Effects of ADMA upon Gene Expression: An Insight into the Pathophysiological Significance of Raised Plasma ADMA

Background Asymmetric dimethylarginine (ADMA) is a naturally occurring inhibitor of nitric oxide synthesis that accumulates in a wide range of diseases associated with endothelial dysfunction and enhanced atherosclerosis. Clinical studies implicate plasma ADMA as a major novel cardiovascular risk factor, but the mechanisms by which low concentrations of ADMA produce adverse effects on the cardiovascular system are unclear. Methods and Findings We treated human coronary artery endothelial cells with pathophysiological concentrations of ADMA and assessed the effects on gene expression using U133A GeneChips (Affymetrix). Changes in several genes, including bone morphogenetic protein 2 inducible kinase (BMP2K), SMA-related protein 5 (Smad5), bone morphogenetic protein receptor 1A, and protein arginine methyltransferase 3 (PRMT3; also known as HRMT1L3), were confirmed by Northern blotting, quantitative PCR, and in some instances Western blotting analysis to detect changes in protein expression. To determine whether these changes also occurred in vivo, tissue from gene deletion mice with raised ADMA levels was examined. More than 50 genes were significantly altered in endothelial cells after treatment with pathophysiological concentrations of ADMA (2 μM). We detected specific patterns of changes that identify pathways involved in processes relevant to cardiovascular risk and pulmonary hypertension. Changes in BMP2K and PRMT3 were confirmed at mRNA and protein levels, in vitro and in vivo. Conclusion Pathophysiological concentrations of ADMA are sufficient to elicit significant changes in coronary artery endothelial cell gene expression. Changes in bone morphogenetic protein signalling, and in enzymes involved in arginine methylation, may be particularly relevant to understanding the pathophysiological significance of raised ADMA levels. This study identifies the mechanisms by which increased ADMA may contribute to common cardiovascular diseases and thereby indicates possible targets for therapies.


A B S T R A C T Background
Asymmetric dimethylarginine (ADMA) is a naturally occurring inhibitor of nitric oxide synthesis that accumulates in a wide range of diseases associated with endothelial dysfunction and enhanced atherosclerosis. Clinical studies implicate plasma ADMA as a major novel cardiovascular risk factor, but the mechanisms by which low concentrations of ADMA produce adverse effects on the cardiovascular system are unclear.

Methods and Findings
We treated human coronary artery endothelial cells with pathophysiological concentrations of ADMA and assessed the effects on gene expression using U133A GeneChips (Affymetrix). Changes in several genes, including bone morphogenetic protein 2 inducible kinase (BMP2K), SMA-related protein 5 (Smad5), bone morphogenetic protein receptor 1A, and protein arginine methyltransferase 3 (PRMT3; also known as HRMT1L3), were confirmed by Northern blotting, quantitative PCR, and in some instances Western blotting analysis to detect changes in protein expression. To determine whether these changes also occurred in vivo, tissue from gene deletion mice with raised ADMA levels was examined. More than 50 genes were significantly altered in endothelial cells after treatment with pathophysiological concentrations of ADMA (2 lM). We detected specific patterns of changes that identify pathways involved in processes relevant to cardiovascular risk and pulmonary hypertension. Changes in BMP2K and PRMT3 were confirmed at mRNA and protein levels, in vitro and in vivo.

Introduction
Asymmetric dimethylarginine (ADMA) is an endogenous inhibitor of all nitric oxide synthase (NOS) isoforms [1]. It is synthesised by the action of protein arginine methyltransferases (PRMTs), and following proteolysis, free ADMA is released into the cell cytosol and thence into plasma. Circulating concentrations of ADMA are increased in patients with renal failure [1], pulmonary hypertension, heart failure, hypercholesterolemia or a wide range of other cardiovascular risk factors [2][3][4][5][6]. In patients with end-stage renal failure, the plasma levels of ADMA predict mortality and cardiovascular outcome [7], and in a cohort of otherwise healthy Finnish men, those with the highest levels of ADMA had an increased risk of acute coronary events [8]. Increased circulating ADMA in pregnant women predicts an increased risk of pre-eclampsia and intrauterine growth retardation [9].
Despite these clinical observations and the increasing excitement surrounding the use of ADMA as a risk marker for vascular disease [3,7], it is still not clear whether ADMA has a causal role in pathophysiology. It has been argued that the concentration of ADMA in plasma is too low to be an effective inhibitor of NOS, and that the usual concentrations of arginine in cells should overcome any inhibitory effects of ADMA on NOS [10]. In order to determine how ADMA might exert effects on endothelial cells and produce pathology, we assessed the effects of ADMA on gene expression in human coronary endothelial cells.

Cell Culture
Human coronary artery endothelial cells (HCAEC) were purchased from Promocell and grown according to the manufacturer's instructions. HCAEC in 75-cm 2 flasks at 70% confluency (passage 3 or 4) were treated for 24 h with complete media supplemented with asymmetric dimethylarginine (N G N G -dimethyl-L-arginine; ADMA; 0, 2, or 100 lM; Merck Biosciences, United Kingdom). This was repeated on three separate occasions with different batches of cells. RNA from each study was used as described below for GeneChip (Affymetrix, Santa Clara, California, United States) analysis. Our strategy for the GeneChip and subsequent analysis is outlined in Figure 1.

GeneChip Experiments
ADMA-treated HCAEC from T75 flasks were harvested in 7.5 ml of TRIzol (Invitrogen, Carlsbad, California, United States), and total RNA was extracted; cDNA and subsequent cRNA synthesis were prepared as previously described [11]. The quality of the biotin-labelled cRNA transcripts was determined using a Bioanalyser 2100 (Agilent Technologies, Palo Alto, California, United States). Purified cRNA (15 lg) was fragmented and hybridised to human U133A GeneChips according to Affymetrix standard protocols (http://www. affymetrix.com). Labelled GeneChips were scanned, using a confocal argon ion laser (Agilent Technologies).

GeneChip Data Analysis
The U133A GeneChip contains oligonucleotides derived from approximately 22,000 human transcripts and includes control bacterial genes bioB, bioC, bioD, and cre. GeneChip data files scaled to 100 and normalised to the median prior to analysis with GeneSpring 7.2 software (Agilent; Figure 1). Genes were excluded if the signal strength did not significantly exceed background values and if expression did not reach a threshold value for reliable detection (based on the relaxed Affymetrix MAS 5.0 probability of detection (p 0.1; [12]) in each of the three separate studies. Finally, genes were excluded if the level of expression did not vary by more than 1.7-fold between ADMA-treated (2 or 100 lM) compared with untreated control HCAEC. The remaining genes were subjected to nonparametric Welch t-tests and are reported with their respective fold changes and p-values. The data have been submitted in a MIAME-compliant format to ArrayExpress at EBI (http://www.ebi.ac.uk/arrayexpress/).

Determination of ADMA Levels
HCAEC cells were grown in 75-cm 2 flasks as described above for 24 h. Methylated arginines in the conditioned medium were quantitated by HPLC as previously described [1,13].

Cell Growth Assay
Cells were seeded into a 96-well microtitre plate at a density of 500 cells per well and grown in complete EC medium in the presence of 0, 2, or 100 lM ADMA. Cell growth was assayed (in triplicate) over a 4-d period using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, Wisconsin, United States).

Confirmation of Gene Changes
For genes of interest selected on the initial GeneChip analysis, further experiments were undertaken, with different batches of cells, to verify changes detected by global expression profiling and using different approaches to assess mRNA or protein.

RT-PCR
For RT-PCR cDNA was synthesised from the total RNA (approximately 1 lg) using Ready-To-Go You-Prime First-Strand Beads (Amersham Biosciences, Little Chalfont, United Kingdom) and supplemented with the 39 gene-specific primers with b-actin as a control. RT-PCR used the primer sequences (Table 1) with b-actin as a control.
The PCR products were separated in 1% agarose gel, and the intensity of the bands was measured; each sample was corrected for b-actin. PCR products used for Northern blotting were excised and purified from agarose gel using Qiaex II (Qiagen, Valencia, California, United States).

Quantitative-PCR
Quantitative PCR (Q-PCR) was carried out using a Light-Cycler (Roche, Alameda, California, United States) and LightCycler software version 3, LightCycler run 4.24. All reagents required for PCR (excluding cDNA and primers) were included in the LightCycler FastStart DNA Master SYBR Green 1 kit (Roche). Reverse transcription was performed as described above for RT-PCR. The PCR cycle settings were 95 8C for 5 min, followed by 45 cycles of 95 8C for 5 s, 58 8C for 10 s (60 8C for b-actin), 72 8C for 40 s, and 77 8C (82 8C for b-actin), where fluorescence was measured at the end of each cycle and is gene-specific. Standard curves were constructed for PRMT3 as described by the manufacturer's instructions and compared to the reference gene b-actin.

Western Blotting
HCAEC were treated with either 0, 2, or 100 lM ADMA for 24 h in 6-well plates and harvested in lysis buffer as described previously [11], the protein concentrations of the lysates were determined by protein assay (Bio-Rad, Hercules, California, United States), and cell lysates were resolved by 12% SDS polyacryamide gel electrophoresis with equal amounts of protein loaded into each lane. Anti-PRMT3 (Upstate Biotechnology, Charlottesville, Virginia, United States) and Anti-BMP2K (Orbigen, San Diego, California, United States) were used with anti-rabbit secondary antibody coupled to horseradish peroxidase and detected with the ECLþ detection system (Amersham Pharmacia, Piscataway, New Jersey, United States). Densitometry of the bands was determined, and results are shown as the mean densitometry, where n ¼ 4 with inset of a typical Western blot.

Pathway Mapping and Gene Ontology Analysis
In order to determine whether ADMA had affected expression of genes in pathways related to the genes identified on the initial analysis, lists of genes that were changed more than 1.7-fold compared to control (irrespective of p-value) were examined using Gene Ontology (Affymetrix) data mining for biological process (at level 3), and Expression Analysis Systematic Explorer (EASE) biological theme analysis were conducted online at http://david. niaid.nih.gov using DAVID [14]. DAVID-EASE [15] generates an EASE score predicting the likelihood of genes mapping to specific biological processes (determined by Gene Ontology consortium) from a given list of changed genes, therefore enabling global themes in gene expression following ADMA treatment to be identified [15].

Dimethylarginine Dimethylaminohydrolase 1 Gene Deletion Mice
ADMA is metabolised to citrulline and dimethylamine by the action of dimethylarginine dimethylaminohydrolase (DDAH). We have created knockout mice that lack DDAH1. DDAH1 heterozygous knockout mice (details to be published elsewhere) have approximately 2-fold higher plasma ADMA levels compared to wild-type litter-mates and thus provide an excellent model to test the effects of moderately raised ADMA levels in vivo. Northern blotting was carried out using RNA extracted from the brain, heart, and kidney of 12-14wk-old DDAH1 heterozygous knockout and wild-type littermates with probes for mBMP2K and mPRMT3, and results are expressed relative to mb-actin.

Statistical Analyses
Q-PCR, Northern blot, and Western blot densitometry data for the treated HCAEC was analysed by one-way analysis of variance (ANOVA) coupled to Bonferroni posttest, and the Bonferroni posttest p-values are reported. The Northern blots from the DDAH1 gene deletion mice were compared with unpaired t-test and the p-values are reported.

Changes in HCAEC Gene Expression in Response to ADMA on U133A GeneChips
A total of 979 genes changed in expression between ADMA-treated and control cells. Following the Welch t-test with a cutoff of p , 0.05, 56 genes were identified as having shown a statistically significant change between the untreated and 2-lM ADMA-treated cells, and 86 genes changed between the untreated and 100-lM ADMA greater than 1.7-fold; 11 genes showed statistically significant changes at both concentrations of ADMA compared to untreated cells (Figure 2A; Tables 3 and 4).

Effects of ADMA upon HCAEC Viability
Basal levels of ADMA and SDMA in HCAEC media were 0.17 6 0.01 lM and 0.22 6 0.02 lM, respectively, and arginine levels exceeded 300 lM. No changes were observed  in HCAEC viability over 72 h in the presence of either 2 or 100 lM ADMA ( Figure 2B).

Confirming Transcriptional Changes
To determine the reliability of changes identified by GeneChip analysis, four genes were selected from those that showed a statistically significant change in either the 2 or 100 lM sample compared to control ( Figure 3A). These were SCAMP1, Calreticulin, (RpL27) and RpS11. In studies on a different batch of HCAEC, Northern blotting confirmed that expression of these genes changed in response to ADMA ( Figure 3B).
In order to elucidate the mechanism of ADMA action, HCAEC were also treated with the potent NOS inhibitor L-NIO and SDMA, which is not a NOS inhibitor or DDAH substrate, but is a naturally occurring methylarginine that competes with arginine for the cationic amino acid transporter [10,16]. Interestingly neither SDMA nor L-NIO elicited significant changes in the expression of RpS11, RpL27, SCAMP1, or Calreticulin ( Figure 3B).

Genes of Specific Interest Identified by GeneChip
Protein arginine methyltransferase. PRMT3 was selected as a gene of interest, because of its involvement in ADMA synthesis [17]. It changed 1.8-fold on the GeneChip untreated versus 100 lM groups (p ¼ 0.0339; Figure 4A). In a separate series of follow-up studies, HCAEC were treated with ADMA, and PRMT3 gene expression was determined by Q-PCR. PRMT3 mRNA levels increased following either low-or highdose ADMA treatment ( Figure 4B untreated versus 2 lM, p , 0.05; and untreated versus 100 lM, p , 0.01, n ¼ 6). PRMT3 protein expression also increased in response to ADMA treatment ( Figure 4C; untreated versus 2 lM; and untreated versus 100 lM, p , 0.05, n ¼ 4). When HCAEC were treated for 24 h with 2 lM SDMA or L-NIO, neither treatment significantly affected PRMT3 expression, whereas 100 lM SDMA or L-NIO caused a small increase in PRMT3 expression ( Figure 4B; n ¼ 6).
Bone morphogenetic protein 2 inducible kinase. We identified that bone morphogenetic protein 2 inducible kinase (BMP2K) changed on the GeneChip in response to ADMA (2 lM and 100 lM ADMA increased expression by 2.304-fold [p ¼ 0.00128] and 2.695-fold [p , 0.001], respectively; Figure 5A). In a separate series of experiments on a different batch of HCAEC this increase in BMP2K expression was confirmed by RT-PCR (data not shown). Western blotting also revealed that BMP2K protein levels were increased in response to ADMA ( Figure 5B); densitometry of these blots indicated that there was a significant increase (untreated versus 2 lM, p , 0.05; and untreated versus 100 lM, p , 0.01, n ¼ 4).

Identification of Genes Involved in Bone Morphogenetic Protein Signalling
Having confirmed changes in BMP2K expression, the total list of 765 genes that changed greater than 1.7-fold in response to 100 lM ADMA, was reexamined to identify additional genes in the bone morphogenetic protein (BMP) signalling pathway affected by ADMA. This analysis identified Smad5 and BMPR1A (Figures 6 and 7). In a separate set of studies Northern blotting confirmed that mRNA was increased in HCAEC after 24 h by either 2 or 100 lM ADMA for Smad5 and BMPR1A (Smad5 untreated versus 2 lM and untreated versus 100 lM; p , 0.05, n ¼ 4; Figure 6A; and BMPR1A untreated versus 2 lM, p , 0.05 and untreated versus 100 lM; p , 0.01, n ¼ 4; Figure 6B).

Identification of Global Changes in Gene Expression
All genes that changed by more than 1.7-fold (irrespective of p-value) were entered into DAVID-EASE. EASE probability scores were generated based upon the number of genes for each biological process altered in response to ADMA (Tables  5 and 6), where these biological processes were defined by enriched Gene Ontology categories. These gene lists indicated that ADMA affects genes involved in metabolism, RNA splicing, transcription, and cell cycle regulation.

Gene Deletion Mice
To examine whether the effects observed in the cell culture model were relevant to the in vivo situation, we determined the expression level of certain genes in DDAH heterozygous knockout mice that have 2-fold elevation in plasma ADMA levels. Total RNA from DDAH1 heterozygous knockout was probed for mBMP2K and mPRMT3 by Northern blotting and corrected for mb-actin expression (Figure 8). Levels of mBMP2K for brain, heart, and kidney, respectively, were 42 6 13.8% (p ¼ 0.047), 33.3 6 18.6% (p ¼ 0.038), 74.0 6 21.4% (p ¼ 0.007), higher in DDAH1 heterozygous mice (n ¼ 9) compared with wild-type litter-mates (n ¼ 5). A similar trend was seen for the expression of mPRMT3 (data not shown).

Discussion
ADMA is an endogenous inhibitor of NOSs [1] and there is an association between increased plasma levels of ADMA and renal disease [1], pulmonary hypertension [5], preeclampsia [9], and the progression of atherosclerosis [18,19]. Whilst the concentration of ADMA in plasma of healthy adults varies between 0.4 and 1 lM, it may increase to 1.45-4.0 lM with certain diseases, and this increase is thought to be causally involved in pathophysiology [1,6,7,9,20]. In the present study we detected substantial changes in gene expression in HCAEC after 24 h of exposure to concentrations of ADMA similar to those reported in pathophysiological states. Furthermore, we identified specific pathways of gene activation that give insight into the mechanisms by which ADMA may contribute to disease. Surprisingly, some of these changes appear to be independent of blockade of the Larginine:nitric oxide (NO) pathway.

Low Concentrations of ADMA Alter Gene Expression
Acute administration of ADMA to healthy individuals elicits a transient fall in heart rate and cardiac output and increases blood pressure [2], but little is known of the potential longer-term effects of raised ADMA. Zoccali et al. reported that a 1-lM increment in ADMA above the upper limits for healthy individuals was associated with increased risk of cardiovascular mortality [7], and levels around 2 lM seem to be associated with a number of diverse cardiovascular pathologies [3,8]. In the current study, HCAEC were treated with 2 or 100 lM ADMA. We repeated GeneChip analysis in three separate studies and observed reproducible changes in the expression of a subset of genes in response to low-and high -dose ADMA. Because there is always a possibility of false positives being identified on arrays, we tested the reproducibility of the GeneChip approach by selecting four genes significantly upregulated by ADMA treatment. In a separate series of studies we confirmed an increase in mRNA levels by Northern blotting at both concentrations of ADMA. Thus it is clear that pathophysiological concentrations of ADMA (2 lM) affect endothelial cell gene expression even in the presence of very high arginine concentrations (.300 lM). Endogenous ADMA is metabolised by DDAH, and DDAH activity is the major determinant of plasma ADMA concentrations [2]. To determine whether the effects we saw in vitro would be reproduced in vivo we examined DDAH1 knockout mice. At least one of the gene changes we saw in vitro also occurs in vivo since DDAH1 heterozygous knockout mice have increased concentrations of ADMA in plasma and showed upregulation of BMP2K in several tissues.
ADMA inhibits NOSs with an IC 50 of about 5 lM, the precise potency depending on the prevailing concentration  of arginine [21]. It is known that adding endogenous NO to endothelial cells alters gene expression [22,23] and that inhibitors of endogenous NO generation can alter expression of specific genes, at least under conditions of endothelial cell activation with cytokines. To determine whether the effects we observed could be accounted for by inhibition of NOS, we treated cells with a highly potent NOS inhibitor (L-NIO) and with SDMA, an endogenous dimethylarginine that has no effect on NOS but which can block arginine transport [10,16]. Neither SDMA nor L-NIO replicated the effects of ADMA on gene expression. We have not undertaken a full GeneChip analysis of responses to L-NIO, so we do not know how much overlap there would be between the effects of L-NIO and ADMA, but of those genes examined we saw a discordance between responses to the two inhibitors. This raises the intriguing possibility that some of the actions of ADMA may be independent of effects of NO, possibly due to other actions such as the ability of ADMA to increase superoxide generation [24,25]. Other authors have reported differences in efficacy of ADMA in cell culture systems compared to other NOS inhibitors that have more potent effects on isolated NOS [26], and further studies will be required to identify to what extent NO-independent effects contribute to the overall action of ADMA. Interestingly, a NOS-independent effect of ADMA on angiotensin-converting enzyme has recently been suggested [25]. NOS-independent effects of ADMA may explain why, in some situations where ADMA concentrations are raised, L-arginine does not restore normal endothelial function [27] and why ADMA can exert effects, even in the presence of high endogenous arginine concentrations.

Patterns of Gene Change
Mapping genes changed by ADMA to identify global changes in biological processes [15], indicated that ADMA treatment may have significant effects on genes involved in cell cycle regulation, cell proliferation, DNA repair, transcriptional regulation, and metabolism. The full biological significance of the range of genes affected is not yet known, but our data demonstrate the potential for elevated ADMA to affect endothelial (and likely other cellular) function in disease. In the present study, we focussed on two pathways of potential importance-BMP pathways and PRMTs.

BMP Pathways and ADMA
Analysis of the U133A GeneChips revealed that BMP2K was induced more than 2-fold in response to either 2 or 100 lM ADMA. The increase in gene expression was mirrored by an increase in BMP2K protein, and the effect was also seen in our high-ADMA mouse model. By relaxing parameters (to exclude false negatives) a search for other genes involved in BMP signalling revealed that Smad5 and BMPR1A were also amongst the transcripts increased by GeneChip analysis, and these changes were confirmed by Northern blotting. The finding of changes in the BMP signalling pathway is important since the ADMA/DDAH pathway seems to be involved in animal models of pulmonary hypertension [28,29], and mutations in the BMP receptor 2 (BMPR2) are associated with familial pulmonary hypertension in humans [30].
The changes in BMP pathways may also be important in understanding some of the effects of renal failure. ADMA accumulates in renal failure and fulfils many of the criteria of a uraemic toxin [1,18]. In addition to effects on cardiovas-  cular risk, ADMA may contribute to renal osteodystrophy, a process in which BMPs have been implicated [31]. Indeed an earlier study that showed that ADMA reduces osteoblast differentiation and decreases osteocalcin expression [32]. The present study confirms osteocalcin as a gene downregulated by ADMA, and since activation of BMP2K attenuates osteocalcin expression and reduces osteoblast differentiation [33], it is possible that the effects of ADMA on osteocalcin may be secondary to induction of BMP2K (Figure 7). Whatever the mechanisms, identification of a link between ADMA and BMP pathways may be relevant to the increased vascular calcification seen in renal disease [31].

ADMA and Arginine Methylation
We observed that PRMT3 gene expression was elevated following exposure to ADMA; this was confirmed by Q-PCR, and PRMT3 protein expression also increased. This is the first report that ADMA may alter the expression of enzymes involved in its own synthesis. There are presently five known PRMTs that asymmetrically methylate arginine residues and two (PRMT5 and PRMT7) that symmetrically methylate arginine residues. PRMT3 has a wide tissue distribution, is expressed in highly vascular tissues, including heart and lung [17], and expression may be increased by oxidised-LDL [34]. Our observations indicate that ADMA can induce a similar increase in PRMT3 expression.
The roles of methylation of arginine residues in proteins are not yet well defined, but studies of PRMT3 in fission yeast have shown that it associates with proteins involved in the translational machinery and that the S2 ribosomal protein, a component of the yeast 40S ribosome, is a specific substrate for PRMT3 [35]. PRMT3 is the only PRMT known to interact with the translational machinery, and it is interesting that we have found several genes, including ribosomal proteins RpS11 and RpL27, involved in translational control, that were also altered in response to ADMA treatment. Amongst the genes changed greater than 1.7-fold in response to ADMA was methionine adenosyltransferase II a, which catalyses the production of S-adenosylmethionine, the methyl donor for the PRMT reaction [36]. The role of ADMA in regulating arginine methylation in protein deserves further study.

Summary
Increased circulating concentrations of ADMA have been reported in cardiovascular and other disorders, and intracellular concentrations may vary independently of circulating levels. In the present study we have demonstrated that relatively small changes in the concentration of ADMA affect gene expression in endothelial cells. Identification of pathways regulated by ADMA may aid our understanding of how ADMA contributes to a wide range of pathologies. Two pathways of specific interest have been identified-BMP signalling and enzymes involved in arginine methylation. The effects on BMP signalling may be particularly important in renal disease and in the link between raised ADMA and pulmonary hypertension.

Accession Numbers
The microarray data have been loaded into the EBI MIAMExpress database (http://www.ebi.ac.uk/miamexpress/) and have been assigned the accession number E-MEXP-377.  Genes that changed by more than 1.7-fold in 100 lM treated compared to untreated HCAEC (765) were analysed by DAVID-EASE [15] software (National Institutes of Health) to determine biological processes likely to have been affected by ADMA treatment. MHC, major histocompatibility complex. DOI: 10.1371/journal.pmed.0020264.t006 Figure 8. BMP2K Is Increased in DDAH1 Gene Deletion Mice Expression of BMP2K mRNA from brain, heart, and kidney is increased in 12-wk DDAH1 heterozygous knockout mice compared to wild-type littermates (p ¼ 0.0473, p ¼ 0.0379, and p ¼ 0.0070, respectively); mRNA was corrected for differences in b-actin mRNA expression, where n ¼ 5 wild type, and n ¼ 9 DDAH heterozygous. DOI: 10.1371/journal.pmed.0020264.g008