Estimating the prevalence of pyruvate kinase deficiency from the gene frequency in the general white population

Pyruvate kinase (PK) deficiency is probably the most common cause of hereditary nonspherocytic hemolytic anemia.1 The true prevalence of this disorder is unknown because diagnoses are made not only in specialized centers throughout the world but also in commercial and hospital laboratories.

The red cells of heterozygotes for PK deficiency generally have approximately half the normal level of enzyme activity. Based on the lowered erythrocyte enzyme activity of heterozygotes, a few efforts have been made to obtain some information regarding the prevalence of this disorder. Assays of red cell PK in 214 normal subjects led Blume et al2 to conclude that approximately 1.4% of the German subjects were heterozygous for the deficiency. Mohrenweiser3 reported that among 697 newborns, 84% of whom were white, 1 (0.15%) manifested red cell PK activity that was more than 3 standard deviations below the mean. In a subsequent study, however, he suggested that the frequency of heterozygous subjects was 18 of 1736 white newborns.4 However, because red cell PK activity varies widely from person to person, there is bound to be overlap between normal and heterozygous levels.

Detection of the existence of mutations based on the DNA analysis is more robust than is the estimation of enzyme levels. Although it is not feasible to perform complete sequence analysis on thousands of persons, it is possible to determine the frequency of those mutations that are the most common causes of PK deficiency. Based on the known frequency of such mutations in patients with the disease, it is possible to extrapolate from the general population to obtain some estimate of the number of persons in the population who may have the PK-deficient genotype.

We now report the genotypes of an additional 20 patients with PK deficiency. Based on the accumulated information of the occurrence of mutations in patients with PK deficiency and a survey of more than 3500 normal subjects, we estimate the prevalence of PK deficiency in the white population at large.

Patients with PK deficiency had documented hemolytic anemia and deficiency of erythrocyte PK activity when assayed by standard methods as modified from previously published methods.5 When the more common mutations were not detected by restriction analysis or direct sequencing, the entire coding region was sequenced as described previously.6 

DNA samples from 3785 anonymous persons, identified only by ethnic origin, were examined for 4 PK mutations: c.1456 C→T(1456T), c.1468 C→T(1468T), c.1484 C→T(1484T), and c.1529 G→A (1529A). These samples were derived from several different sources. Approximately 3500 samples were obtained from patients attending a health-screening clinic. Ethnic origin was based on self-identification. Approximately 300 additional samples from African Americans, identified as such by the phlebotomist, were from discarded diagnostic samples. The segment of DNA containing these mutations was amplified by means of the polymerase chain reaction (PCR) with the following 2 primers: sense 5′-CTCGTTCACCACTTTCTTGC-3′ and antisense 5′-GAGGCAAGGCCCTTTGAGTG-3′. The PCR mixture contained 34 mmol/LTris-HCl, pH 8.8, 8.3 mmol/L ammonium sulfate, 1.5 mmol/L MgCl₂, 85 μg/mL bovine serum albumin, 0.2 mmol/L each dATP, dCTP, dGTP, and dTTP, 120 ng of each oligonucleotide primer listed above, 200 ng genomic DNA, and 1 U Taq polymerase. After a 4-minute denaturation at 98°C, 30 cycles of PCR at 94°C for 30 seconds, 60°C for 30 seconds, and 72°C for 30 seconds were performed.

The amplified DNA was spotted in duplicate on Nytran SuPerCharge membranes (Schleicher and Schuell, Keene, NH) and probed with the wild-type and mutant probes for each mutation studied (Table1). All filters contained controls of normal, heterozygous, and homozygous samples. All positive results were confirmed by restriction analysis or direct sequencing.

The mutations found in the 20 patients with PK deficiency are summarized in Table 2. In one patient, only 1 of the 2 mutations was identified. It is possible that the missing mutation, denoted “?,” was not found because it was not in the coding region of the gene or because of technical reasons. One mutation that has not been reported previously is 1153T. The deduced amino acid change for this mutation is arginine to tryptophan at amino acid 385 (R385W). This new mutation was found with the 1456T mutation in a 1-year-old boy. He was born at full term and had a bilirubin level of 14 mg/dL. His reticulocyte count was 28%, and his hematocrit level was 32%. He did well after exchange transfusion. At 3 months of age he was found to have a hemoglobin level of 5 g/dL and a reticulocyte count of 9.9%. He is transfused every 3 to 4 weeks. His parents are of Irish and mixed European ancestry.

Heterozygotes for each of the 4 mutations for which the population was screened were detected in the population survey except for the 1484T mutation, but there were no homozygotes. The results are shown in Table3. The gene frequency for the 1529A mutation was 0.00203 ± 0.0006 (mean ± 1 SE) in the white population; that of the 1456T mutation was 0.00350 ± 0.0008 in the white population and 0.0039 ± 0.0031 in the African American population. The 1468T mutation was found only among Asians. In the small group of 126 persons examined, it was encountered twice, giving a gene frequency of 0.00794 ± 0.00559. The 1484T mutation was not found in any of the subjects studied.

Estimating the prevalence of uncommon autosomal recessive diseases is a difficult challenge. For example, numerous attempts have been made to estimate the prevalence of Gaucher disease in the Jewish population. One approach has been to attempt to identify all cases in a target population by surveying physicians, hospitals, or both.7 This method depends on accurate and complete ascertainment of cases and thus tends to underestimate the true incidence of the disease. A second means of estimating prevalence in the population is to attempt to identify heterozygotes by measuring the gene product, which tends to be half the normal level in heterozygotes 8,9 However, there is always overlap between heterozygous and normal values and sometimes between those of heterozygous and homozygous subjects. Correcting for this overlap introduces a large error and, indeed, is often ignored, giving inaccurate estimates. The actual documentation of mutations at the DNA level is the most accurate way of identifying heterozygotes, but, unless complete sequence analysis is carried out, only the mutations that are selected for detection will be identified, and correction must be made for patients with other mutations.

In the case of PK deficiency, the only method that has been used to estimate population prevalence is that of detecting heterozygotes on the basis of red cell enzyme activity.2-4Three of the most prevalent mutations in patients with PK deficiency are 1529A, 1456T, and 1468T; 1529A is most common in northern and central Europe,6,10 1456T is most common in southern Europe,11 and 1468T is most common in Asia. Each of these mutations is found in the context of its own haplotype, arguing that each has a unique origin. Each is in the C domain of the enzyme, involved in the intersubunit contact of the homotetramer catalytic unit.12 

If the gene frequencies of the mutations that cause a disease are known, its prevalence may be estimated by applying the Hardy–Weinberg equilibrium. When this is done, the assumption is made that the penetrance of the clinical disorder is 100%. How well the clinical expression approaches this ideal can be estimated by comparing the relative frequency of the mutations that are encountered in the general population with the frequency in the patient population. We have previously studied the ratio of various Gaucher disease mutations in the Jewish population with Gaucher disease on the one hand and in the general population on the other. Here we found marked overrepresentation of the 1226G mutation in the general population compared with the patient population.13 This implied that many of the persons who carry this particular mutation never came to medical attention as patients with Gaucher disease, and this is, indeed, the case.

A similar picture emerges in the case of PK deficiency. Table4 summarizes the frequency of the 1529A mutation among patients with hemolytic anemia who were of European, non-Gypsy ancestry. (Gypsies have been excluded because they have a unique PK deletion12). Among this population, the 1529A mutation is the most common, accounting for 28.3% of all the deficiency-producing alleles; its frequency in the general white population is 0.00203. The 1456T mutation, on the other hand, represents only 15.4% of the disease-producing alleles in the patient population, but its frequency in the general white population, at 0.00350, is appreciably higher than that of the 1529A. It is of interest, in this respect, that the 1456T mutation is only rarely found in the homozygous state in patients with hemolytic anemia and that when the homozygous state has been documented, the anemia is very mild or does not exist at all.14 Further attesting to the relatively mild nature of this mutation, there is a tendency for it to be found together with null (nonsense) mutations rather than missense mutations.

However, within these limitations, population data can be used to estimate the prevalence of PK deficiency in the white population. The approach that we have adopted is to select an “index mutation.” The frequency of this mutation in the patient population is taken to be p_i, and its frequency in the general population under investigation is taken to be g_i. The index mutation should have penetrance that approaches 100%, and it should be one of the more prevalent mutations in the patient population. In the case of PK deficiency, the 1529A mutation has been selected as the index mutation. It is the most common mutation among northern Europeans with PK deficiency, and its penetrance seems to be high because most homozygotes for this mutation have severe hemolytic anemia. Moreover, we know of no instances of siblings found in family studies to be homozygous for this mutation who did not also have hemolytic anemia. If the gene frequency of the index mutation,1529A mutation, in the general population is g_i and the frequency of all other PK mutations is g_o, then the frequency of all homozygotes and compound heterozygotes for PK deficiency mutations is g_i2 + 2 g_ig_o + g_o2, where g_i2 is the population frequency of homozygotes for the 1529A mutation, 2 g_ig_o is that of compound heterozygotes of 1529A mutations and other mutations, and g_o2 is that of homozygotes and compound heterozygotes of all other mutations. In this study we found the value of g_i to be 0.00203, but how do we obtain the value of the other mutations, g_o? If we assume that the representation of all mutations in the patient population is the same as in the general population, then the ratio of the index mutation to all other mutations in the general population will be g_i/g_o = p_i/p_o. Therefore, g_o = p_o×g_i/p_iwhere p_i is the frequency of the index mutation in the patient population and p_o that of all other PK mutations in the patient population. The fractional frequency of the occurrence of the 1529A mutation in the patient population, p_i, was found to be 0.283 (Table 4). The value of p_o, representing all other mutations in the patient population, is then 1 − 0.283 or 0.717, and the value of g_o, the frequency of mutations other than 1529A in the general population, is then calculated to be 0.00514. Using the values of 0.00203 for g_i and 0.00514 for g_o, the value of the expression g_i2 + 2 g_ig_o + g_o2 is 4.12 × 10⁻⁶ + 20.9 × 10⁻⁶ + 26.4 × 10⁻⁶, or 51 per million white population, approximately 10 000 patients in the United States. The standard error of the estimate of 51 per million can be assessed with the delta method,18 which uses a multivariate Taylor series to incorporate the sampling variability in both g_i and p_i into an overall estimate of the variance of g_i²/p_i2 because g_i2 + 2 g_ig_o + g_o2 reduces to this ratio. As expected from the relatively small numbers, this method provides a the standard error is 32.5 per million. This estimate would be increased by inbreeding, as in the Pennsylvania Amish community and in Gypsies, and would be decreased by failure of PK-deficient patients to survive. The number of patients actually identified seems to be smaller. In the past 25 years we have diagnosed PK deficiency in 201 patients. Because we do not know what percentage of all diagnoses in the United States were made in our laboratory, it is not possible to calculate the number of patients actually diagnosed in the United States. However, because there is a limited number of laboratories that carry out these assays, it appears that the actual cases of PK deficiency diagnosed falls far short of the 10 000 case estimate.

As we have pointed out above, the assumption that the distribution of mutations in the patient population is the same as in the general population is not entirely accurate; mild mutations such as 1456T are underrepresented among patients. Nevertheless, the approach we have used has the effect of correcting for those mutations that are underrepresented in the patient population because the estimate of the frequency in the general population depends on representation in the patient population. It thus has the effect of providing a fairly accurate picture of the patients in whom the disease actually develops.

The authors thank Dr James Koziol for his help in calculating the estimate of error.