Mitochondrial Genetics Regulate Breast Cancer Tumorigenicity and Metastatic Potential

Breast cancer affects 230,000 people and claims 40,000 lives annually in the United States (1). Age and ethnicity are both risk factors for breast cancer; interestingly, Caucasian women have the greatest risk of developing breast cancer, but African-American women have the highest risk of mortality, followed by Caucasian then Asian women (1). Although certain nuclear genetic differences have garnered much attention regarding disease susceptibility with the BRCA1 and 2 mutations being leading examples (accounting for ∼10%–15% of patients), breast cancer is a complex disease with no known single cause (2–4). Furthermore, most studies investigating breast cancer susceptibility neglect to account for metastatic potential (5), which is what contributes most significantly to cancer morbidity and mortality. Consequently, a better understanding of both tumorigenesis and metastasis will complement and improve prognosis and treatment for breast cancer.

Several studies have shown differential susceptibility to tumor formation and metastatic spread in certain inbred mouse strains (3, 5–9). Hunter and colleagues were first to show metastatic susceptibility loci in a genetic screen crossing mouse mammary tumor virus (MMTV)–polyoma middle T mice (FVB/N-Tg(MMTV-PyVT)634 Mul/J, hereafter PyMT) on the FVB/NJ genetic background with different mice strains resulting in F₁ progeny with different primary tumor latencies and metastatic potentials (8, 10, 11). Tumors from the PyMT model progress through stages that are comparable to human breast cancer, and biomarker expression in PyMT tumors is similarly consistent with those associated with poor outcome in human disease (12, 13). In work reported by Lifsted and colleagues, F₁ progeny generated from C57BL/6J or BALB/cJ mothers crossed with PyMT males were shown to be more resistant or susceptible, respectively, to advanced disease and metastases (8). Using these models, metastasis-modifier genes have been identified by comparative genomic analysis and were among the first to show inheritance of metastasis susceptibility (14, 15). However, the experimental design—male transgenic mice crossed with females of other strains—left open the possibility that maternal mitochondrial DNA (mtDNA) inheritance was contributing to the observed changes in metastatic behavior.

Recent studies have reported that mtDNA mutations can significantly affect growth and metastatic potential in the tumor cell. For example, Ishikawa and colleagues performed reciprocal transfers of mtDNA from a highly and poorly metastatic mouse tumor cell lines and showed that metastatic potential segregated in a manner consistent with mtDNA background and was associated with reactive oxygen species production and complex I activity levels (16). This has led to suggestions that somatic mutations within the tumor mtDNA have an effect on oncogenic transformation. However, these studies do not address the impact of “normal” nuclear–mitochondrial genetic interactions nor do they consider that cancer development and metastatic potential can be influenced by ordinary mtDNA haplotypes (nonpathogenic mutations).

Our laboratory developed and utilized the Mitochondrial–Nuclear eXchange (MNX) mouse model in conjunction with the PyMT mammary tumor model to directly investigate the role of the mtDNA on breast cancer development (17). Through the process of nuclear transfer, MNX mice receive the intended nuclear and mitochondrial genetic components without a need for multiple backcrosses or exposures to mutagens, like ethidium bromide. Results herein provide direct support to the hypothesis that constitutional mitochondrial genetic background imparts a significant effect on breast cancer tumorigenicity and subsequent metastatic severity.

All animal studies were carried out under the approval of the University of Alabama at Birmingham and University of Kansas Medical Center Institutional Animal Care and Use Committees. MNX mice were utilized in this study and generated as recently described (17). Briefly, pronuclei were removed from pronuclear embryos of FVB/NJ mice and isolated pronuclei implanted into enucleated C57BL/6J (metastasis suppressing; ref. 8) and BALB/cJ (metastasis promoting; ref. 8) embryos, generating reconstructed zygotes having FVB/NJ nuclei but cytoplasms (and therefore mitochondria and mtDNA) of either C57BL/6J or BALB/cJ mice (Fig. 1).

MNX mice were generated having a FVB/NJ nuclear genetic background with either a C57BL/6J mtDNA (FVBⁿ:C57^mt) or a BALB/cJ mtDNA (FVBⁿ:BALBc^mt). The FVB/NJ nuclear and mtDNA backgrounds were used as a control for these studies because this is the background of the PyMT model. All MNX mice were healthy and fertile with no overt physical or behavioral phenotypes. Wild-type FVB/NJ, C57BL/6J, and BALB/cJ animals were purchased from Jackson Laboratories.

The PyMT mouse spontaneous mammary tumor model was utilized for these studies. The PyMT oncogene is transcriptionally controlled by the MMTV promoter, is highly metastatic, and progresses through stages similar to human breast cancer (12). Hemizygous male PyMT mice were used as sires for breeding to control (FVB/NJ strain) and MNX mice (FVBⁿ:C57^mt and FVBⁿ:BALBc^mt) to generate hemizygous PyMT mice having FVB/NJ, C57BL/6J, or BALB/cJ mtDNA.

At weaning (∼3 weeks of age) mice were genotyped for presence of the PyMT transgene as previously described (12, 18). mtDNA haplotypes were also confirmed in all female transgene-positive F₁ progeny by RFLP (restriction fragment length polymorphism) analysis of PCR products screening for diagnostic mtDNA SNPs that differentiate between the FVB, C57BL/6, and BALB/c mtDNA, as previously described (17). Briefly, C57BL/6J and FVB/NJ mtDNA were distinguished by an adenine to cytosine mutation at base pair (bp) 9,461 that results in the loss of a BclI restriction site in the C57BL/6J mtDNA. BALB/cJ and FVB/NJ mtDNA were distinguished by a guanine to adenine mutation at bp 9,348 that results in the loss of an RflfI site in the BALB/cJ mtDNA.

PyMT-positive mice (n = 40) were palpated daily for first detection of tumor onset. Mice were aged an additional week to confirm tumor onset, and aged 40 days past initial primary tumor detection before euthanization and necropsy. Animal age and weight were recorded and tumors removed and weighed to determine total tumor burden. Lungs were resected for metastatic analysis and fixed overnight in Bouin's fixative to facilitate metastasis quantification (19). Areas of surface metastases were determined using a Zeiss dissection microscope (Stemi 2000-C) fitted with an ocular micrometer (#474066-9901-000).

All transgene-positive female F₁ progeny were sacrificed at 70 days (N: FVB/N = 5, FVBⁿ:C57^mt = 7, FVBⁿ:BALBc^mt = 7). Animals were weighed before sacrifice and tumors were removed and weighed to determine total tumor burden. The #3 and #4 mammary glands were digested in Hank's Balanced Salt Solution with Ca²⁺ and Mg²⁺ (GIBCO #24020-117) containing 0.1% collagenase and 0.01% pronase. Tissue in media was minced with scalpel blades, transferred to a 50 mL tube, and then incubated at 37°C on a shaker platform for 2 hours. Samples were then transferred to a 15-mL tube and centrifuged to separate epithelial cells from fatty tissue and red blood cells. After several spins, the remaining epithelial cell fraction was resuspended in growth media (GIBCO #11039-021; phenol red-free DMEM/F12) containing 10% FBS, 20 ng/mL EGF, 0.5 μg/mL hydrocortisone, 100 ng/mL cholera toxin, 10 μg/mL bovine insulin, and 1% Pen- Strep to be plated.

Seahorse Bioscience XF24 Calibrant pH 7.4 (1.0 mL; #100840-000; Seahorse Bioscience) was added to each well of a 24-well-plate (Seahorse Bioscience #100777-004) that was kept overnight at room temperature. Primary tumor epithelial cells were seeded into individual wells (6 × 10⁴ cells per well) on a XF24 cell culture plate prewarmed with unbuffered DMEM (250 μL, 37°C), leaving appropriate temperature correction wells blank (A1, B4, C3, and D6). Epithelial cells were incubated in the XF24 cell culture plate in a 37°C incubator without CO₂ until ready for use in order to allow them to equilibrate to the new media. Injection compound dilutions were made in appropriate running media (unbuffered DMEM) for the bioenergetic profile analysis experiment. Previously, titration experiments were run to determine each effector's appropriate working concentration. For these studies, optimal concentrations were found to be: 10 μg/mL oligomycin, 1 μmol/L FCCP (trifluorocarbonylcyanide phenylhydrazone), an ionophore that acts as a mitochondrial uncoupling agent, and 10 μmol/L antimycin A.

Oligomycin, FCCP, antimycin A, and media were loaded into injection ports A, B, C, and D, respectively, in 75 μL volumes for each experimental well. All wells, including temperature correction wells, had control media or compounds loaded into all ports to ensure uniform injection. Once prepared, the plate containing the effectors was placed atop the plate containing the primary cells and placed into the bioanalyzer. Each oxygen consumption rate (OCR) measurement consisted of 3 minutes of mixing, 2 minutes wait time, and 3 minutes of continuous O₂ level measurement. There were three measurement rounds per injection.

Data analysis was performed as described previously (20, 21). Briefly, baseline average, oligomycin minimum, FCCP maximum, antimycin A minimum, and reserve capacity were calculated as well as ATP-linked and non-ATP-linked baseline respiration and average baseline extracellular acidification (ECAR).

Aconitase activity was determined by measuring the transformation of isocitrate to cis-aconitate at 240 nm in 50 mmol/L TrisHCl (pH 7.4) containing MnCl2 and 20 mmol/L isocitrate at 25°C. Aconitase is specifically inactivated by superoxide (O₂^.−) and peroxynitrite (ONOO⁻) and hence, decreased activity correlates with increased oxidative stress associated O₂^.− and ONOO⁻. Because ONOO⁻ formation is related to superoxide production, decreased aconitase activity can be associated indirectly with increased O₂^.− generation.

Continuous parameters were compared by ANOVA for animal weight, tumor weight, primary tumor latency, and average metastasis area. In instances of nonnormal distributions (e.g., visible lung metastases), a Kruskal–Wallis one-way rank test was used. A P-value <0.05 was considered statistically significant.

mtDNA from C57BL/6J and BALB/cJ mouse strains were chosen to produce MNX mice because F₁ hybrid progeny derived from PyMT males and females of these strains showed differential susceptibility to metastasis severity compared with offspring from PyMT male-FVB female crosses (8). In that study, F₁ progeny having C57BL/6J mothers had less severe metastasis burden relative to PyMT mice from FVB/NJ mothers, although PyMT offspring with BALB/cJ females had increased metastatic burden, as determined by surface area of tumor cells in lung histologic sections (8). For our studies, FVBⁿ:C57^mt and FVBⁿ:BALBc^mt MNX and FVB/NJ control female mice were bred to PyMT males to directly test the effects of mtDNA background upon tumorigenicity and metastatic severity.

Figure 2 reveals that no significant differences in animal weight (FVBwt = 30.08 ± 0.66 g, FVBⁿ:C57^mt = 31.82 ± 0.78 g, FVBⁿ:BALBc^mt = 32.89 ± 1.62 g) or total tumor burden (FVBwt = 6.72 ± 0.46 g, FVBⁿ:C57^mt = 8.18 ± 0.53 g, FVBⁿ:BALBc^mt = 6.90 ± 0.82 g) were detected in PyMT-positive F₁ females 40 days subsequent to initial primary tumor detection. These data indicate that mice with all three mitochondrial genetic backgrounds had similar body weights and tumor growth rates (once tumors developed). In contrast, mitochondrial genetic background significantly affected primary tumor latency. PyMT mice with FVB/NJ, C57BL/6J, and BALB/cJ mtDNA had average latencies of 57.44 ± 0.83, 65.05 ± 1.72, and 53.40 ± 0.97 days, respectively (Fig. 3). These latencies were significantly (P < 0.05) different, suggesting that mtDNA background influenced the early growth properties of PyMT mammary tumors. Interestingly, these results closely mirrored values for mean latencies observed in the matching F₁ progeny reported by Lifsted and colleagues (FVB/NJ = 57.92 days, C57BL/6J = 64.44 days, and BALB/cJ = 56.44 days; ref. 8).

To determine the role of the mtDNA on metastatic severity, surface pulmonary metastases were counted and measured (19). Due to the nonnormal distribution for the number of metastases per lung, the median number of metastases was utilized for statistical comparison. Whereas there were no differences in the median number of metastases between groups of PyMT mice (Table 1; FVB = 11, range: 0–83; FVBⁿ:C57^mt = 12, range: 0–62, FVBⁿ:BALBc^mt = 7, range: 1–58), a statistically significant (P < 0.05) difference in the average size of individual metastases (in cm²) between groups was observed (FVB = 0.128 ± 0.012, FVBⁿ:C57^mt = 0.099 ± 0.009, and FVBⁿ:BALBc^mt = 0.181 ± 0.031; Table 1). These data suggest that, in addition to having an impact on primary tumor latency, mitochondrial genetic background influences the severity of metastatic disease. Given that the number of metastases was unchanged, the differences in average metastasis size confer an overall difference in metastatic burden. This is consistent with previous reports that PyMT F₁ offspring from C57BL/6J mothers had lower metastatic burden; whereas, those from BALB/cJ mothers had more lung metastatic volume compared with PyMT offspring from FVB dams. The findings herein show that presence of the mitochondrial genome alone (i.e., no alteration of the nuclear genome) is sufficient to reproduce this finding.

To assess whether bioenergetic profiles could be contributing to tumor behavior, epithelial cells were isolated from primary mammary tumors and seeded onto a Seahorse XF24 plate. No differences in baseline OCR were observed between FVB/N and FVBⁿ:BALBc^mt cells, but FVBⁿ:C57^mt had a significantly elevated basal OCR (FVB/N = 1.50 ± 0.20, FVBⁿ:C57^mt = 4.45 ± 0.48, FVBⁿ:BALBc^mt = 1.79 ± 0.32 pmol/min/mg). This trend was consistent for OCR measurements following addition of oligomycin, FCCP, and antimycin A (Fig. 4). Reserve capacity is a measure of the difference between the maximum possible OCR (following FCCP addition) and the baseline OCR. This parameter is proposed to be a measure of the ability of cells or tissue to respond to stresses by increasing their metabolic output. Due to the FVBⁿ:C57^mt tumor cells operating basally at a level near their maximum, their reserve capacity was miniscule (0.06 pmol/min/mg). FVB/N cells had an intermediate reserve at 0.18 pmol/min/mg, whereas FVBⁿ:BALBc^mt tumor epithelial cells had a reserve capacity of 0.57 pmol/min/mg, roughly 25% of their maximum OCR.

Baseline oxygen consumption levels are made up of oxygen consumption that is either linked to ATP or not linked to ATP. Along with having different total baseline OCRs, the percent contributions from ATP-linked and non-ATP-linked were different for each group. The baseline OCR for FVB was 44.4% ATP-linked and 55.6% non-ATP-linked, FVBⁿ:C57^mt was 56.4% ATP-linked and 43.6% non-ATP-linked, and FVBⁿ:BALBc^mt was 50.8% ATP-linked and 49.2% non-ATP-linked (Fig. 5A). As a percentage of the maximum potential OCR, these values diverged further. ATP-linked OCR for FVB/N cells fell to 39.6% and non-ATP-linked OCR fell to 49.6% with reserve capacity accounting for 10.8% of the maximum OCR. For FVBⁿ:C57^mt tumor cells, ATP-linked and non-ATP-linked remained nearly the same (55.6% and 43.1%, respectively), as reserve capacity only accounted for 1.3% of the maximum attainable OCR. With the high reserve capacity observed in FVBⁿ:BALBc^mt cells (24.1%), ATP-linked OCR accounted for 38.6% and non-ATP-linked OCR for 37.3% of maximum oxygen consumption. Consistent with the heightened oxygen consumption levels observed in FVBⁿ:C57^mt tumor cells, baseline ECAR was highest in these cells as well (FVB/N = 0.14 ± 0.02, FVBⁿ:C57^mt = 0.27 ± 0.06, and FVBⁿ:BALBc^mt = 0.17 ± 0.02 mpH/min/mg; Fig. 5A). Calculating OCR/ECAR determines the ratio of oxidative phosphorylation relative to glycolysis, and this value was roughly equal for FVB/N and FVBⁿ:BALBc^mt cells (10.48 and 10.51, respectively), although it was significantly higher for FVBⁿ:C57^mt cells (16, 21; Fig. 5B and C). Taken together, these bioenergetic data indicate that primary tumor cells isolated from these animals are metabolizing fundamentally differently.

To assess oxidative stress, the mammary epithelial cells, aconitase activity was assessed 40 days after detection of the first mammary tumors or in animals at 70 days of age. No significant differences between tumor cell lines were detected within either group (Fig. 5D). There was a significant difference comparing the FVBⁿ:C57^mt tumor cells at the different ages. Cells at the 70-day hard cut-off generally had more oxidative stress than cells isolated 40 days after detection. However, the results must be viewed with a consideration that the 70-day hard cutoff animals were always younger than the 40-day postdetection mice. Considering that the FVBⁿ:C57^mt 40-day postdetection were the oldest of all (because they took the longest to get primary tumors), this may simply be an age-related factor. Similar measurements of citrate synthase similarly yielded no detectable differences between cells from the different crosses (unpublished data).

Two of the major challenges facing breast cancer researchers are to expand our knowledge of what contributes to an individual's risk for developing the disease, and to more fully understand influences on metastatic spread. Recent work investigating the roles of acquired mutations and nuclear genetic background on breast cancer susceptibility has provided important new insights; yet, the impact of the “host” mitochondrial genome has rarely been considered in such analyses. Because metabolic reprogramming occurs in cancer progression (21), it is vital to examine the genetic components responsible for providing a significant amount of cellular metabolic instructions.

This is the first study to show that constitutional mtDNA polymorphisms directly affect breast cancer tumorigenicity and metastatic spread in the PyMT model. Our studies show that in animals harboring the same nuclear DNA and subjected to the same oncogenic event (PyMT), primary tumor latency segregates with mitochondrial genetic background. Compared with progeny from control FVB females and PyMT males, offspring of FVBⁿ:C57^mt mothers had significantly delayed primary tumor latency, whereas PyMT mice harboring the BALBcJ mtDNA had an accelerated onset. The finding that there were contrasting primary tumor latencies between groups and no differences in total tumor burdens 40 days after onset further suggests that early stages of tumorigenesis are distinct between groups and are influenced by mtDNA background, whereupon their establishment, mtDNA background in the PyMT model does not appear to dictate primary tumor growth rate.

Metastatic efficiency of PyMT mice was modulated by mtDNA background as well. Our data show that the C57BL/6J mtDNA was associated with less severe metastatic disease while the BALB/cJ mtDNA was linked with more severe metastasis. It is possible that these differences are due to (i) the speed of primary tumor dissemination, (ii) growth rates at the secondary site leading to colonization, or (iii) both. It is important to emphasize that mtDNA differences were not acquired mutations in the tumor cells. Therefore, different mtDNA backgrounds influence tumor development, progression, and steps throughout the metastatic cascade, including at secondary sites. It is possible that the tumor cell's ability to persist in circulation and eventually grow at distant sites is reflected in its capacity to metabolize efficiently while in transit or after arresting at secondary sites. Additionally, it is possible that acquired mutations may superimpose on the somatic differences to further affect tumor progression; however, additional studies will be required to discern these options.

Mitochondria have been implicated in tumor growth rates previously (22–24), and it is evident that the mitochondrial backgrounds are modifying the growth characteristics of these primary mammary tumors, presumptively through direct or indirect modification of cellular metabolic processes or oxidative stress. To explore these possibilities, mammary tumor (epithelial) cells were isolated and OCR and ECAR were measured. Heightened glycolysis is observed in most tumors and can indicate increased aggressiveness. The high ECAR seen in FVBⁿ:C57^mt tumor cells suggests elevated glycolytic activity. That these cells also have a significantly larger OCR implies that they simply have increased metabolic activity without necessarily relying more heavily on aerobic glycolysis. The similarities between FVB/N and FVBⁿ:BALBc^mt tumor cells in baseline OCR and ECAR suggest that other factors may be contributing to observed differences between their tumorigenic phenotype.

One significant difference, in particular, is reserve capacity. Reserve capacity makes up 10.8% of the total possible OCR available to FVB/N tumor cells; yet, it accounts for 24.1% of maximal OCR for FVBⁿ:BALBc^mt mammary tumor cells. Surprisingly, reserve capacity for the FVBⁿ:C57^mt tumor cells only constitutes 1.3% of their maximum potential OCR. When the reserve capacity is this low suggests that the FVBⁿ:C57^mt tumor cells have a baseline OCR nearly equal to their maximum, which, counterintuitively, may indicate that they are operating inefficiently. These cells also display the highest percentage of baseline OCR that is linked to ATP production, which is an often an indicator of poor mitochondrial economy, i.e., mitochondrial damage can render mitochondria less efficient to the point that they require greater oxygen utilization to meet the same metabolic requirements. Considering that reserve capacity reflects a cells ability to persist in response to stress (25), it might also provide insight into the ability for disseminated cells to complete the metastatic cascade. Interestingly, reserve capacities mirrored metastatic severity data previously reported (21). Interestingly, oxidative stress, as measured by aconitase or citrate synthase activities, showed no differences when age was removed as a variable.

The metabolic changes observed could be, at least, partially explained by the specific polymorphisms in the mtDNA. FVB/NJ and C57BL/6J mtDNA are different at sites encoding subunit II of complex IV [nucleotide pair (np) 7,778, a Asp to Tyr missense mutation], as well as subunit III of complex I (np 9,461, a Met to Ile change); whereas, FVB/NJ and BALB/cJ mtDNA are different in sites encoding subunit II of complex IV (np 7,778, a Asp to Tyr missense mutation) and subunit III of complex IV (np 9,348, a Val to Ile missense mutation; ref. 26). Differential complex I and complex IV activities have been recently implicated as contributing to breast cancer progression (27, 28). However, a simple explanation is not likely because polymorphisms within one electron transport complex can potentially affect other complexes or related metabolic processes (e.g., citric acid cycle) because cellular metabolism is a network of tightly coupled processes, and modification of any step might affect the entire process.

Taken together, our data suggest that modulating mitochondrial genetic background has a significant impact on tumor latency, metastasis, and disease progression. The specific causes remain to be fully elucidated; but it is probable that they are a consequence of oncogenic mutations “co-opting” the preexisting metabolic programming of the cell influenced by mtDNA background. To evaluate the mechanisms behind these differences, it will be necessary to further dissect the bioenergetics of PyMT control and MNX animals. Additionally, retrograde signaling to the nucleus that is altered by mitochondrial polymorphisms could be responsible for the changes in cellular behavior observed.

The concept that mtDNA polymorphisms can modify breast cancer susceptibility could be of use for risk assessment and/or prognosis before any malignant transformation has taken place. Because these are not acquired mutations, a blood sample or cheek swab could be sufficient to ascertain an individual's risk for advanced disease as more information is found characterizing metastatic potential versus mtDNA haplotype. Petros and colleagues reported that germline mtDNA mutations contribute to prostate cancer development (29), but ours are the first functional data that directly establish a connection between mtDNA and metastasis. Both studies, however, support roles for mtDNA background as a contributor to outcomes and potential therapies. Furthermore, metabolic pathways in cancer offer a bright future for potential therapeutic targets (30). Delineating specific mitochondrial contributors to altered cancer growth kinetics will only expedite these discoveries.

S.W. Ballinger, D.R. Welch, L.W. Johnson, and R.A. Kesterson have ownership interest in a patent on MNX mice. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.P. Feeley, R.A. Kesterson, S.W. Ballinger, D.R. Welch

Development of methodology: K.P. Feeley, L.W. Johnson, R.A. Kesterson, S.W. Ballinger, D.R. Welch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.P. Feeley, A.W. Bray, D.G. Westbrook, R.A. Kesterson, S.W. Ballinger

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.P. Feeley, D.G. Westbrook, S.W. Ballinger

Writing, review, and/or revision of the manuscript: K.P. Feeley, R.A. Kesterson, S.W. Ballinger, D.R. Welch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.W. Ballinger, D.R. Welch

Study supervision: K.P. Feeley, S.W. Ballinger, D.R. Welch

D.R. Welch is a Kansas Bioscience Authority Eminent Scholar.

This study was financially supported by Susan G. Komen for the Cure (SAC11037 to D.R. Welch); U.S. Army Medical Research and Material Command (W81XWH-07-1-0540d to S.W. Ballinger and D.R. Welch); NCI (CA013148 to D.R. Welch and S.W. Ballinger; CA168524 to D.R. Welch; HL103859 to S.W. Ballinger; and CA134981 to D.R. Welch); National Foundation for Cancer Research (D.R. Welch). A.W. Bray was supported by the predoctoral training program in cardiovascular pathophysiology (T32 HL007918); UAB Transgenic Mouse Facility (R.A. Kesterson) is supported by awards NIH P30 CA13148, P30 AR048311, P30 DK074038, P30 DK05336, and P60 DK079626. The authors are also indebted for the support of the UAB Diabetes Research Center Bioanalytical Redox Biology Core (DK079626).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.