Main

PKU and related forms of HPA result from deficient activity of the hepatic enzyme, PAH, a monooxygenase (EC 1.14.16.1). The principal metabolic fates of L-phenylalanine are hydroxylase-mediated conversion to tyrosine with subsequent oxidation and incorporation during de novo protein synthesis(1). The phenylalanine hydroxylase gene (symbol,PAH) is located on chromosome 12q24.1. Mutations at this locus account for 99% of cases of inherited HPA, with over 300 different phenotype-modifying mutations now recognized(2), including 35 that have been expressed and analyzed for unit-protein activity in an in vitro system. Although deficiency of hydroxylating activity is the major determinant of human HPA phenotypes, the associated impairment of cognitive phenotype in untreated PKU cases does not necessarily correlate with the predicted effect of PAH genotype(3). Because intelligence is a complex trait beyond the control of any one gene or genotype, this finding is not surprising. Because plasma amino acid values also behave as quantitative (complex) traits(4), we elected to study correlations between mutant PAH genotypes and phenylalanine metabolism (plasma phenylalanine levels and whole-body oxidation rates) in vivo in persons with HPA.

We used L-[1-13C]phenylalanine to assay the phenylalanine oxidation rate in vivo; the formation of 13CO2 in expired breath was measured using isotope ratio mass spectrometry. We examined three hypotheses: 1) a gene dosage effect would be seen between homozygous mutant, heterozygous and homozygous normal classes, 2) mutant alleles with “severe” or “mild” effects could be differentiated, and 3) the in vivo metabolic phenotype (an emergent property) would not always be consistent with predictions derived from the in vitro (unit protein) enzymic phenotype. Our findings support these hypotheses.

METHODS

Subjects. We studied 13 children and 4 adults with HPA, 18 obligate heterozygotes, and 8 control subjects (Table 1). The subjects were selected to represent a spectrum of phenotypic expression and whose compliance with dietary restrictions allowed estimates of phenylalanine tolerance. Controls were age- and sex-matched as closely as feasible. All patients had normal growth parameters, indicating adequate intakes of protein and phenylalanine.

Table 1 Subjects enrolled in analysis of in vivo L-[13C]phenylalanine oxidation
Full size table

There are several ways to classify subjects with HPA(1, 5, 6). Here, we have assigned HPA subjects to one of three broad classes: a “severe” class (PKU); a“variant” class, less severe than PKU (variant PKU); and a“mild” class (non-PKU HPA). Our patients were assigned to the PKU category when plasma phenylalanine values, in the absence of treatment, consistently exceeded 1200 μmol/L and the tolerance for dietary phenylalanine (during the first 5 y of life) was less than 400 mg/d; on this intake, phenylalanine levels were brought down to levels generally considered safe or less than 600 μmol/L(7). Patients with variant PKU had pretreatment plasma phenylalanine values lower than those in PKU, and their dietary tolerance for phenylalanine was higher but not normal. Persons with non-PKU HPA had plasma phenylalanine values consistently below 1000μmol/L on an unrestricted diet. Informed consent was obtained from each subject or family, after a detailed explanation of the study and its purpose. Ethical approval was granted by the Institutional Review Board of the Montreal Children's Hospital.

Clinical protocol. Phenylalanine is converted to tyrosine by phenylalanine hydroxylase; this step and subsequent oxidation of tyrosine normally accounts for the main component of phenylalanine catabolism. Phenylalanine oxidation (flux) rates were measured in the subject at rest after an overnight fast; water was given liberally. L-[1-13C]Phenylalanine, 99% enriched (Cambridge Isotope Laboratories, Andover, MA), was given by mouth in a gelatin capsule. Subjects weighing less than 50 kg received 100 mg of labeled substrate; all other subjects received 200 mg. Breath samples were collected at zero time and at 10-min intervals after ingestion for 80 min. The breath samples were taken by blowing a complete exhalation through a straw into an 11-mL blood collection tube(Exetainer). The tube was capped immediately and sent for CO2 isotope ratio content. Venous plasma samples were obtained at noon before eating to measure the corresponding steady state amino acid value.

Mutation analysis. PAH gene mutations were analyzed by PCR, denaturing gradient gel electrophoresis, and DNA sequencing as described previously(8). The mutations identified in this project are shown by position in the PAH gene in Figure 1.

Figure 1
figure 1

The PAH gene mutations identified in the mutant genotypes analyzed in this project. The positions of the mutations in the gene are indicated (x axis); their predicted effect on(unit-protein) phenylalanine hydroxylase activity is indicated on the y axis.

Full size image

In vitro expression analysis. The available data for PAH mutations were obtained from the file prepared by P. J. Waters for the PAH Mutation Analysis Consortium Database on the web site(2) (http://www.mcgill.ca/pahdb). Activity of the mutant enzyme is given as a percent of control (normal) enzyme activity in the corresponding system. Effects of mutation, described as “severe,”“mild,” and so forth are self-defining.

L-Phenylalanine oxidation and13CO2 analysis. To calculate the rate of oxidation, the isotopic composition and the rate of production of carbon dioxide were measured. The isotopic composition was measured using a Europa Scientific ABCAR mass spectrometer. For each breath sample, the atom% 13C was calculated as 13C/(13C +12 C) × 100. The atom% excess above naturally occurring level was obtained by subtracting the baseline percent 13C obtained at time 0 after overnight from the percent 13C at time t. The analytical sensitivity of the atom% excess measurement is influenced by diet and can be as high as 0.003% atom% excess.

The carbon dioxide production for subjects at rest was estimated to be 5 mmol/m2/min(9). The body surface area (BSA) is estimated from height and weight by the following equation(10): The rate of excess 13C in breath recovered from L-phenylalanine oxidation was obtained as:Equation

The dose recovered (PCD) in breath/h is then calculated:Equation

The cumulative recovery (CR) of oxidized dose was obtained by triangulation: Equation

Data were normalized, for comparisons between subjects, by expressing results in the “experimental” subject as a percent of the corresponding mean value for the controls.

Amino acids. Phenylalanine and tyrosine were measured by quantitative elution chromatography on a Beckman 6300 amino acid analyzer. The venous plasma phenylalanine value and the phenylalanine:tyrosine molar ratio(11) were plotted as density functions(12) to display trait values for the heterozygote and homozygous-normal (control) subjects.

Statistical analysis. Analysis of difference was performed by t test.

RESULTS

Plasma Phenylalanine Values

Subjects in the HPA class, whether on treatment (PKU or variant PKU) or not(non-PKU HPA), had modest HPA (Table 1); differences between the three HPA classes are not significant. Obligate heterozygotes and controls had mean values within the normal range. Plasma tyrosine levels in our patients and the heterozygotes were all within the normal range (data not shown).

Phenylalanine Oxidation Rates

Intraindividual variation. We performed three independent measures of phenylalanine oxidation in a control adult male subject (Fig. 2); the intraindividual variation is modest(coefficient of variation, 13%). The analytical error is trivial.

Figure 2
figure 2

Intraindividual variation in L-[1-13C]phenylalanine oxidation in a control subject.13 CO2 production is expressed as progress curves against time for cumulative percent of administered dose converted to CO2 in exhaled gas in each of three separate experiments.

Full size image

Interindividual variation. Trait values are significantly different (Table 2) within the HPA class (PKU versus non-PKU HPA, p < 0.05); between heterozygotes and controls (p < 0.001); and between the pooled HPA group and controls (p < 0.0001). The significant difference between heterozygotes and controls broadly validates sensitivity of the method; it detects a gene dose effect on the phenylalanine oxidation rate (the first hypothesis). Oxidation rates (y axis) by HPA class, and by individual, are related to the corresponding genotypes (x axis) in Figure 3.

Table 2 Cumulative (60 min) rate of L-[1-13C]phenylalanine oxidation (% dose) expressed as primary data (column A) and normalized data (column B)
Full size table
Figure 3
figure 3

Interindividual variation in actual L-[1-13C]phenylalanine oxidation rates (y axis) in probands with classical PKU, variant PKU, and non-PKU HPA phenotypes, obligate heterozygotes (parents) and control subjects (x axis). Data are expressed as cumulative percent dose oxidation during first 60 min. Heterozygous genotypes are indicated (e.g. A104D/+ for mutant allele/normal allele).

Full size image

Genotype-Phenotype Correlations

Consistencies. Class differences in oxidation rates are apparent both in absolute (Fig. 3) and normalized(Table 2) terms in the PKU, variant PKU, and non-PKU HPA classes. The mean value in each class (Table 2) is a measure of a metabolic phenotype that correlates with the overall clinical class of HPA. The finding supports our second hypothesis that genotypes comprising alleles having “severe” or “mild” effects on in vitro enzyme activity will, in general, have the corresponding in vivo phenotypes.

Normalized oxidation values were related to the mutant genotype in homoallelic subjects (Table 3, part A). The R408W mutation, known to be null by in vitro expression analysis(13), totally impaired phenylalanine oxidation in vivo. Allele A104D, with about one-third normal residual activity in vitro(14), conferred the non-PKU HPA phenotype and 8.5% normal phenylalanine oxidation in vivo. The in vivo oxidation rate is informative about the effects of the mutant allele when it is unknown in vitro (Table 3, part B). The E390G allele, co-expressed in vivo with a null allele (IVS12nt1) confers a non-PKU HPA phenotype, inferring that the E390G mutant has significant residual enzyme activity. K42I, a novel mutation, and expressed in vivo with E280K (a so-called “severe” allele) confers a 6% oxidation value, implying that K42I is “mild”(Table 4). Heterozygotes gave results broadly compatible with the allele effect anticipated from in vitro expression and observations in vivo in probands (Table 3, part C).

Table 3 L-[1-13C]phenylalanine oxidation rates (normalized) related to genotype
Full size table
Table 4 Comparisons of in vivo phenylalanine oxidation values, HPA phenotype class, in vitro (unit protein) enzymic activity and PAH genotypes
Full size table

Inconsistencies. Our third hypothesis is supported first:1) by findings that reveal interindividual variation in oxidation rates among heterozygotes with similar genotypes (Table 3, part D); for example, markedly different oxidation rates were found in two different subjects with the A104D/+ genotype, and in two subjects with the R408Q/+ genotype; and 2) failure to find the anticipated correlation between the measured in vitro mutation effect on enzyme activity, the in vivo oxidation rate and clinical HPA class; for example, a patient with the F299C/R158Q genotype, anticipated to have about 5-6% of normal enzyme activity (a significant residual activity), had no measurable phenylalanine oxidation in vivo (13CO2 data) and had the corresponding in vivo PKU (severe) phenotype; a PKU patient with the IVS12nt1/S349P genotype, anticipated to have no PAH enzyme activity, had some in vivo oxidation activity (≈4% normal).

Measures of heterozygosity. Heterozygosity clearly impairs phenylalanine oxidation. Evidence of a gene dose effect with heterozygote values intermediate to affected and control data (Fig. 3,Tables 2 and 3) validates the method. Measurement of phenylalanine oxidation does not improve on the use of noontime plasma phenylalanine and tyrosine values to distinguish carriers and noncarriers (Fig. 4).

Figure 4
figure 4

The metabolic phenotype expressed as a density function(12) for the noontime, preprandial, plasma amino acid values (phenylalanine/tyrosine molar ratio - y axis; plasma phenylalanine level x axis). The ellipse is the boundary beyond which the probability of being heterozygous for a PKU genotype is >0.01. The predicted effect of genotype on the metrical trait (▴, severe effect; •, mild effect) is indicated.

Full size image

DISCUSSION

PKU and related forms of HPA are among the most widely ascertained Mendelian human metabolic disorders, and characterization of the PAH gene has encouraged extensive mutation detection in HPA cases(1, 2). Accordingly, there is considerable interest to relate mutant genotype with the variant HPA phenotype. The in vitro expression of characterized alleles and estimates of in vitro activities of several mutant proteins, particularly in COS cells(15, 16), are likely to overestimate activities in vivo. Accordingly, if the in vitro phenotype is used to anticipate outcomes and dietary requirements (the clinical phenotype) there could be difficulties. Here, we report a novel approach to correlate genotype with in vivo PAH activity.

13C is a stable isotope and ingestion of tracer doses of L-[13C[phenylalanine is simple and harmless. Cumulative13 CO2 formation in the first 60 min after ingestion of the labeled phenylalanine is a measure of immediate phenylalanine uptake and oxidation in the liver by its major pathway(17). Although phenylalanine contains a carboxyl carbon atom, production of CO2 by irreversible removal of this atom (by decarboxylation to form phenylethylamine) is normally negligible and would contribute little to CO2 labeling in the present study. The present method to measure a trait value in PKU differs from other models using i.v. deuterated and13 C-labeled phenylalanine, which are more invasive and may be associated with elevated phenylalanine hydroxylation rates associated with the administration of bolus doses of substrate and upregulation of hydroxylation in vivo(18–22). The latter methods have not yet been used to study genotype-phenotype relationships.

We observed intraindividual variation in phenylalanine oxidation rates in all classes of subjects (Fig. 2). Because analytical error is not a significant contributor to this variation, the main contribution (to trait variation in vivo) is likely to be genotype (in the mutant states), its effect on phenylalanine “outflow” to tyrosine), and physiologic factors affecting hepatic uptake(23) and subsequent oxidation. Differences in absorption rate have been controlled for in this study by the estimation of cumulative oxidation at a point (60 min) when the subjects have reached isotopic plateau. As all the patients were studied under similar conditions (overnight fasting) and on phenylalanine restricted diet with a state of positive nitrogen balance, variables in oxidation that could be attributed to differences in protein intake are controlled.

Phenylalanine pool size may also influence uptake, distribution, and metabolic outflow of labeled substrate. In our study, pool sizes in the three different HPA classes were similar and significantly elevated(Table 1) to levels well above the putative Km value for PAH ≈100 μmol/L). Accordingly, modest differences in substrate concentration above the KM value are less likely, by themselves, to account for the observed differences in oxidation rates. None of the mutations analyzed in our study behaves as a Km mutant according to available analysis data. Another modifier of oxidation rates at any given substrate concentration, and also a contributor to the interindividual (intraclass) variation, is likely to be basal metabolic rate. This may be one of the factors accounting for variation observed between male and female heterozygotes of similar genotype. Physiologic stress affecting metabolic rates will also modulate oxidation(24) as will fasting and feeding(25). In due course, we intend to study the effects of these factors, along with age, gender, and exercise, on phenylalanine oxidation rates.

A gene dose effect on the phenylalanine oxidation rate is clearly seen(Table 2, Fig. 3); this finding validates the method's ability to measure (broadly) the trait value modified by the PAH genotype. The in vivo findings also correspond to direct measurements of PAH enzyme activity in liver biopsy samples(26, 27), where milder forms of PKU were observed to have up to 6% normal activity; and non-PKU HPA, 8-35% normal, suggesting that, within the limits of our protocol, it represents a valid physiologic measurement of phenylalanine oxidation in vivo.

When gene dose effect on PAH enzyme activity is concordant with the phenylalanine oxidation (flux) rate, it implies that hydroxylating activity is the principal determinant of the metabolic “outflow” flux. Accordingly, the rate of 13CO2 production measures pathways and networks conferring phenylalanine oxidation. Our data, both from HPA heterozygotes and probands, imply that phenylalanine hydroxylase has the highest sensitivity coefficient in the system, as predicted(28).

We relied on data from heterozygotes to demonstrate a gene dose effect. On the other hand, the same data reveal something else of equal interest-notably that the whole-body oxidation rate (the metrical trait) is not explained solely by PAH genotype and the corresponding hepatic PAH catalytic activity. For example, heterozygotes with identical mutant PAH genotypes can have greatly different oxidation rates (Table 3). Accordingly, it is unreasonable to expect measures of oxidation rates to improve on noontime plasma phenylalanine and tyrosine values (Fig. 3), another metrical trait used to classify heterozygotes(12).

Correlations of phenotype (oxidation rate) with genotype in HPA probands(homo- and heteroallelic) were both consistent and inconsistent: consistent in the prediction that the anticipated effect of a mutation broadly correlates with the corresponding phenotype in vivo; inconsistent in that this was not always the case. On the side of consistency, we could identify three broad classes of flux rates corresponding to the three HPA classes(Table 2). Mutations anticipated to confer a null enzymic phenotype confer the severe metabolic phenotype (PKU); mutations conferring some degree of residual enzymic activity are usually associated with a clinically “benign” form of HPA (non-PKU HPA)(Fig. 3,Table 4). To this extent, then, mutation analysis to describe genotype has predictive and clinical value, and the oxidation rates described here fill the gap between mutation analysis,“prediction” of enzymic activity (from in vitro expression analysis but recognized to overestimate enzymatic activity) and observed class of HPA, in a way not feasible in earlier studies(13, 16, 29).

However, there are also inconsistencies of two types. 1) Similar mutant PAH haplotypes do not confer consistent oxidative phenotypes. Whole-body phenylalanine oxidation rates are known to vary independently from hydroxylation activity in human subjects(30). Moreover, transport of phenylalanine into rat liver, not under control of the PAH gene, can be as important as hydroxylating activity in the regulation of phenylalanine disposal or runout from its pool in that species(23). 2) The measured unit-protein phenotype(hydroxylase activity) in vitro is not always concordant with the corresponding oxidation rate in vivo. Some of the discrepancy may relate to sensitivity of the 13CO2 method that has yet to be determined and may influence interpretation of very low oxidation values in PKU probands. Some of the discordance is likely attributable to the use of transient over expressing systems for in vitro expression analysis. In these transient systems, the plasmid DNA is gradually lost, therefore protein production can only be temporary. Steady-state levels of hepatic PAH protein in vivo reflect the rates of both synthesis and degradation. The mechanisms and rate of synthesis from a PAH cDNA-containing plasmid in vitro will differ from those involved in protein from genomic DNA under physiologic control systems. The processes regulating protein degradation may also differ between cultured cells and intact liver. Furthermore, the mutant phenotype in vitro is the homozygous expression of only one allele, whereas the majority of probands are heteroallelic, and interallelic complementation in situ may indeed modify the phenotype in vivo.

Thus, our measure of whole-body phenylalanine oxidation provides information different from that obtained from the standard loading test or dietary tolerance tests. The 13CO2 measure is noninvasive, it measures an emergent (whole-body) property, and it is predictive of actual phenylalanine tolerance in the patient. It can also provide information about alleles that have not yet been expressed in vitro. In summary, we present a rapid noninvasive method to measure phenylalanine oxidation in vivo. Whereas phenylalanine oxidation in vivo is largely under the control of PAH locus and the corresponding hydroxylating activity and thus behaves as a Mendelian trait, it is not solely explained by PAH enzymic activity. Accordingly, as expected of an important biochemical and physiologic process, whole-body phenylalanine oxidation, and the corresponding variant metabolic (HPA) traits (emergent properties) have features of a complex (quantitative) trait.