Introduction

The widespread use of genetic biomarkers as surrogate endpoints aiming to describe risks, exposures, intermediate effects of treatments, and biologic mechanisms is a goal that scientists have long been pursuing. The adoption of any genetic test as a surrogate biomarker requires previous demonstration of its analytical and clinical validity as well as its clinical utility, and increasing the predictive capacity of genetic biomarkers is one of the major problems that we have to solve in order to transfer advances in pharmacogenomics to routine clinical practice. Determination of the polymorphic acetylation (NAT2 genotype or phenotype) was initially proposed to predict adverse reactions in patients with tuberculosis receiving isoniazid, prior to the concomitant administration of procainamide and phenytoin, and to analyze the role of NAT2 in drug interactions. These effects, together with the role of NAT2 in cancer risk, in non-malignant spontaneous disorders and in drug response and toxicity, make NAT2 a relevant target for pharmacogenomic testing in clinical practice [1], [2].

Nearly fifty years ago, Evans et al. demonstrated that acetylation of isoniazid was bimodally distributed and that the in vivo acetylation status was inheritable [3], [4]. Since then, traditional phenotype determination by inference from genetic analyses has classified the population in three groups: rapid, intermediate and slow acetylators. Although this classification of individuals into three phenotype categories is widely accepted, it would be desirable to refine further the predictive capacity of acetylation pharmacogenomic testing [5]. Heterologous expression of NAT2 allozymes provided indirect evidence suggesting a differential effect of NAT2 variant alleles and hence heterogeneity within the slow acetylation phenotype (reviewed in [6]).

This study aims to analyze whether this evidence of heterogeneity within rapid and slow acetylators exists in vivo, whether commonly used pharmacogenomic tests are adequate for the inference of phenotype subcategories, and to measure the activities for such phenotype subcategories. Because acetyl metabolites may be pharmacologically active, or function as intermediates in toxic metabolic pathways, further refinement in phenotype prediction may help to identify risks associated to one or more of such phenotype subcategories.

Methods

The subjects were drawn from a study previously described [7], [8], [9]. Briefly, cases (n = 93), of newly diagnosed cancer of the exocrine pancreas were recruited from all hospitals in the 7-county metropolitan area of the Twin Cities, Minnesota and the Mayo Clinic (from the latter, only cases residing in the Upper Midwest of the US were recruited). Controls (n = 411) were randomly selected from the general population and frequency matched to cases by age and sex (Table 1). All were Caucasian. Each participant provided written, informed consent prior to interview and blood draw. The study was approved by the Institutional Review Boards of the University of Minnesota and The Mayo Clinic, USA and by the Ethics Committee of the University of Extremadura, Spain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the individuals included in the study.

https://doi.org/10.1371/journal.pone.0044629.t001

In vivo NAT2 activity was measured with a widely used caffeine-based assay, as described by Butler et al. [10] with minor modifications as described elsewhere [8], [11]. The caffeine assay is highly accurate and reproducible, and it is considered as a gold standard for acetylation phenotyping. Details on accuracy and reproducibility were published elsewhere [10], [12], [13], [14], [15]. In brief, subjects ingested 200 mg of caffeine, following an overnight fast. Subjects refrained from the consumption of caffeine- and methylxanthine-containing foods and beverages from midnight until 5 h after the dose of caffeine. A urine specimen was collected 5 h after the administration of caffeine and samples were acidified and stored as described elsewhere [11]. Regarding HPLC analysis, 200 µl of urine were saturated with 125 mg of ammonium sulfate, and 6.0 ml of chloroform:isopropanol (95∶5) were added. Each sample was vortexed and centrifuged, and the organic phase was removed and evaporated to dryness. The residue was resuspended in 250 µl of 0.05% acetic acid, filtered, and frozen until analysis. Fifty µl of the extract were injected onto a Beckman C18 Ultrasphere octadecylsilane column (25 cm in length, 4.6-mm diameter, 5-µm particle size) and eluted with a 0.05% acetic acid-methanol solvent (flow rate, 1.2 ml/min).

Acetylation phenotype was assigned on the basis of a molar AFMU/1X ratio, which served as quantitative determinant of acetylation capacity with a cut-off value = 0.66 (log AFMU/1X = −0.18) in agreement with previous studies [11].

NAT2 genotyping aimed to identify the signature SNPs for alleles corresponding to the NAT2*5, NAT2*6, NAT2*7 and NAT2*14 clusters, that is, rs1801280 (I114T), rs1799930 (R197Q), rs1799931 (G286E) and rs1801279 (R64Q), respectively. Although several NAT2 alleles have been described (for an updated list of NAT2 alleles and haplotypes see the website http://louisville.edu/medschool/pharmacology/consensus-human-arylamine-n-acetyltransferase-gene-nomenclature/nat_pdf_files/Human_NAT2_alleles.pdf), the SNPs analyzed in this study identify the vast majority of slow NAT2 variant allele clusters [16], [17]. Genotyping was carried out by the use of TaqMan®probes (details available in Table S1). For every SNP analyzed, twenty samples with heterozygous genotypes and up to twenty samples with homozygous genotypes (homozygous non-mutated and homozygous mutated when available), were determined as blind duplicates. In all samples with genotype/phenotype discordance (n = 32) the genotypes were confirmed by the use of PCR-based mutation-specific amplification as described elsewhere [8] or by direct sequencing of the amplified fragments. In all cases the genotypes fully corresponded to those obtained with TaqMan probes. Haplotype assignation and phenotype inference: All possible haplotypes combining the four SNPs analyzed were constructed and their frequencies were analyzed by using PHASE and the NAT2 haplotype table described elsewhere [16]. Phenotype inference was carried out as described elsewhere [16]. Putative departures of Hardy-Weinberg Equilibrium were calculated by using the software Haploview 4.1. Continuous variables (acetylation ratios), expressed as mean (SD), were compared with the Student’ T test, and tests for trend were calculated with the Spearman’s rank correlation by using the statistical software SPSS 15.0 for Windows (SPSS Inc. Chicago, Illinois, USA). A p value <0.05 was considered significant. When multiple comparisons were made, adjustments for multiple comparisons were carried out according to Bonferroni’s procedure.

Results

The SNP frequencies and the genotypes observed in the 504 participants are summarized in Table 2. The degree of phenotype/genotype concordance by using the traditional phenotype classification (i.e. rapid/slow phenotypes), where NAT2*4 containing genotypes are considered a rapid phenotype, and other genotypes a slow phenotype, was equal to 93.7%. We selected 435 individuals with genotypes NAT2*4/*4, *4/*5, *4/*6, *5/*5, *5/*6 and *6/*6 and phenotype/genotype concordance for further analyses. These corresponded to 73 cases and 362 control subjects. Carriers of variant alleles NAT2*14 were not included in the analyses, because these alleles were not present in the study population (Table 2). In addition, carriers of the variant alleles NAT2*7 were not included in the comparisons because these alleles were rare in the population study (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. NAT2 SNP frequencies observed in the present study.

https://doi.org/10.1371/journal.pone.0044629.t002

Table 3 shows the acetylation capacity of the six genotype categories analyzed in the study. The individuals with the genotype categories NAT2*4/*5 and NAT2*4/*6 showed similar acetylation values. However, for the rest of individuals, each genotype category corresponded to a distinct phenotype category, with non-overlapping 95% confidence intervals for the activity, and in all cases the differences between these categories were statistically significant. This provides in vivo evidence that in the absence of NAT2*4 alleles, variant alleles NAT2*5 and NAT2*6 confer different acetylation capacity. In addition, a gene-dose for these variant alleles can be observed within the slow acetylator phenotype, as there is a statistically significant trend to slower acetylation capacity among individuals with the genotypes as follows: NAT2*5/*5>NAT2*5/*6>NAT2*6*6 (Spearman's rank correlation with the number of NAT2*6 alleles, (log AFMU/1X = −0.359); P<0.001). These findings were not influenced by the sex of participants, age, smoking status, pack-years, drinking status, servings of alcohol per week as stated by multivariable linear regression, or by the case-control status (Table 4, Table S2). The effect of NAT2*7 in vivo could not be elucidated because of the low allele frequency in the study population. We identified only two carriers of the alleles NAT2*7 in homozygosity, with metabolic ratios equal to −0.51 and −0.96. The mean value (−0.74), is close to the mean value for carriers of the NAT2*6/*6 genotypes, thus suggesting that the NAT2*7 alleles in homozygosity may confer a very slow acetylation phenotype; although due to the sample size the comparisons of the acetylation phenotype were not statistically significant. Table S3 includes details of the log AFMU/1X ratios of carriers of NAT2*7.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Acetylation ratios (log AFMU/1X) in subjects with different NAT2 genotypes.

https://doi.org/10.1371/journal.pone.0044629.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Effect of the case-control status on the Acetylation ratios (log AFMU/1X) in subjects with different NAT2 genotypes.

https://doi.org/10.1371/journal.pone.0044629.t004

Discussion

Differential effects of acetylation status by different slow acetylation alleles have been suggested previously, but to our knowledge they have not been formally evaluated in vivo. Indirect evidence from in vitro studies and from clinical association studies suggest that NAT2 variant alleles produce different functional effects, implying heterogeneity within the “slow” acetylator phenotype [6]. Antituberculosis drug-induced hepatotoxicity risk is particularly high in carriers of the NAT2*6/*6 allele, thus suggesting that these individuals may constitute a subcategory of “very slow” acetylators [18], [19]. These and other clinical association studies (reviewed in [6]) suggest that the NAT2 slow acetylator phenotype is heterogeneous, and that multiple slow acetylator phenotypes exist [20]. However, no clear association between NAT2 variant alleles and in vivo phenotype categories among slow acetylator individuals has been proved so far. Our findings indicate that the NAT2*6 allele cluster is related with the slowest acetylation capacity in vivo with a gene-dose effect, thus demonstrating the occurrence of a category of “very slow acetylators” with the genotype NAT2*6 in homozygosity. Because of the ethnic origin of the population study, we were unable to dissect the effect of the allele clusters NAT2*7 and NAT2*14; it should, however, be emphasized that these clusters are rare in caucasian populations [21] and that the allele frequencies observed in this study are consistent with those reported for other Caucasian individuals [21], [22].

The effect of NAT2 variant alleles may vary by substrate or with substrate concentration [6]. For instance, it has been shown that the NAT2*7 allele cause a different effect in the N-acetyltransferase activity towards 2-aminofluorene and to sulfamethazine [23]. Therefore the findings obtained in this study should not be extrapolated to other NAT2 substrates without confirmation with every specific substrate. Nevertheless, our findings in vivo agree with findings obtained in vitro which suggests that the protein level expressed by common NAT2 alleles is NAT2*4>NAT2*5>NAT2*6 [6], thus suggesting that the differential effect of NAT2 alleles observed with the probe drug caffeine is likely to be relevant to other NAT2 substrates.

The aims of this study are to refine the phenotype inference of NAT2 genotyping and the identification of clinically relevant associations of the new genotype categories with cancer risk, differential treatment response or clinical outcome are beyond the aims of the study. Although this study included patients with cancer of the exocrine pancreas and control subjects, no association of NAT2 genotype categories with pancreatic cancer risk was observed, in agreement with previous studies [24], [25].

The findings reported in this study show that acetylation capacity in vivo is related to different NAT2 genotypes among slow acetylators, and indicate that variations in the acetylation NAT2 status among slow acetylator individuals result from the co-dominant expression of the NAT2*5 and NAT2*6 alleles or haplotypes, whose diplotypes are related to distinct slow acetylation phenotypes. Additional studies are required to go further in the refinement in phenotype inference, particularly in other human populations with different NAT2 allele frequencies. It may be argued that the difference in function between the variants NAT2*5 and NAT2*6, although statistically significant, is a minor difference compared to the function of any genotype containing at least one NAT2*4 allele and therefore that the clinical relevance of this difference may be limited. However, NAT2*6/*6 homozygotes show roughly a 30% reduction on enzyme activity as compared to NAT2*5/*5 homozygotes. For comparison, the reduction on enzyme activity between NAT2*4 heterozygotes (intermediate acetylators) and NAT2*4/*4 homozygotes (rapid acetylators) in this study is 28%. A 30% reduction in activity among individuals who have a very impaired acetylation capacity may have a higher clinical relevance than a comparable reduction among individuals who have a high acetylation capacity. These findings provide a novel framework for evaluating interactions between NAT2 genotype and adverse drug reactions or cancer risk.

Supporting Information

Acknowledgments

We are grateful to Prof. James McCue for assistance in language editing, and to Ms. Gara Esguevillas, for technical assistance.