Post-transcriptional Regulation of Endothelial Nitric-oxide Synthase by an Overlapping Antisense mRNA Transcript*

Endothelial nitric-oxide synthase (eNOS) mRNA levels are abnormal in diseases of the cardiovascular system, but changes in gene expression cannot be accounted for by transcription alone. We found evidence for the existence of an antisense mRNA (sONE) that is derived from a transcription unit (NOS3AS) on the opposite DNA strand from which the human eNOS (NOS3) mRNA is transcribed at human chromosome 7q36. The genes are oriented in a tail-to-tail configuration, and the mRNAs encoding sONE and eNOS are complementary for 662 nucleotides. The mRNA for sONE could be detected in a variety of cell types, both in vivo and in vitro, but not vascular endothelial cells. In contrast, expression of eNOS is highly restricted to vascular endothelium. Most surprisingly, interrogation of transcriptional events across NOS3/NOS3AS genomic regions, using single- and double-stranded probes for nuclear run-off analyses and chromatin immunoprecipitation-based assessments of RNA polymerase II distribution, indicated that NOS3 and NOS3AS gene transcription did not correlate with steady-state mRNA levels. We found strong evidence supporting a role for NOS3AS in the post-transcriptional regulation of NOS3 expression. RNA interference-mediated inhibition of sONE expression in vascular smooth muscle cells increased eNOS expression. Overexpression of sONE in endothelial cells blunted eNOS expression. Finally, the histone deacetylase inhibitor trichostatin A is known to regulate the expression of eNOS via a post-transcriptional mechanism. We found that trichostatin A treatment of vascular endothelial cells increased expression of sONE mRNA levels prior to the observed decrease in eNOS mRNA expression. Taken together, these results indicate that an antisense mRNA (sONE) participates in the post-transcriptional regulation of eNOS and provide a newer model for endothelial cell-specific gene expression.

Endothelial nitric-oxide synthase (eNOS, 1 NOS3) is a key enzyme in the regulation of vascular wall homeostasis, which is why perturbations in eNOS result in abnormalities of blood pressure, platelet function, and vessel wall remodeling (1)(2)(3). We have observed decreased expression of steady-state mRNA in cells overlying advanced human atherosclerotic plaques, a cellular milieu in which endothelial cells are exposed to varied injurious stimuli (4). Regulation of eNOS gene expression is complex and takes place at multiple levels. Recent work has emphasized the importance of post-translational modification of the eNOS holoenzyme (5,6). NOS3 is a 26-exon gene located on human chromosome 7, which encodes a 4052-nucleotide mRNA (7). Expression of the eNOS mRNA is highly restricted to vascular endothelial cells. Only a few exceptions have been noted (8,9), which contrasts with the broad tissue expression of neuronal NOS (nNOS, NOS1) and inducible NOS (NOS2) (10,11). Most importantly, levels of eNOS mRNA are regulated by both transcriptional and post-transcriptional pathways (12). eNOS mRNA is normally very stable with a half-life of ϳ16 -48 h (13,14). Cytokine activation (14), oxidized low density lipoprotein (15), exposure to lipopolysaccharide (16) or histone deacetylase inhibitors (17), cellular proliferation (18,19), and hypoxia (20,21) all decrease eNOS expression in vascular endothelial cells in a manner dependent, in major part, on decreased mRNA stability. Conversely, treatment of endothelial cells with atheroprotective 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors increases eNOS mRNA levels via transcript stabilization (22,23). Our appreciation that mRNA stability plays an important role in eNOS regulation continues to expand. However, the mechanisms that are involved in this process are poorly understood.
Our understanding of mechanisms of post-transcriptional control of gene expression has matured in recent years. It has recently been appreciated that changes in mRNA half-life and translational regulation can be mediated not only by RNAbinding proteins but also by sense-antisense RNA interactions (24). Here we describe the cloning of a human antisense mRNA (sONE) that is derived from a transcription unit (NOS3AS) on the opposite DNA strand from which the eNOS (NOS3) mRNA * 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  is transcribed. A related genomic organization is also described for the mouse ortholog, Nos3as. Collectively, the results presented here indicate that an antisense mRNA (sONE) participates in the post-transcriptional regulation of eNOS.

EXPERIMENTAL PROCEDURES
cDNA and Genomic Characterization-An oligo(dT)-primed gt11 human testis cDNA library (Clontech) was screened under conditions of high stringency by using a full-length human eNOS cDNA (25). Hybridization conditions and post-hybridization washes were as described (26). For the further characterization of NOS3AS cDNA and genomic sequences, I.M.A.G.E. cDNA clones and human genomic bacteriophage EMBL 3 A and BAC clones were sequenced using an ABI PRISM 377 DNA sequencer (Applied Biosystems, Foster City, CA) and compared with sequences we reported previously (7) and the two assemblies available for human chromosome 7 sequence (27,28). To characterize Nos3as cDNA and genomic sequences, I.M.A.G.E. and new cDNAs were sequenced (see www.chr7.org1). Murine genomic clones for Nos3as were isolated from the RPCI-22 bacterial artificial chromosome library created in the pBACe3.6 vector using female murine 129S6/SvEvTac strain diploid genomic DNA. 5Ј-Rapid amplification of cDNA ends (RACE) and 3Ј-RACE were performed by using total cellular RNA (1 g) derived from various normal human and murine tissues and cell types (human placenta, brain, testis, and vascular endothelial cells and murine ED 11.5 days post-coitum embryos and placenta; Clontech), as described (10). A complete description of the oligonucleotide primers used for human and murine RT-PCR and RACE analyses is available from the authors upon request.
RT-PCR Southern Blot Analyses and Quantitative Real Time RT-PCR-First strand cDNA was synthesized with total cellular RNA (5 g) derived from varied normal human tissues (Clontech) using random or strand-specific primers and SuperScript II reverse transcriptase (Invitrogen). Primers used to prepare strand-specific cDNA were located within 100 nt of the PCR amplicon. The sequence of the eNOS ϩ first-strand primer was 5Ј-CCT GGA GGA ATA ATG CTG AAT C-3Ј and corresponds to the genomic DNA coding strand of eNOS. This primer pair would bind to the complementary antisense strand. The sequence of the eNOS-first-strand primer was 5Ј-AAT GGT AAC GTG CAG GAC ATC-3Ј and corresponds to the antisense genomic DNA strand with respect to eNOS. PCR amplification used a single primer pair located in overlapping exonic regions of the NOS3/NOS3AS transcription units to amplify both eNOS and sONE in the same reaction, specifically eNOS exons 25 and 26 and/or sONE exons 11 and 10 (5Ј-CCT CTG GAC AGA TGT GAG AAG GCA-3Ј and 5Ј-CAT GTT TGT CTG CGG CGA TGT TA-3Ј). Products were size-fractionated by agarose gel electrophoresis and downward transferred to HyBond Nϩ membranes (Amersham Biosciences). Southern blots were hybridized with [␥-32 P]ATP-labeled oligonucleotides positioned internally to flanking PCR primers. The eNOS-specific hybridization primer spanned the exon 25/26 junction (5Ј-TGC TGC GGG ATC AGC AAC GCT ACC A-3Ј), and the sONE-specific hybridization primer spanned the exon 10/11 junction (5Ј-GAA AAG CTC TGG CAC GCC GAT GAC G-3Ј). All quantitative reverse transcriptase-PCR analyses were performed by using an ABI PRISM 7900 HT sequence detection system containing the cDNA equivalent of 100 ng of RNA. We again prepared first-strand cDNA using three approaches. RNA was reverse-transcribed using random primers or strand-specific primers corresponding to either the coding or antisense strand. The eNOS and sONE real time primer pairs were located in nonoverlapping genomic regions of the two transcription units. Strand-specific cDNA primers corresponding to the genomic DNA coding strand or antisense strand for eNOS were eNOSex11ϩ (5Ј-GGA GTG CCG CCA AGG GCA CC-3Ј) and eNOSex12Ϫ (5Ј-TCG GAG CCA TAC AGG ATT GT-3Ј), respectively. Strand-specific cDNA primers corresponding to the genomic DNA coding strand or antisense strand for sONE were sONEex7ϩ (5Ј-GCA CCT CCG CAG CCA GCT CC-3Ј) and sONEex8Ϫ (5Ј-GCT GCC GAT GCT CCC AAC TT-3Ј), respectively. For human tissues, reactions were performed in triplicate on cDNA synthesized from 3 to 6 independent lots of RNA. To quantitate copy number, we used serial 10-fold dilutions of plasmids corresponding to target RNAs to generate standard curves. Normalization of results used the geometric means of GAPDH, cyclophilin, and 18 S RNAs. For quantitation of eNOS, we used primers corresponding to NOS3 exons 11/12, 5Ј-GGC ATC ACC AGG AAG AAG ACC-3Ј, 5Ј-TCA  CTC GCT TCG CCA TCA C-3Ј, and probe 5Ј-FAM TM CCA ACG CCG  TGA AGA TCT CCG C TAMRA TM -3Ј (90-bp amplicon). Quantitation of sONE mRNA levels was performed by using primers spanning NOS3AS exons 7/8 5Ј-CGC CTG ATG AGG AGA AGC C-3Ј, 5Ј-TCT GTG GTC ACC TGA AAC CCT-3Ј, and 5Ј-VIC TM TGC CTC TAG CCC CAG ACA ACA GTG G TAMRA TM -3Ј (123-bp amplicon) (VIC TM , Applied Biosystems). Human GAPDH mRNA levels were assessed by using primers corresponding to exons 2 to 4: 5Ј-GAA GGT GAA GGT CGG AGT C-3Ј, 5Ј-GAA GAT GGT GAT GGG ATT TC-3Ј, and the probe 5Ј-VIC TM CAA GCT TCC CGT TCT CAG CC TAMRA TM -3Ј (226-bp amplicon). Human cyclophilin mRNA levels were assessed using primers corresponding to exons 1 and 2: 5Ј-GAC GGC GAG CCC TTG G-3Ј, 5Ј-TCT GCT TTT GGG ACC TTG T-3Ј, and the probe 5Ј-FAM TM CGC GTC TCC TTT GAG CTG TTT GCA BHQ TM -3Ј (64-bp amplicon). Human 18 S rRNA levels were assessed using the primers 5Ј-AGG AAT TGA CGG AAG GGC AC-3Ј and 5Ј-GA CAT CTA AGG GCA TCA CA-3Ј with Sybr Green chemistry. Primers 5Ј-GAT GAG CTA CCC CCT CAG GT-3Ј and 5Ј-ACA GCC TGA CCT GCT CAT CT-3Ј were used for FLJ22169. For murine tissue RT-PCR, we used primers 5Ј-GCC ACC TGA TCC TAA CTT GC-3Ј and 5Ј-AGA AAA GCT CTG GGT GCG TA-3Ј corresponding to exon 23 and exon 26 for Nos3, primers 5Ј-CCC TTG CAC TGA CTG AGG AT-3Ј and 5Ј-TGC TGC TTG TTT CGT ACA GG-3Ј in exons 8 and 9 for Nos3as, and primers 5Ј-AGA GGA GGG CTC AGA GGA TG-3Ј and 5Ј-GCT GCT CAG CTA TTC CAG GT-3Ј for Flj22169.
Northern Blot Analysis and in Situ cRNA Hybridization-Multiple tissue Northern blots containing ϳ2 g of poly(A) ϩ RNA per lane (Clontech) were hybridized with human cDNA probes corresponding to exons 8 -16 of NOS3 or 4 -9 of NOS3AS. These probes are located in nonoverlapping genomic regions. Murine cDNA probes corresponded to either exon 1 and 2 or exon 9 of Nos3as. Probes were [␣-32 P]dCTPlabeled using Ready-to-Go cDNA labeling beads (Amersham Biosciences) to a specific activity of 1 ϫ 10 9 cpm/g. Hybridization and high stringency wash conditions followed the manufacturer's suggestions. Northern blots were analyzed using a PhosphorImager with Image-Quant software (version 1.3, Amersham Biosciences).
In situ cRNA hybridization studies were performed as described (18). To detect murine Nos3-derived mRNAs, we used a probe representing 992 nt of exons 23-26 of murine eNOS. To detect murine Nos3asderived mRNAs, we used a 481-nt probe that corresponded to exons 2-4 of murine sONE. Briefly, we studied embryonic tissue from ED 9.5, ED 12.5, and ED 15.5, extraembryonic uteroplacental tissue from ED 11.5, ED 12.5, ED 16.5, and ED17.5, and adult heart, aorta, kidney, and testes. Tissues were fixed in 4% paraformaldehyde overnight, dehydrated, and embedded in paraffin. Serial sections (5 m) were mounted on gelatinized slides, deparaffinized in xylene, rehydrated, and postfixed. The sections were digested with proteinase K, treated with triethanolamine/acetic anhydride, washed, and dehydrated. cRNA transcripts were uniformly labeled with 35 S-UTP (Ͼ1000 Ci/mmol; Amersham Biosciences) and hydrolyzed with alkali to a mean size of Ͻ100 nt. Sections were hybridized overnight at 52°C (50% deionized formamide, 0.3 M NaCl, 20 mM Tris-HCl, pH 7.4, 5 mM EDTA, 10 mM NaPO 4 , 10% dextran sulfate, 1ϫ Denhardt's, 50 g/ml total yeast RNA, 50 -7500 cpm/l 35 S-labeled cRNA probe). High stringency washes followed at 65°C in 50% formamide, 2ϫ SSC, 10 mM dithiothreitol prior to treatment with 20 g/ml RNase at 37°C for 30 min. Following washes in 2ϫ SSC and 0.1ϫ SSC for 10 min at 37°C, slides were dehydrated and coated with Kodak NTB-2 nuclear track emulsion and exposed for 4 -6 weeks in light-tight boxes with desiccant at 4°C. Photographic development was carried out in Kodak D-19. Slides were counterstained lightly with toluidine blue and analyzed using both lightfield and darkfield optics of a Zeiss Axiophot microscope. Sense control cRNA probes (see above) from nonoverlapping regions of the Nos3/Nos3as genomic region assessed background levels of hybridization signal.
Chromatin Immunoprecipitation-ChIP was performed essentially as recommended by Upstate Biotechnology, Inc., using the ChIP assay kit. Approximately 3 ϫ 10 6 cells were cross-linked in situ with formaldehyde per ChIP assay. Sonication achieved chromatin fragments ranging in size from 200 to 400 bp. 1% of total chromatin was removed prior to immunoprecipitation to serve as an input sample. Chromatin was precleared with salmon sperm DNA/protein A at 4°C with rotation for 30 min, followed by the addition of 5 g of anti-RNA polymerase II NH 2 -terminal antibody (N20, Santa Cruz Biotechnology, Santa Cruz, CA), as described (30). Immunoprecipitation was performed at 4°C for 16 h with rotation. Controls representing no antibody and control rabbit IgG (SC-2027, Santa Cruz Biotechnology) were performed in parallel. After washing, formaldehyde cross-links were reversed; samples were purified by phenol/chloroform extraction and ethanol precipitation, and DNA was suspended in 25 l of water. Real time PCR was performed as detailed above using primer pairs spanning the NOS3/NOS3AS locus (Table I). Fold enrichment was calculated by subtracting the copy number of each amplicon in the no antibody control or nonspecific control IgG immunoprecipitated fractions from the copy number present in the specific antibody-bound fraction, and dividing this by the number of copies in a dilution of the input chromatin.
siRNA-mediated Gene Silencing-siRNA duplexes targeted to sONE mRNA sequences were designed as described (31) (Table II). sONE targeted duplexes were transfected along with scrambled control siRNA (Scramble II siRNA duplex, Dharmacon) into HAoVSMC using Trans-Messenger (Qiagen, Valencia, CA). Briefly, 16 h before transfection, 1.5 ϫ 10 5 cells were plated prior to transfection with 2 g of siRNA duplexes per sample. Total cellular RNA was extracted after 72 h, reverse-transcribed, and subjected to quantitative PCR as detailed above. To monitor and optimize for transfection and silencing efficiency, cells were transfected in parallel with siRNA duplexes targeted to the nuclear membrane protein Lamin A/C (Table II). Western blots used an anti-lamin A/C antibody (Santa Cruz Biotechnology) and revealed Ͼ95% reduced levels of lamin A/C protein. Relative expression levels were calculated by normalizing target amplicons to co-amplified GAPDH. Data were then compared with similarly normalized expression levels in scrambled control transfected cells (set to 1) by one sample t test.
Overexpression of sONE-Expression plasmids encoding 1555 nt derived from 3Ј-regions of the sONE cDNA that overlap with 662 nt of eNOS sequence (sONE exons 10 -12) were constructed using standard recombinant DNA techniques. Transcription was directed by a 302-bp SV40 promoter/enhancer encoded in the pMACS K K II plasmid (Miltenyi Biotec, Auburn, CA). HUVEC were plated at a density of 2 ϫ 10 5 cells onto gelatin-coated glass coverslips. After 18 -24 h cells were transiently transfected using Effectene (Qiagen). Cells were co-trans-fected with the pDsRed2 N1 plasmid (Clontech) which encoded the Discosoma sp. red fluorescent protein for in situ detection of transfected cells. 48 h post-transfection, cells were washed with phosphate-buffered saline and fixed in 4% paraformaldehyde for 30 min followed by permeabilization in cold 100% methanol for 15 min. The cells were incubated overnight at 4°C in a 1:25 dilution of monoclonal mouse antihuman eNOS COOH-terminal antibody (Transduction Laboratories, Lexington, KY). Coverslips were subsequently washed and incubated in a 1:200 dilution of fluorescein isothiocyanate-labeled goat anti-mouse secondary antibody (Santa Cruz Biotechnology). Cells were visualized on a Bio-Rad Radiance 2100 confocal microscope (Bio-Rad) equipped with argon (514 nm) and Green He-Ne (543 nm) lasers. Images were collected below saturation level and with the use of a Kalman averaging filter and a Nikon E800 upright microscope equipped with an epifluorescent attachment. Images comprising at least 100 red fluorescent protein-positive (i.e. transfected) cells per condition were analyzed using ImageJ version 1.29 (National Institutes of Health). eNOS immunofluorescence is expressed as the mean Ϯ S.E. of the average green channel fluorescence. Overall quantitation between experiments compared the mean of the average green channel fluorescence in sONEoverexpressing (red fluorescent protein-positive) cells normalized to the green fluorescence in empty vector-transfected (red fluorescent proteinpositive) cells. A standard t test was used to test if mean fluorescence was significantly different between conditions. Alternatively, HUVEC were transfected in 100-mm tissue culture dishes. After 24 h cells were trypsinized and immunomagnetically purified using the MACSelect KK.II transfected cell selection kit (Miltenyi Biotec) according to the manufacturer's suggestions. Purified cells were plated on gelatincoated 6-well dishes and grown under standard conditions for a further 48 h. Assessment of immunomagnetic enrichment of transfected cells was performed by monitoring the expression of a co-transfected expression cassette encoding dsRED in live cells and was observed to be Ͼ90%. Total cellular protein and RNA was extracted. RNA levels for eNOS were quantitated with real time PCR as described above. For immunoblots, 20 g of protein was size-fractionated on 10% polyacrylamide gels and transferred onto polyvinylidene difluoride membranes. Protein concentration was determined using the BCA protein assay (Pierce). Western blot analysis used a 1:100 dilution of monoclonal mouse anti-human eNOS COOH-terminal antibody (Transduction Laboratories) and a horseradish peroxidase-linked secondary antibody (Pierce). Detection was performed using the SuperSignal West Pico chemiluminescent substrate (Pierce) and the Fluor-S Max MultiImager (model 170 -7720, Bio-Rad). Quantification was performed using Quan-tityOne software (Bio-Rad).

RESULTS
Characterization of an Antisense Gene to NOS3-Members of the human NOS gene family exhibit expression patterns in gonadal tissue. In many cases, unique mRNA transcripts are expressed (26). Therefore, we screened a human testes cDNA library using a full-length eNOS cDNA probe and identified 12 clones. Five of the clones, containing portions of eNOS exons, did not correspond to transcripts derived from the NOS3 gene. A comprehensive analysis of the structure of these variant mRNA species used searches of the public expressed sequence tag data bases, 5Ј-and 3Ј-RACE, and cDNA and genomic DNA clone analyses and revealed the existence of an antisense mRNA derived from a transcription unit (NOS3AS) on the opposite DNA strand from which the eNOS (NOS3) mRNA was transcribed (GenBank TM accession numbers AY316116 and AY515311). The genes were oriented in a tail-to-tail configuration, and the mRNAs encoding sONE and eNOS were complementary for 662 nt (Fig. 1a). We found that the major 2.9-kb sONE mRNA species was encoded by a 12-exon gene (NOS3AS) spanning 6.5 kb of genomic DNA (Fig. 1a, Tables III-VI). We noted that the human sONE mRNA contains four regions that were perfectly complementary to the eNOS mRNA ( Fig. 1, a  and b) and that sONE is complementary to regions spanning portions of exons 23-26 and representing 182 amino acids of human eNOS. In contrast to NOS3/NOS3AS, the overlap of most sense-antisense RNA interactions mainly involves untranslated regions of the RNAs (24,33). We performed experiments to determine whether RT-PCR analyses would detect both mRNA species. Southern blot analysis of RT-PCR amplifications across exons 25 and 26 of human eNOS using RNA isolated from a variety of human tissues produced the expected 403-bp product when hybridized with an eNOS-specific probe. In some of these tissues (testis, brain, and placenta) these same primers co-amplified a 323-bp fragment that hybridized to a probe specific for sONE (Fig. 1c). When we primed cDNA syn-thesis with strand-specific primers, we detected RNA corresponding to eNOS when cDNA was primed with eNOS-primers but not with eNOS ϩ strand primers (Fig. 1d, upper panel). Conversely, we detected sONE amplification products when cDNA was primed with eNOSϩ strand primers but not eNOSϪ strand primers (Fig. 1d, lower panel). Results of strand-specific RT-PCR Southern blot hybridizations were consistent with amplifications from random-primed cDNA pools both in terms of tissue distribution and abundance.
We quantified eNOS and sONE RNA abundance in tissues using real time PCR. Primer pairs used for detection were located in nonoverlapping regions of the two transcription units. Again, we primed cDNA synthesis using random primers, sense or antisense strand-specific primers (Fig. 1e). We detected significant amounts of eNOS amplification products only in randomly primed or eNOSϪ strand-primed cDNA. sONE amplification products were detected only in randomly primed or eNOSϩ strand-primed cDNA. The distribution of eNOS and sONE mRNA expression in tissues as determined by real time PCR closely mirrored that observed in Southern blot analysis. Taken together, these data indicate the existence of mRNA species arising from two overlapping transcription units located on opposite strands of the same genomic region.
We also characterized murine sONE cDNA and genomic sequences from murine chromosome 5 (GenBank TM accession numbers AY515312 and AY515313) and found that the major 2.1-kb sONE mRNA species was encoded by a 9-exon gene (Nos3as) spanning 3.7 kb of genomic DNA (Fig. 1a and Tables  III and IV). Again the transcript is antisense to Nos3 cDNA sequences (34) and shows a tail-to-tail genomic configuration. We noted two differences between the human and murine transcription units. The region of overlap is smaller in mouse than in humans, up to 189 nt in mouse versus 662 nt in humans. Also, the genomic organization of Nos3as differs from NOS3AS given that overlap regions represent the terminal exon of mouse sONE versus the final three in human sONE.
Open reading frames of 363 and 362 amino acids were evident in the major human and murine sONE cDNAs, which exhibited 87% amino acid identity. Of interest, we noted that the mRNA for human sONE is unusual because the predicted translation termination codon was found within exon 8, with the 3Ј-UTR of sONE comprising five exons. The final three of these exons contain regions of eNOS overlap. This latter finding was surprising given that RNAs with translation termination signals upstream of the terminal exon can be selectively targeted by nuclear mediated nonsense decay pathways (35). We also noted the existence of minor RNA species in varied human and murine tissues. In human tissues, we noted evidence from expressed sequence tags, RACE, and RT-PCR for a longer version of exon 9 and retention of intron 10, both of which would further extend the degree of overlap with NOS3derived transcripts (data not shown). In the current work we have focused upon the functional relevance of the 662-nt overlap region. Most importantly, larger NOS3AS and Nos3as mRNA species were also evident in human and murine tissues, especially testis. This larger mRNA species represented additional sequence at the 5Ј terminus of the mRNA and predicted polypeptides with NH 2 -terminal extensions. Analyses of NOS3AS and Nos3as genomic structure predicted the presence of upstream alternative promoters in both species, producing primary RNA transcripts that spanned 12.3 and 8.0 kb, respectively (Fig 1a, Tables III, and IV). The larger human and murine ORFs predict polypeptides of 924 and 922 amino acids that share 86% amino acid identity. Additionally, the larger NOS3AS protein variant shows 89% identity to a hypothetical Rattus norvegicus protein (XP_231278). The short and long predicted proteins from human, mouse, and rat share conserved domains with Apg9p from yeast (www.sanger.ac.uk/ Software/Pfam/index.shtml). Apg9p/Cvt7p was the first integral membrane protein of the Apg/Cvt pathways identified in the yeast Saccharomyces cerevisiae and is involved in autophagic vesicle formation (36). Similar to yeast Apg9, each of these mammalian proteins appears to have between 6 and 8 hydrophobic transmembrane domains (www.ch.embnet.org/software/TMPRED_form.html). Also, BLASTp (www.ncbi.nlm.nih. gov/BLAST) analysis of the 924-amino acid human NOS3AS protein against the nonredundant data base identified a hypothetical human protein, FLJ22169, found at human chromosome 2q35 that shares 29% identity (37 1. sONE is a naturally occurring cis antisense transcript to eNOS. Schematic representation of the human NOS3/NOS3AS (7q36) (a) and murine Nos3/Nos3as (5A3) genomic loci (b) illustrating the deduced intron/exon structure. NOS3AS and Nos3as are transcribed in an antisense orientation with respect to NOS3 and Nos3. Putative coding and untranslated regions are shown in black and gray, respectively. CpG islands are indicated. c, schematic representation of reverse transcriptase-PCR co-amplification of eNOS and sONE mRNA. Positions of PCR primers with respect to eNOS and sONE exons are indicated by black arrows. Amplicons corresponding to eNOS and sONE mRNA are illustrated schematically above and below the genomic locus. d, PCR products from cDNA pools that were random-primed (random) or primed with oligonucleotides corresponding to the antisense strand (eNOS Ϫ strand) or coding strand (eNOS ϩ strand) of eNOS were size-fractionated, blotted, and probed with 32 P-labeled eNOS-or sONE-specific oligonucleotide probes (e2526 and s1011). The eNOS and sONE amplicons are 403 and 323 bp in length, respectively. e, real time PCR quantification of eNOS and sONE mRNA in cDNA pools primed using random, eNOSϪ, or eNOSϩ strand-specific primers. Abundance of eNOS and sONE mRNA in the pools is expressed as copies per g of input RNA.
to define which tissues and cell types express sONE, and we compared this with the expression of eNOS. The major 2.9-kb sONE mRNA detected on Northern blots containing poly(A) ϩ mRNA and using multiple cDNA probes was consistent with the deduced structure of the sONE mRNA ( Fig. 1a and Tables III-VI), whereas a minor abundance 5.0-kb mRNA was consistent with upstream alternative promoter usage and five upstream exons (AS1-AS5, Tables V and VI). In human tissues, we detected broad but modest expression. sONE mRNA expression was especially prominent in placenta. This overall tissue distribution can be contrasted with eNOS, the results of which are consistent with previously reported findings (Fig. 2b) (25). The majority of the eNOS signal reflects endothelial cell-specific expression. In murine tissues we noted the expression of a 2.0-kb sONE transcript in heart, brain, and placenta and a larger transcript in testis (data not shown). We used real time PCR to quantitate sONE and eNOS mRNA in a variety of human tissues (Fig. 2, c and d). Absolute copy number expressed relative to total cellular RNA, normalized to the geometric mean of a variety of control RNAs (GAPDH, cyclophilin A, and 18 S rRNA), indicated patterns of expression that were consistent with findings from other RNA studies. sONE mRNA was most abundant in placenta, testes, brain, and pituitary, with modest but measurable amounts present in all tissues tested. We used RT-PCR and found expression of the mRNAs derived from the human and murine genes that encode   a Intron type: 0 indicates a splice junction between codons; I indicates a splice junction after the first nucleotide of a codon, and II indicates a splice junction that occurs after the second nucleotide. Amino acids with interrupted codons were assigned to the exon containing two of the three codon nucleotides.
FLJ22169 in a broad number of human and murine tissues (data not shown). Most importantly, we failed to detect any complementarity of the gene that encodes the FLJ22169 protein at 2q35 with NOS3 genomic sequences at chromosome 7q35-36. This indicates that although the NOS3AS and FLJ22169 proteins show sequence relatedness, it would be unlikely that the mRNA expressed from the FLJ22169 gene would interact in trans in a sense/antisense fashion with NOS3 transcripts. a Intron type: 0 indicates a splice junction between codons; I indicates a splice junction after the first nucleotide of a codon, and II indicates a splice junction that occurs after the second nucleotide. Amino acids with interrupted codons were assigned to the exon containing two of the three codon nucleotides. The abundance of sONE and eNOS mRNA was calculated relative to the geometric mean of GAPDH, cyclophilin, and 18 S RNA that was amplified from the same cDNA pools and expressed per g of input RNA. Comparable tissue distribution was also evident when data were not normalized. Data points represent the mean Ϯ S.E. of cDNA pools prepared from 3 to 6 different RNA samples. PBL, peripheral blood lymphocytes.
Mammalian mRNAs that form sense/antisense pairs can exhibit reciprocal expression patterns (24,38). Given that the sONE and eNOS transcripts have significant regions of overlap, and that we detected both sONE and eNOS in tissues, we were interested in defining whether these two mRNA species could co-exist in one cell type. We performed real time RT-PCR analysis of sONE in eNOS-expressing and eNOS-nonexpressing cultured human cell types (Fig. 3, a-c) using random or strand-specific primed cDNA pools. The mRNA for eNOS is known to be endothelially restricted in expression, especially compared with other NOS isoforms (25). We found that endothelial cells expressed abundant eNOS but only very low steady-state mRNA levels for sONE (ratio of mRNAs for sONE/ eNOS abundance very low). In HUVEC, the ratio of sONE/ eNOS was 0.002 Ϯ 0.003 (n ϭ 3). Conversely, sONE mRNA was much more abundant in cultured cells that did not express appreciable amounts of eNOS. For example, the ratio of sONE/ eNOS was 210 Ϯ 35 (n ϭ 3) in HAoVSMC. As expected we only detected significant levels of eNOS mRNA in cDNA pools that were random-primed or primed with eNOSϪ strand primers.  Fig. 2. Abundance is represented as the mean Ϯ S.E. of three or more independently prepared cDNA pools representing independent RNA isolations. For HUVEC and HAoVSMC primary cells, determinations were performed by using cDNA prepared from multiple separate donors. c, eNOS and sONE mRNA abundance was quantified in primary human vascular cells with real time PCR amplification of random and strand-specific primed cDNA pools as in Fig. 1e. d-i, sONE and eNOS mRNA were localized to nonoverlapping cell populations in ED 11.5 (d and e, ϫ360) and ED 12.5 (g and h, ϫ2.5) murine uteroplacental tissues using in situ hybridization with ␣-35 S-UTP antisense riboprobes (dark field images). f, ϫ360, and i, ϫ10, bright field images of d and h, respectively.
Conversely, we only detected significant amounts of sONE in random-primed or eNOSϩ strand-primed cDNA pools.
When we performed RT-PCR using murine tissues (data not shown) or in situ cRNA hybridization with 35 S-labeled eNOS and sONE cRNA probes, we detected expression patterns in murine tissues that paralleled those in human tissues. Most importantly, in all of the murine tissues we examined by in situ, we failed to define co-localization of eNOS and sONE. For example, as shown in Fig. 3, we detected eNOS and sONE mRNA expression in uteroplacental tissues in nonoverlapping cell populations. eNOS mRNA was detected in fetal-derived trophoblast giant cells lining the blood sinuses that contain maternal blood in the murine hemochorial placenta, as described previously by others (Fig. 3h) (39). These cells are analogous to extravillous cytotrophoblasts of the human placenta and are known to express other angiogenic and vasoactive substances, such as vascular endothelial growth factor and adrenomedullin (39). We detected sONE mRNA in the maternal component of uteroplacental tissues in the luminal epithelium of the uterus (Fig. 3d).
Transcription of NOS3 and NOS3AS Does Not Correlate with Steady-state mRNA Expression-To explain the reciprocal expression pattern of sONE and eNOS in cell types, both in vitro and in vivo, we next assessed whether transcription of the sONE and eNOS genes was cell type-specific. We performed nuclear run-off analyses using cDNA and genomic probes derived from overlapping and nonoverlapping regions of the NOS3 and NOS3AS transcription units (Fig. 4, a-c). We detected hybridization signals throughout the NOS3 and NOS3AS transcription units regardless of cell type, indicating that both genes are transcriptionally active in all of the cell types we tested. This was a surprising finding given that none of the cell types we tested, with the exception of endothelial cells, expressed appreciable amounts of eNOS. Most importantly, we found robust signal for probes specific to the nonoverlapping regions of NOS3AS in HUVEC. These cells did not express appreciable levels of sONE.
By using ChIP, we found that the pattern of RNA polymerase II (RNA pol II) distribution over the NOS3 and NOS3AS genes correlated with that of the nuclear run-off analyses in all of the cell types examined (HUVEC, HAoVSMC, HeLa, HepG2, and JEG-3). Data derived from three independent experiments for HUVEC and HAoVSMC are shown in Fig. 4, d and e. The ChIP signal for RNA pol II across the NOS3/NOS3AS transcription units was significantly above background compared with transcriptionally quiescent regions of the genome, control IgG, and no antibody control levels. Findings in HepG2, HeLa, and JEG-3 were similar to HAoVSMC (data not shown). These findings are consistent with the conclusion that the nascent transcripts detected with nuclear run-off studies are derived from the transcriptional activity of RNA pol II.
We then performed nuclear run-off assays in HUVEC and HAoVSMC with single-stranded probes in order to define the orientation of transcription and found that nascent radiolabeled transcripts hybridized to strand-specific probes for both eNOS and sONE. A representative HUVEC experiment is shown in Fig. 4f. Results from three independent experiments performed in HUVEC and HAoVSMC are summarized in Fig.  4g. Our findings provide clear evidence for transcription of both the NOS3 and NOS3AS genes in both cell types. These results also indicate that the double-stranded DNA nuclear run-off signal observed using probes representing NOS3/ NOS3AS overlap regions represent transcription derived from both strands of DNA in both endothelial and nonendothelial cell types. Therefore, transcription of these two genes is not cell-restricted. Most importantly, transcription of the overlapping genes did not correlate with steady-state accumulation of their mRNAs and implicates a post-transcriptional interaction.
Evidence for Functional Interactions between NOS3-and NOS3AS-derived Transcripts-Considering that eNOS-nonexpressing cells can express appreciable levels of sONE mRNA, whereas eNOS-expressing cells contained very little, we hypothesized that sONE might be a regulatory mRNA that functions to silence or down-regulate eNOS expression. We first asked whether decreasing sONE mRNA levels modified eNOS mRNA levels in cells that expressed sONE. We designed double-stranded RNA oligonucleotide duplexes against unique nonoverlapping regions of the sONE mRNA to silence sONE expression by RNA interference. HAoVSMC were transiently transfected with siRNA duplexes targeted to sONE. In parallel studies we transiently transfected control cultures with scrambled control duplexes. As shown in Fig. 5, when we decreased sONE mRNA expression to 48 Ϯ 6% (n ϭ 5, p Ͻ 0.001) of control values we observed a corresponding 2.5 Ϯ 0.7-fold (n ϭ 5, p Ͻ 0.05) increase in eNOS mRNA expression relative to scrambled control siRNA-transfected cells. We take these data to indicate that sONE plays a functional role in down-regulating steady-state levels of eNOS mRNA in nonexpressing cells.
We then overexpressed the sONE cDNA in eNOS-positive HUVEC (Fig. 6a). HUVEC were co-transfected with the pDsRED2N1 plasmid as a transfection marker. Levels of eNOS were quantitated by immunoconfocal microscopy in cells transfected with empty expression vector or an expression vector that encoded 1555 nt derived from overlapping regions of the sONE 3Ј-UTR. The staining for eNOS protein in HUVEC was found to decrease when sONE overlap regions were overexpressed (Fig. 6b), indicating that the sONE mRNA negatively regulates eNOS. Overexpression of sONE overlap regions had little impact on eNOS RNA levels in HUVEC as measured by real time RT-PCR but had marked effects on eNOS protein expression in the same cells as measured by Western blot (Fig. 6c).
To investigate further the translational regulation of eNOS by sONE, we transfected both HUVEC and HepG2 cells with in vitro transcribed chimeric firefly luciferase/eNOS reporter RNAs with or without 1555 nt of in vitro transcribed sONE overlap RNA. Translation of ffLUC/eNOS co-transfected with sONE RNA fell to 39.5 Ϯ 3.9% (n ϭ 3, p Ͻ 0.05) in HUVEC as compared with ffLUC/eNOS co-transfected with a nonspecific, noncomplementary RNA (␤-galactosidase) (Fig. 6d). Comparable findings were evident in HepG2 cells (data not shown). The stability of the transfected reporter RNAs in HUVEC was equivalent between experimental conditions over the 6 h as measured by real time RT-PCR (data not shown). Taken together these studies indicate that the antisense sONE transcript can prevent the translation of the eNOS protein from the eNOS-encoding mRNA or the translation of a reporter protein from a chimeric RNA containing regions of sense-antisense RNA interaction.
Models of regulation of eNOS in vascular endothelium have implicated important contributions from changes in mRNA stability. It has been reported recently that the histone deacetylase inhibitor TSA decreased eNOS mRNA in HUVEC. A post-transcriptional mechanism was implicated in this response although the molecular basis for this alteration has yet to be defined (17). Therefore, we treated HUVEC with maximal doses of TSA (500 nM) and confirmed that eNOS mRNA levels significantly decreased by 24 h (Fig. 7). Although not shown, TSA failed to modify HUVEC eNOS nuclear run-off signal, which is consistent with the data of others (17) that TSA-mediated decreases in eNOS mRNA in endothelial cells are because of post-transcriptional interactions. Of great interest, we noted that TSA reproducibly increased expression of sONE mRNA in HUVEC. This increase was evident as early as 4 h and therefore preceded the observed decrease in eNOS mRNA. Future studies will be necessary to define the relationship between TSA-induced decreases and increases in eNOS and sONE, respectively. FIG. 4. Transcriptional activity of the NOS3/NOS3AS locus. a, representative nuclear run-offs from HUVEC and HAoVSMC. 32 P-Labeled nascent nuclear transcripts prepared from HUVEC or HAoVSMC were hybridized to immobilized probes spanning the eNOS/sONE locus. b and c, nuclear run-off signals from HUVEC and HAoVSMC. Signals from a minimum of three independent nuclear run-off assays per cell type were quantitated using digital densitometry. Signals were corrected for background (hybridization to pBluescript) and normalized to GAPDH. d and e, RNA pol II occupancy of genomic regions was assessed using real time PCR amplification of DNA purified from anti-RNA pol II ChIP. Fold enrichment represents the mean Ϯ S.E. of determinations from a minimum of three anti-RNA pol II ChIPs from each cell type. f, representative single-stranded nuclear run-offs from HUVEC. 32 P-Labeled nascent nuclear transcripts prepared from HUVEC or HAoVSMC were hybridized to single-stranded probes spanning the eNOS/sONE locus. g, schemata of single-stranded DNA probes used to address transcription events on separate genomic DNA strands over the NOS3/NOS3AS locus and a summary of transcriptional activity representing data from 3 to 5 separate single-stranded run-off experiments.

FIG. 5. Reduction of sONE mRNA in
HAoVSMC increases eNOS. a, sONE mRNA levels were assessed after transfection of siRNAs targeted to sONE by amplification of cDNA prepared from transfected cells from independent donors. mRNA levels were normalized to coamplified GAPDH and compared with normalized sONE levels in cells transfected with scrambled sequence control siRNA duplexes. * indicates statistical significance at p Ͻ 0.05. b, increase in eNOS mRNA transcript levels in HA-oVSMC after siRNA-mediated knockdown of sONE mRNA. eNOS mRNA levels were assessed and normalized as in a. * indicates statistical significance at p Ͻ 0.05. Experiments were performed five times using HAoVSMC prepared from multiple independent donors.
FIG. 6. Reduction of eNOS expression in endothelial cells overexpressing sONE overlap region. a, representative confocal photomicrographs of HUVEC co-transfected with pDsRed2N1 and either an sONE expression vector or empty vector (ϫ20 magnification). b, representative decrease of eNOS immunofluorescence assessed by digital image analysis of green channel fluorescence in transfected (red) cells. eNOS immunofluorescence is expressed as the average Ϯ S.E. of the mean green immunofluorescence in Ͼ100 cells per condition. * indicates statistical significance at p Ͻ 0.05. c, eNOS mRNA abundance was quantified by real time PCR of randomly primed cDNA prepared from pools of immunomagnetically purified HUVEC transfected with either empty vector or an sONE expression vector. mRNA levels were normalized to co-amplified GAPDH and compared with cells transfected with empty vector. Data are presented as the mean Ϯ S.E. of four experiments. Representative Western blot analysis of total protein isolated from purified transfected HUVEC is shown in the inset. d, translation efficiency of in vitro transcribed, capped luciferase/eNOS reporter RNAs transiently translated in HUVEC was assessed in the presence or absence of cotransfected sONE overlap RNA (sONE overlap). Firefly luciferase (ffLUC) activity was normalized for protein content and for Renilla luciferase (rrLUC) activity and expressed relative to the activity of the respective reporter co-transfected with nonspecific, noncomplementary RNA (non-spec RNA). Data are expressed as the means Ϯ S.E. of three independent experiments, each performed in triplicate. * indicates statistical significance at p Ͻ 0.05.

DISCUSSION
We have identified a natural antisense transcription unit (NOS3AS) that is oriented in a tail-to-tail overlapping genomic configuration with NOS3. Most surprisingly, rates of transcription from the NOS3 and NOS3AS genes did not correlate with steady-state mRNA levels thus implicating post-transcriptional processes in the regulation of gene expression. In this study, we provide several lines of evidence to implicate functional sense-antisense RNA interactions in the regulation of eNOS expression. First, levels of eNOS mRNA increased in vascular smooth muscle cells following RNA interference knockdown of sONE. Second, heterologous expression of sONE overlap regions in vascular endothelial cells decreased eNOS protein levels. Finally, because eNOS transcripts are known to be regulated at the post-transcriptional level in vascular endothelial cells by a variety of exogenous cellular stimuli, we assessed the expression of the sONE mRNA in a representative model. TSA is a histone deacetylase inhibitor that has a number of biologically important effects on vascular endothelial cells (40,41). We considered the possibility that TSA-induced decreases in eNOS transcripts might be associated with changes in sONE expression. We found that TSA increased sONE mRNA expression in endothelial cells, a response that preceded the observed decrease in eNOS mRNA transcripts. Our findings provide a newer perspective on the mechanisms of regulation of the eNOS mRNA, a transcript that is known to be regulated at the post-transcriptional level by a variety of cellular stimuli. Mechanism-based studies have, to date, focused upon the effect of these stimuli on cis-RNA interactions with ribonucleoprotein complexes (12). Taken together the data presented in this paper support a model for a functional interaction between NOS3 and NOS3AS at the post-transcriptional level. Because eNOS is predominantly expressed in the endothelium, we conclude that cell-specific expression of eNOS (encoded by the NOS3 gene) is regulated by sONE, an antisense mRNA transcript (encoded by the NOS3AS gene).
How can antisense RNAs influence expression of the sense RNA? Antisense RNAs can influence the earliest nuclear events in mRNA biogenesis, including control of chromatin structure, transcription initiation, and splicing (42)(43)(44). For example, in resting peripheral blood human T-cell levels of eIF2␣ mRNA and protein are rate-limiting for translation. During mitogenic T-cell activation, levels of eIF2␣ mRNA increase 50-fold. Transcription of an antisense mRNA that initiates within intron 1 of the eIF2␣ gene negatively correlates with expression of eIF2␣ and has been implicated in the intranuclear destabilization of nuclear eIF2␣ RNA species (45). In contrast to eIF2␣, where transcription of the antisense gene negatively correlates with mRNA levels of the sense gene, we found that NOS3 and NOS3AS are both transcriptionally active in vascular endothelial cells. The genes are also transcrip-tionally active in other cell types, such as vascular smooth muscle cells.
Because transcription from the NOS3 and NOS3AS genes did not correlate with steady-state mRNA levels, this implicated post-transcriptional processes in the regulation of eNOS gene expression. This is a newer concept in how an endothelial gene is expressed in a cell-restricted fashion. It is of interest that eNOS promoter/reporter constructs exhibit endothelial cell-specific expression in insertional transgenic murine models (46,47). These random integration events only assess the function of eNOS genomic regions. The constructs would be free from tail-to-tail transcriptional interactions with the antisense gene and also do not include RNA sequences that would interact with sONE mRNA species because native eNOS 3Ј-ORF and 3Ј-UTR sequences were not included in the genomic constructs (46,47). It is possible that NOS3 transcription in vascular endothelial cells is qualitatively, but not quantitatively, unique in this cell type. Studies with other genes have suggested that during transcription initiation, or preinitiation complex formation, RNA pol II can recruit and differentially associate with factors that accompany RNA pol II during the elongation and termination phases of transcription (48,49). Some of these RNA pol II-associated factors show variations in their level of expression within and between cell types (50). Also, post-translational modifications of the carboxyl-terminal domain of RNA pol II can show cell-type specificity (51). Future studies will need to address the cell-type specificity of RNA poll II modifications over the NOS3/NOS3AS transcription units, especially at the eNOS proximal promoter.
A further model to contemplate when addressing the relationship between sense-antisense RNA interactions and cellspecific gene expression would be differences between cell types in transcriptional elongation. We do not favor this consideration given that we found no important differences between cell types in nuclear run-off signals or pol II distribution along genomic regions that spanned exon 2 to the terminal exon 26 of NOS3. These findings also provide evidence against cell-specific attenuation of transcriptional elongation, as described for adenosine deaminase (52,53), where pol II pauses or arrests in proximal portions of the gene in nonexpressing cells. Here nuclear run-off analysis revealed the existence of positive signals with 5Ј-but not 3Ј-probes. In a further model, an antisense mRNA has been implicated in the regulation of alternative splicing events for cErbA␣, which encodes the 3,5,3Ј-triiodo-Lthyronine (T3) thyroid hormone receptor (54). The Rev-ErbA␣ mRNA overlaps with and negatively regulates one of two transcripts produced from the cErbA␣ gene (cErbA␣2) by inhibiting splicing of cErbA␣2.
We have no evidence to indicate that sense-antisense RNA interactions participate in the regulation of eNOS mRNA expression through monoallelic effects, as described for other FIG. 7. sONE and eNOS mRNA expression in endothelial cells is reciprocally regulated by histone deacetylase inhibition. Confluent monolayers of HUVEC were treated with TSA (500 nM) for 4 and 24 h. eNOS (a) and sONE mRNA (b) expression were measured by real time PCR. One of three representative experiments is shown. Shown are the mean Ϯ S.E. of triplicate determinations. * indicates statistical significance at p Ͻ 0.05 versus vehicle-treated control.
genes (33). This concept was an important one to consider given that recent surveys of human gene expression have demonstrated that allelic differences in gene expression are more common than previously appreciated (55)(56)(57). Results reported here also indicate that cell type-specific monoallelic expression cannot explain the cell type-specific expression patterns in steady-state mRNA levels of eNOS versus sONE. Using singlestranded probes we found evidence for NOS3 and NOS3AS transcription in the same genomic region in all of the expressing and nonexpressing cell types we studied.
Finally, we will need to consider a model involving the terminal stages of transcription and primary RNA transcript processing because the antisense transcription unit (NOS3AS) is oriented in a tail-to-tail overlapping genomic configuration with NOS3. This model predicts that eNOS transcripts would accumulate preferentially in endothelial cells, whereas sONE transcripts predominate in nonendothelial cell types because the terminal processing of the RNAs differ in a cell-specific fashion. In this regard, studies have described mRNA species that are actively stabilized in a cell-type specific fashion. For example, modulation of Xist mRNA stability is the primary determinant of developmental up-regulation of the X i -derived transcript (58). Xist is transcribed at approximately equivalent rates in two cell types, namely XX embryonic stem cells and XX somatic cells. However, because the Xist RNA is highly unstable in XX embryonic stem cells relative to XX somatic cells, with respective half-lives of 30 -45 min and 5-7 h (58), the RNA only accumulates in the one cell type. In the case of ␣2-globin, ribonucleoprotein complexes protect the mRNA species from degradation in the cytoplasm (59,60). The RNAbinding proteins functionally implicated in this process (e.g. heterogeneous nuclear ribonucleoprotein E1 and E2 family members) are reported to associate with the RNA transcript prior to nuclear-cytoplasmic trafficking. It is not known how sense-antisense RNA interactions would be modified by these ␣2-globin ribonucleoprotein complexes.
Here we have described a related genomic organization for the mouse ortholog Nos3as. However, the studies reported here addressed interactions between the two overlapping human genes. Therefore, it remains to be determined whether the murine genes also interact at the post-transcriptional level, especially because the human and murine genes for sONE show key differences in the organization of genomic regions that encode overlapping 3Ј-UTR regions.
We anticipate that the human NOS3AS gene has two functional roles. The current work focused upon the novel finding that RNA transcripts derived from NOS3AS functionally interact with NOS3-derived transcripts. Our analyses indicated that the NOS3AS and Nos3as genes are highly conserved in upstream regions and that the genes produce coding mRNAs, a property shared by most antisense partners (24). Future studies will be required to define whether the short and long versions of the predicted NOS3AS-and Nos3as-derived proteins are translated and, if so, whether they have a functional role in autophagy given the conservation of protein domains with yeast Apg9p. Autophagy is an intracellular degradation or recycling process that mediates the turnover of cellular components, which is mediated by membrane trafficking of the unique double-membrane structures known as autophagosomes. The process is broadly relevant in eukaryotic cells and is induced by a number of stimuli including cellular nutrient deprivation, exogenous stimuli (e.g. hypoxia), and developmental signals, among others (61). Because autophagy is an evolutionarily conserved pathway for degrading long lived proteins and cytoplasmic organelles, the finding that NOS3AS encodes proteins related to yeast Apg9p is of interest. Consistent with the convention of the Human Genome Nomenclature Committee, we have designated the NOS3AS-derived protein and the related FLJ22169 protein the aliases APG9L2 and APG9L1, respectively, and await future definition of their functional properties (62).
The number of overlapping but oppositely directed transcription units in the human genome is increasing, and the number now annotated is over 300 (33), 38 of which are on human chromosome 7 (27). Lehner et al. (24) found that ϳ52% of the human sense-antisense exist in cis, with 72% representing tail-to-tail genomic configurations. Analysis of the regions of complementarity between cis or trans sense/antisense pairs revealed median lengths of 139 and 121 nt, respectively. Therefore, the 662-nt overlap between NOS3 and NOS3AS is uncommonly large. Of the natural antisense transcription units, only a limited number have experimental evidence to support a functional role for sense-antisense interactions (33,38,42). However, the variety of regulatory roles that has been attributed to these sense/antisense pairs continues to grow as follows: imprinting or monoallelic expression, RNA interference, RNA editing, translational regulation, and activation of innate defense signaling pathways, such as PKR among others (24,33,63). In the snail Lymnaea stagnalis it is of interest that a transcribed pseudogene for an nNOS equivalent participates in sense-antisense RNA interactions. By using an in vitro translation assay, it was demonstrated that an antisense region of the pseudogene transcript prevented the translation of nNOS protein from the nNOS-encoding mRNA (64,65). Also, it is now known that RNA can feed back to direct epigenetic modifications of the genome, at least in lower species (66,67). The results of our study indicate that an antisense mRNA (sONE) participates in the post-transcriptional regulation of eNOS. This is the first example of an antisense mRNA implicated in the regulation of an endothelial cell-specific gene.