Background: Previous evaluations of HPLC as a tool for detection of hemoglobin variants have done so within newborn-screening programs and/or by use of stored samples. We describe a 32-month prospective study in a clinical diagnostic laboratory in which we evaluated the imprecision of HPLC retention times and determined the retention times for hemoglobin variants seen in a multiethnic setting.
Methods: We analyzed all samples on the Bio-Rad Variant II HPLC system. For normal hemoglobin fractions and hemoglobin variants, we recorded and analyzed their retention times, their proportion of the total hemoglobin (%), and the peak characteristics. We compared the imprecision of retention time with the imprecision of retention time normalized to the retention time of hemoglobin A0 (HbA0) and to the retention time of HbA2. Alkaline and acid hemoglobin electrophoresis, and in certain cases globin chain electrophoresis, isoelectric focusing, and DNA analysis, were performed to document the identities of the hemoglobin variants.
Results: The mean (SD) imprecision (CV) of the retention time was 1.0 (0.7)% with no statistical difference compared with the imprecision for normalized retention times. Among 60293 samples tested, we encountered 34 unique hemoglobin variants and 2 tetramers. Eighteen variants and 2 tetramers could be identified solely by retention time and 3 variants by retention time and proportion of total hemoglobin. Four variants could be identified by retention time and peak characteristics and eight variants by retention time and electrophoretic mobility. One variant (HbNew York) was missed on HPLC. Retention time on HPLC was superior to electrophoresis for the differentiation and identification of six members of the HbJ family, four members of the HbD family, and three variants with electrophoretic mobilities identical or similar to that of HbC. Six variants with electrophoretic mobilities identical or similar to that of HbS could be differentiated and identified by retention time and proportion of total hemoglobin. HPLC detected two variants (HbTy Gard and HbTwin Peaks) missed on electrophoresis.
Conclusions: The retention time on HPLC is reliable, reproducible, and in many cases superior to conventional hemoglobin electrophoresis for the detection and identification of hemoglobin variants. Confirmatory testing by electrophoresis can be eliminated in the majority of cases by use of retention time, proportion of total hemoglobin, and peak characteristics of HPLC.
The laboratory diagnosis of hemoglobinopathies and thalassemias is of growing importance, particularly because of an increasing requirement for antenatal diagnosis of significant disorders of globin chain synthesis. It has been recommended that all individuals of all ethnic groups except Northern European Caucasians be screened for variant hemoglobins, all ethnic groups for β-thalassemia trait, and selected ethnic groups for α-thalassemia trait (1). The identity of a hemoglobin variant is generally inferred from its electrophoretic mobility, its quantity, and the patient’s ethnic background. Family studies can be of considerable importance in elucidating the nature of disorders of hemoglobin synthesis, but definite identification can be achieved only by DNA analysis or amino acid sequencing (1)(2)(3).
Alkaline and acid hemoglobin electrophoresis are the most widely used methods for investigating hemoglobin variants and hemoglobinopathy. Alkaline electrophoresis is rapid, reproducible, and capable of separating common hemoglobin variants, such as hemoglobin A (HbA), HbF, HbS, and HbC, but HbS, HbD, HbG, and HbLepore are unresolved from each other, as are HbC, HbA2, HbO-Arab, and HbE. In addition, there are other variants with electrophoretic mobilities identical or similar to those of HbS and HbC. Consequently, acid electrophoresis is needed for the identification of the aforementioned variants. Nevertheless, these electrophoretic methods are still not able, in most cases, to separate HbD from HbG and HbLepore and, in some cases, HbE from HbO-Arab (2).
Hemoglobin fraction analysis by cation-exchange HPLC has the advantage of quantifying HbF and HbA2 along with hemoglobin variant screening in a single, highly reproducible system, making it an excellent technology to screen for hemoglobin variants and hemoglobinopathies along with the thalassemias (1),(4)(5)(6)(7). The simplicity of the automated system with internal sample preparation, superior resolution, rapid assay time, and accurate quantification of hemoglobin fractions makes this an ideal methodology for the routine clinical laboratory (5)(6). Numerous automated HPLC systems are now commercially available, and evaluations have been published (8)(9)(10)(11). The use of HPLC technology in the clinical laboratory setting has increased ∼12.5-fold in the past 10 years (12).
Much of the published literature on the use of HPLC for the investigation of hemoglobinopathies and thalassemias has evaluated its effectiveness in newborn-screening programs (13)(14)(15)(16). Other publications have evaluated its performance, in comparison with various other technologies (6)(17)(18), in the analysis of complicated α-thalassemia and β-thalassemia syndromes in Southeast Asia (19) and in the analysis of a small patient population and a reference collection of rare hemoglobin variants (9). We report here the results for HPLC performance in a large prospective study of 60293 samples over a 32-month period in a multiethnic population.
Material and Methods
Specimens were drawn into tubes containing dipotassium EDTA (Becton Dickinson Vacutainer Systems). All specimens were analyzed on the Bio-Rad Variant II HPLC system with use of the Variant II β-Thalassemia Short Program Reorder Pack (Bio-Rad Laboratories) as described in the instruction manual for the assay. Briefly, in this system the samples are mixed by the Variant II sampling station, diluted with the specific hemolyzing/wash buffer, and injected into an assay-specific analytic cartridge. The Variant II dual pumps deliver a programmed buffer gradient of increasing ionic strength to the cartridge, where the hemoglobin fractions are separated based on their ionic interaction with the cartridge material. The separated hemoglobin fractions pass through a flow cell, where absorbance is measured at 415 nm; background noise is reduced with the use of a secondary wavelength at 690 nm. The raw data are integrated by the Clinical Data Management software (Bio-Rad Laboratories), and a chromatogram/sample report is generated. The integrated peaks are assigned to manufacturer-defined windows derived from the retention time, i.e., the time in minutes from sample injection to the maximum point of the elution peak, of normal hemoglobin fractions and common variants (Table 1⇓ ). If a peak elutes at an retention time not predefined, it is labeled as an unknown.
Over a 32-month period, 60293 samples were analyzed in the Special Hematology Laboratory at Bellevue Hospital Center for quantification of hemoglobin fractions and screening for hemoglobin variants. For specimens that showed chromatogram patterns consistent with sickle trait, the presence of HbS was confirmed by use of the sodium metabisulfite reduction test (20). All non-A non-S variants were confirmed by alkaline and acid electrophoresis on the Helena Hemoglobin Electrophoresis System or the Helena SPIFE 3000 (Helena Laboratories) according to the manufacturer’s recommendations. The presence of HbH was confirmed by use of the brilliant cresyl blue test for inclusion bodies (21). Certain specimens were forwarded to the Mayo Medical Laboratories (Rochester, MN) for confirmation, where additional testing was performed by isoelectric focusing, globin chain electrophoresis, and unstable hemoglobin screen. Amino acid sequencing was performed on selected specimens.
For hemoglobin elution peaks with retention times <0.63 min, the proportion of total hemoglobin (%Hb) was determined by densitometry of the alkaline electrophoresis gel using the QuickScan 2000 (Helena Laboratories) and/or manual calculation. Manual calculation was accomplished by determining the peak area of each elution peak, i.e., the product of the height and the width at half-height as measured with a Vernier caliper. The %Hb was calculated by determining the area of a peak as a fraction of the total area of all hemoglobin peaks seen on the HPLC chromatogram.
All data analyses were performed with Minitab Statistical Software (Minitab Inc.). Student’s t-test was used to calculate statistical significance where n was ≥3 for each group being analyzed.
retention times and proportions of hemoglobin variants
Shown in Table 2⇓ are the number of observations, retention times, %Hb, interval period in months between the first and last observation for each hemoglobin variant, and the type of variant with the specific amino acid substitution for the 34 observed variants, the 2 tetramers, HbF, HbA0, and HbA2. The mean retention times and %Hb for HbF, HbA0, and HbA2 were determined by use of the first five normal specimens seen in each month of the study period.
evaluation of the retention times
The CVs of the retention times for HbF, HbA0, HbA2, and each of the 15 variants seen on three or more occasions were calculated. HbNew York was not included in these calculations because it coelutes with HbA0. It was postulated that normalization of the retention time of the hemoglobin variant to that of either HbA0 or HbA2 at the time of assay might minimize the imprecision of the assay attributable to changes in lots of reagents and columns. The CVs of the retention time normalized to the retention time for HbA0 or the retention time for HbA2 at each occurrence for each of the 15 variants were calculated. For each incident of the six variants with an retention time in the HbA2 window, the mean retention time for HbA2 in the preceding five samples and the following five samples was determined and used in the calculation. The mean (SD) of the CVs for the retention time, the retention time normalized to the retention time for HbA0, and the retention time normalized to the retention time for HbA2 were calculated. There was no statistical difference in imprecision between the mean (SD) CV for the retention time [1.0 (0.7)%] and the mean CV for the retention time normalized to the retention time for HbA0 [1.3 (0.5)%; P = 0.10] or normalized to the retention time for HbA2 [1.0 (0.4)%; P = 0.51]. For this reason, subsequent analysis of the data was done with only the retention time.
The SD for the retention times of HbF, HbA0, HbA2, and the 15 variants seen on three or more occasions did not correlate with either the retention time (P = 0.889) or the %Hb (P = 0.288), demonstrating that the SD is independent of these two variables. The SD, the measure of the variation around the mean for the retention time, was therefore used to predict the statistical difference of the retention time of a variant seen fewer than three times from that of another variant. The mean (SD) for the individual SDs of the retention times observed in these different hemoglobins was 0.026 (0.016) min. A difference (d) in the retention time of two hemoglobins greater than the mean of the individual SDs + 2 SD [0.026 + (2 × 0.016) = 0.058 min] was considered significant.
hemoglobin variants with retention times <0.63 min
The Clinical Data Management software does not integrate elution peaks that occur at <0.63 min. The tetramers HbBarts (γ4) and HbH (β4), and HbF1, the acetylated form of HbF, all elute before chromatogram integration; they therefore are not indicated on the chromatogram report. The elution peaks are detected only by visual analysis of the chromatogram. HbBarts was seen in newborns at risk for at least two gene deletions of the α-globin. Of the 12 cases of HbH disease seen, 3 did not show a discernible fast-moving band on electrophoresis. HbF1 was seen mostly in newborns and was ∼10–15% of the total HbF present.
hemoglobin variants with retention times in the p1 window (0.63–0.85 min)
No hemoglobin variants were detected in this window.
hemoglobin variants with retention times in the f window (0.98–1.20 min)
A survey of the Globin Gene Server (22) revealed that at least seven hemoglobin variants (four β- and three α-variants) are expected to elute in this window, all in quantities >10%. To eliminate the possibility of incorrectly designating an elution peak as HbF, 16 months into the study the laboratory began performing electrophoresis on all samples with a HbF fraction >7%, which was not age appropriate. Of the 288 specimens analyzed by electrophoresis, all were confirmed to be HbF.
hemoglobin variants with retention times in the p2 window (1.24–1.40 min)
HbA1C eluted in the P2 window. When the elution peak was >7% of the total hemoglobin, the patient records were checked for indication of diabetes and HbA1C quantification. If no quantification was available, the HbA1C was quantified in the hospital’s chemistry laboratory by phenol-borate affinity HPLC (Primus Corporation). If the %Hb in the P2 window and the HbA1C values were concordant, i.e., within 15% of each other, no further studies were performed. The only hemoglobin variant found to elute in this window was HbHope, which had a mean (SD) %Hb [45.9 (2.2)%] much greater than would be expected for HbA1C.
hemoglobin variants with retention times in the p3 window (1.40–1.90 min)
Nine hemoglobin variants (four α- and five β-variants) had elution peaks in the P3 window. It is predicted that HbCamden (d = 0.10 min from HbHope) and HbJ-Oxford (d = 0.11 min from HbCamden) can be differentiated and identified based solely on their retention times. HbAustin, HbN-Baltimore, and HbFukuyama could not be differentiated from each other by their respective retention times; hemoglobin electrophoresis was required (Fig. 1⇓ , lanes 3, 6, and 4, respectively). The %Hb was sufficient, however, to distinguish these variants from HbFannin-Lubbock, HbJ-Anatolia, and HbJ-Mexico. To distinguish these latter three variants, however, retention times and electrophoresis were required (Fig. 1⇓ , lanes 9, 5, and 8, respectively). It is predicted that HbJ-Meerut can be differentiated and identified based solely on its retention time (d = 0.12 min from HbJ-Mexico).
hemoglobin variants with retention times in the a0 window (1.90–3.10 min)
Six hemoglobin variants (two α- and four β-variants) had elution peaks in the A0 window. HbJ-Toronto (d = 0.06 min from HbJ-Meerut), HbJ-Bangkok (d = 0.08 min from HbJ-Toronto), and HbTy Gard (d = 0.18 min from HbJ-Bangkok) could all be differentiated and identified solely by their retention times. HbKöln (d = 0.06 min from HbTy Gard), an unstable hemoglobin, could be identified by its retention time along with its characteristic chromatogram. In addition to the elution peak for the intact hemoglobin at retention time 2.26 min, there was a secondary peak at retention time ∼4.90 min representing the denatured HbKöln (Fig. 2B⇓ ). HbTy Gard (data not shown) and HbTwin Peaks (Fig. 3⇓ , lane 4) did not separate from HbA on hemoglobin electrophoresis and were detected only by HPLC. HbTy Gard appeared to have a unique retention time, whereas HbTwin Peaks had a characteristic chromatogram in which there was a hump on the downward slope of the HbA0 elution peak (Fig. 2C⇓ ). HbNew York appeared to have a retention time identical to that of HbA0. Alkaline electrophoresis was required to detect this β-variant; it moved more anodal than HbA under these conditions (Fig. 3⇓ , lane 6).
hemoglobin variants with retention times in the a2 window (3.30–3.90 min)
Five hemoglobin variants (one δβ-hybrid, one α-, and three β-variants) had elution peaks in the A2 window. HbLepore could be differentiated and identified based solely on its retention time. The %Hb and the characteristic hump on the downward slope of the elution peak (Fig. 2D⇑ ) were additional distinguishing features of HbLepore. The retention time for HbD-Iran appeared to be significantly different from those of both HbLepore (d = 0.12 min) and HbA2 (d = 0.14 min). In addition, the %Hb of HbD-Iran (47.7%) was significantly greater than either of these variants [HbLepore, 12.1 (1.5)%; HbA2, 3.63 (0.04)%]. The retention times and %Hb for HbA2 and HbE were significantly different (P = 0.001 for both). The retention time for HbOsu-Christiansborg appeared to be significantly different from that for HbE (d = 0.08 min) in addition to the %Hb [44.0% vs 30.3 (4.0)%, respectively]. HbG-Honolulu appeared to have a retention time (d = 0.09 min) and %Hb significantly different from those of HbOsu-Christiansborg (27.4% vs 44.0%, respectively). Although HbG-Honolulu and HbKorle-Bu appeared to have significantly different retention times (d = 0.06 min), the lower %Hb [27.4% vs 46.5 (3.7)%, respectively] and characteristic chromatogram of HbG-Honolulu allowed further differentiation of the two variants. HbG-Honolulu, an α-variant, showed the presence of the characteristic minor HbA2 variant peak (α2G-Honoluluδ2) immediately after the variant peak, which was missing in HbOsu-Christiansborg, a β-variant.
hemoglobin variants with retention times in the d window (3.90–4.30 min)
Three hemoglobin variants had elution peaks in the D window, all of which were β-variants. The retention times, along with the %Hb, were statistically different for HbKorle-Bu vs HbD-Punjab (P <0.001, respectively) and HbKorle-Bu vs HbG-Philadelphia (P <0.001, respectively). Although the retention times for HbD-Punjab and HbG-Philadelphia were statistically different (P = 0.015), there was no statistical difference in %Hb (P = 0.21). The mean HbA2 values for HbD-Punjab trait [1.4 (0.4)%] and HbG-Philadelphia trait [1.3 (0.4)%] were significantly lower (P <0.001, respectively) than the range for HbA2 in the normal specimens. The chromatogram for HbG-Philadelphia (Fig. 2F⇑ ), an α-variant, showed the presence of the characteristic minor HbA2 variant peak (α2G-Philadelphiaδ2) in all heterozygous cases. Because the range of retention times for HbKorle-Bu straddled both the A2 and D windows, one-half of the specimens seen in this series fell in the A2 window although the mean retention time was in the D window.
hemoglobin variants with retention times in the s window (4.30–4.70 min)
Six hemoglobin variants (three α-, two β-, and one δ-variant) had elution peaks in the S window. HbE-Saskatoon and HbS appeared to have significantly different retention times (d = 0.18 min). Although HbManitoba, HbMontgomery, and HbA2′ all appeared to have identical retention times, their retention times and %Hb were statistically different from those of HbS (P <0.001 for all). The %Hb values for HbManitoba and HbMontgomery did not appear to be statistically different [16.5% vs 15.7 (2.2)%, respectively]; however, the %Hb for HbA2′ was statistically different from these two variants [1.2 (0.1)%; P <0.001]. The retention time for HbQ-Thailand appeared to be different from the retention times of HbManitoba, HbMontgomery, and HbA2′ (d = 0.09, 0.09, and 0.08 min, respectively). At least in this single case, the %Hb for HbQ-Thailand appeared to be greater than those for the other three variants (Table 2⇑ ). HbManitoba and HbMontgomery showed different mobilities on electrophoresis, which was necessary for identification (data not shown).
The incidence for HbA2′ (0.76%) was determined in a 6-month prospective study in samples submitted to this laboratory for hemoglobin analysis. In patients heterozygous for HbA2′, the %Hb for HbA2 [1.64 (0.17)%] was significantly lower (P <0.001) than the value for the normal specimens.
hemoglobin variants with retention times in the unknown window (4.70–4.90 min)
The elution peak for HbHasharon, an α-variant, fell in the time interval for unknowns. Although its retention time appeared to be significantly different from that of HbQ-Thailand (d = 0.16 min), the retention time and %Hb were statistically different (P <0.001 for both) from those for HbO-Arab. This variant also had a characteristic chromatogram. In addition to the expected HbA2 variant peak (α2Hasharonδ2), immediately after the variant peak there were two small peaks in all examples of HbHasharon seen in this laboratory (Fig. 2G⇑ ). One minor peak appeared at a retention time of ∼4.27 min, presumably glycated or degraded Hb, and another appeared immediately preceding the elution peak.
hemoglobin variants with retention times in the c window (4.90–5.30 min)
Three hemoglobin variants (three β-variants) had elution peaks in the C window. As reported previously (23), HbO-Arab and HbC had statistically different retention times (P <0.001), whereas the %Hb values were not statistically different (P = 0.84). In addition, all examples of HbO-Arab seen in this laboratory had a minor peak in the D window (Fig. 2H⇑ ), which was not seen in HbC trait. The retention time for HbG-Siriraj appeared to be statistically different from that of HbO-Arab (d = 0.17 min) and that of HbC (d = 0.10 min). HbG-Siriraj, a β-variant, is reported to be 33–40% of the total hemoglobin in heterozygotes (22). This patient appeared to have a concomitant α-thalassemia, which would account for the lower %Hb for the variant.
Hb J family
Six members of the HbJ family (five α- and one β-variants) were identified in this series of hemoglobin variants. Three α-variants (HbJ-Oxford, HbJ-Anatolia, and HbJ-Mexico) had identical electrophoretic mobilities (Fig. 1⇑ , lanes 2, 5, and 8, respectively). Although the retention times for HbJ-Anatolia and HbJ-Mexico did not appear to be significantly different (d = 0.01 min), the retention time for HbJ-Oxford appeared to be significantly different (d = 0.15 min and 0.16 min, respectively). Two other α-variants (HbJ-Meerut and HbJ-Toronto) with identical electrophoretic mobilities (Fig. 1⇑ , lanes 10 and 11, respectively) had significantly different retention times (d = 0.06 min). The β-variant (J-Bangkok), with yet a different electrophoretic mobility (Fig. 1⇑ , lane 12), appeared to have a unique retention time and a significantly higher %Hb [43.6% vs 22.6 (2.4)% for the α-variants].
Hb D family
Three members of the HbD family were identified in this series, all of which were β-variants. Whereas HbD-Punjab and HbD-Iran had identical electrophoretic mobility (Fig. 4⇓ , lanes 6 and 7, respectively), the retention times were quite different (d = 0.69 min). HbOsu-Christiansborg, which had alkaline electrophoretic mobility identical on the Helena Hemoglobin Electrophoresis System (data not shown) and similar on SPIFE (Fig. 4⇓ , lane 9), and an acid electrophoresis mobility identical to these two hemoglobins, had a retention time apparently significantly different from both of them (d = 0.41 and 0.28 min, respectively).
Hb G family
Three members of the HbG family were identified in this series (two α- and one β-variant). HbG-Honolulu and HbG-Philadelphia had mobilities that were similar, but not identical, on electrophoresis (Fig. 4⇑ , lanes 4 and 2, respectively). Their retention times, however, appeared to be significantly different (d = 0.36 min); they eluted in different windows. HbG-Siriraj had a significantly different retention time from the other two members (d = 1.22 and 0.86 min, respectively) and exhibited a unique mobility pattern on electrophoresis. It was slightly cathodal to HbS on alkaline electrophoresis (Fig. 4⇑ , lane 3) and slightly cathodal to HbC on acid electrophoresis (data not shown).
Hb C family
Two β-variants (HbE and HbE-Saskatoon) were identified, each of which had alkaline mobilities on the Helena Hemoglobin Electrophoresis System identical to that of HbC (data not shown) and SPIFE mobilities (Fig. 3⇑ , lanes 8 and 10, respectively) similar to that of HbC. On acid electrophoresis, the two hemoglobin variants, however, had mobilities identical to HbA but not HbC. The two variants had significantly different retention times (d = 0.65 min), and the relative amount of HbE-Saskatoon appeared to be greater than that of HbE [39.9% vs 30.3 (4.0)%, respectively]. An additional β-variant (HbO-Arab) with alkaline electrophoretic mobility identical to that of HbC (Fig. 3⇑ , lane 11) had an acid electrophoretic mobility slightly cathodal to HbA (data not shown).
hemoglobin variants with alkaline electrophoretic mobility slightly anodal from HbA0
Four β-variants in this series (HbHope, HbCamden, HbFukuyama, and HbNew York) all had similar alkaline electrophoretic mobilities (Fig. 3⇑ , lanes 2, 3, 5, and 6, respectively), but their retention times were significantly different (1.39, 1.49, 1.73, and 2.43 min, respectively).
The laboratory diagnosis of hemoglobinopathies and thalassemias, both of which are common, may be required (a) to confirm a provisional diagnosis, such as significant sickling disorders or β-thalassemia major; (b) to explain a hematologic abnormality such as anemia, microcytosis, or polycythemia; (c) to identify an abnormality in the presymptomatic phase, as in neonatal screening; (d) to predict serious disorders of globin-chain synthesis in the fetus and offer the option of termination of pregnancy; (e) to permit genetic counseling of prospective parents; and (f) to allow preoperative screening for the presence of sickle cell hemoglobin (1).
The most common investigative tools in the clinical laboratory are alkaline and acid electrophoresis for hemoglobin variants and hemoglobinopathies, HbA2 quantification by ion-exchange column chromatography, and HbF quantification by alkali denaturation or radial immunodiffusion for thalassemia. Whereas the more common sickling disorders (HbSS, HbSC, HbSD-Punjab, HbSE, HbSG-Philadelphia, HbSHope, HbSLepore, HbSO-Arab, and HbS/β-thal) are all clinically significant, these combinations do present different manifestations and degrees of severity (24)(25), making precise identification important. None of these can be conclusively identified by a single electrophoretic technique (2). The identification of hemoglobin variants is often presumptive, based on characteristic electrophoretic mobility, quantity, and/or ethnic origin. Definite identification usually requires DNA analysis or amino acid sequencing.
HPLC has been shown to be a sensitive, specific, and reproducible alternative to electrophoresis. Its use has been dramatically expanded, especially with the development of rapid, well-resolving, and fully automated analyzers. In the past decade HPLC, with its automation and its quantitative power, has appeared to be an appropriate candidate for direct identification and sensitive quantification of major and minor, normal and abnormal, hemoglobin fractions (6)(9),(13)(14)(15)(16)(17)(18)(19).
To date, evaluations of the performance and use of HPLC technologies in the diagnostic laboratory have been in relation to newborn screening (13)(14)(15)(16), screening specific ethnic populations (19), evaluation of patients studied because of the presence of an abnormal hemoglobin component, and evaluation of stored library samples (9). There has been no large prospective study, however, evaluating its use in the clinical diagnostic laboratory setting.
Over a 32-month period, 60293 samples were analyzed in the Special Hematology Laboratory at Bellevue Hospital Center for quantification of hemoglobin fractions and screening for hemoglobin variants. Twenty β-variants, 12 α-variants, 2 tetramers, 1 δ-variant, and 1 δβ-fusion globin were observed. The mean (SD) interval period between the first and last observations was 23.3 (6.6) months for the 15 variants seen on three or more occasions. HbS (5.95%), HbC (1.60%), HbA2′ (0.76%), and HbE (0.14%) were the most common variants encountered. Thirty unique additional variants were encountered with various incidences in 104 samples (Table 2⇑ ).
Different reports have addressed the precision of the retention times obtained with stored normal samples (17); specimens containing HbS, HbC, and HbE (9); and liquid controls (14). Only two reports have tabulated the retention times for various hemoglobin variants (6)(9). None of these reports, however, addressed the feasibility of using the retention time as a diagnostic tool. The mean (SD) imprecision (CV) of the retention time for the variants seen on three or more occasions over the 32-month observation period was 1.0 (0.7)%, confirming and extending the data of Eastman et al. (14). Because it has been suggested that retention times of hemoglobin peaks differ slightly, but significantly, with different columns or reagent lot numbers (9), the retention time for each variant was normalized to the retention time of HbA0 or to the retention time of HbA2 in the hope of eliminating the effects of these variables. There was no statistical difference (P = 0.10 and 0.51, respectively), and all subsequent data analyses were performed with use of the retention time. During our observation period, three different lot numbers of columns and 10 different lot numbers of reagents were used. Analysis of the retention times and %Hb for HbA0, HbA2, and HbS showed no statistical difference (data not shown) with regard to column and/or reagent changes. This was further confirmed by the CVs of the retention times for HbF, HbA0, and HbA2 (2.5%, 1.7%, and 0.96%, respectively) determined from the retention time of the first five normal patient values for each month of observation. In addition, the CVs for those variants seen over the 32 months (i.e., HbHope, HbE, HbS, HbMontgomery, HbHasharon, and HbC) ranged from 0.24% to 1.9%. Finally, the use of a normalized retention time did not allow the identification of any additional variants not identified by the retention time alone.
To predict whether the 17 hemoglobin variants seen on fewer than three occasions had retention times significantly different from other hemoglobins eluting in the same windows, the SD, a measure of the variation around the mean, of the retention times was used. The SD for the retention times of HbF, HbA0, HbA2, and the 15 variants seen on three or more occasions showed no correlation with the retention time or the %Hb, demonstrating that it is independent of these two variable. The mean (SD) for the individual SDs of the retention times was 0.026 (0.016) min. A difference in retention times of two hemoglobins >0.058 min was, therefore, considered significant. This calculation is probably conservative. For example, although the retention times for HbD-Punjab and HbG-Philadelphia were significantly different (P = 0.015) according to Student’s t-test, the difference in retention time was 0.04 min. This suggests that accumulation of further data for those hemoglobin variants seen on fewer than three occasions may give additional statistically significant different retention times, e.g., HbAustin and HbFukuyama (d = 0.045 min).
The retention time alone (n = 18) or in conjunction with either the %Hb (n = 3) or the peak characteristics (n = 4) could identify 25 of the hemoglobin variants seen in this series. The %Hb can generally be the initial predictor whether the detected variant is an α- or β-variant. In this series, the α-variants all had mean %Hb values <30%, whereas the β-variants all had mean %Hb values >34% except for HbE [30.3 (4.0)%] and HbG-Siriraj (24.2%).
The Clinical Data Management software does not integrate peaks with elution time shorter than 0.63 min, a weak point for the Bio-Rad Variant II HPLC system. Visual examination of the elution pattern for each specimen is required and is particularly important for laboratories with a patient population at high risk for α-thalassemia, HbH disease, and HbBarts hydrops fetalis. When HbH or HbBarts is noted, the %Hb must be determined by either densitometry or manual calculation. Twelve cases of HbH disease with a %Hb range of 5–20% were observed. Interestingly, three of these specimens did not show an easily discernible band on alkaline electrophoresis when performed according to the manufacturer’s recommendations. The band was noted only after a more concentrated solution was analyzed by electrophoresis. HPLC was, therefore, more sensitive to the presence of HbH. Although HbBarts and HbH appear to have identical retention times, they can be differentiated based on the age of the patient. HbBarts, the γ4 tetramer, will be present when HbF production is increased, i.e., generally in the newborn period, whereas HbH, the β4 tetramer, will be present when β-chain production is significant, i.e., >1 year of age.
HbA2′ is an example of a hemoglobin variant that can be identified by the retention time and %Hb. Although two variants, HbManitoba and HbMontgomery, have retention times identical to that of HbA2′, it is easily identified by its statistically significantly reduced %Hb.
The elution peak characteristic in addition to the retention time can be used to identify certain hemoglobin variants. The retention times for HbD-Punjab and HbG-Philadelphia were statistically significantly different. In addition, both of these variants were associated with decreased HbA2, confirming a previous report (26) in which a similar range (0.9–2.5%) in the presence of HbD-Punjab was reported. It was postulated in that report that the decrease in HbA2 may be attributable to either coelution with the HbA0 or the HbD-Punjab peaks or the mutation itself influencing the amount of δ-chain. This same rationale probably does not account for the decreased HbA2 seen in the presence of HbG-Philadelphia. All cases of heterozygous HbG-Philadelphia demonstrated the minor HbA2 variant peak (α2G-Philadelphiaδ2; Fig. 2F⇑ ) falling in the S window, with no such minor peak seen in any of the cases of HbD-Punjab. The true HbA2 [2.7 (0.9)%] is the sum of the normal and variant peaks and is not statistically different (P <0.001) from the proportion seen for the normal specimens. Certain other variants present consistent and distinctive characteristics in their elution patterns that can be incorporated into an algorithm for hemoglobin variant identification. Such characteristics are demonstrated in Fig. 2⇑ for HbKöln, HbTwin Peaks, HbLepore, HbHasharon, and HbO-Arab.
Six hemoglobin variants with retention times of 1.68–1.78 min could be divided into two groups (group I, HbAustin, HbN-Baltimore, and HbFukuyama; group II, HbFannin-Lubbock, HbJ-Anatolia, and HbJ-Mexico). Although the retention time was not sufficient to distinguish the two groups, the groups could be distinguished by the %Hb. Within the groups, electrophoresis was required to distinguish the members of each group (Fig. 1⇑ ). Additionally, HbManitoba and HbMontgomery, both α-globin variants with statistically identical retention times, %Hb values, and peak characteristics, required electrophoresis (data not shown) to be differentiated.
Two hemoglobin variants, HbTy Gard (data not shown) and HbTwin Peaks (Fig. 3⇑ , lane 4), did not separate from HbA on electrophoresis but could be easily detected by HPLC. HbTy Gard appeared to have a retention time significantly different from that of HbJ-Bangkok (d = 0.18 min). It can be distinguished from HbKöln by both the retention time (d = 0.06 min) and the characteristic secondary peak seen with this unstable variant (Fig. 2B⇑ ). HbTwin Peaks did not fully separate from HbA0 on HPLC, presenting a characteristic hump on the downward slope of the elution peak. Although HbTwin Peaks is hematologically and clinically insignificant, HbTy Gard is a high-oxygen-affinity hemoglobin that can lead to erythrocytosis, which is of potential clinical importance (22).
HbNew York, a β-globin chain variant, did not separate on HPLC and migrated slightly anodal to HbA on alkaline electrophoresis (Fig. 3⇑ , lane 6). This slightly unstable variant is the second most common variant found among the Chinese, but it has been described infrequently in other ethnic groups. It has been found in the heterozygous state or in combination with HbE, α-thalassemia, and β-thalassemia (22). Heterozygotes have normal hematologic indices with HbNew York representing 40–45% of the total hemoglobin. βNY-chains have a higher turnover than βA-chains and, in contrast to some β-chain variants, such as βS, βC, or βE, have a higher affinity for α-chains. If there is a co-inheritance of HbH and HbNew York, preferential synthesis of the unstable HbNew York, the small number of α chains available to synthesize HbA, and the high excess of βA- and βNY-chains, which precipitate in the erythrocyte precursors and cause ineffective erythropoiesis, lead to aggravation of the anemia. Although carriers of this variant are usually not anemic, they have occasional erythrocytes with inclusion bodies and a reduced α-/non-α-globin chain ratio attributable to rapid turnover of the βNY-chain (27)(28). For a 12-month period, all samples from patients with an Asian-appearing name were analyzed by alkaline electrophoresis in addition to HPLC. Of 632 samples analyzed, 4 (0.63%) cases of HbNew York (3 heterozygotes and 1 HbNew York/β-thalassemia) were detected.
HbA2′ is a δ-globin variant commonly found in ∼1–2% of the African-American population with West African heritage (29). HbA2′ is easily detectable by HPLC, producing a minor peak [1.2 (0.1)%] in the S window [retention time, 4.59 (0.030) min; Fig. 2⇑ E], which is consistent with previously published results (30). Among laboratories participating in a recent proficiency survey by the College of American Pathology, only 32% of the laboratories using HPLC correctly identified the presence of HbA2′ (31). HbA2′, whether heterozygous or homozygous, is clinically and hematologically silent. Its sole importance is that it may cause an underestimation of the HbA2 concentrations in the work-up for thalassemia. An accurate HbA2 value for this purpose represents the sum of the HbA2 and HbA2′ peaks (29)(30). In our own laboratory, we have seen a case of homozygous HbA2′ (HbA2 = 0.0%; HbA2′ = 2.3%) and β-thalassemia trait in conjunction with HbA2′ (HbA2 = 2.5%; HbA2′ = 2.2%; total HbA2 = 4.7%). This latter case might have been mistakenly diagnosed as α-thalassemia if only the HbA2 fraction had been reported.
The HbJ family consists of a large group of α- and β-variants that migrate faster than HbA on alkaline electrophoresis because of an amino acid substitution, resulting most often in the gain of a negative charge and less often in the loss of a positive charge on the Hb molecule (2). The α- and β-variants can often be predicted by the relative amounts of the variants present; however, they exhibit similar or identical alkaline electrophoretic mobilities, as shown in Fig. 1⇑ . The diagnosis of these members is often only inferred by the electrophoretic mobility, %Hb, and ethnicity of the patient. However, the retention time alone was superior to electrophoresis in the identification of members of the HbJ family encountered in this series of specimens.
Finally, HPLC was able to avoid misidentification of two hemoglobin variants having benign interaction with HbS. HbD-Punjab and HbD-Iran exhibited identical electrophoretic mobilities (Fig. 4⇑ , lanes 6 and 7, respectively), but their retention times appeared to be unique and significantly different (d = 0.069 min). Similarly, HbC and HbE-Saskatoon exhibited identical mobilities on the Helena Hemoglobin Electrophoresis System (data not shown) and very similar mobilities on alkaline SPIFE (Fig. 3⇑ , lanes 9 and 10, respectively). HbE-Saskatoon appeared to have a unique retention time (d = 0.11 min compared with HbG-Philadelphia and d = 0.18 min compared with HbS). These situations are clinically important because HbSD-Punjab and HbSC are both significant sickling disorders, whereas HbSD-Iran and HbSE-Saskatoon are clinically benign (24)(25). The misdiagnosis of HbD-Iran and HbE-Saskatoon may lead to incorrect genetic counseling in addition to undue anxiety for the family.
Electrophoresis of hemoglobin variants with similar mobilities has inherent limitations. The identification of variants is dependent on the technical performance of electrophoresis, which has many variables, e.g., hemoglobin concentration, amperage, running temperature, and length of electrophoresis run. These variables can affect the quality of separation and relative positioning of the bands. Variants that migrate identically or similarly (see Figs. 1⇑ , 3⇑ , and 4⇑ ) would be very difficult, if not impossible, to evaluate without the unknown sample being electrophoresed directly adjacent to the reference hemoglobin mixture or adjacent to several known stored specimens. HPLC, on the other hand, has been shown to have a high degree of reproducibility and precision.
In this study the influence of hemoglobin stability testing and patient ethnicity were not included. Hemoglobin stability testing is an uncommon and subjective test for which controls are not readily available. The collection of ethnicity information is very difficult for the clinical laboratory, especially a reference laboratory. It requires cooperation from the requesting physician to collect from the patient information that is frequently subjective.
In conclusion, this is the first report of a large prospective study on the use of HPLC for determining the presence and identities of hemoglobin variants in a clinical laboratory setting. These data demonstrate that HPLC is an excellent, powerful diagnostic tool for the direct identification of hemoglobin variants with a high degree of precision in the quantification of major and minor, normal and abnormal, hemoglobin fractions. HPLC is suitable for the routine investigation of hemoglobin variants, hemoglobinopathies, and thalassemia. With the integration of proper algorithms (see the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol50/issue10/) involving retention time, %Hb, and peak characteristics, a clinical laboratory is capable of identifying ∼75% of the common variants encountered without the need for confirmatory studies such as alkaline and acid electrophoresis. More importantly, identification of the common variants (i.e., HbC, HbD-Punjab, HbE, HbG-Philadelphia, HbHope, HbLepore, HbO-Arab, and HbS) that in combination with HbS lead to a clinically significant sickling disorder can be quickly and accurately accomplished by use of such algorithms without the need for confirmatory testing.
We express our gratitude to Margaret Karpatkin, MD, and David Hart, MD, for helpful discussions during the preparation of this manuscript and to the laboratory staff of the Special Hematology Laboratory at Bellevue Hospital Center for their technical expertise. We particularly wish to acknowledge Robert Boorstein, MD, PhD, for his constant support and encouragement, and the Health and Hospital Corporation of New York City for providing these state-of-the-art laboratory facilities, allowing excellent patient care.
- © 2004 The American Association for Clinical Chemistry