Main

Cystic fibrosis (CF) is an autosomal recessive disease caused by mutations in the CF transmembrane conductance regulator (CFTR) gene.1 To date, more than 1,900 unique mutations have been described in the CFTR gene.2 The observed phenotypic heterogeneity of CF is due to the variable impact of the underlying mutations, in combination with the action of genetic modifiers and environmental factors.3 The phenotypic spectrum of CF ranges from a mild disease, found in ~15% of CF patients, to a progressive, multisystem disease primarily involving the pulmonary, pancreatic, and gastrointestinal systems.1,4 Individuals with a mild form of the disease currently have a median survival of 56 years, whereas those with classic CF have a median survival of ~37 years, with the most common cause of death being respiratory failure.1

Although CF is most prevalent in Caucasians (1 in 2,500), it is a panethnic disease in North America, found in all races, including African-American (1 in 15,000), Hispanic-American (1 in 13,500), and Asian-American (1 in 35,000) individuals.5,6 This reflects the heterogeneity of the North American population. The incidence of CF in the North American population is primarily due to the higher-frequency alleles present mainly in the Caucasian population. Alleles found in relatively high frequency in other races are less likely to be found in high frequency in the general population due to the lower prevalence of CF in these populations.4,7

In 2001, the American College of Medical Genetics and Genomics (ACMG) and the American College of Obstetricians and Gynecologists (ACOG) recommended a 25-mutation panel for CF carrier screening, which was later refined to one with 23 mutations.8 Selection of CF variants was based on pathogenicity and an allele frequency of ≥0.1% of the general affected US population.6 The panel did include variants predicted to be panethnic. This panel has performed well in the past decade and is predicted to have detected close to 90% of CF carriers in the white or Ashkenazi Jewish populations.4 However, the panel does not detect CF carriers equally for all races.4,7 Although the detection rate for non-Hispanic whites is 88%, the detection rates for Hispanic-American (72%), African-American (64%), and Asian-American (49%) CF carriers are significantly lower.4 In 2011, the ACOG recognized the fact that assigning a single ethnicity to individuals with CF had become increasingly difficult and expanded its 2001 recommendations to recommend offering CF carrier screening to all women regardless of ethnicity.4 A number of studies have identified apparent race-specific alleles in affected individuals. Those alleles in highest frequency have been recommended as additions to the current 23-mutation panel as a means to increase CF carrier detection rates in ethnic populations to levels similar to that of non-Hispanic whites.9,10,11,12,13 Several commercial reference laboratories now offer extended panels consisting of all variants from the 23-mutation panel, as well as an increasing number of race- or region-specific alleles.9,14 An analysis of the performance of these extended panels has been difficult due to the paucity of published data. Heim et al.14 examined a total of 5,840 chromosomes from a variety of races and demonstrated that a detection rate in the range of 70–95% could be achieved for all races, except Asians, with a panel of 50–70 CFTR mutations. Rohlfs et al.9 reported on the use of a 97-mutation panel among 364,890 individuals referred for CF carrier screening. They confirmed that the mutation frequency distribution for Caucasians was significantly different from that observed for other races. The purpose of the current study was to compare the mutation frequency distribution observed with a 32-mutation panel versus a 69-mutation panel used for CF carrier screening. The CF carrier detection rate of each panel for different ethnic groups was also determined, and the implications of these data to the composition of future screening panels are discussed.

Materials and Methods

Patients

In this study, we performed testing of samples from 1,711,138 individuals referred to LabCorp’s Center for Molecular Biology and Pathology for CF screening from January 2005 through December 2010. Samples were tested using either a 32- or 69-mutation CFTR panel. The CF32 panel, which includes all 23 mutations recommended by the ACMG and the ACOG, was used to test 1,601,308 cases. A total of 109,830 individuals were tested using the expanded CF69 mutation panel, which also includes all 23 mutations recommended by the ACMG/ACOG. The majority of samples for CF screening (95%) were submitted by obstetricians/gynecologists who followed the ACMG/ACOG guidelines. The remainder of the samples came chiefly from in vitro fertilization clinics, family practitioners, internists, urologists, pulmonologists, and gastroenterologists. Samples were received from all 50 US states. Institutional review board approval was not required because testing was performed for clinical purposes. Patient ethnic background was provided by the referring physicians in 40% of cases. Ethnicity was categorized as Native American, Asian, African American, Caucasian, or Hispanic. Because the indication for testing was not routinely captured, we presume the majority of cases were referred for population-based carrier screening. However, we realize that similar to Strom et al.,7 in some cases we could have tested known carriers, symptomatic individuals, or patients’ relatives. We excluded all cases wherein two mutations were identified. A paired sample z-test was used to test the difference in detection rates within different ethnic groups.

Molecular testing

DNA was extracted from peripheral blood and/or buccal swabs using standard nonorganic protocols. DNA samples were screened for CFTR gene mutations using the 32-mutation panel (oligonucleotide ligation assay; Celera Diagnostics, Alameda, CA) or the 69-mutation (Universal Array Platform, Luminex, Toronto, Ontario) panel approved by the US Food and Drug Administration. The 69-mutation panel was a combined panel that included variants approved by the Food and Drug Administration with 10 additional variants (R117C, R352Q, S364P, 3120G>A, 2869insG, G480C, 405+3A>C, 1812-1G>A, 444delA, and F311del) added on the basis of their published frequencies and relevancy to CF.

Oligonucleotide ligation assay

The oligonucleotide ligation assay was performed according to manufacturer recommendations (Celera Diagnostics). In brief, genomic DNA was amplified in a multiplex PCR. Next, the multiplex oligonucleotide ligation was performed on CFTR amplicons. The final oligonucleotide ligation assay products were size fractionated using an ABI3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA). The multiplex signal was analyzed with Gene Mapper Software (Applied Biosystems).

Luminex Universal Array Platform

Patient DNA was amplified with multiplex PCR, then allele-specific primer extension was used to discriminate among the alleles. The allele-specific primer extension products, 5′-tagged with 24-mer sequences selected from the Tm 100 Universal Sequence Set (Luminex), were captured by their tag complements, which were chemically coupled to polystyrene beads (Luminex). The xMAP instrument (Luminex Corporation, Austin TX) was used to examine the beads for both normal and mutation sequences. The data were analyzed by TDAS software (Luminex).

Results

The 32-mutation panel is composed of the 23 variants from the ACMG/ACOG screening panel combined with 9 additional mutations, namely, 3876delA, R347H, S549N, 3905insT, 1078delT, V520F, 394delTT, S549R, and 2183AA>G. Table 1 gives a frequency distribution of the mutations detected by the 32-mutation panel. Thirty of the 32 mutations on the panel were detected with a frequency that exceeded the 0.1% ACMG/ACOG prevalence threshold among affected individuals. Even the 1078delT mutation that was removed from the ACMG/ACOG screening panel achieved 0.1% prevalence.8 Only the S549R variant was observed at a lower frequency (0.07%). The frequency of the 2183AA>G mutation could not be established because this variant was added to the 32-mutation panel less than a year before the data collection end point. The argument can be made that the original ACMG/ACOG panel’s 0.1% prevalence threshold was derived from the panethnic cohort of the affected individuals, not the carriers routinely discussed in the literature and in the current study. We would like to emphasize that in our study we consistently compared allelic frequencies of the mutations detected by the 32- and 69-mutation panels with frequencies of ACMG/ACOG–selected variants detected in the carrier population.

Table 1 CFTR allele frequency identified by the CF32 mutation panel
Full size table

The 69-mutation panel incorporated all variants from the 32-mutation panel along with additional pathogenic variants present with relatively high frequency in Caucasian, Hispanic, and African-American populations. Genotyping data presented in Table 2 demonstrate that the frequency rate of the F508del variant detected by the 69-mutation panel is lower (60.5%) as compared with the rate obtained with the 32-mutation panel (68.7%; P < 0.01). This may be the result of population stratification because usage of the 69-mutation panel appears to be biased for testing among non-Caucasian individuals.

Table 2 CFTR allele frequency identified by the CF69 mutation panel
Full size table

The R117H variant was the second most common variant identified by both panels ( Tables 1 and 2 ). Among 5,198 carriers of the R117H variant detected by the 32-mutation panel, two copies of 5T alleles were detected in only 23 cases (0.44%). In 4,708 cases (90.56%), the 5T allele was not detected ( Table 3 ). The 5T status of the R117H variant could not be determined in 468 cases (9%) for carriers of 5T/7T or 5T/9T alleles. The 69-mutation panel identified two cases with two copies of the 5T allele (0.73%) among 274 R117H carriers, whereas the majority of carriers (89.41%) did not carry a 5T allele ( Table 4 ). In 27 occurrences (9.86%) the 5T background of the R117H variant could not be identified.

Table 3 Frequency of 5T/7T/9T genotypes as a result of R117H reflex testing
Full size table
Table 4 Carrier frequencies by ethnic group identified by the 32- and 69-mutation panels
Full size table

The variants D1152H (4.0%) and L206W (2.4%), each associated with a variable CF phenotype, were the most common variants not present on the ACMG/ACOG 23-mutation panel. The Hispanic-specific variant 3876delA (1.0%) and the panethnic variant F311del (0.8%) were the next most common variants from the extended panel.

Overall, 53 mutations were detected by this panel with a frequency that met or exceeded the ACMG-proposed threshold of 0.1% ( Table 2 ). This threshold could not be reached by 10 variants: the 2143delT, A455E, S549R, Y122X, and M1101K mutations, typically observed in Caucasians; 935delA, 2869insG, and Q890X in Hispanics; and 405+3A>C and G480C in the African-American population. In addition, six variants were not identified in our target population. These were 1898+5G>T, 444delA, G330X, S364P, K710X, and S1196X. The low frequency of the A455E mutation (0.07%) from the ACMG/ACOG panel may be explained by patient preselection through the use of the 32-mutation panel, or random drift. Although the relative frequencies of test populations screened using either the 32-mutation panel or the 69-mutation panel were generally similar ( Table 4 ), there were more individuals with reported Caucasian ethnicity in the 32-mutation panel (27.4%) than in the 69-mutation panel (14.8%). The carrier frequencies detected in Caucasians by both panels were the same, 1 in 28 for the 32-mutation panel and 1 in 27 for the 69-mutation panel. There was higher representation of the Hispanic population in the 69-mutation panel (12.7%) as compared with the 32-mutation panel (5.5%). Consequently, the carrier frequency observed in Hispanics from the 69-mutation panel screening (1 in 48) was higher than that observed for Hispanics screened using the 32-mutation panel (1 in 69). By contrast, the carrier frequency detected by the 32-mutation panel for the Asian-American population (1 in 261) was higher than the frequency identified by the 69-mutation panel (1 in 313). The Native-American population demonstrated higher carrier frequency in a group tested by the 32-mutation panel (1 in 59) than in the group screened by the 69-mutation panel (1 in 67; Table 4 ). In general, the number of observations for the Asian-American and Native-American populations that underwent CF screening was low, especially for the group screened by the 69-mutation panel, and thus the significance of these observations is limited.

To evaluate the performance of the ACMG/ACOG panel, we analyzed groups of carriers from the 32- and 69-mutation panels using only 23 ACMG/ACOG variants. Then we compared carrier frequencies and detection rates in different ethnic groups using the 23 ACMG/ACOG variants versus the 32- and 69-mutation panels. Of the carriers detected by the 32-mutation panel, the ACMG/ACOG panel would identify 97.4% in the Native Americans, 95.5% in Asian Americans, 97.3% in African Americans, 98.9% in Caucasians, 94.4% in Hispanic Americans, and 98.3% in two groups without reported ethnicity that were screened using the 32-mutation panel (see Supplementary Table S1 online).

The ACMG/ACOG panel detects only 78.3% of carriers in the African-American ethnic group identified by the 69-mutation panel (P < 0.01). The CF69 panel identified a further 21.7% of CF mutations (12 variants) that are not present on the ACMG/ACOG 23-mutation panel (see Supplementary Table S2 online). Six of these variants were specific to African Americans (R75X, G480C, A559T, 2307insA, 3791delC, and S1255X). The 69-mutation panel did not significantly improve the detection rate demonstrated by the 23 ACMG/ACOG mutations (93.5%) in the Caucasian population (P = 0.123), although an additional 16 variants were identified (see Supplementary Table S2 online).

In Hispanic individuals, the ACMG/ACOG panel would have detected only 68.2% of CFTR mutations identified by the 69-mutation panel. The CF69 panel identified 15 additional variants (31.8%; P < 0.01; see Supplementary Table S2 online). Two variants, L206W and D1152H, accounted for 19.1% of the mutant alleles detected in this ethnic group. There were two variants identified only in Native Americans (40%), and none were unique to the Asian-American group.

Discussion

Universal CF carrier screening panels must balance detection of CF carriers in the highest-prevalence populations, Caucasians and Ashkenazi Jews, with the need to adequately screen for a panethnic autosomal recessive disease within a racially heterogeneous North American population. The introduction of the ACMG/ACOG panel in 2001 was a huge step forward in universal CF carrier screening. The 23-mutation panel has performed well over the years and has met the expectations of a minimum CF carrier screening panel.5,15 However, the panel does not perform as well in regions with a higher percentage of African Americans, Hispanic Americans, or members of other ethnic groups.9,13,14,16 The ACMG/ACOG panel was originally designed for the Caucasian and Ashkenazi Jewish population and excludes mutations that are important in other ethnic groups “that are significant components of the American mosaic.”16 One may reduce bias in CF carrier screening by using a screening panel containing CFTR variants representing the population to be screened. However, much of the CF carrier screening is performed at reference laboratories, where tailored CF screening panels are not tenable. Instead, extended CF carrier screening panels are used. These panels are composed of the ACMG/ACOG CFTR variants and additional variants that represent mutations found in high frequency in a number of racial or ethnic groups.9,13

The most common variants from the 32-mutation panel are high-frequency alleles found on the ACMG/ACOG panel ( Table 1 ). The nine additional variants detected only an additional 0.9% of CFTR mutations. At least one CFTR mutant allele was identified for 63 of the 69 mutations (88.6%) making up the extended 69-mutation panel ( Table 2 ). Forty variants not present on the 23-mutation panel accounted for 14.4% of all mutations, similar to the results reported by Rohlfs et al.9 using a 97-mutation panel. These authors identified 64 mutations that were not part of the 23-mutation panel, accounting for 13% of the mutations identified.

Several alleles not found on the ACMG/ACOG panel were found at relatively high frequency ( Table 2 ), including D1152H (4.0%), L206W (2.4%), c.3744delA (0.9%), F311del (0.8%), R117C (0.7%), and S549N (0.6%). Clearly, the variants found on the 23-mutation panel are not the only variants present at a relatively high frequency among CF carriers.

The disparity in our study between the proportions of Caucasian individuals in the populations screened using the 32-mutation panel (27.4%) and the 69-mutation panel (14.8%) probably results from a clinician bias toward utilization of expanded panels for non-Caucasians and a relaxation in ethnicity data collection as CF carrier screening test volumes increased and other ethnic groups began to be included in the screening. For example, Strom et al.7 used the 32-mutation screening panel and observed ~60% Caucasians in their test population, whereas Rohlfs et al.,9 using a 97-mutation panel, observed only 43% Caucasians in their test population.

The observed carrier frequencies for both the 32-mutation and the 69-mutation panel screening populations were very similar to those previously observed for Caucasians (1/28), individuals without ethnicity data provided (1/33), and individuals from large heterogeneous test populations (1/36).7,9

The carrier frequency observed for the test population as a whole is expected to be lower than that observed for Caucasians because the presence of ethnicities in which CF is less prevalent dilutes the overall carrier frequency. The similarity in the carrier frequencies between the overall test population and individuals for whom ethnicity data were not collected suggests an absence of bias in terms of ethnic makeup of the latter group, as reported by Rohlfs et al.9

The carrier frequency for Hispanics was greater in the population screened with the 69-mutation panel (1 in 48) than the frequencies obtained with either the 32-mutation panel (1 in 69) or the previously reported 23-mutation panel (1 in 58).17 This suggests greater mutation detection by the 69-mutation panel due to the addition of variants specific to the Hispanic population. In fact, eight non–ACMG/ACOG variants were observed only in Hispanics screened by the 69-mutation panel (see Supplementary Table S2 online). These variants are 3876delA, S549N, 406-1G>A, 3199del6, W1089X, R1158X, R352Q, and 2183AA>G, and they account for 8.1% of the mutations detected in the Hispanic population. In addition, these variants include three of the five variants that Sugarman et al. (2004)18 identified as accounting for 5.6% of CFTR mutations in Hispanic CF carriers and 7.6% of mutations in Hispanic CF patients. Three of these variants are also among the seven alleles recommended by Schrijver et al.13 to increase carrier detection rates in Hispanics. These data also suggest that the CF carrier frequency in Hispanics may be greater than previously anticipated.

The carrier frequency for African Americans was higher in the 69-mutation panel population (1 in 79) relative to the 32-mutation panel (1 in 105) or the 23-mutation-panel-derived data (1 in 107). These data reflect greater mutation detection by the 69-mutation panel. The data for Native Americans and Asian Americans are difficult to interpret due to the low number of race-specific variants in the panel and the low prevalence of CF in these populations.

Overall, use of the ACMG/ACOG panel would have detected only 78.3% of mutations in the African-American ethnic group screened by the 69-mutation panel. The extended panel detected 21.7% more mutations (P < 0.01) using six additional race-specific (R75X, G480C, A559T, 2307insA, S1255X, and 3791delC) and seven panethnic variants (see Supplementary Table S2 online). Among Hispanics, the 69-mutation panel detected 31.8% more pathogenic variants than the ACMG/ACOG panel (P < 0.001). Four ethnicity-specific variants (R352Q, 406-1G>A, 3199del6, and W1089X) and 12 panethnic variants were detected. Two variants, D1152H and L206W, account for 19.1% of all mutations identified in the Hispanic group and deserve special attention. In a combined African-American and Caucasian group, the relative frequencies for the two variants are only 1.6 and 3.2%, respectively (see Supplementary Table S2 online). Their phenotypic variability has created controversy around their inclusion in extended CF panels. Strom et al.7 advocate against adding these variants to CF panels, stating that detection of D1152H and L206W during carrier screening may increase the rate of pregnancy termination among parents who fear having a child with classic CF, not to mention inflating the mutation detection rate among Hispanic Americans.7 It is acknowledged now in the literature that the D1152H variant belongs to both the CF-causing and CFTR-related disorder groups,19,20,21 and, in conjunction with an established CF-causing mutation, it could still manifest as typical CF.19,22 Burgel et al.23 studied 42 patients with D1152H mutations and reported that the variant, in conjunction with a CF-causing mutation, can cause significant pulmonary disease, albeit with longer survival. One of the authors of this article observed similar results (K.J.F., unpublished data). Sosnay et al.24 did not find enough evidence for the D1152H variant to meet clinical and functional criteria consistent with disease and hence categorized the variant as indeterminate. We agree with the authors that variants regarded by some as indeterminate warrant periodic review “to quantify the probability of causing or not causing disease,”24 but their inclusion in conjunction with accurate counseling should not be summarily dismissed. Even without the D1152H variant, the use of the 69-mutation panel will improve the detection rate for the Hispanic population by 22.2% (P < 0.04).

The same disagreement exists with regard to the L206W variant. Clain et al.25 analyzed 36 clinical cases of CF patients who were compound heterozygous for L206W and a severe CF mutation. Twelve of the cases were from the authors’ own data, whereas the other 24 cases were obtained from the literature. Using clinical and functional studies, the group demonstrated that L206W is a disease-causing mutation with variable phenotype ranging from CF with pancreatic insufficiency to congenital bilateral absence of the vas deferens. Data from the CFTR2 project indicated that the L206W variant was observed in 82 CF patients worldwide and should be considered a disease-causing mutation.26 As such, we support this conclusion.

Data from screening of two different populations for different sets of mutations allow for examining the role of screening population sample size on mutation detection. Screening a population of 1.6 million individuals with the 32-mutation panel detects a set of unique mutations for each ethnic group. As expected for mostly rare alleles, as population size decreases, the probability of detecting a set of mutations also declines. However, even with the population size remaining small, the number of mutations detected increases as the panel size increases. These are expected results, but they shed light on the relative improvement in race representation for a given panel. Among Caucasians, the increase in panel size across the same relatively small sample population reflects an observation of ACOG in its latest guidelines: “it is becoming increasingly difficult to assign a single ethnicity to affected individuals.”4 Thus, the number of mutations detected in Caucasians across the same population increases when an extended panel is enriched with other ethnicity-specific variants. Another factor may be that extended panels may be used preferentially by physicians for diagnostic testing. Admixture and uncertain ethnicity may also contribute to higher detection rates for extended panels.

Based on our data alone, there were no obvious advantages of using the 69-mutation panel instead of the 32-mutation panel for a purely Caucasian population. Rohlfs et al.9, however, reported improved mutation detection with a 97-mutation panel as compared with the 23-mutation ACMG/ACOG panel in a well-defined Caucasian population. This discrepancy may arise from the contribution of 28 variants not included in the 69-mutation panel, especially 2184insA and CFTRdele2,3, which were the 15th and the 16th most common variants among Caucasians in the study by Rohlfs et al.9 The apparent higher representation of Caucasians (43%) reported by Rohlfs et al.9 in their target population as compared with our Caucasian group screened for 69 mutations (14.8%) is also probably a contributing factor.

Recently, Sosnay et al.24 extensively characterized 127 CF variants in a cohort of 36,696 mostly Caucasian CF patients (95%). According to the authors, use of those 127 variants for carrier screening would significantly increase sensitivity of testing for couples undergoing carrier screening, thus indicating possible advantages of this expanded panel in a primarily Caucasian population.

Our results demonstrate that extended panels have definable value for reference laboratories that analyze samples from many different regions of the country, with directors and genetic counselors well versed in results interpretation, and will benefit ethnic groups beyond the standard recommendation. There is also specific utility for the screening of partners of known CF carriers and equivocal diagnostic cases.

In conclusion, although the clinical utility of extended CF panels remains a topic of active discussion, their use improves the mutation detection rate for specific mutations within specific target populations.

Disclosure

V.V.Z. and M.E. are current full-time employees at Laboratory Corporation of America. The other authors declare no conflict of interest.