Common ADRB2 Haplotypes Derived from 26 Polymorphic Sites Direct β2-Adrenergic Receptor Expression and Regulation Phenotypes

Background The β2-adrenergic receptor (β2AR) is expressed on numerous cell-types including airway smooth muscle cells and cardiomyocytes. Drugs (agonists or antagonists) acting at these receptors for treatment of asthma, chronic obstructive pulmonary disease, and heart failure show substantial interindividual variability in response. The ADRB2 gene is polymorphic in noncoding and coding regions, but virtually all ADRB2 association studies have utilized the two common nonsynonymous coding SNPs, often reaching discrepant conclusions. Methodology/Principal Findings We constructed the 8 common ADRB2 haplotypes derived from 26 polymorphisms in the promoter, 5′UTR, coding, and 3′UTR of the intronless ADRB2 gene. These were cloned into an expression construct lacking a vector-based promoter, so that β2AR expression was driven by its promoter, and steady state expression could be modified by polymorphisms throughout ADRB2 within a haplotype. “Whole-gene” transfections were performed with COS-7 cells and revealed 4 haplotypes with increased cell surface β2AR protein expression compared to the others. Agonist-promoted downregulation of β2AR protein expression was also haplotype-dependent, and was found to be increased for 2 haplotypes. A phylogenetic tree of the haplotypes was derived and annotated by cellular phenotypes, revealing a pattern potentially driven by expression. Conclusions/Significance Thus for obstructive lung disease, the initial bronchodilator response from intermittent administration of β-agonist may be influenced by certain β2AR haplotypes (expression phenotypes), while other haplotypes may influence tachyphylaxis during the response to chronic therapy (downregulation phenotypes). An ideal clinical outcome of high expression and less downregulation was found for two haplotypes. Haplotypes may also affect heart failure antagonist therapy, where β2AR increase inotropy and are anti-apoptotic. The haplotype-specific expression and regulation phenotypes found in this transfection-based system suggest that the density of genetic information in the form of these haplotypes, or haplotype-clusters with similar phenotypes can potentially provide greater discrimination of phenotype in human disease and pharmacogenomic association studies.


Introduction
The b 2 -adrenergic receptor (b 2 AR) is a member of the G-protein coupled receptor superfamily (GPCR), and is expressed on virtually all cell-types [1]. The catecholamines epinephrine and to a lesser extent norepinephrine are the endogenous agonists which activate these receptors as part of the sympathetic nervous system. Of particular interest for therapeutic purposes have been b 2 AR expressed on smooth muscle cells of the airway which promote bronchodilatation and those expressed on cardiomyocytes which promote cardiac contractility. bagonists acting on airway smooth muscle b 2 AR, and non-selective bblockers acting on b 2 -and b 1 AR on cardiomyocytes, are standard treatments of obstructive lung disease and heart failure, respectively.
The b 2 AR gene (ADRB2), which is localized to 5q31-32, is intronless and encodes a protein of 413 amino acids that has a seven transmembrane topology typical of the superfamily. The amino-terminus is extracellular while the carboxy-terminus is intracellular. Ligand binding occurs in a ''pocket'' formed by residues of the transmembrane domains. The substantial interindividual variability in the responses to b-agonists [2] and antagonists [3] has prompted examination of the gene for common polymorphisms. Our identification [4] of coding polymorphisms of the b 2 AR in 1993 represented the first report of discreet nonsynonymous polymorphisms of any GPCR and prompted a host of subsequent clinical association studies [5,6]. The two common non-synonymous single nucleotide polymorphisms (SNPs, Table 1) are at nucleic acid 46 resulting in amino acid position 16 being either Arg or Gly, and at nucleic acid 79 resulting in amino acid position 27 being Glu or Gln [4]. These polymorphisms are localized to the extracellular amino-terminus, and have a subtle effect on agonist-promoted downregulation of receptor expression when studied in a cell-based system where transfections of vectors containing only the open reading frames were carried out [7]. Virtually all association studies for pharmacogenomic effects have utilized these coding polymorphisms with variable or inconsistent results (reviewed in [5,8]). These inconsistent associations have prompted examination of the promoter, 59UTR, and 39UTR of ADRB2 for additional polymorphisms. Initial studies by our group indicated multiple polymorphisms in these other regions, many of which were not in significant linkage disequilibrium with the coding polymorphisms [9][10][11]. Thus there is the potential for a more precise ADRB2 genetic signature for association studies, which may be particularly important if these other polymorphisms affect receptor expression or agonist mediated regulation of the receptor. Of interest is the potential for interaction between polymorphisms, such as promoter polymorphisms that alter transcription and coding polymorphisms that alter protein stability, so that the net effect could be a composite of multiple variants on the phenotype. Based on our previous studies of this intronless gene, there appear to be 17 common SNPs in the 59 upstream region (3500 bases 59 of the initiator ATG), 7 coding SNPs (3 nonsynonymous), and one SNP and a variable poly-C tract in the 39UTR (Table 1). In subjects of European or African ancestry, these are found to be arranged into eight haplotypes with frequencies of .0.05 in at least one racial group [11]. These common haplotypes have been named based on our initial haplotypes that only considered the 59-flanking and coding regions [9], and are denoted I-1, II-1, IV-1, IV-2, IV-3, IV-4, VI-1 and VI-2 ( Table 1). The cellular phenotypes of these combinations of SNPs, based on the common haplotypes, have not been determined. Because of the intronless nature of ADRB2, ''whole-gene'' transfections are possible and provide an opportunity to ascertain the effects of the combination of SNPs, as they appear in nature, on cellular phenotypes. The current work utilizes this approach, where we identify phenotypes which have not been previously appreciated by studying individual SNPs or limited groups of SNPs.

Results and Discussion
The constructs utilized for transfection consisted of 5,580 bp of ADRB2 (23685 to +1895) which represents contiguous promoter, 59UTR, coding and 39UTR sequence up to the poly-A termination site. Site-directed mutagenesis and digestion-ligation reactions were utilized to construct the eight haplotypes, which were verified by sequencing ( Table 1). The final constructs were cloned into a modification of the pcDNA3.1(+) expression vector in which the CMV promoter was removed (denoted pcDNA3.1(+)/DCMV, see Methods). Thus expression of the b 2 AR was driven by its own promoter, and influences of promoter, 59UTR, coding and 39UTR polymorphisms, in their appropriate context (i.e., the eight haplotypes) could be ascertained. Studies were performed in transfected COS-7 cells. Two major phenotypes were considered: baseline b 2 AR protein expression, and, agonist-promoted downregulation of receptor protein expression. b 2 AR expression was ascertained using a highly quantitative 125 I-cyanopindolol ( 125 I-CYP) radioligand binding assay. In the absence of transfection, such binding revealed ,10 fmol/mg expression in COS-7 cells, while transfection with the whole-gene vectors provided for ,1,000 fmol/mg human b 2 AR expression. Table 1. Localization of the common polymorphisms of ADRB2.
Shown are the common haplotypes (allele frequency.0.05 in either white subjects of European descent or black subjects of African descent. Haplotype names, and their prevalence in these populations, are given in references [8] and [10]. doi:10.1371/journal.pone.0011819.t001 The results from these transfection studies are shown in Figure 1. Haplotypes I-1, II-1, IV-1 and IV-3 had higher expression levels than the other four haplotypes. To verify the radioligand binding method for ranking expression by haplotypes, Western blots were performed as shown in Figure 2 using a monoclonal antibody directed to a non-polymorphic region of the receptor protein. As indicated, haplotypes IV-1 and IV-3 had greater immunoreactivity at the expected molecular weight compared to haplotypes IV-4 and VI-2, consistent with the results from radioligand binding. The ADRB2 mRNA expression, ascertained by quantitative RT-PCR [12], did not necessarily mirror b 2 AR protein expression for each haplotype ( Figure 3A). The relationship between protein expression and mRNA (essentially showing the ratio for each haplotype) is shown in Figure 3B. As shown, there is a trend towards a statistically significant overall correlation between ADRB2 mRNA and b 2 AR protein (r 2 = 0.47, p = 0.06). However, when formally tested two haplotypes (I-1 and VI-1) had b 2 AR/ ADRB2 ratios that differed by 1 standard deviation from the mean ratio. Indeed, removal of either of these outliers increases the r 2 value to ,0.60 with significance at p = 0.03. The basis for this difference in relative levels of mRNA vs. protein expression is not readily apparent by inspection of haplotype sequence. Haplotype VI-1 differs from its closest haplotype (VI-2, a haplotype that is consistent with the regression line) by a single nucleotide within the 59-flanking region at position 23159 (C or T). On the other hand, VI-1 differs from the regression-conforming haplotype II-1 by 25 SNPs. A similar analysis with I-1 reveals multiple differences in sequence between it and all other seven haplotypes. Taken together, these results further support the notion that, for most haplotypes, multiple SNPs are likely interacting to ultimately establish phenotype.
We next assessed the agonist-promoted downregulation phenotype, which may be relevant to tachyphylaxis during prolonged exogenous or endogenous agonist exposure [13]. Eighteen hours after transfection, cells were exposed to vehicle (0.1 mM ascorbic acid) or the agonist isoproterenol (10 mM) for 48 hours, washed, and cell membranes prepared. 125 I-CYP radioligand binding was performed, and the results are shown in Figure 4 compared to untreated cells. Two haplotypes (I-1 and IV-3) showed greater extents of agonist-promoted downregulation compared to the other five. One potential concern regarding defining the downregulation phenotype based on a percentage of the untreated (baseline) expression is a confoundment from differences in baseline expression. However, we found no evidence for any relationship between baseline levels of expression and the % downregulation ( Figure 5).
In terms of haplotypes predicting clinical responsiveness to bagonists in asthma, our results suggest that these phenotypes are complex and can segregate as shown in Figure 6. For purposes of this grouping, expression is defined as ''reference'' or ''increased,'' and agonist-promoted downregulation as ''reference'' or ''increased.'' As shown, then, the haplotypes segregate into three groups: reference expression and downregulation, increased expression with reference downregulation, and increased expression with increased downregulation. These are designated as groups A (haplotypes IV-2, IV-4, VI-1 and VI-2), B (haplotypes II-1 and IV-1) and C (haplotypes I-1 and IV-3). Using group A as the reference, patients in group C would be predicted to have a high initial bronchodilating response to b-agonist, but would experience the greatest tachyphylaxis to chronic administration. On the other hand, patients in group B would also have a high initial response, but would experience less tachyphylaxis, so the overall best clinical response would be with these patients, which have haplotypes II-1 and IV-1. Although no studies have utilized these full haplotypes, there are some that have assessed potential associations between limited haplotypes, genotype combinations, or individual SNPs and b-agonist phenotypes in asthma [5], which can be interpreted to conform to the dual phenotypic groups of Figure 6. However, each study showed variability within the limited geneticallydefined groups, consistent with additional genetic variations potentially represented by the extended haplotypes, contributing to phenotypic noise.
A phylogenetic analysis of the b 2 AR haplotypes was undertaken with an overlay of the expression phenotypes, and the dual phenotypes (Figure 7). Two ancestral lineages are noted (L1, L2). L1 subsequently evolved to the two indicated haplotypes, both of which have the increased expression phenotype. L2 underwent a more complex evolution resulting in two major lineages, L2(1) and L2 (2). The former consists of two haplotypes both with the reference expression phenotype. L2(2) segregates into a single, distinct, increased-expression haplotype, and a three-haplotype clade consisting of two reference-expression haplotypes and one increased-expression haplotype. This latter haplotype (IV-1) differs from its nearest neighbor (IV-2) by one additional C in a poly-C tract in the 39UTR. This may represent an ancestral replication error along this repetitive sequence of up to 14 C's. While there appears to be some consistency with the tree and expression phenotype, there was no readily discernible pattern with the dual phenotypic groups A, B and C of Figure 6. And furthermore, the increased agonist-promoted downregulation haplotypes I-1 and  . Western blots confirm the radioligand binding method for ranking expression. Monoclonal antibody to a nonpolymorphic region of the b 2 AR protein was utilized to confirm the radioligand binding method for ranking expression phenotypes of the b 2 AR haplotypes. Shown is a single experiment revealing higher expression of haplotypes IV-1 and IV-3 compared to IV-4 and VI-2 b 2 AR protein at the expected molecular weight (,79 kDa). These results are consistent with those from radioligand binding of Figure 1. The control antibody was to Na + /K + -ATPase, a cell membrane protein.  (Table 1). Taken together, this may imply that ADRB2 evolution was driven by b 2 AR expression rather than downregulation or the dual phenotype. This interpretation is consistent with most b 2 AR in the body being activated by circulating epinephrine, which shows variability in plasma concentration [14] but rarely reaches the high concentrations necessary for downregulation. Nevertheless, the high concentrations of b-agonist achieved during pharmacologic therapy do promote b 2 AR downregulation in lung cells [15], so this component of the dual phenotype should be considered in grouping haplotypes for potential clinical associations with therapeutic response.
In conclusion, we have utilized the most common ADRB2 haplotypes (allele frequency of 0.05 or greater) to ascertain expression and agonist-promoted downregulation. The full intronless gene was cloned into a promoterless expression vector, so that phenotypes could be manifested by polymorphisms in the promoter, 59UTR, coding, or 39UTR regions. The great majority of association studies in asthma (and other diseases) which have considered the ADRB2 as a candidate gene for risk, severity or other clinical phenotype, or as a pharmacogenomic locus, have utilized one or both of the nonsynonymous coding polymorphisms. The results of these association studies range from highly significant associations [16] to the lack of any association [17]. While the study designs are virtually always different, which could explain these inconsistencies, we have been concerned that these limited polymorphisms do not provide a sufficient density of genetic information to discriminate phenotypes. Indeed, there are   known pairs of polymorphisms with low linkage-disequilibrium within ADRB2 [9]. To begin to coalesce the .26 polymorphisms in ADRB2 with cellular phenotypes, in the specific combinations that are observed in the population, we constructed the eight common haplotypes. We note that there are little or no ''spare b 2 AR'' in airway smooth muscle in relation to relaxation [1], so the ,20-40% lower expression observed between some haplotypes could be related to clinically significantly lower bronchodilatory responses to b-agonist. A similar close relationship between receptor expression and function is present in cardiomyocytes where the b 2 AR subtype increases inotropy and acts to inhibit certain apoptosis events in progressive heart failure when catecholamine levels are elevated [18]. Thus non-selective bblocker efficacy could be affected by b 2 AR haplotype. The other phenotype examined, receptor downregulation, is also relevant to therapeutic efficacy since this represents a major mechanism of tachyphylaxis during chronic agonist exposure. Association studies using full b 2 AR haplotypes with b-agonists in obstructive lung disease, or b-blockers in heart failure, have not been carried out to date. Given the phenotypes that we have defined, it would appear appropriate to carry out such studies with the haplotypes, or groups of haplotypes with similar cellular phenotypes. Such an approach may improve the predictivenesss of ADRB2 variation with relevant clinical phenotypes and resolve discrepancies between studies.

Expression Vectors
The expression vector pcDNA 3.1(+) (Invitrogen, Carlsbad, CA, USA) was mutated to remove the CMV promoter by sequential digestion with MluI and NheI followed by blunt-end ligation. This modified promoterless vector is denoted pcDNA 3.1(+)/DCMV and the full ADRB2 haplotypes (with the ADRB2 promoter) were inserted into the Hind III/Xba1 sites. To generate the eight ADRB2 haplotypes, the bacterial artificial chromosome RP11-44B19 was used as the template to amplify ADRB2 (the haplotype IV-1 form). Several rounds of site-directed mutagenesis by methods previously described [19] were utilized to generate mutations at the 26 variable sites, so as to construct the eight   Table 1. Full-length sequencing of the final construct was performed to verify the nucleotide changes and the integrity of the ADRB2.

COS-7 cells (obtained from American Type Culture Collection)
were grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum, 100 units/ml penicillin and 100 mg/ml streptomycin and for all conditions were maintained at 37u in a 95% air 5% CO2 environment. Transfections were carried out using methods previously described [20]. Briefly, 5.0 mg haplotype construct with 15 ml Lipofectamine 2000 (Invitrogen) were added to 10 7 cells and incubated for 6 hours. Then, fresh media was added and the media changed the next day. Eighteen hours after transfection the cells were treated with vehicle (0.1 mM ascorbic acid, representing baseline) or the b-agonist isoproterenol in the media for 48 hours, with an exchange of media and fresh isoproterenol after 24 hours.

Radioligand Binding and Western Blots
Attached cells were washed three times with PBS and then scraped in 5 mM Tris (pH 7.40) 2 mM EDTA at 4u and then centrifuged at 33,0006g for 15 min. Membranes were resuspended in 75 mM Tris (pH 7.40), 12 mM MgCl 2 , 2 mM EDTA and radioligand binding with 125 I-CYP carried out in triplicate as described [21]. Coincubations with propranolol (10 mM) were used to define nonspecific binding. Reactions were terminated by dilution in cold buffer and bound radioligand separated from free radioligand by vacuum filtration over glass fiber filters. The filters were counted in a gamma counter, and specific binding calculated as total minus nonspecific binding normalized to protein and expressed as fmol/mg. For Western blots, 15 mg of protein was electrophoresed through 10% SDS-polyacrylamide gels and transferred to nitrocellulose membranes as described [19]. Membranes were incubated with antibodies to b 2 AR (1:200 dilution, Santa Cruz) or Na + /K + -ATPase (1:200 dilution, Santa Cruz) for 1 hour and processed using enhanced chemiluminescence (GE Healthcare).

Quantitative RT-PCR
RNA was prepared using TRIZOL (Invitrogen) as described [22]. Reverse transcriptase reactions were carried out with Moloney murine leukemia virus reverse transcriptase (MultiScribe, Applied Biosystems, Foster City, CA, USA), 500 ng of extracted RNA, and random hexamer primers. Real time PCR was carried out with 2 ml of diluted reverse transcription reaction using methods previously described [12]. The TaqMan probe and primers sets were from Applied Biosystems for ADRB2 (Hs00240532_s1), which provided an amplicon of 66 bp representing nucleotides 500-565 of the coding region. For the control gene, probe and primer sets for phosphoglycerate kinase 1 (PGK2) were utilized (Applied Biosystems 4326318E) which provides for an amplicon of 75 bp. Reactions were carried out in an Applied Biosystems 7300 RT-PCR system. Threshold cycle (Ct) values were obtained (,15 for ADRB2 and ,23 for PGK2) and relative mRNA levels calculated using a 2 2DCt method [23].

Phylogenetic modeling
A phylogenetic tree was constructed using methods similar to those that we have previously described [24,25]. The unweighted pair group method with arithmetic mean was utilized with the maximum likelihood model to estimate evolutionary distance; 1,000 bootstrap sampling was performed.

Statistical Analysis
Data are shown as mean 6 standard error. Statistical comparisons were by ANOVA with post-hoc 2-way t-tests. Significance was considered when p,0.05.