Mitochondrial polymerase gamma dysfunction and aging cause cardiac nuclear DNA methylation changes

Abstract

Cardiomyopathy (CM) is an intrinsic weakening of myocardium with contractile dysfunction and congestive heart failure (CHF). CHF has been postulated to result from decreased mitochondrial energy production and oxidative stress. Effects of decreased mitochondrial oxygen consumption also can accelerate with aging. We previously showed DNA methylation changes in human hearts with CM. This was associated with mitochondrial DNA depletion, being another molecular marker of CM. We examined the relationship between mitochondrial dysfunction and cardiac epigenetic DNA methylation changes in both young and old mice. We used genetically engineered C57Bl/6 mice transgenic for a cardiac-specific mutant of the mitochondrial polymerase-γ (termed Y955C). Y955C mice undergo left ventricular hypertrophy (LVH) at a young age (∼94 days old), and LVH decompensated to CHF at old age (∼255 days old). Results found 95 genes differentially expressed as a result of Y955C expression, while 4,452 genes were differentially expressed as a result of aging hearts. Moreover, cardiac DNA methylation patterns differed between Y955C (4,506 peaks with 68.5% hypomethylation) and aged hearts (73,286 peaks with 80.2% hypomethylated). Correlatively, of the 95 Y955C-dependent differentially expressed genes, 30 genes (31.6%) also displayed differential DNA methylation; in the 4,452 age-dependent differentially expressed genes, 342 genes (7.7%) displayed associated DNA methylation changes. Both Y955C and aging demonstrated significant enrichment of CACGTG-associated E-box motifs in differentially methylated regions. Cardiac mitochondrial polymerase dysfunction alters nuclear DNA methylation. Furthermore, aging causes a robust change in cardiac DNA methylation that is partially associated with mitochondrial polymerase dysfunction.

cardiomyopathy (CM) is an intrinsic weakness of the myocardium that leads to myocardial contractile dysfunction and congestive heart failure (CHF) (15, 29). Mechanistically, it has been suggested that CM results from failure of cardiac mitochondria to maintain sufficient energy production for cardiac contraction (1, 30). We and others have shown left ventricle (LV) mitochondrial (mtDNA) depletion and mitochondrial dysfunction in human CM, thus supporting this hypothesis of energy starvation (17, 22).

The mitochondrial theory of aging posits that mitochondrial dysfunction leads to the accumulation of free radicals that damage biochemical moieties (4, 5, 10). This mechanism for aging is consistent with mitochondrial dysfunction in CM, arguing for shared or related molecular events in the pathophysiology of aging and CM.

Epigenetics may be defined as the control of nuclear gene expression not mediated by DNA sequence. A number of epigenetic mechanisms exist, such as DNA methylation and histone modifications (6, 12, 32). We previously showed that DNA methylation changes occur in the LV of human CM, and mtDNA depletion in the LV was an associated marker of CM (19, 22). The link between DNA methylation and mitochondrial dysfunction pointed to changes in mtDNA nucleotide substrates being altered by epigenetic changes to nuclear DNA. Here, we examined the causality of mitochondrial dysfunction on nuclear DNA methylation in aging in genetically engineered mice.

C57Bl/6 transgenic mice expressing a cardiac-specific mutant of the mtDNA polymerase (pol)-γ (termed Y955C) undergo LV hypertrophy (LVH) (21, 25). Using this model, we explored the changes in DNA methylation from cardiac Y955C transgenic expression. We operationally defined “young” (∼94 days old) and “old” (∼255 days old) mice. Young Y955C mice undergo LVH that progresses to decompensation and CHF at old age. Gene expression microarrays demonstrated 95 differentially expressed genes in Y955C hearts compared with 4,452 genes associated with aging. Differential DNA methylation resulted from both Y955C (4,506 peaks with 68.5% hypomethylation) and aging (73,286 peaks with 80.2% hypomethylated). Of the Y955C-dependent differentially expressed genes, 30 genes (31.6%) also displayed differential DNA methylation. Among the age-dependent differentially expressed genes, 342 genes (7.7%) displayed DNA methylation changes. Cardiac mitochondrial polymerase dysfunction and aging each affect nuclear DNA methylation, with significant enrichment of CACGTG-associated E-box motifs.

METHODS

Reagents.

All reagents were analytical grade and purchased from Sigma Aldrich (St. Louis, MO) unless otherwise indicated.

Mouse care.

Congenic Y955C mice were generated by back crossing (6 generations to C57Bl/6 mice) our original lines created on the inbred FVB/n background and that displayed cardiac-specific expression of mutant mitochondrial pol-γ as previously described (21, 25). Transgenic founders were mated with wild-type (WT) C57Bl/6 mice (Jackson Labs, Bar Harbor, ME) to produce offspring on a C57Bl/6 background, and mice for all experiments were at least six generations from the founding lines. WT C57Bl/6 littermates were used as control mice in all experiments. For genotyping, genomic DNA was extracted from mouse tail clippings, and genotype was determined by PCR. All mice were housed at the Emory University Vivarium, an Association for Assessment and Accreditation of Laboratory Animal Care-certified vivarium, and in accordance with Institutional Animal Care and Use Committee protocols and National Institutes of Health (NIH) guidelines.

Experiments employed mice of ∼95 days old for the young group and mice of ∼250 days for the old group. In a “two-by-two” factorial design, WT and transgenic mice were analyzed for physiological, histopathological, and molecular changes.

Echocardiography and electrocardiogram.

Mice were anesthetized with Avertin (0.25 mg/g body wt) and weighed to determine body mass. Echocardiography was performed using VisualSonics Vevo 770 (VisualSonics, Toronto, Ontario, CA). Results from two-dimensional M-mode analysis along the short axis (at the level of the largest LV diameter) were used to determine LV mass, LV end diastolic dimension (LVEDD), wall thickness, ejection fraction, and LV fractional shortening. LV mass values were normalized to mouse body weight. Upon completion of echocardiography, mice were terminated by cervical dislocation under Avertin anesthesia and wet heart weight was measured.

Gene expression analysis.

Gene expression analysis was performed as previously described (19, 22). Briefly, total RNA was extracted from at least three individual mouse hearts from each 2×2 groups using the Fibrous Tissue RNeasy kit (Qiagen, Germantown, MD). Each individual mouse heart RNA was analyzed by itself and not pooled. Double-stranded cDNA was synthesized using the SuperScript Double-Stranded cDNA Synthesis Kit (Life Technologies, Grand Island, NY). cDNA was then labeled with Cy3 and hybridized overnight to a 12 × 135 kb mouse expression array (Roche Nimblegen). Expression arrays were washed and scanned on a Roche Nimblegen MS200 scanner. Images were analyzed using Nimblescan software as directed by the manufacturer. Expression results were analyzed by two-way ANOVA using Bioconductor for R. Differentially expressed genes were identified as those with a false discovery rate (FDR) < 0.2 and 1.5-fold change in gene expression compared with controls. Gene ontology was performed using DAVID bioinformatics database (13). Microarray array data (raw and processed) were deposited into NIH/National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) and can be accessed under GEO series GSE72890.

DNA methylation analysis.

DNA methylation analysis was performed as previously described (19, 22). DNA was extracted at least three individual mouse hearts, and each individual mouse heart DNA was analyzed by itself and not pooled. Total cellular DNA was sonicated to an average fragment size of 200–500 bp. A sample of DNA was set aside for later normalization (denoted “Input”), and then a portion of the sonicated DNA was enriched using the MethylCollector Ultra kit (Active Motif, Carlsbad, CA) following the manufacturer's directions. Both the methylated and input DNA were amplified using whole genome amplification (Sigma-Aldrich). Samples of the methylated and input DNA were validated for enrichment of methylated DNA using PCR. For DNA methylation analysis, Roche Nimblegen 2.1M Mouse Deluxe Promoter Arrays were utilized (Roche Nimblegen). DNA was labeled with Cy3 and Cy5 dyes to distinguish methylated and input DNA, and DNA was allowed to hybridize to arrays overnight. Arrays were scanned using a Roche Nimblegen MS200 scanner. Images were analyzed using Nimblescan software as directed by the manufacturer (which included normalizing to the input DNA), resulting in a final analysis including a P score of the detected methylated DNA peak. A two-way ANOVA was performed on only the differentially expressed gene promoters to identify DNA methylation changes that correlate with gene expression changes. DNA methylation results were analyzed using Bioconductor for R. Differentially methylated gene promoter regions were identified as those with an FDR < 0.2 and peak score change > 1. Gene ontology was performed using DAVID bioinformatics database. Microarray array data (raw and processed) were deposited into NIH/NCBI GEO and can be accessed under GEO series GSE72890.

Motif analysis.

Motif analysis was performed using modules available through Meme Suite (2). DNA regions exhibiting altered DNA methylation were analyzed for motif enrichment using both the MEME and MEME-Chip modules. Enriched motifs were compared with known motifs using the Tomtom module. This module compares each enriched motif to known motifs in multiple databases, and P values and FDRs are calculated for each motif based on seven statistical analyses built into the Tomtom module (7). An FDR < 0.1 was deemed significant for motif enrichment. Enrichment was further verified by calculating the number of times a significant motif appeared in each experimental group to the number of time the motif was observed in the CpG-containing probes on the array (which served as a pseudo-random set). Chi-square test (P < 0.05) with Bonferroni-adjusted P values were used to identify significant enrichment.

Statistical analysis.

Statistical analyses not performed in R were performed using GraphPad Prism 5.0 (Graphpad, La Jolla, CA). Each experiment was analyzed using a one-way or two-way ANOVA where appropriate, with a P < 0.05 deemed statistically significant. The experimental results are displayed as means ± SE, with all experiments performed at least three times each.

RESULTS

Pathophysiology.

We assessed the changes in cardiac function and structure with echocardiography. Young mice were 94 ± 1 days old at time of experiments, and old mice were 255 ± 45 days old (Fig. 1A). The survival of Y955C transgenic mice was similar between young and old mice (Fig. 1B). We observed decreased Y955C survival by day 95 (33%) compared with the number of live-born transgenic pups, which was similar to survival at day 255 (22%). We found increases in LV mass and LVEDD in young Y955C mice compared with WT controls (Fig. 1, C and D, respectively). No change in LV mass or LVEDD was observed in old Y955C mice compared with WT controls. LV wall thickness was significantly diminished in old Y995C mice compared with old WT controls (Fig. 1E). There was no change in ejection fraction in any experimental group (Fig. 1F). Results are consistent with LVH at a young age (LV mass increase, LVEDD increase, ejection fraction normal) and progression to CM in old mice (LV wall thickness decreases).

Fig. 1.

Cardiac physiology. Y955C and wild-type (WT) control hearts were analyzed for structural and functional changes. A: “young” mice were analyzed at ∼94 days, while “old” mice were analyzed and ∼255 days. B: Y955C mice exhibited a decreased survival during the first 90 days of life, with 33% of live-born Y955C positive mice surviving to 94 days for young mice experiments and 22% surviving to old mice experiments. Vertical dotted lines denote young and old ages. C: left ventricular (LV) mass was significantly increased in the young Y955C mice compared with WTs. Aged Y955C mice did not show a significant increase in LV mass, with age significantly depressing the hypertrophic response in Y955C. D: left ventricular end diastolic dimension (LVEDD) was significantly increased in young Y955C mice compared with WTs, while no change was observed in older mice. E: LV wall thickness was significantly increased in Y955C young mice compared with WT young controls, but LV wall thickness was significant decreased in Y955C old mice compared with WT old mice. LV wall thickness was also significantly decreased in Y955C old mice compared with Y955C young mice. F: ejection fraction was unchanged in all experimental groups. *P < 0.05 by 1-way ANOVA and Tukey post hoc. #P < 0.05 by 2-way ANOVA for interaction between Y955C and age.

Gene expression.

Gene expression microarrays were used to identify expression differences between Y955C and aging mice. Genes were defined as differentially expressed if they exhibited a 1.5-fold change in expression compared with controls and an FDR < 0.2 (and P < 0.1 as a secondary threshold) (Supplemental Table S1).¹ We identified 95 genes differentially expressed due to Y955C and 4,452 genes differentially expressed due to aging (Fig. 2A). For the Y955C genes, 46 were upregulated and 49 were downregulated (Fig. 2B); for the age-associated genes, 1,652 were upregulated and 2,800 were downregulated (Fig. 2C). Of these genes, 39 genes were shared in both groups, suggesting a common mechanism of action for these genes (Fig. 2A). Furthermore, 37 of the 39 shared genes demonstrated similar expression directionality between Y955C and aging. While these genes were present in both sets, they did not necessarily represent true interactions between Y995C and aging. Using a two-way ANOVA, we found no genes where Y995C and aging interacted to alter gene expression either synergistically or antagonistically. Among the 95 genes differentially expressed by Y955C, cardiotrophin (CTF1), glutathione S-transferase-α-1 (GSTA1), monoamine oxidase A (MAOA), and mitofusin 1 (MFN1) were all differentially expressed.

Fig. 2.

Differential gene expression. Y955C hearts were analyzed for gene expression changes mediated by both the transgene and age. Genes were deemed significant if they exhibited a 1.5-fold change compared with controls and a false discovery rate (FDR) < 0.2. A: microarrays identified 95 genes differentially expressed as a result of Y955C expression; 4,452 genes were differentially expressed as a result of age in cardiac samples; 39 genes were present in both experimental sets, of which 37 shared similar directionality of expression change. B: heat map of 95 genes differentially expressed by Y955C. C: heat map of 4,452 genes differentially expressed by age. All heat-map expression levels have been log₂ transformed. Y, young; O, old.

DNA methylation.

DNA promoter microarrays were utilized to identify changes in DNA methylation. Each array analyzed 2.1 M DNA regions of average length of 50 bases that were equally spaced along the promoters of genes starting 5 kb prior to transcriptional start site (TSS) and ending 1 kb after TSS. Global DNA methylation patterns were not evident using scatter of DNA peak scores in Y995C young, WT old, and Y955C old mice compared with WT young controls, with both hyper- and hypomethylation changes seen (Fig. 3, A, B, and C, respectively). Interestingly, a shift toward hypermethylation was noticeable in the Y955C old mice compared with WT old mice (Fig. 3D). Significant peaks were identified as those with peak score difference > 1 compared with controls and exhibited an FDR < 0.2 (and P < 0.05). Individual group comparisons to WT-Young mice found that Y955C-Young, WT-Old, and Y955C-Old hearts displayed significant differential DNA methylation with hypomethylation being the overall trend (Fig. 4A). According to a two-way ANOVA, Y955C expression caused 4,506 differentially methylated peaks, while aging was associated with 73,286 peaks (Fig. 4B). For Y955C, 1,419 peaks hypermethylated and 3,087 hypomethylated (31.5% hyper, 68.5% hypo) (Fig. 4C). Aging caused 14,492 regions to become hypermethylated and 58,794 hypomethylated (19.8% hyper, 80.2 hypo) (Fig. 4D). Only five differentially methylated regions displayed significant interactions between Y955C and aging, suggesting little interaction between both factors on DNA methylation changes. Results show that both mitochondrial pol-γ dysfunction and aging lead to DNA methylation changes, with hypomethylation the predominant effect of each. Results also demonstrate that aging causes an order of magnitude more DNA methylation changes than mutant pol-γ in the heart.

Fig. 3.

DNA methylation changes. DNA microarrays were used to asses DNA methylation changes in gene promoters staring 5 kb upstream of the transcriptional start site (TSS) to 1 kb downstream of the TSS. Scatterplots show global changes in DNA methylation in Y955C-young (A), WT-old (B), and Y955C-old hearts (C) compared with WT-young hearts. Most notable feature is bowing (widening of the scatter) in the scatterplot in the aging hearts independent of transgene. D: scatterplot of Y955C-old methylation compared with WT-old methylation.

Fig. 4.

Directionality of DNA methylation changes. A: DNA methylation changes were first analyzed pairwise to WT-young hearts to determine DNA methylation changes. Y955C-young hearts demonstrated 19,748 significant methylation changes with 79.2% of peaks exhibiting hypomethylation. WT-old hearts demonstrated 34,864 significant methylation changes with 83.6% exhibiting hypomethylation. Y955C-old hearts demonstrated 30,809 significant methylation changes with 79.1% exhibiting hypomethylation. B: using a 2-way ANOVA, 4,506 peaks with differentially methylated as a result of Y955C transgene expression, while 73,286 peaks demonstrated differential methylation as a result of age. C: of the 4,506 peaks differentially methylated by Y955C, 68.5% were hypomethylated. D: of the 73,286 peaks differentially methylated as a result of age, 80.2% were hypomethylated.

The relationship between DNA methylation changes and gene expression was determined. For the 95 genes differentially expressed by Y955C, 50 significant peaks were identified in 30 genes. This result shows that 31.6% of differentially expressed genes also exhibited changes in promoter DNA methylation. Of these peaks, 16 were hypermethylated (32.0%) and 34 were hypomethylated (68.0%). For the 4,452 genes differentially expressed by age, 982 peaks were identified in 342 genes, or 7.7% of differentially expressed genes. Of these peaks, 189 were hypermethylated (19.2%) and 793 were hypomethylated (80.8%).

Motif analysis.

We utilized Meme Suite to identify sequence motifs enriched within the differentially methylated DNA regions (2). Prior to motif analysis, only enriched regions with a CpG present in the sequence were selected for analysis; this is done because the sliding window test may identify peaks neighboring a CpG peak as significantly enriched though no CpG is present. This selection resulted in a smaller sample set of sequences for motif analysis (Table 1). We were unable to identify significant enrichment of any motif using only the sequences of the regions with significant peaks associated with differentially expressed genes (i.e., the 50 peaks for Y955C and 982 peaks for aging). Using the complete list of differentially methylated peaks for Y955C (4,506 peaks) and aging (73,286 peaks), we identified one motif enriched in the Y955C DNA sequences containing a CG site and three motifs in the aging DNA sequences containing a CG site (Fig. 5). Interestingly, three of these motifs (1 for Y955C and 2 for aging) shared sequence similarity. Mapping these three sites back to transcription factors identified transcription factors targeting the CACGTG E-box motif.

Table 1.

Peaks with CpG sites

Group	Significant Peaks	Peaks with CpG
TG-50	50	25
TG-4506	4,506	2,388
Age-982	982	546
Age-73,286	73,286	41,333
Complete set	2,191,584	1,001,506

TG, transgenic.

Fig. 5.

Motif enrichment analysis. Differentially methylated DNA regions containing a CpG site were analyzed for motif enrichment. Four motifs were significantly enriched (1 in Y955C sequences and 3 in age-related sequences). Of these, three motifs showed sequence similarities to the CACGTG E-box motif.

We analyzed the peaks to identify the probability of finding this CACGTG motif (and associated sequence variations) in our sequence compared with the random chance of observing the same motif in the CpG-containing promoter probes on the array (Table 2). Only the CACG motif was significantly enriched in the Y955C differentially methylated peaks; this was not observed in the peaks related to differential gene expression likely due to low samples size. Significant enrichment of all CACGTG-related motifs was seen in the age-related differentially methylated peaks, while CACGT and CACGTG motifs were also enriched in the methylated peaks associated with age-associated differentially expressed genes.

Table 2.

Motif enrichment in differentially methylated peaks

	TG-25	TG-2388	Age-546	Age-41333	CG-probes-1,001,506
CACGT/ACGTG	16.0% (1.99)	9.0% (1.12)	11.4% (1.41)*	8.9% (1.11)*	8.0% (1.0)
ACGT	16.0% (1.38)	11.7% (1.01)	13.6% (1.17)	11.9% (1.03)*	11.6% (1.0)
CACGTG	4.0% (2.56)	1.7% (1.07)	2.2% (1.41)*	1.8% (1.17)*	1.6% (1.0)
CACG/CGTG	32.0% (1.14)	30.9% (1.10)*	34.8% (1.24)	33.0% (1.18)*	28.0% (1.0)

^*P < 0.05 by χ² test.

DISCUSSION

Our results show that dysfunction of mitochondrial pol-γ and aging can lead to gene expression and nuclear DNA methylation changes in the heart. These findings provide new insights into a complex interplay between mitochondrial function and epigenetic control in cardiac cells. Our results show that DNA methylation, an epigenetic marker important in developmental maturation and coded directly on genomic DNA, can be modulated by mitochondrial dysfunction and suggests a pathogenetic link to CM.

We identified 95 genes that were differentially expressed in Y955C hearts. Of these, genes related to cardiac hypertrophy (e.g., CTF1, MAOA) and mitochondrial dysfunction and oxidative stress (e.g., GSTA1, MFN1) were differentially expressed as a result of Y955C. In the aging hearts, 4,452 genes were differentially expressed, but only 39 genes were shared with those in the Y955C experimental group. These results suggest that mitochondrial dysfunction alone does not account for the aging heart expression changes, and a more complex response to aging is at play.

Our investigation of DNA methylation revealed significant changes in methylation caused by Y955C and aging. Interestingly, results show 31.6% of Y955C-associated differentially expressed genes also showed significant DNA methylation changes. It is important to remember that additional epigenetic markers may work in tandem or in opposition to DNA methylation changes to alter expression, which may increase or decrease the number of differentially expressed genes associated with differential DNA methylation. With respect to aging, only 7.7% of differentially expressed genes exhibited DNA methylation changes, suggesting that DNA methylation changes are not a single determinant for age-associated cardiac decline but rather may play a combinatorial role. Further studies would be required to delineate the role of cardiac DNA methylation in the context of other epigenetic changes.

The mitochondrial theory of aging hypothesizes that increasing mitochondrial dysfunction and subsequent free radical generation is a primary cause of aging (3, 10, 16). The Y955C pol-γ mutation decreases polymerase activity, depletes mtDNA abundance, and increases ROS production (21, 25–28). In our analyses, we anticipated that the effects of Y955C and aging would synergize to accentuate their individual effects. Echocardiographic results support this hypothesis, with acceleration of LVH in young Y955C and CM in old transgenic mice (Fig. 1). DNA methylation results reinforced the overall trends toward DNA hypomethylation in both Y955C and aging (Fig. 4). These suggest a loss of DNA methylation is characteristic of aging in the heart. The amplitude of gene expression changes did not correlate with DNA methylation, a result readily attributed to the complex epigenetic regulation of gene expression through chromatin remodeling and other epigenetic mechanisms.(24) The mechanism by which pol-γ alters nuclear DNA methylation remains unclear by the nature of this study because we focused on classifying nuclear DNA methylation changes as a product of pol-γ mutation. The first issue is how pol-γ exerts its effects on nuclear DNA methylation. We've previously shown that the Y955C mutation increases cardiac oxidative stress; moreover, pharmacological pol-γ inhibitors such as the HIV-1 antiretroviral drug AZT also cause oxidative stress (21, 25–28). Our recent work showed that AZT alters nuclear DNA methylation in cardiac samples, but AZT-induced methylation changes could be attenuated by transgenically expressed mitochondrial catalase (18). Those results support mitochondrial ROS as a DNA hypomethylating agent, an effect observed here as Y955C and aging-induced hypomethylation. Taken together with findings here, mitochondrial-derived oxidative stress is a key mediator linking mitochondrial dysfunction to DNA methylation changes, likely DNA hypomethylation, through site-specific modifications in DNA methylation may be regulated on a per-gene basis.

Another mechanistic issue is how DNA methylation changes are altered by mitochondrial dysfunction. If mitochondrial derived ROS is the active moiety, ROS could oxidize 5-methylcytosine in nuclear DNA yielding 5-hydroxymethylcytosine and subsequent DNA hypomethylation. This represents a credible hypothesis for a mechanism (11, 14). In face of decreasing DNA methylation, DNA methyltransferase activity may be expected to methylate those hypomethylated regions with S-adenosylmethionine as the methyl-donating agent. For DNA hypomethylation caused by Y955C or aging, impaired DNA methyltransferase activity or diminished S-adenosylmethionine levels may contribute to the observed hypomethylation. Our recent results support alterations in S-adenosylmethionine and DNA methylation following AZT-induced mitochondrial ROS production (18).

Motif analysis identified enriched CACGTG-related motifs in differentially methylated regions in Y955C and aging mice. The CACGTG motif represents the canonical sequence of an E-box, an enhancer element identified by basic helix-loop-helix-containing transcription factors (8, 9). This motif is often used by members of the circadian rhythm family of transcription factors such as CLOCK, ARNTL, and BHLHB2. BHLHB2 is important in cardiac development and cardiomyocyte differentiation (23, 31). In support of a link between DNA methylation and circadian rhythm dysfunction, we recently showed the drug MDMA (“ecstasy”) altered DNA methylation and circadian gene expression in the murine heart (20). It is currently unknown if the binding of these transcription factors is affected by the DNA methylation state of the CpG moiety in CACGTG motifs. This may suggest circadian gene expression patterns are altered by oxidative stress mediated by DNA methylation changes, though further investigation is warranted.

GRANTS

Supported by National Institute on Drug Abuse Grant 1R01DA-030996 to W. Lewis.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

Author contributions: C.A.K. and W.L. conception and design of research; C.A.K., I.L., E.J.F., Z.J., T.L., and R.R. performed experiments; C.A.K., E.J.F., and Z.J. analyzed data; C.A.K. and E.J.F. interpreted results of experiments; C.A.K. prepared figures; C.A.K. and W.L. drafted manuscript; C.A.K. and W.L. edited and revised manuscript; C.A.K. and W.L. approved final version of manuscript.