of insufficient autophagy correlated with impaired mitochondrial function and more severe organ damage. In contrast, key players in mitochondrial fusion/fission or biogenesis were not significantly different regarding survival status. Therefore, we focused on autophagy to study the impact of preventing hyperglycemia. Both after 3 and 7 days of illness, autophagy was better preserved in normoglycemic than in hyperglycemic rabbits, which correlated with improved mitochondrial function and less organ damage. Stimulation of autophagy in kidney with rapamycin correlated with protection of renal function. Conclusions: Our findings put forward insufficient autophagy as a potentially important contributor to mitochondrial and organ damage in critical illness and open perspectives for therapies that activate autophagy during critical illness. (Crit Care Med 2013; 41:182-194) Key Words: autophagy; critical illness; hyperglycemia; insulin; mitochondria; multiple organ failure Insufficient Autophagy Contributes to Mitochondrial Dysfunction, Organ Failure, and Adverse Outcome in an Animal Model of Critical Illness* organ failure independent of oxygen delivery. Patient and animal studies have found a clear association of the degree of mitochondrial dysfunction with severity of organ dysfunction and mortality risk (3,4). Prevention of hyperglycemia with insulin has been shown to limit the extent of mitochondrial and vital organ damage in patients and animals (5)(6)(7)(8). These benefits appear explained by prevention of hyperglycemia and not insulin administration per se (6,7). The threshold at which hyperglycemia becomes deleterious remains debated (9,10).
Potential mechanisms leading to mitochondrial dysfunction during critical illness have been intensely investigated, focusing on how to prevent direct mitochondrial damage such as provoked by oxidant species or cellular glucose overload (11). Another potential cause of mitochondrial dysfunction that has drawn less attention in this context is a defective mitochondrial repair. Indeed, maintaining a healthy mitochondrial population requires sophisticated damage repair systems (12). Three main processes are in-volved: mitochondrial fusion/fission, (macro)autophagy, and biogenesis ( Fig. 1).
In postmortem liver biopsies of prolonged critically ill patients, we previously observed lack of mitochondrial repair in the face of severe mitochondrial damage (13,14). We now hypothesized that survival from critical illness requires sufficient activation of mitochondrial repair mechanisms to prevent mitochondrial dysfunction and organ failure. We further hypothesized that mitochondrial dysfunction evoked by hyperglycemia could be partly attributed to suppressed mitochondrial repair (15)(16)(17)(18) and that stimulation of mitochondrial repair may improve outcome. We addressed these questions in a validated animal model of critical illness (19,20). First, we studied whether hepatic and renal mitochondrial repair pathways are differentially affected in surviving and nonsurviving hyperglycemic critically ill animals in relation to mitochondrial function and organ damage. Next, we investigated the impact of preventing hyperglycemia over time and of administering rapamycin as a well-documented autophagy activator. Simplified scheme highlighting the investigated mediators of mitochondrial repair. Mitochondrial biogenesis requires an orchestrated coordination between the nucleus and the mitochondria. Expression of nuclear-encoded mitochondrial proteins and transcription factors is mediated by peroxisome proliferator-activated receptor-γ coactivator-1-α (PGC-1α) and nuclear respiratory factors 1 and 2 (NRF-1 and NRF-2). The majority of mitochondrial proteins, among which nicotinamide adenine dinucleotide reduced form (NADH)-dehydrogenase-[ubiquinone]-1α-subcomplexsubunit-9 (NDUFA9), is encoded by nuclear deoxyribonucleic acid (nDNA) and imported into the mitochondria after synthesis. nDNA-encoded mitochondrial transcription factors such as mitochondrial transcription factor A (TFAM) control the expression of mitochondrial DNA (mtDNA)-encoded proteins such as NADH dehydrogenase subunit 6 (ND6) (27). Healthy mitochondria undergo continuous cycles of fusion and fission. Consecutive fusion and fission processes allow both exchange and dilution of molecular damage as well as bundling of damage into one irreversibly damaged daughter organelle that is subsequently targeted for removal by autophagy (15,28). Mitofusin-2 (Mfn2) and optic-atrophy-1 (OPA1) are two key proteins involved in mitochondrial fusion and are located on the outer, respectively, inner mitochondrial membrane. Dynamin-related protein 1 (Drp1) is a cytosolic protein and a key effector of mitochondrial fission. Autophagy is a general degradation process for organelles and macromolecules and is the only clearance mechanism for damaged mitochondria (15,29). It starts with the formation of an isolation membrane that engulfs portions of cytoplasm with or without organelles. The isolation membrane elongates toward a vesicular structure (autophagosome), which subsequently fuses with a lysosome to form an auto(phago) lysosome in which the sequestered content is degraded and recycled to the cytoplasm. On the molecular level, a complex consisting of Class 3 phosphatidylinositol-3-kinase (PI3KCl3) and Beclin-1 can activate autophagy. A crucial step in the elongation of the isolation membrane is the incorporation of microtubule-associated protein light-chain-3 (LC3). To enable incorporation, the cytosolic form of LC3, LC3-I, needs to be converted into the lipidated form LC3-II. LC3-II probably acts as a scaffold protein that supports membrane elongation and is also thought to have a role in cargo recruitment through p62 (also called sequestosome-1). Because p62 has binding sites for LC3 and ubiquitin, it can provide a molecular link between LC3-II on the growing isolation membrane and ubiquitinated substrates (e.g., mitochondria, protein aggregates). p62 is known to accumulate in conditions when autophagy is absent or insufficient (25,26). nl ψ m = normal mitochondrial membrane potential; ↓ψ m = decreased mitochondrial membrane potential.

Mitochondrial Repair in Relation to Survival Status.
Eighteen 3-month-old male New Zealand white rabbits were anesthetized on day −1 at 10 am ± 1 hr by intramuscular injection of 20 mg/kg ketamine (Eurovet, Heusden-Zolder, Belgium) and 0.05 mg/kg medetomidine (Orion, Espoo, Finland) and maintenance by isoflurane inhalation (Schering-Plough, Brussels, Belgium) (Fig. 2). Catheters were placed for repeated blood sampling (right carotid artery) and administration of fluids and insulin (right jugular vein). After instrumentation, alloxan monohydrate (150 mg/kg; Sigma-Aldrich, Bornem, Belgium) was slowly administered intravenously to eliminate endogenous insulin release. This allows exogenous control of glucose and insulin administration and minimizes interindividual differences in insulin levels on glucose load. During emergence from anesthesia, 0.15 mg/kg piritramide was injected intramuscularly (Janssen-Cilag, Beerse, Belgium). Basal fluid resuscita-tion (9 mL/hr), consisting of Hartmann solution (Baxter, Lessines, Belgium) enriched with 5% glucose, was given to prevent hypoglycemia.
On day 0, at 4 pm ± 1 hr, animals were anesthetized. After performing a paravertebral block (5 mL lidocaine 1%; Astra-Zeneca, Brussels, Belgium), a third-degree burn wound was applied on both flanks (15%-20% body surface area, painless by itself because sensory nerve endings are destroyed). Fluid resuscitation was increased (16 mL/hr Hartmann glucose), and animals were targeted to hyperglycemia (250-350 mg/dL, 13.9-19.4 mmol/L) with concomitant hyperinsulinemia (insulin [Actrapid; Novo Nordisk, Bagsvaerd, Denmark] infused at minimum rate of 2 U/kg/24 hr). This condition is relevant because most critically ill patients develop severe hyperglycemia unless treated and are hyperinsulinemic regardless of insulin treatment (21). During emergence from anesthesia, 0.15 mg/kg piritramide was administered intramuscularly. Animals were deprived of regular chow but received water and hay ad libitum.
At 1:30 pm ± 1 hr on day 7, animals were anesthetized with ketamine (7 mg/kg intravenous bolus) and isoflurane maintenance. After performing a tracheostomy, samples of liver and kidney (cortex) were taken under normoventilation (Small animal ventilator KTR-5; Hugo Sachs Elektronik, March-Hugstetten, Germany), frozen in liquid nitrogen, and stored at −80°C. Animals were euthanized by cardiectomy. Animals entering the (pre)terminal state before day 7 (n = 6), as evidenced clinically and by blood gas analysis, were anesthetized, sampled, and euthanized at that time. If death occurred unwitnessed (n = 4), organs were not harvested to avoid bias by postmortem artefacts.
To establish a reference range for the measured parameters, samples were harvested from healthy animals (n = 15) receiving regular rabbit feeding ad libitum.

Impact of Preventing Hyperglycemia on Mitochondrial
Repair Over Time. This study was performed in critically ill rabbits for which we previously studied the impact of preventing hyperglycemia on mitochondrial function (6,7). After the burn wound, animals were randomized to hyperglycemia or normoglycemia (80-110 mg/dL, 4.4-6.1 mmol/L) with concomitant hyperinsulinemia. Animals were fluid-resuscitated and parenterally fed throughout the study. Organs were sampled from animals surviving their prespecified 3-day (n = 5 per group) or 7-day (n = 8 per group) time course of illness and from healthy animals (Fig. 2).

Impact of Rapamycin Administration on Vital Organ Damage.
Immediately after burn injury, rapamycin-treated animals (n = 6) received an intramuscular injection of 0.06 mg/kg rapamycin (Rapamune; Pfizer Limited, Sandwich, UK; Fig. 2) followed by a 0.05-mg/kg injection after 24 and 48 hrs. Animals were fluid-resuscitated, parenterally fed, and targeted to hyperinsulinemia-hyperglycemia throughout the 3-day study period. This group was compared with a similarly treated group not receiving rapamycin (n = 6) and with healthy animals (n = 4).

Evaluation of Organ Damage and Mitochondrial Function and Measurement of Plasma Insulin
Alanine aminotransferase (ALT), aspartate aminotransferase (AST), and creatinine were measured in plasma and urine (creatinine) using Modular Roche and specific reagents (Roche/Hitachi, Bern, Switzerland; Algemeen Medisch Laboratorium, Antwerp, Belgium). Creatinine clearance was calculated. Activities of the mitochondrial enzyme complexes I-V were measured using spectrophotometry (5). Plasma insulin was analyzed by radioimmunoassay (antibody provided by Prof. R. Bouillon, KU Leuven, Belgium) (22).

Statistics
Data are presented as mean ± sd or median (interquartile range). Differences between groups were analyzed using the Kruskal-Wallis and Mann-Whitney U test (nonnormally distributed data) and factorial one-way analysis of variance and Fisher's protected least-significant-difference post hoc testing (normally distributed data) after verifying homoscedasticity (F-test for equal variance). Correlations were studied by Pearson and Spearman's rank-correla tion. Nonnormally distributed data were analyzed with parametric tests after log or square root transformation if this resulted in a normal distribution. Two-sided p values ≤ 0.05 were considered statistically significant. No corrections were made for multiple comparisons. Analyses were performed using Statview 5.0.1 (SAS Institute, Cary, NC).

Mitochondrial Repair in Relation to Survival Status
Mortality, Nutrition, Metabolic Status, and Organ Damage. Mortality of the critically ill rabbits was 56% (10 of 18). Amounts of infused nutrients-within the physiological range-levels of hyperglycemia and infused insulin doses were not significantly different between survivors and nonsurvivors ( Table 1). However, nonsurvivors had higher plasma insulin levels, reflecting greater insulin resistance. Especially nonsurvivors revealed multiple organ damage, as illustrated by increased last day ALT and AST levels (cytolytic liver injury) and decreased mean creatinine clearance (kidney dysfunction) ( Table 1).

Mitochondrial Function.
Hepatic respiratory chain complexes I-V activities were lowered by critical illness with a further decrease of complex V activity in nonsurvivors (Fig. 3). Renal complexes I, II, III, and V activities were not significantly different from controls in survivors but significantly decreased in nonsurvivors, except for complex III. Renal complex IV activity was increased in survivors but not in nonsurvivors.
Autophagy. In the liver and kidney of nonsurvivors, we observed a marked accumulation of p62 protein (typical finding in conditions of inhibited/insufficient autophagy [25,26]), which was much less pronounced in survivors (Fig. 4). The p62 rise was not the result of increased synthesis, because mRNA levels did not increase. Critical illness increased the levels of LC3-I (unprocessed cytosolic form of LC3) in both organs, not significantly different between survivors and nonsurvivors. LC3-II levels (autophagosome-associated form of LC3) were lower in nonsurvivors than in survivors. The LC3-II/LC3-I ratio (marker of autophagosome formation) was substantially lower in the liver and kidney of nonsurvivors as compared with critically ill survivors and healthy animals. No significant differences were found in the levels of Class 3 phosphatidylinositol-3-kinase (upstream activator of autophagy), except for a decrease in the kidney of survivors. Beclin1 (involved in Class 3 phosphatidylinositol-3-kinase activation) was elevated in the liver but not in the kidney irrespective of survival status.
Hepatic p62 protein levels correlated inversely with the activities of complexes I, II, III, and V in the liver and positively with plas ma ALT and AST (Table 2). Renal p62 protein cor-  related inversely with the activities of complexes I, II, and V in the kidney and had a mean daily creatinine clearance.
Mitochondrial Fusion and Fission. Critical illness induced a decrease in hepatic mitofusin-2, a protein involved in outer mitochondrial membrane fusion (Fig. 4), whereas the inner membrane fusion protein optic-atrophy-1 showed a marked increase. The fission mediator dynamin-related protein-1 was not significantly different from the normal range. Critical illness did not significantly affect the renal expression of these proteins. In both tissues, levels of these proteins were not significantly different between survivors and nonsurvivors.
Mitochondrial Biogenesis. Expression of mitochondrial transcription factor A, a key transcription factor in mitochondrial biogenesis, was lowered in the liver of nonsurviving critically ill rabbits, whereas in survivors it remained low-normal (Fig. 4). Two complex I subunits (mtDNA-encoded NADHdehydrogenase-subunit-6 and nDNA-encoded NADHdehydrogenase-[ubiquinone]-1α-subcomplex-subunit-9) were significantly decreased in the liver of survivors but not in nonsurvivors. In the kidney, these proteins were not significantly affected by illness or survival status.

Mortality, Organ Damage, and Mitochondrial Function.
We previously described improved 7-day survival, attenuation of severe liver and kidney damage, and protection of mitochondrial function by prevention of hyperglycemia (6,22). Three days of critical illness did not lead to premature deaths, but already mildly affected renal mitochondrial and organ function of hyperglycemic rabbits (7). We now evaluated hepatic mitochondrial function after 3 days. Mitochondrial com-plexes I-V were severely affected in hyperglycemic rabbits but preserved in normoglycemic rabbits (Supplemental Digital Content 1, http://links.lww.com/CCM/A521, which shows the mitochondrial respiratory chain enzyme activities in the liver of hyperglycemic and normoglycemic rabbits after 3 days of critical illness).
Autophagy. Because only autophagy was differentially affected in relation to survival, we focused further on this pathway (Fig. 5). In liver of hyperglycemic rabbits, p62 protein levels (but not mRNA, Supplemental Digital Content 2, http:// links.lww.com/CCM/A522, which illustrates the relative p62 mRNA expression level in liver and kidney of hyperglycemic and normoglycemic critically ill animals after 7 days of illness) increased after 3 and 7 days of illness, which was largely prevented in normoglycemic rabbits. Observations were similar for kidney. Increases in hepatic LC3-I were attenuated by normoglycemia after 3 but not after 7 days. LC3-II levels had increased only after 7 days. Thus, the hepatic LC3-II/LC3-I ratio decreased substantially after 3 days, which was partially prevented by normoglycemia and in both groups returned to a low-normal range after 7 days. In the kidney, LC3-I significantly increased after 7 days only, whereas LC3-II levels were not substantially affected at either time point. Hence, the LC3-II/ LC3-I ratio in kidney was not significantly altered after 3 days but decreased after 7 days. Normoglycemia had no major impact on renal LC3. Class 3 phosphatidylinositol-3-kinase and beclin1 were not significantly altered in the liver or kidney, except for an increase in beclin1 after 3 days in the liver and after 7 days in the kidney of hyperglycemic animals, and were not substantially affected by maintaining normoglycemia.
In the liver and kidney, p62 protein correlated inversely with several mitochondrial enzyme activities and directly with the  degree of organ damage after 3 and/or 7 days of critical illness, as assessed by plasma AST, ALT, and creatinine ( Table 3).

Impact of Rapamycin on Vital Organ Damage
To establish a role for insufficient autophagy in vital organ damage, we administered rapamycin as a well-validated autophagy activator. All animals survived the preset 3-day study period. In the kidney, but not in the liver, rapamycin prevented the rise in p62 protein (Fig. 6). Rapamycin had no systematic impact on LC3. Hepatic complex V activity was low irrespective of rapamycin treatment and correlated inversely with the elevated p62 ( Table 4). In the kidney, complex V activity appeared better preserved by rapamycin. Rapamycin did not significantly affect AST or ALT levels but prevented the rise in plasma creatinine. In the liver and kidney, markers of insufficient autophagy correlated strongly with more severe organ damage.

DISCUSSION
In an animal model of hyperglycemic prolonged critical illness, we documented insufficient activation of autophagy in the liver and kidney, which correlated with impaired mitochondrial function, more severe organ damage, and death. Suppression of autophagy occurred already early in the disease course after 3 days. Autophagy activation appeared modifiable by maintaining normoglycemia, correlating with improved mitochondrial function and less organ damage with this intervention. Activation of autophagy in the kidney with rapamycin prevented the decline in renal function. Unlike autophagy, key mediators of mitochondrial fusion/fission and biogenesis were largely comparable for surviving and nonsurviving animals, suggestive of a less determining impact on mitochondrial and organ function and ultimate outcome in our model. Deficient mitochondrial repair or failure to upregulate these mechanisms in conditions of external mitochondrial damage leads to accumulation of damaged mitochondria, compromises energy provision, and can ultimately lead to cellular dysfunction and death (12). The importance of intact mitochondrial repair is underscored by the severe (often lethal) phenotype of mice with (tissue specific) inactivation of the pathways (25)(26)(27)(28)(29)(30). Inactivation of core autophagic proteins in specific tissues of adult mice resulted in accumulation of p62 protein as a marker of compromised autophagic flux, hampered formation of LC3-II as a readout for autophagosome formation, and exerted deleterious effects on mitochondrial and organ function (25,26,30). The abnormalities we observed in critically ill rabbits thus closely resemble this phenotype. Importantly, autophagy appeared more suppressed in nonsurviving animals. Furthermore, p62 protein accumulation correlated with the degree of mitochondrial and organ damage. Signs of inadequate autophagy were already present at an early stage of the disease (after 3 days). Importantly, administration of the autophagy inducer rapamycin improved autophagy in the kidney as well as renal function. Altogether, our data implicate insufficient autophagy in the pathogenesis of organ failure and risk of death during critical illness. In contrast, initial observations from patient and animal studies suggesting increased autophagosome formation were interpreted as a detrimental role of autophagy contributing to adverse outcome (31)(32)(33)(34). However, the number of autophagosomes also increases when autophagy is impaired at the level of fusion with lysosomes (35). The same applies for LC3-II levels (35). Thus, a major limitation of those studies was that substrates that are normally cleared by autophagy such as p62 were not quantified. Hence, the possibility that activation of autophagy could be insufficient was not considered. Recent studies that documented the consequences of insufficient autophagy or interfered with the pathway support the notion of our present work that autophagy is required for organ protection and may be insufficiently activated during critical illness (13,36,37). Numerous mediators tightly regulate the mitochondrial repair processes according to (patho)physiological needs. This implies the need for a more intense upregulation of the damage removal system whenever the organ threat or damage is more severe. Although the underlying mechanisms remain speculative, our data showing signs of insufficient autophagy in vital organs of especially nonsurviving animals may point to a failure of certain regulatory circuits. Pathways that are clas- Figure 6. Effect of rapamycin administration on autophagy and complex V activity in the liver and kidney and on markers of respective organ damage. Relative expression levels of key proteins in autophagy are shown for healthy rabbits and rapamycin-untreated and treated critically ill rabbits (n = 6 per sick group + four healthy rabbits). Plasma markers of organ damage represent the plasma level at day 3. Boxplots indicate median, interquartile range, 10th and 90th percentile. p values ≤ 0.1 are shown. p values above boxplots represent comparisons with healthy controls; p values above a solid line represent comparisons between rapamycin-untreated and treated sick rabbits. LC3 = light-chain-3; ALT = alanine aminotransferase; AST = aspartate aminotransferase.
sically activated in stress conditions such as hypoxia, inflammation, oxidative and endoplasmic reticulum stress are known to stimulate autophagy (15,38) and hence cannot explain the current observations. Two powerful, physiological suppressors of autophagy are nutrients and insulin (15). All critically ill animals were parenterally fed, which may have hampered organ recovery (39). However, both surviving and nonsurviving animals received the same amount of nutrients. In contrast, circulating insulin levels during critical illness were higher in nonsurviving than in surviving animals and thus may have played a role. Recently, a number of previously unknown autophagy-regulating systems have been discovered, including acetylation status of intracellular proteins, epigenetic deoxyribonucleic acid modifications, and changes in microRNA patterns (15,40). Some of these pathways may be disrupted in lethal critical illness (41). This needs further study.
Although insulin is a well-known suppressor of autophagy, our data suggest improved efficiency of autophagy by preventing hyperglycemia with insulin. The similar degree of hyperinsulinemia in both groups may have played a role (22). Because increased cellular glucose per se also can compromise autophagy (42), the benefit of avoiding hyperglycemia may have overruled any deleterious insulin effect. Indeed, two animal models of diabetes showed that prolonged severe hyperglycemia, leading to extensive cellular damage, overwhelmed the autophagic pathway (16) and that treatment of hyperglycemia by insulin or islet transplantation reversed the diabetesinduced decline in autophagic vacuoles in the kidney (43). The impact of glucose-lowering on autophagy appeared more pronounced after 3 days than in rabbits surviving 7 days of critical illness. This may be explained by the fact that by day 7, the sickest, nonsurviving hyperglycemic animals, with the most suppressed autophagy, were no longer included. The impact of survival status may also clarify why we did not observe improved autophagy with glucose-lowering in human postmortem biopsies (13). Also, the higher level of hyperglycemia, more severely overloading the autophagic pathway, in the rabbit model as compared with the critically ill patients may have played a role. Nevertheless, less mitochondrial and organ damage by preventing hyperglycemia in patients and animals (5-7) clearly suggest that during critical illness, the imbalance between damage and damage removal by autophagy is beneficially affected by preventing severe hyperglycemia with insulin.
In contrast to autophagy, mitochondrial fusion/fission and biogenesis were largely unaffected by survival status and thus appear less important for outcome in our model. The data may at first sight contradict the available literature. A study on postmortem biopsies from critically ill patients reported an increase in markers of mitochondrial fusion/fission (14). Although species differences and model-specific alterations cannot be excluded, the higher degree of supportive care and ensuing a more protracted time course of illness before death in patients vs. animals may be an explanation. Likewise, studies on mitochondrial biogenesis put forward an important role for this pathway in mitochondrial recovery, but the timing of the response is not fully clear. A study on muscle biopsies found an early biogenesis response in surviving but not in nonsurviving patients with severe sepsis (44). However, the biogenesis program was mainly activated upstream and not accompanied by a rise in mitochondrial transcription factor A or a clear difference in mitochondrial proteins. Similarly, in another study on muscle biopsies from septic patients-the majority ultimately surviving their illness-found unaltered in vivo mitochondrial protein synthesis despite increased mRNA levels of mitochondrial transcription factors (45). In the liver from patients who died in the ICU, attempts to activate mitochondrial biogenesis appeared insufficient to renew damaged mitochondria (14). Experimental data in septic rodents suggest that mitochondrial biogenesis is crucial for mitochondrial recovery. However, with sublethal insults, effective biogenesis and mitochondrial recovery were found to occur later in the disease course, after an initial phase of mitochondrial depletion (46,47). Altogether, these data suggest that, for effective mitochondrial recovery, both autophagy and mitochondrial biogenesis need to be intact. Our present data only to a limited extent support a role for mitochondrial biogenesis in mediating recovery. Mito- For assessing correlations with circulating markers of organ function, healthy control animals were excluded, because the studied markers of organ function are affected by the parenteral fluid administration and resulting dilution (our unpublished data), which is absent in control animals. The "+" or "-" sign between brackets indicate the deflection of the corresponding r value. For correlations of tissue markers, all animals were included. a p value (determined by Pearson correlation) where Spearman rank correlation coefficient was ≤ 0.05.
chondrial transcription factor A levels were lowest in the liver of nonsurviving animals. However, this decrease was not accompanied by decreases in the measured complex I subunits. This suggests that complex I levels were mainly determined by lowered clearance rather than by insufficient biogenesis. Some limitations of our study need to be addressed. Although the animal model is well validated and closely mimics human critical illness (19,20), one should always be cautious to extrapolate experimental findings to the human setting. However, data on human liver and muscle (13,14) largely corroborate the current findings. Like in every animal study, the number of included animals was relatively small, which may have increased the chance of Type I errors. However, the autophagy-deficient phenotype was consistently present in the consecutive studies, which largely opposes this possibility. Second, although there were clear indications of multiple organ dysfunction/damage in nonsurviving rabbits, the exact cause of death is unclear. Also in the clinical setting of critical care medicine, where patients usually die with nonrecovering multiple organ failure, such a causal relationship often remains elusive (48). Although difficult to exclude, there were no clinical signs of sepsis causing death. Third, we studied the impact of preventing severe hyperglycemia, which may be less pronounced with smaller differences in blood glucose.

CONCLUSIONS
Our observations in a well-validated animal model of critical illness put forward insufficient autophagy as an important mechanism possibly contributing to persistent mitochondrial dysfunction and organ failure as well as mortality. Markers of insufficient autophagy were much more elevated in nonsurviving animals and correlated with the degree of mitochondrial and organ damage in the liver and kidney. Pharmacological autophagy activation improved autophagy in the kidney, which correlated with improved renal function. Preventing hyperglycemia may better preserve the efficiency of autophagy. The underlying mechanisms leading to insufficient autophagy in critical illness should be further investigated. These data open perspectives for therapies that activate autophagy during critical illness.