Acute kidney injury (AKI) is being increasingly shown to be a risk factor for chronic kidney disease (CKD), but little is known about the possible mechanistic links. We hypothesized that analysis of the genomic signature in the repair stage after AKI would reveal pathways that could link AKI and CKD. Unilateral renal pedicle clamping for 45 min was performed in male C57BL/6J mice. Mice were euthanized at 3, 10, and 28 days after ischemia-reperfusion injury (IRI). Total RNA was isolated from kidney and analyzed using an Illumina mouse array. Among 24,600 tested genes, 242, 146, and 46 genes were upregulated at days 3, 10, and 28 after IRI, and 85, 35, and 0 genes were downregulated, respectively. Gene ontology analysis showed that gene expression changes were primarily related to immune and inflammatory pathways both early and late after AKI. The most highly upregulated genes late after AKI were hepatitis A virus cellular receptor 1 (Havcr1) and lipocalin 2 (Lcn2), which code for kidney injury molecule-1 (KIM-1) and neutrophil gelatinase-associated lipocalin (NGAL), respectively. This was unexpected since they are both primarily potential biomarkers of the early stage of AKI. Furthermore, increases observed in gene expression in amiloride binding protein 1, vascular cell adhesion molecule-1, and endothelin 1 could explain the salt-sensitive hypertension that can follow AKI. These data suggested that 1) persistent inflammation and immune responses late after AKI could contribute to the pathogenesis of CKD, 2) late upregulation of KIM-1 and NGAL could be a useful marker for sustained renal injury after AKI, and 3) hypertension-related gene changes could underlie mechanisms for persistent renal and vascular injury after AKI.
- ischemia-reperfusion injury
- gene expression profiling
- chronic kidney disease
ischemia-reperfusion injury (IRI) is a major cause of acute kidney injury (AKI) (36). and despite the advances in renal replacement therapy, the mortality rate still remains high (1). AKI has largely been considered reversible (6). However, recent studies have demonstrated persistent and permanent structural and functional changes in the postischemic kidney, and some reports suggested that postinflammatory renal scarring caused by IRI might be an important contributor to the development of chronic kidney disease (CKD) (13, 41). Moreover, clinical evidence has shown that initial delayed graft function from prolonged ischemic time is associated with impairment in long-term graft function (32). A significant body of experimental work has focused on the early events that occur during IRI or in the early recovery phase, and little is known about the long-term effects after IRI and mechanisms that link AKI to CKD. As in other etiologies of CKD, there is a need for effective biomarkers for detecting the progressive renal injury following IRI.
We hypothesized that genomic analysis of kidneys during the later stages of repair after IRI would help us to understand the pathophysiology of the repair stage and long-term consequences in the kidney following IRI. This, in turn, could provide us with surrogate markers for detecting sustained kidney damage and novel treatment targets for modulating persistent injury. We used an experimental model where moderate, persistent renal structural changes are seen after IRI and analyzed changes in gene expression in an established murine IRI model at 3, 10, and 28 days. Significant changes in transcriptional activity were revealed in ischemic kidneys compared with nonischemic kidneys. Both up- and downregulated genes were related to immune and inflammatory responses at all time points, implicating persistent inflammation as a potential mechanism for progressive renal injury after IRI. Unexpectedly, hepatitis A virus receptor 1 [Havcr1; T cell immunoglobulin mucin protein 1 (TIM-1)-producing gene in mice; KIM-1, human homolog] and lipocalin 2 [Lcn2; neutrophil gelatinase-associated lipocalin (NGAL)], both well-known and promising early biomarkers for AKI, were the top dysregulated genes late after AKI. Increased expression of genes [e.g., vascular cell adhesion molecule 1 (VCAM-1), amiloride binding protein 1 (Abp1), and endothelin 1 (Edn1)] related to salt-sensitive hypertension, which can occur late after experimental IRI in rats (35), was also found 28 days after IRI in this model. Increased protein expression of candidate genes was validated, and sustained changes in genes expression were accompanied by progressive renal fibrosis after IRI. These findings suggested the possible role of candidate genes as a surrogate marker for detecting progressive renal injury following IRI.
MATERIALS AND METHODS
Male C57BL/6J mice (7–9 wk old) were purchased from Jackson Laboratories (Bar Harbor, ME). For inducing ischemic injury, a model of moderate renal IRI in mice was used. Briefly, 22 to 26-g mice were anesthetized with intraperitoneal pentobarbital sodium (75 mg/kg), abdominal incisions were made, and then the left renal pedicle was bluntly dissected. A microvascular clamp was placed on the left renal pedicle for 45 min while the animal was kept at a constant temperature (38–40°C) and well hydrated. After 45 min, the clamps were removed, wounds were sutured, and the animals were allowed to recover. Preliminary data demonstrated minimal histological changes in kidney 28 days after 30 min of ischemia, with very severe changes after 60 min (7). Therefore, 45 min of ischemia was chosen for moderate long-term changes. The same number of animals (n = 3) was maintained for 3, 10, and 28 days, respectively, before sacrifice. All animal studies were approved by Institutional Animal Care and Use Committee and in accordance with National Institutes of Health guidelines.
Tissue histological analysis.
Kidneys were harvested after exsanguination. Tissue sections were fixed with 10% formalin followed by paraffin embedding and stained with Masson's trichrome. Renal fibrosis was scored by a renal pathologist who was blinded to the experimental groups.
Purification and preparation of RNA.
Total RNA was extracted using the TRIzol reagent method (Invitrogen, Carlsbad, CA). Additional purification was performed on RNeasy columns (Qiagen, Valencia, CA). The quality of total RNA samples was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). RNA samples were labeled according to the chip manufacturer's recommended protocols. In brief, for Illumina, 0.5 μg of total RNA from each sample was labeled by using the Illumina TotalPrep RNA Amplification Kit (Ambion, Austin, TX) in a two-step process of cDNA synthesis followed by in vitro RNA transcription. Single-stranded RNA (cRNA) was generated and labeled by incorporating biotin-16-UTP (Roche Diagnostics, Mannheim, Germany). Biotin-labeled cRNA (0.75 μg) was hybridized (16 h) to Illumina Sentrix Mouse Ref8_v1.1 BeadChips (24,611 transcripts, Illumina, San Diego, CA). The hybridized biotinylated cRNA was detected with streptavidin-Cy3 and quantitated using Illumina's BeadStation 500GX Genetic Analysis Systems scanner.
Identification of significant transcriptional changes.
We used our well-tested filtering approach (17) updadated (18–19) and modified (12) for the Illumina platform. Briefly, the significance of hybridization signals was tested, and “Present” [BeadStudio detection (BSD) < 0.05] and “Absent” (BSD ≥ 0.05) transcripts were identified. The chip background and brightness were computed using a high quartile and a whole set of absent hybridization signals, respectively. The expression data were stratified by experimental conditions, and hybridization of each transcript was evaluated for each cluster. The transcript that was present and produced a signal at least twice as high as that of background in at least 66% hybridizations was considered expressed. To normalize expression values between different BeadChips, each background-adjusted value was divided by a corresponding chip brightness coefficient. Significance Analysis of Microarrays (SAM 2.20) (38) was conducted using 1,000 permutations of 3 controls and ischemic kidney samples without application of arbitrary restrictions (28). Fold-change (FC) cutoffs were calculated based on the power prediction analysis for minimum array requirements (21). An identified high quartile of SDs of log2 expression values of all probe sets (σ = 0.390) was submitted to the microarray sample size-identifying formula (42) using n = 3, imposing 95% of detecting power (1 − β). The significance level (α) was individually calculated for each dataset and represented 5% of the highest false discovery rate (FDR) for a given data set. The calculated cutoff values were FC = 3.93, FDR = 2.1; FC = 3.86, FDR = 2.2; and FC = 3.56, FDR = 2.8, for days 3, 10, and 28, respectively. The raw data and processed data were submitted to the NCBI Gene Expression Omnibus (GEO) database (NCBI tracking system no. 15723305).
Gene ontology analysis.
The Unigene IDs for each Illumina probe were retrieved from the Illumina Oligator MEEBO Mouse database generated by The Gene Index Project of Computational Biology and Functional Genomics laboratory at the Harvard School of Public Health (http://compbio.dfci.harvard.edu). Oligator and Illumina probe IDs were matched using Microsoft Access, and corresponding Unigene IDs were assigned to probes that comprise Mouse_Ref-8_V1 microarray. The MAPPFinder-compatible files were prepared using the GenMAPP-converting tool as we described previously (17). This approach identified 20,972 Unigene-associated probes of which 10,592 were detectable in mouse kidney tissues and 7,159 linked to gene ontology (GO) terms. Significant bioprocesses were selected by choosing GO terms containing >5% significantly changed genes of the total number of genes linked by MAPPFinder. A Z-score >2 and first GO node >0 were also applied as filtering criteria. Pathway analysis that identifies the biological functions that were significantly associated with identified candidate genes was conducted using the Ingenuity Pathways Knowledge Base tool (http://www.ingenuity.com). Fisher's exact test was used to calculate a P value determining the probability that each biological function is assigned to our candidate genes. Transcripts were matched against “inflammation,” “immune response,” “chronic kidney disease,” “fibrosis,” and “proteinuria” terms in PubMed using PubMatrix, the automated literature search engine (http://pubmatrix.grc.nia.nih.gov) (4). Hierarchical clustering analysis of candidate genes identified by all three approaches was conducted using the MultiExperiment Viewer (MeV, http://www.tm4.org/mev.html) of the microarray software suite TM4 system for microarray data management and analysis. The clustering was performed with application of the uncentered Pearson correlation and average linkage algorithm.
Validation of gene expression data.
Reverse transcription was performed using total RNA isolated from mouse tissues and processed (Applied Biosystems, Foster City, CA). A High-Capacity cDNA Archive kit first-strand synthesis system was used for quantitative real-time (QRT)-PCR according to the manufacturer's protocol. QRT-PCR was performed using the TaqMan assay system from Applied Biosystems. All PCR amplifications were carried out in duplicate on an ABI Prism 7300 Sequence Detection System, using a fluorogenic 5′-nuclease assay (TaqMan probes). Probes and primers were designed and synthesized by Applied Biosystems: mu_GAPDH (Mm99999915_g1); mu_Actb (Mm00607939_s1); mu_pgk1 (Mm00435617_m1); mu_Abp1 (Mm00504051_m1); mu_Edn1 (Mm00438656_m1); mu_Havcr1 (Mm00506686_m1); mu_SOCS3 (Mm00545913_s1); mu_VCAM-1 (Mm01320970_m1). Relative gene expressions were calculated by using the 2−ΔΔCt method, in which Ct indicates cycle threshold, the fractional cycle number where the fluorescent signal reaches detection threshold (29). The normalized ΔCt value of each sample was calculated using a total of three endogenous control genes (GAPDH, Actb, and pgk1). Fold-change values were presented as average fold-change = 2−(average ΔΔCt) for genes in treated relative to control samples. Error bars represented the SE for multiple biological replicates.
Renal tissue fixed with 10% buffered formalin was sliced in 4-μm sections and placed on microscope slides. For antigen retrieval, specimen slides were heated by microwave in a pressure cooker with a preheated 10-mmol sodium citrate solution (pH 6.0) for 5–10 min. After cooling, the slides were immersed in 3.0% H2O2 for 20 min to block endogeneous peroxidase and incubated with an avidin-, biotin-, and serum-free protein block (DAKO, Carpinteria, CA) at room temperature for 10–15 min, respectively, to prevent nonspecific detection. Then, the slides were incubated overnight at 4°C with rabbit polyclonal anti-mouse suppressor of cytokine signaling 3 (SOCS3) antibody (1:100, Abcam, Cambridge, MA) and goat monoclonal anti-mouse VCAM-1 antibody (1:200, R&D Systems, Minneapolis, MN). After rinsing in PBS, slides were incubated in secondary antibody for 1 h. For colorization, an avidin-biotin-horseradish peroxidase complex (ABC peroxidase; Vector Laboratories, Burlingame, CA) and 3,3′-diaminobenzidine tetrahydrochloride (DAB; Vector Laboratories) were applied to the slides at room temperature, and then the slides were counterstained with hematoxylin (Sigma, St. Louis, MO). Negative controls were stained under identical conditions, with rabbit or goat serum substituting for the primary antibody.
Protein was extracted from tissue, and concentration was determined using a Bradford solution (Bio-Rad, Hercules, CA). One hundred micrograms of protein was electrophoresed on a SDS-PAGE precast gel (Bio-Rad) under denaturing conditions. The protein was transferred onto a polyvinylidene difluoride membranes (Millipore, Bedford, MA) over 40 min at 400 mA with semidry methods. After the membrane were incubated with 5% skim milk for 1 h at room temperature, the membrane was hybridized with rat monoclonal anti-mouse TIM-1 antibody (1:100, R&D Systems), rat monoclonal anti-mouse lipocalin-2/NGAL antibody (1:250, R&D Systems), goat polyclonal anti-mouse amiloride binding protein antibody (1:100, Santa Cruz Biotechnology, Santa Cruz, CA), and rabbit polyclonal anti-mouse endothelin-1 antibody (1:200, Abbiotec, San Diego, CA) overnight at 4°C. The membranes were reacted with horseradish peroxidase-conjugated secondary antibody (1:1,000, Santa Cruz Biotechnology) for 90 min at room temperature. The detection of specific signals was performed using ECL methods (Thermoscientific, Waltham, MA), and signal density was measured by ImageJ (http://rsbweb.nih.gov/ij) was compared between groups. Equal amounts loading of protein was confirmed by GAPDH (1:1,000, Abcam).
Group means were compared using ANOVA followed by Newman-Keuls post hoc analysis in GraphPad Prism version 4 (GraphPad Software, La Jolla, CA). Statistical significance was determined when the P value was <0.05.
Progressive renal fibrosis in the repair phase after kidney IRI.
Fibrosis in the repair phase in the kidney examined with Masson's trichrome staining was increased even after a single episode of IRI (Fig. 1A). Histological scoring revealed that tubular fibrosis in the ischemic kidney was significantly increased compared with the contralateral kidney at days 10 and 28 (Fig. 1B).
Genomic responses in the repair phase after kidney IRI.
SAM identified 327, 181, and 46 transcripts which were changed by IRI and showed significant transcriptional activities at days 3, 10, and 28 after IRI, respectively. Among the lists, 242, 146, and 46 genes were upregulated at each time point, and 85, 35, and 0 genes were downregulated at days 3, 10, and 28, respectively (Fig. 2). The supervised hierarchical clustering of selected candidate genes separated control and ischemic kidneys clearly (Fig. 3), which supported the existence of specific transcriptional activities in the kidney after IRI. Up- and downregulated genes of ischemic groups showed similar expression between time points.
Specific gene expression changes in the repair phase in the kidney following IRI.
To select the candidate genes representing characteristics of the repair and remodeling phase of kidney following IRI, cross-referencing was done between data from days 3, 10, and 28 gene lists which were significantly affected by IRI (Fig. 2). Thirty-one genes showed persistent transcriptional activities at all time points (Table 1). Hepatitis A virus receptor 1 and lipocalin 2 were the most upregulated genes during the whole duration of the experiment (fold-changes: 63.0, 39.1, 8.4, and 50.7, 26.7, 11.6, respectively at days 3, 10, and 28, P < 0.01), and both were known to code biomarkers for AKI such as KIM-1 and NGAL, respectively. SOCS3, which was revealed as one of the ischemic transcripts in the kidney, showed high and constant activity in all measurements at three different time points after IRI. VCAM-1, Abp1, and Edn1 were among the top differentially expressed genes in the repair phase in the kidney after IRI, which are known to be implicated in endothelial dysfunction.
Identification of specific biological processes involved with the repair phase in the kidney after IRI.
To identify relevant biological processes potentially involved in persistent renal injury following IRI, we performed GO analysis using MAPPFinder. GO analysis identified predominant activation of immune and inflammatory responses at all time points of the repair phase in the kidney after IRI (Table 2). Ingenuity pathway analysis (IPA) about important functional pathways associated with genes showing significant changes also demonstrated that cellular responses involved with immune and inflammation such as cellular movement and immune cell trafficking were activated at all time points (Fig. 4). Apoptosis, represented as cancer, was common to all time points as well. To validate the potential biological relevance of candidate genes involved in these bioprocesses, a qualitative analysis was conducted using PubMatrix, the automated literature search engine. Most of the candidate genes have been implicated in inflammation and immune response and also displayed the association with CKD, fibrosis, and proteinuria according to the current PubMed database (Table 1).
Identification of network based on interaction between candidate molecules.
The IPA was then used to build functional networks between the candidate molecules. When we conducted pathway network analysis based on 6 candidate molecules, it revealed significant functional networks related to cell cycle, connective tissue development and function, and cellular movement, which scored 17. The network score means the probability that a network would be assembled by chance, which is statistically significant (P < 0.001), with a level of >3. Their interaction was mediated with two molecules, phosphatidylinositol 3-kinase (PI3-kinase) and epidermal growth factor (Fig. 5).
Validation of gene expression profiling.
To validate the results of our candidate gene expression, QRT-PCR measurements of selected individual genes were conducted using the TaqMan assay. There was overall agreement between gene expression by microarray analysis and by the TaqMan assay, and the fold-changes for the selected genes (Havcr1, Lcn2, SOCS3, VCAM-1, Abp1, and Edn1) are demonstrated in Fig. 6.
Validation of protein products from selected genes in the repair phase in the kidney after IRI.
Further validation for the manifestation of these genes in kidney tissue was pursued by immunohistochemical analysis and Western blotting. The distribution of SOCS3 proteins was evaluated by immunohistochemistry (Fig. 7). Contralateral kidneys during all time points expressed very low levels of SOCS3. After IRI, SOCS3 immunoreactivity was manifested in tubular epithelium (long arrow), and interstitial infiltrated inflammatory cells (short arrow, Fig. 7, B, D, and F). VCAM-1 also showed little immunoreactivity in the contralateral kidney, except a positive reaction in endothelium and parietal epithelium (arrowheads, Fig. 8C). IRI induced VCAM-1 expression mainly on damaged tubular epithelium and peritubular regions of the interstitium (Fig. 8, B, D, and F).
For verification of TIM-1/KIM-1 (Havcr1), NGAL (Lcn2), Abp1, and Edn1 expression in tissue, Western blotting was performed. Expression of NGAL and Edn1 was significantly higher in ischemic kidneys compared with contralateral kidneys at all three time points. An increase in TIM-1/KIM-1 expression was significantly different between ischemic and control kidneys at days 10 and 28 after IRI. Abp1 was expressed more in ischemic kidneys at day 10 after IRI (Fig. 9).
These data demonstrated that progressive renal fibrosis and concomitant robust transcriptional changes following kidney IRI were sustained during the late repair phase, and the regulated genes are largely involved with inflammation and immune responses, which can also participate in progressive renal injury. KIM-1 (Havcr1) and NGAL (Lcn2), both well-known promising early biomarkers of AKI, showed the highest activity during the whole experimental period, suggesting that those genes may also be useful as surrogate markers for progressive renal injury following IRI, and maybe even candidate mediators of the AKI-to-CKD transition. The increase in gene expression products associated with hypertension and endothelial dysfunction (VCAM-1, Abp1, and Edn1) can give more insight into persistent injury and the salt-sensitive hypertension found after AKI.
Acute kidney injury from IRI is common in critically ill patients related to high mortality and morbidity (1). It is well established that patients in intensive care units who require renal replacement therapy have an in-hospital mortality rate exceeding 50% (37). However, long-term outcomes of survivors of AKI remain poorly described. It was demonstrated that significant numbers of survivors from AKI remained dialysis dependent at long-term follow-up (1–10 yr), and 19–31% of the population had CKD (15). A recent study reported that an AKI episode requiring dialysis increased the risk of maintenance of dialysis treatment more than three times (41). According to the United States Renal Data System, acute tubular necrosis without recovery increased from 1.2% in 1994–1998 to 1.7% in 1999–2003 (5). The incidence could continue to rise with an aging population. Progressive renal fibrosis even after a single episode of IRI in our experiment demonstrated that IRI could be one of the reasons for the transition from AKI to CKD. There are no known markers detecting progressive renal injury after IRI, and little has been known about the mechanisms.
Genomic analysis combined with bioinformatics facilitates discovery of mechanisms of complex biological processes and has been a useful tool for identifying biomarkers for diagnosis of various renal diseases (9, 22, 30). When we examined expression profiles of ∼24,000 kidney genes during the repair phase after IRI, significant transcriptional activities were found as in a previous study (2), and they showed similar characteristics between time points. The candidate genes were largely related to immune and inflammatory responses, and some of them revealed an association with CKD, fibrosis, and proteinuria according to the current PubMed database. The most highly upregulated genes were KIM-1 (Havcr1) and NGAL (Lcn2). Although they are well-known promising early biomarkers for AKI, sustained expression of those genes even in the late repair phase following IRI was not revealed before. Our findings suggest their role as surrogate markers detecting progressive renal injury late after IRI. However, expression of these genes, although useful from the marker perspective, may not automatically correlate with pathogenic activity. TIM-1/KIM-1, coded from Havcr1, increases dramatically in the postischemic kidney (24) and plays a role in CD4 (+) T cell expansion. TIM-1 is also involved in the conversion of the regulatory T cell phenotype (Tregs) (10), and blockade of TIM-1 prolonged allograft survival (39). Based on the fact that effector T cell recruitment and depletion of Tregs worsen tissue injury in the repair phase in the kidney after IRI (8, 14), increased TIM-1 expression may contribute to continued inflammation after IRI. However, other investigators suggested the possibility that TIM-1 mitigated inflammation, promoting tissue repair. TIM-1 in macrophages recognized apoptotic bodies, which facilitated clearance of apoptotic cells (26), and TIM-1 expression in tubular cells after IRI led to transformation of tubular epithelium into the phagocytic phenotype, which also might serve to clear debris (23). NGAL, produced from Lcn2, may be induced from interactions between inflammatory cells or cytokines and the epithelial lining after IRI (20). NGAL administration conferred to attenuate ischemic AKI (31), and blockade of NGAL resulted in reduced renal regeneration (40), which reflect the beneficial role of NGAL in AKI. However, NGAL was demonstrated to induce inflammation in adipocytes and macrophages (44), which suggests that NGAL might also trigger tissue damage following IRI. Constant activation of Havcr1 (TIM-1) and Lcn2 (NGAL) expression in both the early and late phases after IRI could be involved in the continuum of AKI to CKD (3, 25), and the exact role of both genes in the repair phase in the kidney following IRI should be investigated further.
An important finding regarding recovery after IRI is that the susceptibility to sodium-dependent hypertension developed during the repair phase following IRI (35). Salt-sensitive hypertension could contribute to AKI progression to CKD. Among the candidate genes, enhanced transcriptional activities of VCAM-1, Abp1, and Edn1 could be related to this phenomenon. VCAM-1, which is related to endothelial dysfunction and an influx of inflammatory cells, was investigated as one of the candidate genes involved in salt-dependent hypertension (16). Abp1 and Edn1 were implicated in ischemic AKI in mice (43) and could be associated with sodium-dependent hypertension based on the functions of those genes. The same group also showed that immune suppression blocked sodium-sensitive hypertension in the recovery phase of ischemic AKI, which indicated that infiltrating immune cells mediate or participate in the development of hypertension (33).
SOCS3 is a key physiological regulator of the immune system. Although it was known to induce negative feedback on the production of proinflammatory cytokines and costimulatory pathway for T cell activation, the role is controversial (27). Elevated SOCS3 expression was linked with severity of autoimmune (11) and allergic disease (34). Thus SOCS3 may regulate the inflammatory reaction in diverse ways, and SOCS3 overexpression in the late phase in the kidney following IRI could have a strong impact on the immune response.
When we conducted pathway network analysis including six candidate molecules, they were related to cell cycle, connective tissue development and function, and cellular movement. It implied that these molecules might be associated with controlling cell death such as necrosis or apoptosis, tissue fibrosis, and chemotaxis associated with cellular movement. Two molecules revealed in network analysis mediating interaction, PI3-kinase and epidermal growth factor, might be next candidate molecules in the repair phase of IRI, and they should be studied further in the future.
Despite the limitation of number of time points and the need for detailed mechanistic studies on the regulation of the genes in question, our data demonstrate novel and potentially important changes in the repair phase of kidney IRI associated with progressive renal fibrosis. Persistent immune and inflammatory responses during the repair phase might explain progressive renal injury after IRI and underlie progression to CKD in susceptible groups. Changes in endothelial dysfunction and sodium-sensitive hypertension could also participate in CKD development. Candidate genes such as Havcr1 (KIM-1), Lcn2 (NGAL), VCAM-1, Abp1, Edn1, and SOCS3 could be useful biomarkers for detecting persistent renal injury after IRI. The exact mechanistic role of each specific gene change for mediating progressive injury requires further study.
G. J. Ko was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science, and Technology (E00029).
No conflicts of interest are declared by the authors.
- Copyright © 2010 the American Physiological Society