Genotype–Phenotype Correlations of Glucose-6-Phosphate–Deficient Variants Throughout an Activity Distribution

Abstract

Background

Glucose-6-phosphate dehydrogenase (G6PD) deficiency is an X-linked disorder that may manifest as neonatal jaundice or acute hemolytic anemia. Quantitative assessment of G6PD activity in erythrocytes is required to definitively diagnose a deficiency. Most males and homozygous females have low enzyme activities, whereas heterozygous females may have a range of activities. We sought to examine G6PD genotype–phenotype associations to identify an activity cutoff above which G6PD deficiency is unlikely.

Methods

Ninety-five residual samples were randomly selected to represent the various regions of a G6PD activity distribution. DNA was isolated from the leukocyte fraction and sequenced using the Sanger method. ROC curves were used to establish cutoffs.

Results

Thirteen variant alleles were identified, including 1 not previously reported. In the very deficient activity range, we found males and homozygous females of both class II and III variants. In the deficient category, we found predominantly class III males and heterozygous females. The presumed deficient category contained class III and IV variants and nonvariants. An activity cutoff of <7.85 U/g hemoglobin (Hb) was 100% sensitive and 94% specific for identifying a G6PD-deficient male, and a cutoff of <8.95 U/g Hb was 90% sensitive and 82% specific for a deficient female.

Conclusions

The observed activity groupings were not because of a particular variant class. Cutoffs to identify the presence of a deficiency variant for males and females may be useful when trying to decide whether to recommend genetic analysis.

Impact Statement

Although most glucose-6-phosphate dehydrogenase (G6PD)–deficient males and homozygous females have low G6PD activities, heterozygous females and some males have activities that may or may not suggest deficiency. After sequencing of the exon region of samples within each area of an activity distribution, we prepared ROC curves to identify activity cutoffs above which a deficiency is unlikely. A cutoff with 100% sensitivity and 94% specificity was identified for males, whereas a cutoff with 90% sensitivity and 82% specificity was identified for females.

Glucose-6-phosphate dehydrogenase (G6PD)² deficiency is an X-linked recessive disorder of the pentose phosphate (hexose monophosphate) pathway. In erythrocytes, this pathway supplies NADPH to maintain sufficient levels of reduced glutathione. If NADPH and reduced glutathione are low, erythrocytes may lyse because of oxidative damage, resulting in clinical manifestations such as acute hemolytic anemia or jaundice. Several review papers give an overview and/or a historical perspective (1–4). Deficiency phenotypes have been grouped according to severity and clinical manifestations (classes I–IV) by the WHO: class I, severe enzyme deficiency and causing nonspherocytic hemolytic anemia; class II, severe (<10% normal activity); class III, moderate (10%–60% normal activity); and class IV, mild or no deficiency (60%–100% normal activity) (5, 6). More than 140 mutations have been identified, and most are single nucleotide variants that occur throughout the coding regions (7). Online databases of these variants are available (http://www.bioinf.org.uk/g6pd/db/).

Neonatal screening has long been recommended in areas where the prevalence of G6PD deficiency is relatively high, such as malaria-endemic areas of Africa and Asia. However, deficiency alleles are now also prevalent in North and South America and in parts of northern Europe (3). Many different screening tests are available, including qualitative, semiquantitative, and quantitative assays (8, 9). Various screening strategies for identification of G6PD-deficient newborns have been discussed (10, 11). In some geographic areas, DNA-based screening for targeted mutations may be useful (8, 11, 12); however, individuals with a rare variant would be missed. Enzymatic activity screening tests generally have a high diagnostic sensitivity and specificity for detecting severely deficient hemizygous males and homozygous females but a lower sensitivity for detecting G6PD  ³ mutation heterozygous females who have a range of G6PD activities because of random X-inactivation. Higher activity cutoffs may allow identification of more heterozygous female neonates (13).

For definitive diagnosis of G6PD deficiency, a quantitative analysis of G6PD activity is required. G6PD-deficient individuals are then advised to avoid certain drugs, foods, or other oxidizing agents that may precipitate a hemolytic crisis (2, 3). A variety of drugs have been implicated, including some antimalarial drugs, sulfonamides, and rasburicase. These drugs directly or indirectly produce H₂O₂, which can deplete the available reduced glutathione (2). In the case of rasburicase, a recombinant enzyme used to lower uric acid levels in plasma, H₂O₂ is produced as uric acid is converted to allantoin. However, because a variety of factors may affect whether a crisis occurs, clinical risks for moderate G6PD phenotypes are not easily predicted. Some studies to predict risk of neonatal hyperbilirubinemia group infants into 3 risk categories: low, intermediate, and high G6PD activities (14, 15). The Clinical Pharmacogenetics Implementation Consortium Guideline (www.pharmgkb.org) has 3 risk categories of hemolytic anemia (low risk, at risk, and unknown risk) regarding rasburicase administration to G6PD-deficient patients. Understanding what dose of primaquine might be safe for use for various deficiency types in malaria transmission-blocking programs is still being investigated (16, 17).

The objective of this study was to examine the G6PD genotype–phenotype associations of selected individuals with G6PD activities both within and below an established reference interval to identify an activity cutoff above which G6PD deficiency is unlikely for both males and females of our population.

MATERIALS AND METHODS

Samples

Whole blood samples submitted to ARUP Laboratories for G6PD activity measurement were selected based on the result's location in a frequency distribution created from retrospective G6PD test results extracted from the laboratory information system. Based on that distribution, 95 residual samples were randomly chosen to represent 1 of 4 possible classification categories (very deficient, deficient, presumed deficient, and sufficient). The ages of patients from whom samples were used ranged from 0 to 81 years, but race and ethnicity information was not available. The University of Utah Institutional Review Board approved this project and its protocols.

G6PD activity and hemoglobin concentration

Whole blood samples were collected in tubes containing acid citrate dextrose, heparin, or EDTA. A hemolysate was made with 2% saponin, and the lysate was centrifuged to remove cellular debris. The G6PD activity and hemoglobin (Hb) concentrations in each lysate were determined using commercially available methods (Trinity Biotech and Roche Diagnostics, respectively) on a Cobas c501 chemistry analyzer (Roche). G6PD activity was measured at 37 °C and reported as units per gram of Hb.

G6PD gene sequencing

DNA was extracted from leukocytes isolated from aliquots of the same whole blood samples used for enzyme activity measurement using a Roche MagNA Pure System and kit. Sanger sequencing of the entire G6PD (RefSeq number NM_001042351.1) coding region and intron/exon boundaries was performed using an Applied Biosystems PCR System 9700 and Capillary Electrophoresis System 3730 (PCR primers and cycling parameters available on request). Mutation Surveyor version 4.0.6 (Softgenetics) was used to identify mutations. Samples were designated as nonvariant if there was no mutation found within the G6PD coding regions.

RESULTS

The frequency distribution created from 19717 retrospective G6PD test results extracted from the laboratory information system is shown in Fig. 1. The distribution is trimodal with the majority of results being greater than the lower limit of the established reference interval of 9.9 to 16.6 U/g Hb. Minor populations of 1.4 to 4.5 U/g Hb and <1.4 U/g Hb represented the 2 other modes. There was a continuum of results between 4.6 and 9.8 U/g Hb. Accordingly, the 95 residual samples used for G6PD gene sequencing were classified into 1 of 4 categories: very deficient (<1.4 U/g Hb; n = 15), deficient (1.4–4.5 U/g Hb; n = 20), presumed deficient (4.6–9.8 U/g Hb; n = 41), and sufficient (≥9.9 U/g Hb; n = 19). DNA extraction failed from 2 presumed deficient samples. Of the 93 samples that were sequenced, 50 were obtained from males and 43 from females. Fig. 2 shows the distribution of G6PD activities for variants that occurred more than once in our population, as well as the distribution of nonvariants examined as part of this study. To compare our results with the various WHO classifications, we have taken the phrase “normal activity” to be the median value (13.2 U/g Hb) of our established reference interval.

A frequency distribution of 19717 G6PD activity results reveals 3 distinct modes with the majority of results occurring between 9.9 and 16.6 U/g Hb (sufficient), and 2 minor populations of 1.4 to 4.5 U/g Hb (deficient) and <1.4 U/g Hb (very deficient).

Fig. 1.

There is a continuum of results between 4.6 and 9.8 U/g Hb (presumed deficient).

Distribution of G6PD activities by variant type.

Fig. 2.

Only variant groups in which there were ≥2 representatives are shown. Males are represented by squares, homozygous females by filled circles, and heterozygous females by half-filled circles. Nonvariant males and females are also included. Dotted horizontal lines indicate 10% and 60% of our median G6PD activity, representing WHO classifications.

Table 1 summarizes the ranges of activities for the various types of deficiency variants and nonvariants determined in this study. There were 13 different variant alleles identified, 1 of which (p.Arg454Pro) has not been previously reported. The variant p.Asp194Glu is found in the National Center for Biotechnology Information dbSNP database (rs145247580) as a variant of uncertain significance (18). Of the 93 samples, 74 (80%) had G6PD activities below the lower reference limit of 9.9 U/g Hb. Of these, 44 (59%) were class III A− variants, 4 (5%) were class II Mediterranean variants, and 4 (5%) were Viangchan variants (variably characterized as either a class II or III variant). Nine (12%) samples with activities <9.9 U/g Hb contained no variants. Other variants accounted for the remaining 13 samples. Of the 19 samples with activities >9.9 U/g Hb, 16 (84%) were nonvariants, 2 (10%) were from females with a class III A− variant (10.3 and 12.0 U/g Hb), and 1 (5%) was from a male with the class IV A+ variant (11.2 U/g Hb).

Table 1.

Observed ranges of G6PD activity with variant analysis by sex.

Sex	Zygosity	N	G6PD, U/g Hb	Protein variant designation	Variant	WHO classification
M	Hemizygous	1	0.4	p.Arg454Cys	Union	II
M	Hemizygous	1	0.7	p.Asn126Asp, p.Asp181Val	Santa Maria	II
M	Hemizygous	1	0.7	p.Arg454Pro	Previously unreported	II
M	Hemizygous	3	0.5–0.8	p.Ser188Phe	Mediterranean	II
F	Homozygous	1	0.6	p.Ser188Phe	Mediterranean	II
M	Hemizygous	1	0.9	p.Arg459Leu	Canton	II
F	Homozygous	1	1.2	p.Arg459Leu	Canton	II
M	Hemizygous	3	0.9–2.7	p.Val291Met	Viangchan	II, III
F	Heterozygous	1	7.8	p.Val291Met	Viangchan	II, III
M	Hemizygous	20	1.1–7.8	p.Asn126Asp, p.Val68Met	A−	III
F	Homozygous	1	1.0	p.Asn126Asp, p.Val68Met	A−	III
M	Hemizygous	1	1.2	p.Asn126Asp,p.Leu323Pro	A−	III
F	Heterozygous	1	3.1	p.Asn126Asp, p.Val68Met, p.Leu323Pro	A−	III
F	Heterozygous	19	2.4–9.4	p.Asn126Asp, p.Val68Met	A−	III
F	Heterozygous	2	10.3–12	p.Asn126Asp, p.Val68Met	A−	III
F	Heterozygous	2	5.8, 8.7	p.Asn126Asp, p.Leu323Pro	A−	III
M	Hemizygous	1	4.2	p.Ala335Thr	Chatham	III
F	Heterozygous	1	4.2	p.Val68Met	Asahi	III
M	Hemizygous	1	6.5	p.Asp282His	Seattle	III
F	Heterozygous	1	6.7	p.Asn126Asp, p.Val68Met, p.Asp350His	A−/Mira d'Aire	III/IV
F	Heterozygous	1	7.5	p.Asn126Asp, p.Asp181Val, p.Asp350His	Santa Maria/Mira d'Aire	II/IV
F	Heterozygous	1	5.3	p.Asn126Asp	A+	IV
F	Heterozygous	1	9.2	p.Asp194Glu	Nonea	Nonea
M	Hemizygous	2	9.4, 11.2	p.Asn126Asp	A+	IV
M	NA	6	7.4–9.8	Nonvariant	NA	NA
M	NA	9	9.8–10.8	Nonvariant	NA	NA
F	NA	3	8.8–9.4	Nonvariant	NA	NA
F	NA	7	10–12.7	Nonvariant	NA	NA

^a

Variant of unknown significance reported without activity or name (18).

NA, not applicable.

Of the 15 samples randomly selected to represent the very deficient category of G6PD activity (<1.4 U/g Hb), 3 were from females and 12 were from males. The females included 1 each of the class II Mediterranean and Canton variants and a class III A− variant. The males included 1 each of the Union, Canton, and Santa Maria class II variants, 3 class II Mediterranean variants, 2 class II or III Viangchan variants, 3 class III A− variants, and 1 previously unreported variant. The previously unreported variant, p.Arg454Pro, had a G6PD activity of 0.7 U/g Hb. The 4 Mediterranean variants and the Union and Santa Maria variants had activities of 0.4 to 0.8 U/g Hb, whereas the A− variants had activities of 1.0 to 1.2 U/g Hb, and the Canton and Viangchan variants had activities of 0.9 to 1.3 U/g Hb.

Twenty samples in the deficient category (1.4 to 4.5 U/g Hb) were sequenced. An attempt was made to select samples from both males and females across this region, but a greater number of males (n = 14) were available. Of the entire group, 17 of 20 (85%) were class III A− variants, with 1 female containing the less common p.Leu323Pro variant. All females in this range were heterozygous. Other variants included a class II or III Viangchan variant (male, 2.7 U/g Hb), as well as the class III Asahi (female, 4.2 U/g Hb) and Chatham variants (male, 4.2 U/g Hb).

Thirty-nine samples (25 females and 14 males) were sequenced with activities in the presumed deficient region of 4.6 to 9.8 U/g Hb. Of these, 23% (6 males and 3 females) had no mutations and 59% (6 males and 17 females) were class III A− variants. One A− variant was the less common p.Asn126Asp, p.Leu323Pro. The remainders were class III or IV variants and 1 variant previously reported in the National Center for Biotechnology dbSNP database (rs145247580) but without a corresponding name or activity given (18). These named variants included 1 Viangchan, 1 Santa Maria in combination with Mira d'Aire, 1 A− in combination with Mira d'Aire, 1 Seattle, and 2 A+ variants.

Nineteen samples in the sufficient region of ≥9.9 U/g Hb were sequenced. Two (11%) females were heterozygous for the class III A− variant, and 1 male had the class IV A+ variant that is often not considered to be a deficiency variant. The remainder (84%) had no G6PD mutations.

Because a better understanding of sensitivity and specificity at various activity cutoffs might serve to guide additional testing or treatment recommendations, ROC curves were generated for males and females by comparing those with deficiency alleles (class II and III variants) to those with only nondeficiency alleles (nonvariant and class IV variants). The variant of uncertain significance, rs145247580, was not included in the analysis. Results are shown in Fig. 3. For the males (32 deficient and 17 nondeficient alleles), a cutoff of <7.3 U/g Hb would be 97% sensitive and 100% specific for identifying a G6PD-deficient male, whereas a cutoff of <7.85 U/g Hb would be 100% sensitive and 94% specific. For females (31 deficient and 11 nondeficient), a cutoff of <8.95 U/g Hb would be 90% sensitive and 82% specific, whereas a cutoff of <8.75 would be 87% sensitive and 91% specific at identifying a G6PD deficiency.

ROC curves for males (A) and females (B).

Fig. 3.

The area under the curve was 0.998 for males (n = 49) and 0.891 for females (n = 42). To maximize sensitivity, a cutoff of <7.85 U/g Hb was chosen for males (sensitivity 100%, specificity 94%) and <8.75 U/g Hb for females (sensitivity 91%, specificity 87%).

DISCUSSION

Distributions of G6PD activities in neonates reported in literature show separation into 2 groups (deficient and nondeficient) for males but no clear separation for females (13, 15, 19). However, in studies of broader age-groups (9, 20) or more diverse populations (14), the separation between groups is less defined even for males. Our distribution of G6PD activities from an age-diverse group of patients seemed to suggest 2 slightly overlapping groups at the low end of G6PD activities in Fig. 1 and categorized by us as very deficient or deficient, respectively.

The very deficient region (<1.4 U/g Hb) contained a variety of class II variants and some A− variants. Although the A− variant is considered a class III (moderate deficiency) variant (4, 7), it is well known that affected males and homozygous females typically have lower activities than heterozygous females because of the X-linked nature of the disorder. The deficient region (1.4–4.5 U/g Hb) contained mostly A− variants but also 1 each of the class III Asahi and Chatham variants and a Viangchan variant (previously identified as either a class II or class III variant). Our results are similar to what has been reported before for these variants. Finding class II variants (<10% normal activity) such as Union, Mediterranean, Santa Maria, and Canton in the very deficient region was expected (4, 7) because our reference interval is 9.9 to 16.6 U/g Hb (median, 13.2). However, 1 older review mentions both class II and III activities for the Canton variant (21), and a more recent article shows a range of activities for Mediterranean heterozygous females similar to what they observed for heterozygous female A− variants (22).

The single Asahi variant in our study had a G6PD activity of 4.2 U/g Hb (31% of the reference interval median). This agrees with its previous classification as a class III variant (7). For the sole Chatham variant male in this study, an activity of 4.2 U/g Hb (32% of the reference interval median) meets expectations for a class III variant. Some literature mentions Chatham variants as class III (4), but most reports seem to be of class II variants (7, 23–25) with activities much lower than we observed. Our results for the 4 Viangchan variants ranged from 0.9 to 2.7 U/g Hb for hemizygous males (n = 3) and 7.8 U/g Hb for a heterozygous female, suggesting that it spans both WHO classifications. Although the literature often describes this as a class II variant (4, 7), a study with a larger number of heterozygous females shows a broad range of activities (26). We suggest that for many class II and III variants, if enough heterozygous females are examined, a wider range of activities will be found. Thus, it may not be appropriate to always think of a particular variant as either class II or class III. This may also be true of males with the Viangchan variant as suggested by our data (see Fig. 2).

The samples designated as presumed deficient were expected to contain class III and IV variants as well as nonvariants because activities spanned the range of 4.6 to 9.8 U/g Hb. Two of the 3 class IV A+ variants were found here, as was a single Seattle variant, described as class III previously (21) and with activities that are 8% to 21% of normal (6). In our study, the observed activity of 6.5 U/g Hb (49% of the reference interval median) of the Seattle variant is within expectations for class III variants. The p.Asp194Glu variant reported once before (rs145247580) was classified as a variant of uncertain significance (18). Our sample had a G6PD activity of 9.2 U/g Hb (70% of the reference interval median), which would suggest it is a class IV variant. The compound heterozygous females, which contained 2 variant alleles, also fell into this region of activity, as did 9 samples (3 females and 6 males) with no variants. Thus, there is no clear separation of variant and nonvariant activities for either males or females in this region.

The range of A− variants (44% of our population) was the largest range observed, extending across all categories of the deficient regions and into the sufficient regions. Although there is molecular heterogeneity in this variant (27), it does not account for the variety in activities observed in Table 1. Our slightly wider range of activity for hemizygous A− males compared with a study by LaRue and colleagues (9) is likely because of the larger number of males in our study. As with the LaRue study, the wider range of G6PD activities for A− females than for A− males is expected because of X-inactivation in females.

Before establishing cutoff values for testing, the population and purpose should be considered. Various studies have dealt with cutoffs to maximize the percentage of female heterozygotes detected in neonatal screening protocols (13, 15) or for risk of neonatal bilirubinemia (14) and suggest that higher cutoffs allow detection of more female heterozygotes. For adults, activities are usually measured on incidence of a hemolytic crisis or before administration of certain drugs (e.g., rasburicase, primaquine). However, a study by Maffi (24) reports G6PD screening of Italian blood donors without any incidence or suspicion of G6PD deficiency. They found 33 subjects (25 male, 8 female) with low G6PD activity of 0.27 to 5.98 U/g Hb, which revealed 48% as class II alleles and 43% as class III alleles. Further studies of cutoffs for these purposes may be warranted.

Because our laboratory routinely receives inquiries regarding the meaning of G6PD activities that fall within the 4.6 to 9.8 U/g Hb range, we sought to determine whether a cutoff value for deficient vs nondeficient individuals would be useful in our patient population. For our ROC analysis, we considered those with no variant in the exon region and those with an A+ allele to be nondeficient. Our population of deficient samples was either class II or class III variants. ROC analysis for males suggested that if the G6PD activity is <7.3 U/g Hb, males can be expected (97% sensitivity, 100% specificity) to have a mutation that causes a deficiency. However, for screening purposes, one might choose the cutoff of <7.85 to achieve 100% sensitivity for this group. For females, a cutoff of <8.75 or <8.95 U/g Hb would give useful sensitivity (87% or 90%) while still having a reasonable specificity. This type of cutoff could be useful in deciding whether to recommend genetic sequencing in geographic areas where a variety of mutations might be present in the population.

The newly discovered variant (c.1361G>C, p.Arg454Pro) identified in our study, and hereby referred to as G6PD Salt Lake, is predicted to be class II based on having an activity of 0.7 U/g Hb (5% of the reference interval median). Interestingly, the class II Union variant is a modification of the same amino acid (p.Arg454Cys, c.1360C>T) and appears to be a severe deficiency variant with activities <10% of the reference interval median in our study and others (28, 29). However, the G6PD Andalus variant (p.Arg454His, c.1361G>A) also shows low activities (29, 30) and has apparently been associated with chronic nonspherocytic hemolytic anemia, making it a class I variant by definition (30). Studies of recombinant versions of these mutants showed similar kinetics, however, and 3-dimensional structures suggest that Arg 454 forms a salt bridge with Asp 286 (30). The p.Arg454Pro would not be able to form this salt bridge either, thus explaining the greatly reduced activity we observed for this variant. Because our samples were deidentified for this study, we were not able to correlate this newly discovered variant with clinical symptoms for further classification as a class I or class II variant.

A limitation of our study was the use of only 95 samples for genotyping and the lack of associated clinical history. Because sample size was limited, we have not attempted to establish a separate cutoff for adults vs neonates, which would be needed for neonatal screening purposes. Our laboratory was interested in understanding the usefulness of a cutoff value for identifying G6PD deficiency when activities are below our reference interval but still well above that for severely deficient males and homozygous females. Sensitivity and specificity of our proposed cutoffs are better for males than females but may be useful in our laboratory setting for both. However, the possibility of false-positive or false-negative findings will remain, especially for heterozygote females.

2 Nonstandard abbreviations

 
• G6PD

  glucose-6-phosphate dehydrogenase

 
• Hb

  hemoglobin.

3 Human gene

 
• G6PD

  glucose-6-phosphate dehydrogenase.

Author Contributions:  All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved.

Authors' Disclosures or Potential Conflicts of Interest:  No authors declared any potential conflicts of interest.

Role of Sponsor: No sponsor was declared.

Previous presentations/abstracts: Oral presentation at 51st annual meeting of the Academy of Clinical Laboratory Physicians and Scientists as “Genotype-Phenotype Associations for Low Glucose-6-Phosphate Dehydrogenase Activity in Red Blood Cells” and published abstract in Am J Clin Pathol 2017;147:S175–6.

Acknowledgments

The authors thank the technical staff in the ARUP Special Chemistry laboratory for G6PD activity testing, the ARUP Genetics Sequencing laboratory for DNA extraction and Sanger sequencing, and Kim Kalp for obtaining and deidentifying samples.

Author notes

Current address: Core Laboratory for Clinical Studies, Washington University in St. Louis, St. Louis, MO 63110.

Current address: TriCore Reference Laboratories, 1001 Woodward Place NE, Albuquerque, NM 87102.