Glucose regulates the intrinsic inflammatory response of the heart to surgically induced hypothermic ischemic arrest and reperfusion

Abstract

We investigated the isolated working rat heart as a model to study early transcriptional remodeling induced in the setting of open heart surgery and stress hyperglycemia. Hearts of male Sprague Dawley rats were cold-arrested in Krebs-Henseleit buffer and subjected to 60 min normothermic reperfusion in the working mode with buffer supplemented with noncarbohydrate substrates plus glucose (25 mM) or mannitol (25 mM; osmotic control). Gene expression profiles were determined by microarray analysis and compared with those of nonperfused hearts. Perfused hearts displayed a transcriptional signature independent from the presence of glucose showing a more than twofold increase in expression of 71 genes connected to inflammation, cell proliferation, and apoptosis. These transcriptional alterations were very similar to the ones taking place in the hearts of open heart surgery patients. Prominent among those alterations was the upregulation of the three master regulators of metabolic reprogramming, MYC, NR4A1, and NR4A2. Targeted pathway analysis revealed an upregulation of metabolic processes associated with the proliferation and activation of macrophages and fibroblasts. Glucose potentiated the upregulation of a subset of genes associated with polarization of tissue reparative M2-like macrophages, an effect that was lost in perfused hearts from rats rendered insulin resistant by high-sucrose feeding. The results expose the heart as a significant source of proinflammatory mediators released in response to stress associated with cardiac surgery with cardiopulmonary bypass, and suggest a major role for glucose as a signal in the determination of resident cardiac macrophage polarization.

cardiopulmonary bypass (CPB) is used in over 80% of adult cardiac surgeries performed in the United States (4). CPB allows performing surgery on a motionless and bloodless heart, while simultaneously providing physiological support for the patient. However, this technique amplifies the inflammatory response to open heart surgery, which can lead to multiple organ dysfunction and increased morbidity and mortality in the postoperative period (34). The pathophysiological mechanisms contributing to this inflammatory response syndrome involve contact of the blood components with the synthetic material of the extracorporeal circuit as well as reperfusion injury upon restoration of the normal circulation (35, 67). Substantial evidence implicates the reperfused heart itself as a primary source for the circulating proinflammatory mediators released in the postoperative period (15, 64, 68). Global gene expression analyses of human atrial and ventricular tissues have confirmed that on-pump cardiac surgery is associated with a distinct transcriptional signature in the heart. This signature centers on inflammation, apoptosis, and stress-related remodeling (19, 54). However, the respective contribution of resident cardiac immune cells, compared with infiltrating peripheral blood mononuclear cells, remains uncertain.

Stress hyperglycemia and insulin resistance frequently accompany open heart surgery and are further aggravated by the use of CPB (33). Both conditions develop independently from any pre-existing metabolic disorder and are risk factors for increased perioperative morbidity and mortality (13, 20, 57). Glucose metabolism plays a central role in the activation, proliferation, and maturation of immune cells (18). In cultured monocytes, high glucose levels increase the expression of several proinflammatory cytokines, including tumor necrosis factor (TNF)-α and interleukin (IL)-1β, as well as the expression of cell adhesion molecules such as β₂-integrin. Hyperglycemia may thereby contribute to increased leukocyte-endothelial cell interaction and leukocyte vascular infiltration in vivo (56, 59). The impairment of neutrophil degranulation is another potential mechanism linking hyperglycemia to an increased risk of wound infection after surgery (40, 72). Despite these studies reporting on the consequences of hyperglycemia on the activity of peripheral blood mononuclear cells, the effects that glucose has on the myocardial response to surgery and the modulation of these effects by insulin resistance have yet to be investigated.

The isolated perfused mammalian heart preparation is a simple, versatile, and reproducible way to study a broad spectrum of cardiac parameters independently of the effects of neurohormonal factors, of other organs, or cell systems (36, 61, 62). Cold cardioplegic arrest and normothermic reperfusion of Langendorff-perfused isolated rat hearts trigger the expression of the proto-oncogenes c-Jun, c-Fos, and Egr1 in a similar way to what is observed in the heart of patients undergoing cardiac surgery with CPB (1). Using isolated working rat hearts, we have already provided evidence that an increase in intracellular levels of glucose and its metabolites may act as a signal to induce gene expression in the stressed heart (71). Therefore, we propose that the isolated perfused rat heart provides a well-suited and unique approach to study the myocardial-specific response to hypothermic ischemic arrest and reperfusion and the effects of glucose on this response.

The goal of the present study was to investigate the effect of exogenous glucose on transcriptional remodeling of the isolated working rat heart, in the presence or absence of a pre-existing state of insulin resistance. We hypothesized that glucose promotes the activation of resident cardiac immune cells to generate a proinflammatory environment.

MATERIALS AND METHODS

Animals.

Animals were kept on a 12 h light/12 h dark cycle in the University of Texas Health Science Center (UTHealth) McGovern Medical School Animal Care Center or in the Center for Comparative Research Animal Facilities of the University of Mississippi Medical Center (UMMC). Animal experiments were conducted in accordance with the National Institutes of Health's Guide for the Care and Use of Laboratory Animals with all animal protocols approved by the Institutional Animal Care and Use Committees at UTHealth and UMMC.

Male Sprague Dawley rats (200–224 g) were obtained from Envigo (Indianapolis, IN). For ex vivo heart perfusion studies, rats were fed ad libitum a standard laboratory chow (Laboratory Rodent diet 5001; LabDiet, St. Louis, MO) or a high-sucrose diet (sucrose 67% of total calories; diet D11725; Research Diets, New Brunswick, NJ) for 8–10 wk. We and others have previously demonstrated that 8 wk on the high-sucrose diet (HSD) are sufficient to significantly impair systemic and myocardial insulin sensitivity (24, 25, 47). Moreover, the abnormalities in myocardial insulin signaling resemble the ones observed in hearts from Type 2 diabetic individuals and other rodent models of Type 2 diabetes (11, 24). To investigate further the regulation of cardiac gene expression by glucose in vivo, we induced hyperglycemia in another set of rats by administering two low doses of streptozotocin (STZ, 40 mg/kg ip) at 24 h intervals. Control animals were injected with vehicle (citrate buffer pH 4.0). Rats were anesthetized with thiobutabarbital (120 mg/kg ip) and killed 96 h after initiation of STZ treatment. Thiobutabarbital was used as the anesthetic due to its lack of effect on glycemia in the first 15 min following injection (28). The maintenance of normal glycemia after anesthesia was confirmed by measuring blood glucose levels from the tail vein with OneTouch Ultra test strips (LifeScan, Milpitas, CA).

Male C57BL/6J mice (8 wk old) were obtained from the Jackson Laboratory (Bar Harbor, ME). Mice were rendered hyperinsulinemic and insulin resistant using subcutaneous injections of increasing doses of neutral protamine Hagedorn insulin (Novolin N; Novo Nordisk, Bagsværd, Denmark) for 15 days as described previously (23). All mice were killed by cervical dislocation and exsanguination at the time of tissue sample collection.

Perfusion buffers.

The perfusion medium consisted in Krebs-Henseleit (KH) buffer containing (in mmol/l) 118.5 NaCl, 4.75 KCl, 1.18 KH₂PO₄, 1.18 MgSO₄, 2.54 CaCl₂, and 25 NaHCO₃, and equilibrated with 95% O₂, 5% CO₂. All isolated heart perfusions were performed in the presence of the noncarbohydrate substrates DL-β-hydroxybutyric acid (10 mM), acetoacetate (1 mM), and propionate (2 mM). These substrates enter the Krebs cycle directly without being further metabolized in the cytoplasm and therefore provide energy for contraction without producing metabolic intermediates that could potentially alter gene expression (71). To determine whether glucose has an effect on gene expression in the isolated working heart, KH buffer was supplemented with glucose (25 mM). Mannitol (25 mM) was added as an osmotic control to the KH buffer used for the “no glucose” condition. All substrates were purchased from Sigma-Aldrich (St. Louis, MO). The KH buffer composition (glucose or mannitol) was alternated every day, and chow-fed and HSD-fed rats were randomly picked over the 2 wk of heart perfusion to control for time and age-dependent effects on myocardial gene expression.

Isolated heart perfusion.

Hearts were isolated from 16 to 18 wk old rats anesthetized with chloral hydrate (500 mg/kg ip). Following the onset of deep anesthesia, a median longitudinal incision was performed through the ventral abdominal wall to reveal the inferior vena cava. Heparin (200 USP units) was injected directly into the blood vessel to reduce any risk of thrombus formation in the coronary vasculature. A median thoracotomy was then performed to expose the content of the chest cavity. Both heart and lungs were excised and transferred to a dissection dish containing ice-cold KH buffer to arrest the heart and rinse it free of blood.

Hearts were perfused in the working mode using the system described by Taegtmeyer et al. (62). In brief, the hearts were rapidly cleansed from noncardiac tissues, transferred to the aortic perfusion cannula, and immediately perfused on a Langendorff apparatus with a constant pressure of 100 cmH₂O. The left atrium was then cannulated, and the heart perfused in the working mode with nonrecirculating KH buffer using a filling pressure and an afterload of 15 and 100 cmH₂O, respectively. We allowed hearts to stabilize for 5 min before taking measurements every 10 min for 60 min. Heart rate was determined by placing a 2-French pressure catheter (Millar, Houston, TX) in the aortic line. Cardiac power (watts) was calculated as the product of cardiac output (coronary flow plus aortic flow, m³/s) and the afterload (pascals). Myocardial oxygen consumption (MV̇o₂; μmol/min) was measured with a 5300A biological oxygen monitor (YSI, Yellow Springs, OH), using 1.06 mmol/l for the concentration of dissolved O₂ at 100% saturation. A total of 10 hearts from each feeding group were used per perfusion condition. The five hearts from each group that presented the most stable parameters over the whole perfusion period were selected for subsequent analyses.

At the end of perfusion, hearts were quickly removed from the cannula and arrested in ice-cold KH buffer. Approximately 200 mg of the apical region of the left ventricle was immediately dissected and snap-frozen in liquid nitrogen for further molecular analyses. Control nonperfused hearts (n = 5 per feeding group) were dissected and frozen following the same method. The apparatus was completely dismantled at the end of each perfusion day to prevent bacterial contamination. The glassware was immersed in a chromic/sulfuric acid solution, nonglassware parts were soaked in a detergent solution, and the tubing was renewed every day. All parts of the apparatus were thoroughly rinsed with ultrapure water before use.

RNA isolation.

Total RNA was extracted with the use of RNeasy fibrous tissue kit (QIAGEN, Valencia, CA). Residual DNA contamination was eliminated by treatment with DNase I according to the manufacturer’s instructions. RNA concentrations and purity were determined on a NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, Waltham, MA); each sample yielded a minimum concentration of 0.1 µg/µl with an A₂₆₀/A₂₈₀ ratio ranging between 2.0 and 2.1. All samples had a RNA integrity number ≥ 8.5 as determined on a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA).

Microarray processing.

Target preparation and array hybridization were carried out at the Quantitative Genomics and Microarray Facilities Laboratory at UTHealth at Houston. Briefly, Cyanine-3 (Cy3) labeled cRNA was prepared from 150 ng RNA using the One-Color Low RNA Input Quick Amp Labeling kit (Agilent Technologies) according to the manufacturer's instructions and followed by RNeasy column purification (QIAGEN). A total of 1.65 µg of labeled cRNA per sample was hybridized to single-color SurePrint G3 Rat Gene Expression v2 8x60K slides (Agilent Technologies). After washing and drying the slides, we scanned the fluorescent signals and stored them as a high-resolution fluorescence intensity data file. The data were quality-controlled and preprocessed with Agilent Feature Extraction Software (v10.3). Microarray data were deposited in the Gene Expression Omnibus (GEO) database (accession No. GSE85811) and follow the Minimum Information About a Microrray Experiments (MIAME) requirements.

Statistical analysis of microarray data.

Preprocessed data were uploaded into BRB-ArrayTools (v4.5.0 Beta 1), quantile normalized, and analyzed by class comparison. Probes with >40% data points missing were excluded from the analysis. Differentially expressed genes between perfused and nonperfused hearts were filtered on the basis of ≥1.2-fold or 2-fold difference using multivariate analysis test with a false discovery rate set at 0.01 and a 90% confidence level. Data sets were corrected to account for the greater loss of RNA-containing blood cells in perfused hearts when compared with nonperfused hearts.

Gene sets were imported into the Protein ANalysis THrough Evolutionary Relationships (PANTHER v10.0) Classification System and searched to detect statistically overrepresented pathways using the Bonferroni correction for multiple testing. The degree of similarity between the gene expression signatures from different perfusion groups was assessed using the rank-rank hypergeometric overlap algorithm developed by Plaisier et al. (50).

Lists of cardiac genes differentially regulated after on-pump cardiac surgery were retrieved from the literature (19, 54). The published human data were compared with the present results using the rat-human gene orthology data selection as previously described (9). In brief, a list composed of the human genes with a Rattus norvegicus ortholog was retrieved with the Ensembl BioMart tool. The total number of orthologs available at the time of the analysis was 20,046. The genes identified as differentially regulated in perfused rat hearts were retained for gene list overlap analyses only if present in the list of orthologs and on the DNA chips used in the human studies. The number of overlapping genes x between two gene lists was visualized on a Venn diagram. The significance of the overlap was determined by calculating the exact hypergeometric probability of observing x overlapping genes between the two sets of genes. The expected random overlap y from two independent gene lists A and B was calculated with the formula y = (n*D)/N, where n is the number of genes in list A, D the number of genes in list B, and N is the total number of orthologs included in the analysis.

Real-time PCR quantification.

Total RNA was reverse-transcribed into complementary DNA using SuperScript III Reverse Transcriptase (Life Technologies, Carlsbad, CA). Relative quantification of target mRNA levels was performed on a ViiA 7 Real-Time PCR system (Applied Biosystems, Foster City, CA) using custom primers and TaqMan probes synthesized by Eurofins Genomics (Louisville, KY). Primer and probe sequences are available on demand.

Immunoblotting.

Antibodies were purchased from Cell Signaling Technology (CST, Danvers, MA) or Abcam (Cambridge, MA). Heart tissue samples were homogenized in presence of cOmplete Mini protease inhibitor cocktail (Roche Life Science, Indianapolis, IN). Protein concentration was determined using the bicinchoninic acid assay, and 30 µg aliquots were separated by electrophoresis on Criterion TGX precast gels (Bio-Rad, Hercules, CA). Proteins were transferred to 0.45 µm pore size polyvinylidene difluoride membranes and incubated overnight at 4°C with primary antibodies against FOS (CST #2250), SOCS3 (CST #2932), JUN (CST #9165), HSP60 (CST #12165), and TNF-α (Abcam ab6671). Protein detection was carried out using horseradish peroxidase-conjugated anti-rabbit IgG and chemiluminescence. Densitometric analyses were performed with ImageJ 1.48v.

Statistical analysis.

Statistical analysis was performed with GraphPad Prism v6.01 (GraphPad Software, La Jolla, CA). Multiple comparisons of hemodynamic data were performed by two-way ANOVA and Tukey post hoc test. Real-time PCR data were analyzed using one-way ANOVA and Holm-Sidak’s multiple comparisons test. A P value < 0.05 was considered statistically significant.

RESULTS

Transcriptional remodeling in the isolated working rat heart occurs independently from the presence of glucose or insulin resistance.

Fluid-dynamic and functional parameters of the isolated hearts from chow-fed rats perfused with or without glucose remained steady throughout the perfusion experiment, thereby confirming the stability of the perfused heart preparation (Fig. 1A). The functional parameters for insulin-resistant hearts from HSD-fed rats perfused with glucose were not different from the parameters measured in hearts of chow-fed rats. However, hearts from HSD-fed rats perfused without glucose were unable to maintain a stable pump function, and both aortic flow and cardiac power were significantly decreased after 60 min when compared with the other groups (Fig. 1B). This suggests that, unlike the normal rat heart, the insulin-resistant heart is unable to maintain sufficient energy generation for contraction when supplied with the noncarbohydrate substrates only.

Fig. 1.

Fluid dynamic parameters and oxygen consumption in the isolated working rat hearts. Hearts of chow-fed (A) and high-sucrose diet (HSD)-fed rats (B) were perfused with DL-β-hydroxybutyric acid (10 mM), acetoacetate (1 mM), propionate (2 mM), and glucose (25 mM) as substrates (closed symbols) or with the 3 noncarbohydrate substrates and mannitol (25 mM) as an osmotic control (open symbols). Aortic flow, coronary flow, and myocardial oxygen consumption rates were determined every 10 min until end of perfusion. Data are means ± SE of 5 hearts per group. *P < 0.05 and **P < 0.01 vs. hearts of HSD-fed rats perfused with glucose. #P < 0.05 and ###P < 0.01 vs. hearts of chow-fed rats perfused with glucose. $P < 0.05 and $$P < 0.01 vs. hearts of chow-fed rats perfused without glucose.

In hearts of chow-fed rats perfused with the noncarbohydrate substrates and mannitol, 72 genes were differentially expressed when compared with nonperfused hearts of chow-fed rats (fold change ≥ 2; Supplemental Table S1). (The online version of this article contains supplemental data.) In hearts of chow-fed rats perfused with the noncarbohydrate substrates and glucose, 107 genes were differentially expressed when compared with nonperfused hearts of chow-fed rats (fold change ≥ 2; Supplemental Table S1). The vast majority of those genes (71 for the hearts perfused without glucose and 103 for the hearts perfused with glucose) were upregulated, and all 71 genes upregulated in hearts perfused without glucose were also among the genes upregulated in hearts perfused in the presence of glucose (Fig. 2A and Supplemental Table S1). A rank-rank hypergeometric overlap analysis confirmed the strong similarity between the two transcriptional profiles, with the highest degree of overlap found between the 28 top-ranking genes in both lists (10.9-fold enrichment over expected overlap; Fig. 2B).

Fig. 2.

Transcriptional remodeling of isolated working rat hearts occurs independently from glucose or insulin resistance. Venn diagrams representing the degree of overlap between genes differentially regulated (fold change ≥ 2) in presence or in absence of glucose in perfused hearts of chow-fed (A) and high-sucrose diet (HSD)-fed rats (C). A rank-rank hypergeometric overlap analysis was performed on the 72 genes in common differentially regulated among perfused hearts of chow-fed rats (B) and on the 67 genes in common differentially regulated in perfused hearts of HSD-fed rats (D) to assess the degree of similarity between gene expression signatures obtained in presence or in absence of glucose. Each hypergeometric heat map is accompanied by the cluster of genes located in the region of most significant overlap.

In hearts of HSD-fed rats perfused in the absence of glucose, 111 genes were differentially expressed when compared with nonperfused hearts of HSD-fed rats (fold change ≥ 2; Supplemental Table S2). In hearts of HSD-fed rats perfused in presence of glucose, 70 genes were differentially expressed when compared with nonperfused hearts of HSD-fed rats (fold change ≥ 2; Supplemental Table S2). As for the hearts of chow-fed rats, most of the genes differentially expressed in isolated working hearts were upregulated (109 in hearts perfused without glucose; 69 in hearts perfused with glucose), and there was a high degree of overlap in the transcriptional remodeling occurring with 67 genes shared between both groups (Fig. 2C, Supplemental Table S2). A rank-rank hypergeometric overlap analysis demonstrated that not only the transcriptional signature of heart perfusion was shared between the two groups of HSD-fed rats (8.6-fold enrichment over expected overlap), but it was also highly similar to the one observed in chow-fed rats, with 23 out of 24 of the top ranking genes being identical among all four groups of perfused hearts (Fig. 2, B and D). Thus, the results indicate that the presence of glucose or pre-existing insulin resistance does not alter the nature of the early transcriptional remodeling events affecting isolated working rat hearts.

Isolated working rat hearts produce an injury-related response similar to the one observed in patients undergoing on-pump cardiac surgery.

The Protein ANalysis THrough Evolutionary Relationships (PANTHER) ontology-based pathway database was used to identify the top biological pathways activated in perfused hearts from chow-fed rats (42). In accordance with the rank-rank hypergeometric overlap analysis, overrepresented pathways were identical in hearts perfused in absence or in presence of glucose. These included the cholecystokinin receptor pathway, the apoptosis signaling pathway, the gonadotropin releasing hormone receptor pathway, and the inflammation mediated by chemokine and cytokine signaling pathway (Table 1).

Table 1.

Biological pathways overrepresented with a >5-fold enrichment in perfused hearts of chow-fed rats

	Reference List	Perfusion − Glucose			Perfusion + Glucose
PANTHER Pathway	Mapped IDs, n	n	Expected	P Value	n	Expected	P Value	List of Genes
CCKR signaling map	166	8	0.51	7.19E-06	10	0.73	6.52E-07	Myc, Mcl1, Serpine1, Fos, Ier3, Cxcl1, Nr4a1, Jun, Crem,* Hbegf*
Apoptosis signaling pathway	127	6	0.39	4.23E-04	6	0.56	3.63E-03	Tnf, Mcl1, Fos, Atf4, Jun, Atf3
Gonadotropin releasing hormone receptor pathway	233	8	0.71	9.22E-05	11	1.03	1.28E-06	Irs2, Junb, Fos, Map3k8, Jund, Nr4a1, Jun, Atf3, Dusp1,* Nab2,* Tgif1*
Inflammation mediated by chemokine and cytokine signaling pathway	253	6	0.77	1.96E-02	6	1.12	1.45E-01	Junb, Ccl3, Jund, Ccl7, Ccl2, Jun

CCKR, cholecystokinin receptor. The reference list for Rattus norvegicus included 22,656 mapped IDs at the time of analysis.

^*Genes identified in hearts perfused with glucose only.

The cholecystokinin and gonadotropin signaling pathways play an important role in the regulation of cell growth, proliferation, and apoptosis in various cell systems and especially cancer cells (16, 21). Indeed, the genes classified under these overrepresented pathways included several proto-oncogenes and transcription factors playing a major role in cell proliferation and cellular transformation (Myc, Fos, Jun, Junb, Jund), as well as regulators of apoptosis (Mcl1, Ier3). The upregulation of several chemokines (Cxcl1, Ccl3, Ccl7, Ccl2) was also indicative of the activation of nonmyocytes, particularly tissue-resident macrophages and fibroblasts. The same biological processes have been reported to be upregulated in the hearts of patients undergoing cardiac surgery, especially when CPB was used (19). The proinflammatory cytokines TNF-α and IL-1β, which are known to be increased early in the systemic inflammatory response syndrome associated with cardiac surgery (35), were also strongly upregulated in the perfused rat hearts (Supplemental Table S1). In summary, the overrepresented pathways are indicative of the activation of an immediate stress response following hypothermic ischemic arrest and normothermic reperfusion of the rat heart.

To further assess the similarity between the stress responses elicited by the isolated working rat heart and the heart of patients undergoing cardiac surgery with CPB, we performed a systematic comparison of the genes upregulated by more than twofold in hearts from chow-fed rats (Supplemental Table S1) with previously published human data (19, 54). Of the 71 genes upregulated in perfused rat hearts, 54 had a human ortholog present on the microarray chip used in the study from Ruel et al. (54). The comparison of these 54 genes with the 31 genes upregulated in the human right atrium in response to cardiac surgery with CPB revealed a highly significant overlap of 10 genes (P < 2.34E-16; Fig. 3A). Of the 68 genes identified by Ghorbel et al. (19) as exhibiting a twofold or greater increase in expression in the left ventricle of on-pump patients, 65 were found to have a rat ortholog. There was also a highly significant overlap of 25 genes between these 65 human genes and the 71 genes upregulated in perfused rat hearts (P < 4.83E-46; Fig. 3B). These results confirm that the isolated working rat heart closely mimics the transcriptional remodeling events induced by cardiac surgery with CPB in the human heart.

Fig. 3.

Stress responses induced in isolated rat hearts and in cardiac surgery patients are highly similar. The list of genes upregulated in perfused hearts of chow-fed rats (Supplemental Table S1) was compared with the lists of genes previously reported as being activated in response to surgery with cardiopulmonary bypass (CPB) in human right atrium [A (54)] or left ventricle [B (19)]. Only the human-rat orthologs present on both DNA microarray chips were included in the analysis, which represented a total of 8,318 and 19,561 genes, respectively. The numbers of distinct and overlapping orthologs are shown in the proportional Venn diagrams. A detailed list of the distinct and overlapping orthologs for each comparison is provided in Supplemental Table S3.

Top biological pathways activated in perfused hearts of HSD-fed rats were similar to the ones detected in hearts from chow-fed animals (Table 2), thereby confirming that the global response of insulin-resistant hearts to hypothermic ischemic arrest and normothermic reperfusion does not differ from the response induced in insulin sensitive hearts.

Table 2.

Biological pathways overrepresented with a >5-fold enrichment in perfused hearts of HSD-fed rats

	Reference List	Perfusion − Glucose			Perfusion + Glucose
PANTHER Pathway	Mapped IDs, n	n	Expected	P Value	n	Expected	P Value	List of Genes
CCKR signaling map	166	11	0.75	5.41E-08	9	0.48	2.38E-07	Myc, Mcl1, Serpine1, Fos, Ier3, Cxcl1, Nr4a1, Jun, Hbegf, Crem,† Birc3†
Apoptosis signaling pathway	127	7	0.58	3.32E-04	5	0.37	5.73E-03	Mcl1, Fos, Atf4, Jun, Atf3, Birc3,† Tnf†
Gonadotropin releasing hormone receptor pathway	233	10	1.06	2.01E-05	9	0.68	4.32E-06	Irs2, Junb, Fos, Map3k8, Jund, Nr4a1, Tgif1,* Jun, Atf3, Dusp1† Nab2†
Inflammation mediated by chemokine and cytokine signaling pathway	253	7	1.15	2.59E-02	6	0.74	1.53E-02	Junb, Ccl3, Jund, Ccl7, Ccl2, Jun, Rgs1†
TGF-beta signaling pathway	93	4	0.42	1.40E-01	4	0.27	2.57E-02	Junb, Jund, Gdf15, Jun
Blood coagulation	55	4	0.25	1.95E-02	3	0.16	8.99E-02	Plaur, Serpine1, Procr, Plat†
Plasminogen activating cascade	18	3	0.08	1.29E-02	2	0.05	2.02E-01	Plaur, Serpine1, Plat†

HSD, high-sucrose diet. The reference list for Rattus norvegicus included 22,656 mapped IDs at the time of analysis.

^*Genes identified in hearts perfused with glucose only.

^†Genes identified in hearts perfused without glucose only.

Isolated working rat hearts display a metabolic reprogramming characteristic of immune cells and fibroblasts activation.

Real-time PCR was used to confirm upregulation for 12 of the genes identified by microarray. Genes with a moderate (2 ≤ fold change < 6; Tnf, Socs3, Mcl1, Dusp1), high (6 ≤ fold change < 15; Ccl7, Myc, Ccl2, Btg2), and a very high (15 ≤ fold change; Nr4a1, Nr4a2, Fos, Adamts1) increase in expression were investigated. Among the selected genes independently quantified, 10 out 12 (83%) were significantly increased in hearts perfused without glucose, and all 12 genes were confirmed to increase in hearts perfused with glucose (Table 3, Fig. 4). Next, we investigated whether the increase in mRNA levels was accompanied by a similar increase in protein levels. Although neither SOCS3 nor TNF-α could be detected in the heart tissue samples (data not shown), FOS and JUN levels were confirmed to increase dramatically in perfused hearts (Fig. 4B).

Table 3.

Real-time PCR confirmation of genes with ≥2-fold increase in perfused rat hearts

		Fold Change Average vs. Nonperfused
Diet	Gene Symbol	−Glucose	P Value	+Glucose	P Value
Chow	Adamts1	14.38	<0.0001	21.06	<0.0001
	Btg2	10.01	<0.0001	12.91	<0.0001
	Ccl2	8.46	0.1878	12.86	0.0096
	Ccl7	8.23	0.0956	13.24	0.0013
	Dusp1	3.43	0.0003	4.57	<0.0001
	Fos	32.67	<0.0001	43.79	<0.0001
	Mcl1	2.91	<0.0001	3.99	<0.0001
	Myc	11.58	0.0005	18.05	<0.0001
	Nr4a1	36.85	<0.0001	48.03	<0.0001
	Nr4a2	27.15	<0.0001	47.65	<0.0001
	Socs3	16.67	0.0111	29.77	<0.0001
	Tnf	5.58	<0.0001	6.20	<0.0001
HSD	Adamts1	11.48	<0.0001	12.10	<0.0001
	Btg2	9.25	<0.0001	10.01	<0.0001
	Ccl2	9.94	0.1877	15.50	0.0066
	Ccl7	8.09	0.0956	15.00	0.0001
	Dusp1	4.19	<0.0001	4.30	<0.0001
	Fos	64.04	<0.0001	78.40	<0.0001
	Mcl1	2.90	<0.0001	3.10	<0.0001
	Myc	9.27	0.0040	12.30	0.0001
	Nr4a1	39.53	<0.0001	46.60	<0.0001
	Nr4a2	27.88	0.0002	37.8	<0.0001
	Socs3	14.41	0.0305	21.90	0.0006
	Tnf	5.13	<0.0001	5.50	<0.0001

Fig. 4.

Increased expression of transcription and signaling factors involved in cell proliferation, inflammation, and apoptosis. A: real-time PCR quantification was performed to confirm the change in expression of 12 genes identified by microarray as upregulated (fold change ≥ 2) in hearts perfused in presence (▲) or in absence (□) of glucose when compared with nonperfused hearts (●). For clarity, P values reporting differences between perfused hearts and nonperfused hearts are reported in Table 3. B: immunoblotting was used to determine protein expression level of JUN (exposure time: 16 min) and FOS (exposure time: 20 min) proto-oncogenes. Data were normalized to 60 kDa heat-shock protein (HSP60) levels (exposure time: 20 s). All samples were derived at the same time and processed in parallel. n.d. , Nondetectable.

Three of the highly to very highly upregulated genes that were confirmed to increase in all groups of perfused hearts (Myc, Nr4a1, and Nr4a2) encode transcription factors known to act as master regulators of energy homeostasis and metabolic reprogramming in response to various stress stimuli (48, 66). Because the isolated hearts were perfused for only 60 min, we reasoned that any metabolic reprogramming induced as part of the stress response may only have been in a very early phase of activation at the time the tissue samples were recovered. Therefore, the microarray data were reanalyzed with lowered filtering parameters to detect genes with a ≥1.2-fold change in expression compared with the nonperfused hearts. We specifically searched for metabolic genes that were upregulated in at least one of the four groups of perfused hearts. Analysis of the identified genes with Reactome revealed an upregulation of membrane transport proteins and enzymes controlling flux of glucose through the glycolytic pathway (SLC2A3, SLC16A3, HK2) and the hexosamine biosynthetic pathway (GFPT2, UAP1, GNPNAT1), control of pyruvate utilization by mitochondria (PDK4, PDP2), glutamine and glutamate transport and catabolism (SLC1A5, SLC1A1, GLS), collagen and polyamine biosynthesis (ARG1, SAT1), and fatty acid synthesis and lipolysis (OLAH, ACSL4; Fig. 5). The metabolic pathways identified play a central role in the activation and proliferation of immune and stromal cells, and most particularly fibroblasts and macrophages (18). These data are consistent with the results from the PANTHER ontology-based pathway database analysis and with the profile of upregulated chemokines and cytokines, and suggest stress-induced metabolic reprogramming associated with activation and proliferation of fibroblasts and resident cardiac macrophages.

Fig. 5.

Increased expression of enzymes and membrane transport proteins involved in biomass production and cell proliferation. A: metabolic genes upregulated in at least 1 group of perfused rat hearts as determined by microarray analysis (fold change ≥ 1.2) are identified with a red star. Intermediary metabolites from the represented metabolic pathways are not all shown. B: real-time PCR quantification was performed to confirm the change in expression of 12 metabolic genes in hearts perfused in presence (▲) or in absence (□) of glucose when compared with nonperfused hearts (●).

Cardiac expression of the transcription factors NR4A1, NR4A2, and MYC is regulated by glucose and insulin signaling in rat heart both ex vivo and in vivo.

In hearts of chow-fed rats, the presence of glucose in the perfusate potentiated the upregulation of the three master regulators of metabolic reprogramming NR4A1 (+30%), NR4A2 (+76%), and MYC (+56%). Interestingly, the enhancing effect of glucose on the expression of these transcription factors was abrogated in perfused hearts from HSD-fed rats (Fig. 4A). Next, we investigated the consequences of impaired glucose homeostasis on expression of NR4A1, NR4A2, and MYC in rat heart in vivo. In STZ-treated rats, fed blood glucose levels increased from 125 ± 6 to 512 ± 37 mg/dl (P < 0.0001), which was accompanied by a significant increase in cardiac gene transcript levels for all three transcription factors (Fig. 6A).

Fig. 6.

Transcriptional regulation of Nr4a1, Nr4a2, and Myc by glucose and insulin in the rodent heart. A: male Sprague Dawley rats were injected with 2 low doses streptozotocin (STZ; 40 mg/kg ip) or vehicle (Control) given at 24 h interval and killed 96 h after the first injection to determine blood glucose levels and cardiac gene expression. B, C: male C57BL/6J mice were injected with glucose (2 g/kg ip; B) or insulin (1 U/kg ip; C) and killed at the indicated time points to determine blood glucose levels and cardiac gene expression. Data are means ± SE of 4–8 animals per time point. **P < 0.01 and ***P < 0.001 vs. time 0 min. #P < 0.05 and ##P < 0.01 vs. time 120 min. D: myocardial gene expression in C57BL/6J mice rendered hyperinsulinemic and insulin resistant by 15 days of carbohydrate-enriched diet alone (CARB) or in combination with repeated insulin injections (INS + CARB). *P < 0.05 and **P < 0.01 vs. untreated.

To determine whether the regulation of NR4A1, NR4A2, and MYC expression is also affected by glucose and insulin signaling in the mouse heart in vivo, C57BL/6J mice were first subjected to acute glucose and insulin treatments. Cardiac NR4A1 and NR4A2 mRNA expression increased in response to a bolus injection of glucose. The increase was transient and mirrored the rise in blood glucose levels (Fig. 6B). Following acute insulin stimulation, the expression of NR4A1 and NR4A2 in the mouse heart also increased transiently and concomitantly to blood glucose disappearance, reaching a respective 2.1-fold and 1.7-fold increase from baseline 10 min after the injection (Fig. 6C). However, the expression of MYC remained unchanged (P > 0.9). We previously reported that myocardial insulin signaling is impaired in C57BL/6J mice maintained on a carbohydrate-enriched diet for 15 days and that the signaling defect is further aggravated under chronic insulin treatment (23). These experimental conditions caused a downregulation of NR4A1 and NR4A2, but not MYC, in the mouse heart (Fig. 6D).

Altogether, the results suggest that the expression of the nuclear receptors NR4A1 and NR4A2 and of the proto-oncogene MYC is rapidly upregulated by glucose in rat heart and that this increase is additive to other stimuli such as the stress engendered by hypothermic ischemic arrest and normothermic reperfusion. Moreover, alterations in glucose homeostasis and insulin sensitivity may affect the expression of these transcription factors. Lastly, although we found evidence of some overlap in the mechanisms regulating the expression of NR4A1 and NR4A2 in the mouse heart, the upregulation of MYC in response to glucose seemed more specific to the rat heart.

Glucose enhances the expression of a subset of genes associated with alternative activation of macrophages in perfused rat hearts.

Because increased expression of the transcription factors MYC, NR4A1, and NR4A2 in the isolated working rat heart is most likely involved in the induction of the stress-induced metabolic reprogramming of cardiac fibroblasts and resident macrophages, we decided to investigate further whether the presence of glucose was also associated with enhanced expression of a subset of metabolic genes linked to that reprogramming. Among the metabolic genes previously identified by microarray, Hk2, Slc1a1, Slc16a3, Gfpt2, Uap1, Arg1, and Olah only increased in hearts of chow-fed rats perfused with glucose. Increased expression for Uap1, Arg1, Slc16a3, and Olah was confirmed by real-time PCR (Fig. 5B). Interestingly, many of those genes encode enzymes rate limiting for metabolic pathways that play a central role in macrophage M2 polarization (17, 29). NR4A2 has recently been reported to directly activate the transcription of Arg1 and to promote the alternative polarization of macrophages (37). Moreover, the proto-oncogene MYC also activates the expression of a cluster of M2-related genes during macrophage polarization (49), while NR4A1 deletion polarizes macrophages toward a proinflammatory M1 phenotype (22). Therefore, the data suggest that glucose specifically enhances the expression of a subset of transcription factors linked to metabolic reprogramming of alternatively activated resident cardiac macrophages.

The transcriptional regulatory networks involved in both classical and alternative activation of macrophages have recently been investigated by DeepCAGE sequencing (53). By crossing the DeepCAGE-generated data with the present microarray results, transcription factors and transcriptional regulators linked to the M1 phenotype (ARID5A, ATF3, CREM, JUN, MAFF, MAFK, ZFP281), M2 polarization (BHLHE40, FOSL2, ZC3H12A), and both M1 and M2 activation (EGR2, FOS, IRF1, NFIL3) were found to be upregulated in the isolated working rat hearts (Fig. 7A). The transcription factor CCAAT/enhancer binding protein (C/EBP)-β was also identified as another potential inducer of M2 polarization upregulated in isolated working rat hearts (55). Real-time PCR analysis confirmed increased expression for six selected transcription factors in perfused rat hearts when compared with nonperfused controls (Fig. 7B). For perfused hearts of chow-fed rats, glucose further increased the expression of the M2-associated transcriptional regulators ZC3H12A (+35%; P < 0.05), FOSL2 (+39%; P = 0.004), and C/EBP-β (+25%; P = 0.015) but had no significant additive effect on transcriptional regulators associated with the M1 phenotype ATF3 [+21%; not significant (n.s.)], MAFK (−3%; n.s.) and MAFF (+10%; n.s.). Here again, the enhancing effect of glucose on the expression of M2-associated transcriptional regulators was absent in isolated working hearts from HSD-fed, insulin-resistant rats (Fig. 7B).

Fig. 7.

Glucose enhances the expression of transcription factors associated with alternative activation of macrophages. A: several genes encoding transcription factors and transcriptional regulators linked to the activation and polarization of macrophages were identified by microarray analysis as upregulated (fold change ≥ 1.2) in isolated perfused rat hearts. B: real-time PCR quantification confirmed increased expression of those genes in hearts of chow-fed or HSD-fed rats perfused in presence (▲) or in absence (□) of glucose when compared with nonperfused hearts (●). The M2-related transcription factors increased significantly more in hearts of chow-fed rats perfused with glucose, but not in hearts of HSD-fed rats. n.d. , Nondetectable.

DISCUSSION

The goal of this study was to evaluate the transcriptional profile of the isolated rat heart subjected to a brief period of hypothermic ischemic arrest followed by normoxic and normothermic reperfusion in the working mode. Several significant findings came to light. First, the heart itself, in complete absence of neurohormonal or humoral signals, was a significant source of proinflammatory mediators, most of which are likely to be released upon activation of tissue-resident macrophages. Second, glucose did not participate in the induction of the proinflammatory cascade but, rather, potentiated the expression of transcription factors and enzymes associated with the activation of a M2-like/anti-inflammatory phenotype. Third, the transcription factors MYC, NR4A2, and NR4A1 acted as glucose sensors in the adult heart and may mediate part of glucose effects in the stressed myocardium. Lastly, the state of myocardial insulin resistance completely abrogated the stimulatory effect of glucose on this transcriptional remodeling.

Since the inception of DNA microarray technology, the isolated rodent heart has been used many times to perform cardiac gene expression profiling under various conditions. The cardiac transcriptional response to ischemia reperfusion injury and its modulation by protective interventions, including preconditioning (7, 46), postconditioning (12), pharmacological treatments (3, 32), or cardioplegic solutions (58), have been particularly well studied. Most studies relied on hearts perfused under normoxic conditions to establish baseline gene expression levels and, therefore, failed to report the effect of perfusion itself on the cardiac transcriptome. To our knowledge, the study from Farago et al. (14) is the only previously published work to investigate the time-dependent effect of normoxic and normothermic perfusion on gene expression in the isolated rodent heart. The investigators identified 36 genes differentially expressed after 60 or 120 min perfusion clustering in biological pathways associated with the regulation of oxidative and nitrosative stress, cholesterol synthesis, muscle contraction, and regulation of cell cycle. Intriguingly the majority of these genes was downregulated, and none of them overlap with our current findings (14). The reasons for such a difference are unclear but may involve variations in the perfusion technique (Langendorff vs. working heart) or a difference in the concentration of calcium, substrates, or osmolarity of the perfusion buffer. Indeed, the induction of many immediate early genes, including the proto-oncogenes Fos, Jun, and JunB, which were all upregulated in our model, depends more upon variations in intracellular calcium concentrations and temperature rather than ischemia (1, 52). Moreover, while the gene expression profiling performed by Farago and colleagues was limited to 8,000 mouse probes, the use of SurePrint G3 Rat Gene Expression microarrays enabled the analysis of >30,000 Entrez Gene unique IDs in the present study. The coverage of the whole rat transcriptome revealed a striking similitude with the alterations occurring in cardiac surgery patients with CPB, thereby confirming the usefulness of our model to study cardiac remodeling in this particular stress condition.

Expression of the proto-oncogene MYC and of the nuclear receptors NR4A1 and NR4A2 was rapidly induced in the isolated working rat heart, and this induction was potentiated by glucose. We now also provide evidence that these transcription factors may act as cardiac glucose sensors in vivo. These transcription factors play a crucial role in the activation of cellular metabolic processes required for cell growth and proliferation (10, 43). Thus, the induction of MYC in the adult myocardium promotes cardiomyocyte hypertrophy as well as glucose uptake and utilization (2, 69). MYC expression has long been known to be sensitive to glucose concentration in cultured cells (6), and its induction is required for carbohydrate-responsive element-binding protein-dependent activation of glucose-responsive genes (73).

The NRA4 family of nuclear receptors is composed of three members (NR4A1/NUR77, NR4A2/NURR1, and NR4A3/NOR1) that are expressed in a wide variety of tissues and cell types. In contrast to other nuclear receptors, the activity of the NR4A family members is ligand independent and mostly regulated at the level of gene expression (38). In the rodent heart, NR4A1 plays a complex regulatory role in pathological remodeling with either a pro- or an antihypertrophic function depending on the nature of the initial insult (41, 70). Myocardial β-adrenergic signaling transiently upregulates all three NR4A nuclear receptors, which is accompanied by the induction of the genes encoding hexokinase 2 (Hk2), pyruvate dehydrogenase kinase 4 (Pdk4), and peroxisome proliferator-activated receptor gamma coactivator 1-alpha (Ppargc1a) (44). Similarly in skeletal muscle β-adrenergic receptor-mediated activation of NR4A1 leads to increased expression of glycolytic genes (8). In the present study, we demonstrate that a rapid induction of myocardial NR4A1 and NR4A2 and of metabolic genes encoding enzymes promoting glycolysis (HK2, PDK4) can occur in complete absence of neurohormonal stimulation. Moreover, increased expression of key enzymes in the hexosamine biosynthetic pathway (GFPT2, UAP1, GNPNAT1) suggests enhanced mechanisms of protein O-GlcNAcylation in the perfused rat heart. For adult cardiac myocytes, the switch from oxidative phosphorylation to glycolysis and increased protein O-GlcNAcylation may represent an adaptive mechanism to promote immediate cell survival as suggested in other stress conditions (30, 63). Prior work proposed that the rerouting of glucose into the hexosamine biosynthetic pathway increases the glycosylation of transcription factors to favor the reinduction of the fetal gene program to protect the stressed heart (63, 71).

Although changes in glucose utilization may take place in cardiac myocytes, the global activation of metabolic pathways involved in cell proliferation rather suggests a reprogramming of stromal cells (Fig. 5A). The metabolic remodeling is reminiscent of cancer cell reprogramming, with increased glycolysis promoting rapid energy generation, while glutamine catabolism provides ornithine and feeds into the Krebs cycle for biosynthesis of macromolecules (5, 18). Many of the genes identified in Fig. 5 encode rate-limiting enzymes or regulators of rate-limiting steps in their respective metabolic pathways. Thus, PDK4 and PDP2 are respectively inhibitor and activator of pyruvate dehydrogenase, the rate-limiting enzyme of glucose oxidation. GLS controls the rate-limiting step in glutaminolysis, and its upregulation correlates with increased glutamine utilization in cancer cells. Likewise, the concentration of the glutamine transporter SLC1A5 is a regulator of mTOR activation and cell growth (26). ACSL4 controls intramitochondrial arachidonic acid generation and export, while GFPT2 is the first and rate-limiting enzyme in hexosamine biosynthesis. In macrophages, the expression of ARG1 is rate limiting for polyamine synthesis, while SAT1 is the rate-limiting enzyme in polyamine catabolism (31).

Because several M1-like/proinflammatory markers (IL-1β, TNF-α, SOCS3, CCL2, CCL3, CCL7) were among the most upregulated genes in the perfused rat hearts, we further propose that this metabolic reprogramming is indicative, at least in part, of the activation of resident cardiac macrophages. In support of this hypothesis a substantial number of factors, activators, and regulators of transcription involved in macrophage activation were also induced in the isolated working rat heart. In mice, the population of cardiac resident macrophages has been estimated at a few hundred per milligram of tissue, and these cells can self-renew locally through proliferation (27, 45). Interestingly, cardiac tissue macrophages are not overtly polarized and express both M1 and M2 markers under quiescent conditions. However, these cells are highly plastic and can adopt a classical proinflammatory M1 phenotype, a tissue-reparative M2 phenotype, or an intermediate phenotype, depending on the nature of the signals they encounter within tissue (45).

Intracellular metabolic remodeling plays a crucial role in the acquisition of particular phenotypic traits by activated macrophages. Thus, M1-like/proinflammatory macrophages are characterized by high glycolytic rates and an increased number of glucose transporter GLUT3 at the plasma membrane (18, 60). Conversely, M2-like/anti-inflammatory macrophages rely more on fatty acid oxidation and oxidative phosphorylation for energy generation (18). Contrary to our initial hypothesis, glucose did not affect the expression of proinflammatory markers in the isolated working rat heart. Conversely, the hexose had an enhancing effect on the upregulation of M2-associated transcription factors and activators (including MYC, NR4A1, and NR4A2), which was accompanied by increased expression of a specific subset of metabolic genes implicated in polyamine biosynthesis (ARG1), glutamate transport (SLC1A1), protein glycosylation (UAP1, GFPT2), and fatty acid synthesis (OLAH). The expression of ARG1 is increased upon activation of NR4A1, NR4A2, and C/EBPβ (37, 51), and endogenous polyamine biosynthesis potentiates macrophage M2 polarization (65). MYC-driven glutaminolysis also fuels polyamine synthesis upon immune cells activation (66), and macrophage M2 polarization is dependent on the activation of glutamine catabolism and hexosamine biosynthesis (29). In addition, increased fatty acid synthesis is required to sustain elevated membrane biosynthesis activity in phagocytic M2 macrophages (18). Collectively, these observations strongly suggest that tissue-resident macrophages are activated in our ex vivo model of surgically induced hypothermic ischemic arrest and reperfusion. While M1-like/proinflammatory markers were upregulated to the same level in all four groups of perfused hearts, glucose specifically promoted the induction of an M2-like/tissue reparative phenotype, an effect that was lost in the insulin-resistant heart (Fig. 8). Whether this corresponds to the acquisition of an intermediate phenotype by macrophages or to the activation of a subpopulation of cells remains to be determined.

Fig. 8.

Transcriptional remodeling in the isolated rat heart and its modulation by glucose and insulin resistance. The gene expression signature of the isolated working rat heart is characterized by the activation of immediate early genes (IEGs), transcription factors (TFs), and comodulators of transcription, most of which have been associated with inflammation and M1 and/or M2 macrophage polarization. This signature is completed by a metabolic reprogramming indicative of increased cell growth and proliferation. The supply of glucose modifies the gene expression signature of the isolated working heart by further increasing the expression of M2-related transcription factors and coactivators of transcription (gene symbols in boldface), thereby promoting the expression of metabolic genes associated with M2 macrophage polarization and wound healing. The presence of insulin resistance completely inhibits this enhancing effect of glucose on M2-associated markers. *Genes whose upregulation was confirmed by real-time PCR quantification.

In conclusion, the isolated working rat heart is a fitting model to study, at the transcriptional level, the early stress response associated with cardiac surgery with CPB. The balance between proinflammatory and anti-inflammatory molecules plays a key role in determining the clinical course of patients following cardiac surgery (39). Our results indicate that glucose-enhanced metabolic reprogramming may locally accelerate collagen formation, tissue repair, and wound healing in the stressed heart (Fig. 8). Altogether, these results reveal a potential mechanism by which the insulin resistance that develops with stress hyperglycemia is associated with increased morbidity and mortality following cardiac surgery with CPB.

GRANTS

This work was supported by National Institutes of Health Grants R01 HL-075360 (to M. L. Lindsey), R01 HL-061483 (to H. Taegtmeyer), K99/R00 HL-112952 (to R. Harmancey), P01 HL-051971, and P20 GM-104357. The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

A.S.B. and R.H. analyzed data; A.S.B., M.L.L., H.G.V., H.T., and R.H. approved final version of manuscript; M.L.L., H.G.V., and H.T. edited and revised manuscript; H.G.V. and R.H. performed experiments; R.H. interpreted results of experiments; R.H. prepared figures; R.H. drafted manuscript.