Polymerization of three hemoglobin A2 variants containing Val6 and inhibition of hemoglobin S polymerization by hemoglobin A2.

To understand determinants for hemoglobin (Hb) stability and Hb A2 inhibition of Hb S polymerization, three Valδ6 Hb A2 variants (Hb A2 δE6V, Hb A2 δE6V,δQ87T, and Hb A2 δE6V,δA22E,δQ87T) were expressed in yeast, and stability to mechanical agitation and polymerization properties were assessed. Oxy forms of Hb A2 δE6V and Hb A2 δE6V,δQ87T were 2- and 1.6-fold, respectively, less stable than oxy-Hb S, while the stability of Hb A2 δE6V,δA22E,δQ87T was similar to that of Hb S, suggesting that Alaδ22 and Glnδ87 contribute to the surface hydrophobicity of Hb A2. Deoxy Hb A2 δE6V polymerized without a delay time, like deoxy Hb F γE6V, while deoxy Hb A2 δE6V,δQ87T and deoxy Hb A2 δE6V,δA22E,δQ87T polymerized after a delay time, like deoxy Hb S, suggesting that β87 Thr is required for the formation of nuclei. Deoxy Hb F γE6V,γQ87T showed no delay time and required a 3.5-fold higher concentration than deoxy Hb S for polymerization, suggesting that Thr effects on Valδ6 Hb A2 and Valγ6 Hb F variants are different. Mixtures of deoxy Hb S/Hb A2 δE6V,δQ87T polymerized, like deoxy Hb S, while polymerization of Hb S/Hb A2 δE6V mixtures was inhibited, like Hb S/Hb F γE6V mixtures. These results suggest α2βSδ6 Val, 87 Thr hybrids and Hb A2 δE6V,δQ87T participate in Hb S nucleation, while only 50% of α2βSδ6 Val hybrids and none of the Hb A2 δE6V participate. These findings are in contrast to those of mixtures of Hb S with Hb F γE6V or Hb F γE6V,Q87T, which both inhibit Hb S polymerization. Our results also suggest participation in nucleation of some α2βSδ hybrids in A2S mixtures but not α2βSγ hybrids in FS mixtures.

Hb F and Hb A 2 inhibit hemoglobin (Hb) 1 S polymerization. Mixtures of Hb A 2 and Hb S, like mixtures of Hb F and Hb S, have a higher solubility than Hb S alone and result in inhibition of Hb S polymerization (1,2). The primary inhibitory effects of Hb A 2 and Hb F in these mixtures are to exclude the asymmetrical ␣ 2 ␤ S ␦ and ␣ 2 ␤ S ␥ hybrids, respectively, as well as Hb A 2 and Hb F from initiation of polymerization with Hb S (3)(4)(5)(6)(7)(8)(9)(10)(11). Studies comparing minimum gel concentrations for naturally occurring hemoglobin variants, including the Lepore (␦␤ hybrid) hemoglobins, suggested that Gln ␦87 and Ala ␦22 in Hb A 2 are important sites potentiating inhibition of polymerization of deoxy Hb S (5). In addition, inhibition of polymerization by Hb A 2 and Hb F is primarily in trans to the Val ␤6 contact in Hb S polymers. Gln ␥87 in Hb F is also an important site for inhibition of Hb S polymerization by Hb F (5).
We previously engineered and isolated Hb S ␤T87Q as well as Hb F ␥E6V and Hb F ␥E6V,␥Q87T, using a yeast expression system, and characterized the polymerization properties of these variants to clarify the role of Gln at ␥87 in Hb F-mediated inhibition of deoxy Hb S polymerization (12,13). Oversaturated deoxy Hb F ␥E6V and Hb F ␥E6V,␥Q87T polymerized similarly without a delay time at a higher concentration than Hb S, even though studies of mixtures of Hb S and Hb S ␤T87Q indicated that the Thr 3 Gln difference at 87 (F3) in the non-␣-globin chain is a key amino acid required for Hb F inhibition of Hb S polymerization. Furthermore, changing Gln to Thr at ␥87 in Hb F ␥E6V (Hb F ␥E6V,␥Q87T) slightly inhibited rather than promoted polymerization compared with Hb F ␥E6V (13).
Recent studies on high levels of expression of human Hb A 2 with Hb S in red blood cells from transgenic mice showed that overexpressed ␦ chains interact with red blood cell membranes, which results in drastic modification of their properties (14). To further understand Hb A 2 -and Hb F-mediated inhibition of Hb S polymerization as well as hydrophobicity of Hb A 2 , three Hb A 2 variants containing Val ␦6 (1) Hb A 2 ␦E6V; 2) Hb A 2 ␦E6V,Q87T; and 3) Hb A 2 ␦E6V,␦A22E,␦Q87T) were expressed in yeast. We reasoned that since the ␦ chain is more homologous to the ␤ chain than to the ␥ chain, then important sites in inhibition of Hb S polymerization by Hb F or Hb A 2 could be more readily identified by studying the polymerization properties of these Val ␦6 Hb A 2 variants. In this report, we characterized surface hydrophobicity by monitoring stability to mechanical agitation. In addition, we assessed the polymerization properties of Hb A 2 ␦E6V (␣ 2 ␦ 2 6 Glu 3 Val ), Hb A 2 ␦E6V,␦Q87T (␣ 2 ␦ 2 6 Glu 3 Val, 87 Gln 3 Thr ) and Hb A 2 ␦E6V,␦A22E,␦Q87T (␣ 2 ␦ 2 6 Glu 3 Val, 22 Ala 3 Glu; 87 Gln 3 Thr ) to understand the role of Ala ␦22 and Gln ␦87 in Hb A 2 -mediated inhibition of Hb S polymerization as well as the role of these sites in the polymerization of these Val ␦6 Hb A 2 variants.

MATERIALS AND METHODS
Full-length human ␦-globin cDNA (15) was isolated by reverse transcription-polymerase chain reaction using mRNA from COS-1 cells that had been transfected with the ␦-globin expression vector pSVK3 (5Ј␦ SalI-PstI). The ␤-globin cDNA in the shuttle vector pGS188 was replaced by the ␦-globin cDNA to create pGS188 ␦ (16, 17) after isolation of human ␦-globin cDNA. Three ␦-globin cDNAs containing 1) Val ␦6 , 2) Val ␦6 and Thr ␦87 , and 3) Val ␦6 , Glu ␦22 , and Thr ␦87 were constructed using polymerase chain reaction mutagenesis and subcloned as described previously (17) using the ␦-globin cDNA containing Val ␦6 as a template to introduce the Thr ␦87 change. The double mutant ␦-chain cDNA containing Val ␦6 and Thr ␦87 was then used as a template to introduce the Glu ␦22 change. The complete coding sequence of wild-type and ␦-globin cDNA variants, flanked by the GGAP promoter and MF * This research was supported by grants from the National Institutes of Health (HL38632 and DK16691) and the March of Dimes Birth Defects Foundation (FY95, 96-0942). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Division of Hematology, The Children's Hospital of Philadelphia, Abramson Ctr., 34th St. and Civic Center Blvd., Philadelphia, PA 19104. Tel.: 215-590-3576; ␣3Ј-UN regions in pGS188 ␦, was determined using site-specific primers and fluorescently tagged terminators in a cycle-sequencing reaction in which extension products were analyzed on an automated laseractivated, fluorescence emission DNA sequencer (18). The entire ␦-globin cDNA region was then excised by XhoI digestion and used to replace the entire ␤-globin cDNA region in the expression vector pGS389 (16,17). The resultant vectors contain full-length human cDNAs coding for ␣and ␦-globin variants under transcriptional control of dual pGGAP promoters, as well as a partially functional yeast LEU2d gene and the URA3 gene for plasmid amplification and selection in yeast (16,17,19).
Expression in yeast and isolation of Hb A 2 variants were as described previously (17,19). Purified Hb A 2 variants were subjected to electrospray mass analysis (Fisons Instruments, VG Biotech, Altricham, United Kingdom) using the multiply charged ion peaks from the ␣-globin chain (M r ϭ 15, 126.4 Da) as an external reference for mass scale calibrations (20). The Val ␦6 amino acid change and the N-terminal amino acid sequence of purified ␣ and ␦ chains were directly confirmed by Edman degradation with a pulsed liquid protein sequencer (ABI 477A, Applied Biosystems, Inc., Foster City, CA). Native human Hb A 2 was purified from normal hemolysates on a Mono S column (Pharmacia Biotech, Inc.) using fast performance liquid chromatography with slight modification of the method described by Ou et al. (21).
Cellulose acetate electrophoresis of hemoglobins was performed at pH 8.6 using Supre-Heme buffer (Helena Laboratories, Beaumont, TX), and hemoglobin concentration was determined spectrophotometrically on a Hitachi U2000 spectrophotometer using millimolar extinction coefficients of 13.5 at 541 nm for oxyhemoglobin and 13.4 at 540 nm for carbon monoxyhemoglobin (22). Methods for determination of kinetics of polymerization and polymer formation of deoxyhemoglobins in 1.8 M phosphate buffers using the temperature jump method as well as the mechanical stability of oxyhemoglobins were as reported previously (23,24). The temperature jump method using 1.8 M phosphate buffer differs from a simple "salting out" in high phosphate (2.2 M) in which a hemoglobin sample is injected directly into deoxygenated high phosphate buffer and results in rapid formation of amorphous aggregates (13,25).

RESULTS
Characterization of Hb A 2 Variants-Fast performance liquid chromatographic elution patterns of the two Hb A 2 variants, Hb A 2 ␦E6V and Hb A 2 ␦E6V,␦Q87T, were very similar; however, elution of the triple variant, Hb A 2 ␦E6V,␦A22E,␦Q87T, occurred slightly after the other two variants. Both Hb A 2 ␦E6V and Hb A 2 ␦E6V,␦Q87T tetramers migrated with less negative charge than native Hb A 2 following cellulose acetate electrophoresis, while the triple variant migrated in a manner similar to that of native Hb A 2 (Fig. 1). Differences in migration between native Hb A 2 and Hb A 2 ␦E6V,␦A22E,␦Q87T or between Hb A 2 ␦E6V and Hb A 2 ␦E6V,␦Q87T are similar to that comparing Hb A and Hb S. These results indicate that Glu 3 Val or Ala 3 Glu substitu-tions at the 6th and 22nd positions in Hb A 2 and Hb A, respectively, contribute similarly to surface charge differences.
The absorption spectra of the three Hb A 2 variants in the CO form were virtually identical with those of native CO Hb A 2 . In addition, N-terminal sequence analysis to 15 residues for each globin chain showed identical results compared with native ␣and ␦-globins except for the Glu 3 Val ␦6 change. Electrospray mass spectrometry analysis of ␣ and ␦ chains from the three recombinant Hb A 2 variants and native Hb A 2 showed that the ␣ chains from native Hb A 2 and the variants had the same mass as native ␣-globin (M r 15,126 Da), while ␦ chain analysis of the three variants showed expected mass values (Table I).
Mechanical Stability -The oxy form of Hb S denatures ϳ10 times faster than oxy Hb A during mechanical agitation (23), which is related to differences in amino acid surface hydrophobicity of dimeric and tetrameric hemoglobins (26). The physical basis of mechanical instability is predicated on the fact that mechanical agitation generally increases the rate of unfolding of proteins at the air-water interface and provides mixing to allow new, undenatured material to reach the surface from the bulk solution (2). The rate of precipitation during mechanical agitation also depends on the hydrophobicity of exposed amino acids on the surface as well as the tertiary structure of proteins (2,23,26). The oxy form of Hb A 2 is about 3-fold more unstable than Hb A during mechanical agitation (27). Our results show that native oxy Hb A 2 was less stable than oxy Hb A and that the oxy form of the Val ␦6 variant was about 7-fold less stable than native oxy Hb A 2 (Fig. 2). These differences in mechanical stability are similar to those comparing Hb S and Hb A and are consistent with our previous findings showing a direct relationship between mechanical instability and hydrophobicity at the ␤6 position (26,28). The stability of Hb A 2 ␦E6V was 2-fold less than Hb S, while that of the double mutant Hb A 2 ␦E6V,␦Q87T was 1.6-fold less. The mechanical stability of the triple mutant Hb A 2 ␦E6V,␦A22E,␦Q87T was, however, similar to that of Hb S. These results suggest that Ala ␦22 and Gln ␦87 in Hb A 2 promote instability to mechanical agitation by increasing surface hydrophobicity.
Polymerization Properties-Polymerization of deoxy Hb S in high (Ͻ1.8 M) as well as low phosphate buffer using the temperature jump method is characterized by a delay time prior to polymer formation the length of which depends on hemoglobin concentration: the lower the concentration, the longer is the delay time ( Fig. 3A) (24). Polymer formation in vitro can be assessed using low and high phosphate buffers, both of which result in nucleation-controlled formation of ordered polymers (29,30). This process clearly differs from a simple "salting out" in high phosphate (Ͼ1.8 M) in which hemoglobin sample is directly injected into deoxygenated high phosphate buffer and results in rapid formation of amorphous aggregates (25). High phosphate buffer conditions facilitate comparative studies of polymerization properties of sickle and non-sickle hemoglobins, since the hemoglobin concentrations required for these studies are much lower than those required in low phosphate buffer. For example, experiments in 1.8 M phosphate buffer require about 100 mg/dl for Hb S and Hb A 2 variants, which is less than 1/100 of the hemoglobin concentration required for studies in low phosphate buffer. The relative order of inhibition of Hb S polymerization by Hb A, Hb F, and Hb A 2 in 1.8 M phosphate buffer is also similar to that in low phosphate buffer (4, 5, 9 -11, 13), and those studies show that one-half of AS hybrids and no FS hybrids participate in the initiation of Hb S polymerization (10,11,13). Recent comparative studies of deoxy Hb S polymer structures made in high phosphate buffer with those made in low phosphate buffer using electron microscopy showed that oversaturated deoxy Hb S first formed fibers, which then began to form bundles, macrofibers, and crystals. These structures had the same appearance in low or high phosphate buffers (31), indicating that deoxy Hb S fibers formed in high phosphate buffer are similar or identical with those formed in low phosphate buffer. Furthermore, contact sites of deoxy Hb S crystals made in polyethylene glycol or high phosphate buffer correspond to those of deoxy Hb S polymers in low phosphate buffer (32,33). Polymerization of the three Val ␦6 Hb A 2 variants alone was also performed using the temperature jump method in 1.8 M phosphate buffer. Hb A 2 ␦E6V in the deoxy form polymerized at slightly higher hemoglobin concentrations than deoxy Hb S, while the other two variants polymerized at concentrations similar to that of deoxy Hb S. Polymerization of Hb A 2 ␦E6V, like Hb F ␥E6V, however, was not accompanied by a delay time before polymerization (Fig. 3B) (13) and may be explained by a linear polymerization mechanism (12,29). In contrast, polymer formation by Hb A 2 ␦E6V,␦Q87T or Hb A 2 ␦E6V,␦A22E,␦Q87T was accompanied by a delay time before polymerization like that of Hb S (Fig. 3, C and D). The length of the delay time for these two variants depended on hemoglobin concentration, and logarithmic plots of delay time versus concentration for Hb A 2 ␦E6V,␦Q87T or Hb A 2 ␦E6V,␦A22E,␦Q87T showed a straight line which was identical with the line for Hb S (Fig. 4). These results indicate that Thr ␦87 in Hb A 2 ␦E6V,␦Q87T facilitates the formation of nuclei as evidenced by a delay time prior to polymerization, while Ala ␦22 in Hb A 2 ␦E6V,␦Q87T has a minimal or no effect on the formation of nuclei.
Total polymer formation as a function of hemoglobin concentration was also determined at the plateau of the polymerization curves to define critical concentrations required for polymerization of Hb A 2 ␦E6V, Hb A 2 ␦E6V,␦Q87T, and Hb A 2 ␦E6V,␦A22E,␦Q87T. These results were then compared with our previously reported results for polymerization of Hb S and Hb F ␥E6V,␥Q87T (Fig. 5). Polymer formation increased linearly with increases in hemoglobin concentration. Critical concentrations for all three Val ␦6 Hb A 2 variants were slightly higher than that of Hb S. In contrast, values for deoxy Hb F ␥E6V,␥Q87T were more than 3.5-fold higher than that for deoxy Hb S (13). The slope of the line for all three Val ␦6 Hb A 2 variants was similar and about one-half of that for Hb S.
We also studied polymerization of mixtures of Hb S and two of the Hb A 2 variants to further clarify whether these differences in polymerization between Hb S and Hb A 2 ␦E6V could be explained by differences in the amino acid at position 87 (F3) in the non-␣ chains. 1:1 mixtures of deoxy Hb S/Hb A 2 ␦E6V polymerized with a delay time when the concentration was about 2-fold greater than that required for deoxy Hb S polymerization (Fig. 6A). 1:1 mixtures of Hb S/Hb A 2 ␦E6V,␦Q87T also polymerized, but the concentration required for polymer formation was similar to that of Hb S (Fig. 6B). Logarithmic plots of delay time versus concentration for the Hb S/Hb A 2 ␦E6V mixture showed a straight line shifted right ϳ0.25 unit from the line for Hb S, which was similar to the line for 1:1 mixtures of Hb S/Hb F ␥E6V,␥Q87T (Fig. 7).
Polymerization of 1:1 mixtures of deoxy Hb A 2 and Hb S under the same conditions showed a delay time prior to polymerization that was slightly shorter than that of deoxy Hb F/Hb S mixtures at the same hemoglobin concentration. The length of the delay time also depended on hemoglobin concentration: the lower the concentration, the longer is the delay time. Log-arithmic plots of delay time versus concentration for Hb A 2 /Hb S mixtures showed a straight line like Hb F/Hb S mixtures which was shifted right 0.42 unit from the line for Hb S (Fig. 7). The value for Hb F/Hb S mixtures was right shifted 0.5 unit from Hb S, indicating that the initiation of polymerization in these mixtures is controlled by Hb S, which represents about one-fourth of the total hemoglobin concentration (3,10). These results also suggest that Hb F, Hb A 2 , and FS hybrids are excluded from the initiation of polymerization with Hb S, while some A 2 S hybrids in the A 2 S mixture can participate in the formation of nuclei with Hb S.
Total polymer formed as a function of hemoglobin concentration was also determined in order to define critical concentrations required for polymerization of Hb S/Hb A 2 ␦E6V and Hb S/Hb A 2 ␦E6V,␦Q87T mixtures. These results were then compared with those for Hb S, Hb S/Hb A 2 , and Hb S/Hb F mixtures. Polymer formation increased linearly with increases in hemoglobin concentration (Fig. 8). Critical concentrations for polymerization of Hb S/Hb A 2 ␦E6V mixtures were slightly higher than that of Hb S, while critical concentration for Hb S/Hb A 2 ␦E6V,␦Q87T mixtures was similar to that of Hb S. Critical concentrations for polymerization for Hb A 2 and Hb F, determined by extrapolation of the lines to zero, are 2.8-and 3.3-fold higher than that of deoxy Hb S, respectively. These results reinforce the contention that exclusion of A 2 S hybrids from Hb S polymerization in Hb S/Hb A 2 mixtures is primarily in trans to the Val ␤6 contact and that Thr ␤87 near the EF helix hydrophobic acceptor pocket made by Leu ␤88 and Phe ␤85 in the F helix affects the formation of nuclei and polymers.

Polymerization of Hb A 2 Variants
Containing Val ␦6 -Hb A 2 represents about 2% of the total hemoglobin in adult erythrocytes, and like Hb F, Hb A 2 inhibits polymerization in vitro of deoxy Hb S (5). The primary sequence of the ␦ chain of Hb A 2 differs from that of the ␤ chain of Hb A in only 10 of 146 residues. Previous studies comparing naturally occurring hemoglobin variants suggested that ␦87-Gln and ␦22-Ala in Hb A 2 were important amino acids for inhibition of Hb S polymerization (5). Our previous studies on polymerization of the double mutant deoxy Hb S ␤T87Q also suggested that Gln ␤87 is a key amino acid involved in exclusion of A 2 S as well as FS hybrids from Hb S nucleation (12). Our present results show that the critical concentration for polymerization of Hb A 2 ␦E6V was 1.3-fold higher than that of deoxy Hb S and that critical concentrations for Hb A 2 ␦E6V,␦Q87T and Hb A 2 ␦E6V,␦A22E,␦Q87T were similar to that of deoxy Hb S. These results are similar to those comparing Hb S ␤T87Q and Hb S and indicate that Gln at position 87 in Hb A 2 ␦E6V or Hb S ␤T87Q play comparable roles in the inhibition of Hb A 2 ␦E6V and Hb S ␤T87Q polymerization, respectively. These results in low and high phosphate buffers using recombinant and naturally occurring Hb variants reinforce the role of Gln at position ␦87 as a key amino acid involved in Hb A 2 inhibition of Hb S polymerization.
Recently, recombinant human Hb A containing Gln ␤87 and Ala ␤22 was produced in transgenic mice (30). Evaluation of polymerization of mixtures (25:75) of this variant and Hb S in a high phosphate buffer showed that polymer formation was the same as that of FS mixtures containing 25% Hb F. These results suggested that this Hb A variant, which contains Gln ␤87 and Ala ␤22 like Hb A 2 , inhibits Hb S polymerization as effectively as Hb F and Hb A 2 (30). These findings also corroborate our previous results on polymerization of Hb S/Hb S ␤T87Q mixtures which showed that substitution of Gln for Thr at ␤87 is sufficient to promote exclusion of hemoglobin tetramers (12).
Although the critical concentrations for polymerization of the three Hb A 2 variants in the deoxy form were similar to that of deoxy Hb S, the kinetics of polymerization are different: Hb A 2 ␦E6V polymerized without a delay time; while polymerization of Hb A 2 ␦E6V,␦Q87T and Hb A 2 ␦E6V,␦A22E,␦Q87T was accompanied by a delay time. These results suggest that Hb A 2 ␦E6V and the two other variants (e.g., Hb A 2 ␦E6V,␦Q87T and Hb A 2 ␦E6V,␦A22E,␦Q87T) polymerize by different mechanisms, the former by a linear and the latter two by a nucleation-controlled polymerization mechanism like that for deoxy Hb S (29). Our recent results with Hb F variants containing Val ␥6 show that both deoxy Hb F ␥E6V and Hb F ␥E6V,␥Q87T polymerized similarly without a delay time at a higher concentration than deoxy Hb S in low and high phosphate buffers (13). In addition, changing Gln to Thr at ␥87 in Hb F ␥E6V (Hb F ␥E6V,␥Q87T) slightly inhibited rather than promoting polymerization compared to Hb F ␥E6V (13). These results suggest that the effect of Thr at position 87 in ␦ and ␥ chains of the Val ␦6 -Hb A 2 and Val ␥6 -Hb F variants in promoting nucleationcontrolled polymerization is different.
McCune et al. (30) explained the anti-sickling effect of Hb A containing Gln ␤87 in mixtures with Hb S using a computer graphic simulation and proposed that the larger side chain of Gln compared with Thr at ␤87 prevents insertion of Val ␤6 into the acceptor pocket of the F helix. However, our results show that the critical concentration for polymerization of deoxy Hb A 2 ␦E6V was only about 1.3-fold higher than that of deoxy Hb A 2 ␦E6V,␦Q87T, which is similar to the difference in critical concentrations comparing deoxy Hb S ␤T87Q with deoxy Hb S. Our previous results also showed that inhibition of polymer formation by substitution of larger bulky amino acids like Phe ␤88 for Leu in the acceptor pocket resulted in a 10-fold increase in the concentration required for polymerization compared with deoxy Hb S (34). In addition, Hb S/Hb Quebec-Chori (Hb ␤T87I) mixtures, in which the larger side chain of Ile is substituted for Thr at ␤87, actually accelerates nucleation and polymerization (35), suggesting that substitution of a larger side chain than Thr at ␤87 can, in fact, promote nucleation and polymerization. We speculate that the EF helix acceptor site for Val ␤6 in Hb S ␤T87Q has a similar conformation to the EF helix of deoxy Hb A 2 . Even though Hb S ␤T87Q polymer formation was preceded by a delay time prior to polymerization, the kinetic progress curve of polymerization after the delay time was linear and not sigmoidal like deoxy Hb S (12). These results suggest that Gln at position 87 in the ␦ or ␤ chains of Hb A 2 ␦E6V or Hb S ␤T87Q, respectively, affects the environment of the acceptor pocket for Val ␤6 during polymerization, inhibits the formation of nuclei and polymers, and may also affect protein-protein interactions between ␦87 or ␤87 and other amino acids during polymerization.
Thr ␤87 in 1-␤ 1 of deoxy Hb S is not a direct contact site for Val ␤6 but is involved in lateral contacts with Ser ␤9 , Ala ␤10 , and Ala ␤13 in 1-␤ 2 of Hb S polymers. Thr ␤87 in Hb S is involved in interactions between parallel double strands in crystals or fibers and also forms strong hydrogen bonds with 1␣ 1 -139Lys and 2␣ 2 -81Ser (32,33). From these results, we propose that Val ␦6 in Hb A 2 ␦E6V and Val ␤6 in Hb S ␤T87Q are able to insert into the acceptor pocket made by Phe ␦85 and Leu ␦88 , even though this F helix contains Gln at ␦87. Previous results with deoxy Hb S crystals indicated participation of Glu ␤22 in an axial contact to form an ionic bond with His B1 (20) ␣ 2 (32). Even though differences between Glu and Ala at position ␦22 in Hb A 2 ␦E6V,␦Q87T or Hb A 2 ␦E6V,␦A22E,␦Q87T were expected to affect polymerization properties, our results showed minimal effect if any. These results suggest that His B1 (20) ␣ 2 at neutral pH did not affect axial contacts of the polymers and/or that Ala ␦22 was not involved in protein-protein interactions in polymers of Hb A 2 ␦E6V,␦Q87T or Hb A 2 ␦E6V, ␦A22E,␦Q87T. Other amino acid differences between Hb A 2 and Hb A such as Ser ␤9 versus Thr ␦9 , His ␤116 versus Arg ␦116 , and Val ␤126 versus Met ␦126 , which may be involved in contact sites (34), may also be contributing to differences in polymerization properties between Hb S and Hb A 2 ␦E6V,␦Q87T.
Our present results with the Val ␦6 Hb A 2 variants suggest that changing Thr to Gln at ␤87 of Hb A is critical for antisickling, while changing Glu to Ala at ␤22 has little or no effect on exclusion of hybrid hemoglobins in mixtures with Hb S. Therefore, it may not be necessary to substitute Ala for Glu at ␤22 when preparing anti-sickling hemoglobins for gene therapy of sickle cell disease (30). In fact, Asp is present in the ␥ chain of Hb F instead of Glu, and Hb F inhibits polymerization of deoxy Hb S like Hb A 2 . Furthermore, the rate of assembly of ␣ and non-␣ chains to form ␣␤ or ␣␤-variant dimers depends on electrostatic attraction (36). More negatively charged ␤ chains are expected to have a higher affinity than ␤ S chains for ␣ chains; therefore, hemoglobin with Glu rather than Ala at ␤22 may, in fact, be a more efficient anti-sickling hemoglobin for gene therapy for sickle cell disease. X-ray analysis of these Hb A 2 variants is now required to evaluate the structural effects of the Gln for Thr substitution in the Val ␤6 acceptor pocket as well as at other interaction sites.
Inhibition of Hb S Polymerization by Hb A 2 -Our previous results using mixtures of Hb S/Hb S ␤T87Q suggested that Hb S ␤T87Q and one-half of the asymmetrical ␣ 2 ␤ S ␤ S, T87Q hybrids were excluded from the formation of nuclei with Hb S (12). Logarithmic plots of delay time versus concentration for the Hb S/Hb A 2 ␦E6V mixture showed a straight line shifted right ϳ0.25 unit from the line for Hb S, which was similar to the line for Hb S/Hb S ␤T87Q mixtures (Fig. 6). In contrast, the line for the Hb S/Hb A 2 ␦E6V,␦Q87T mixture was similar to that of Hb S. Equal mixtures of Hb S and Hb A 2 ␦E6V contain 25% Hb S, 50% asymmetrical ␣ 2 ␤ S ␦ 6 Val hybrid, and 25% Hb A 2 ␦E6V. Nucleation in these mixtures is calculated to be controlled by 56% of the total hemoglobin concentration based on the 0.25unit difference comparing the lines for Hb S and Hb S/Hb A 2 ␦E6V mixtures. These results suggest that Hb S and about one-half of the asymmetrical ␣ 2 ␤ S ␦ 6 Val hybrid contribute to the formation of nuclei with Hb S in these mixtures and that Val ␦6 can interact with an EF acceptor pocket containing Thr ␤87 in hybrid hemoglobins as well as Hb S, just like Val ␤6 . These results are consistent with our previous findings on polymerization of Hb S/Hb S ␤T87Q mixtures (12).
The primary inhibitory effect of Hb A 2 on Hb S polymerization is to exclude both the asymmetrical hybrid ␣ 2 ␤ S ␦ in A 2 S mixtures like ␣ 2 ␤ S ␥ in FS mixtures from the Hb S polymer (5,6). Thus, inhibition of polymerization by Hb A 2 is primarily in trans to the ␤6 Val contact of Hb S polymers. The 0.1-unit difference in left shift on the X axis of the line for A 2 S mixtures compared with that of FS mixtures in the experiments in a high phosphate buffer (Fig. 7) indicates that Hb A 2 inhibits Hb S polymerization slightly less than Hb F, which corresponds to results in 0.1 M phosphate buffer reported previously by Benesch et al. (9). These differences can be explained by participation of some ␣ 2 ␤ S ␦ hybrids in the formation of nuclei with Hb S at about one-eighth the efficiency of Hb S. Copolymerization with Hb S occurs because Thr ␤87 in ␤ S in the ␣ 2 ␤ S ␦ hybrids can interact with Val ␤6 in Hb S. In contrast, ␣ 2 ␤ S ␥ hybrids in FS mixtures are completely excluded from Hb S nucleation. Our results also offer an explanation of the slight differences between the effects of Hb A 2 and Hb F on the inhibition of Hb S polymerization. These results again suggest that hydrophobic interactions between ␤6 donor/acceptor sites as well as communication of other interaction sites with Thr ␤87 are critical for the formation of nuclei. These findings also reinforce our conclusions from polymerization studies of mixtures of Hb S and the Val ␦6 Hb A 2 variants, suggesting differences in inhibition of deoxy Hb S polymerization by Hb A 2 and Hb F.
Even though Hb A 2 inhibits polymerization of Hb S like Hb F in vitro, recent studies in transgenic mice show that expression of high levels of ␦-globin with ␤ S -globin chains resulted in severe red blood cell shape abnormalities. These findings suggest that when overexpressed, ␦ chains can interact with red blood cell membranes, drastically modify their properties, and generate red blood cell membrane abnormalities (14). However, Hb A 2 is more resistant to thermal denaturation than Hb A because of an additional contact at the ␣ 1 ␦ 1 interface (37). Our present results showed more instability to mechanical agitation for Hb A 2 ␦E6V and Hb A 2 compared with Hb S and Hb A, respectively. These results suggest that high hydrophobicity of ␦-globin chains may promote increased hydrophobicity of the ␣ 2 ␤ S ␦ hybrid, which may result in acceleration of interaction of the hybrid hemoglobin with red blood cell membrane proteins. These results discourage the use of the anti-sickling properties of ␦ chains in gene therapy even though Hb A 2 inhibits Hb S polymerization like Hb F. Although deoxygenation-induced sickling is the most easily demonstrated property of sickle erythrocytes, numerous membrane abnormalities have also been described (38). It seems likely that some of these membrane and red blood cell abnormalities in transgenic mice expressing high levels of ␦ and ␤ S chains may be related to the instability of oxy Hb S and Hb A 2 that becomes apparent during mechanical agitation.