The Hb A Variant (β73 Asp→Leu) Disrupts Hb S Polymerization by a Novel Mechanism

J. Mol. Biol. (2006) 362, 528–538

doi:10.1016/j.jmb.2006.07.047

The Hb A Variant (β73 Asp→Leu) Disrupts Hb S Polymerization by a Novel Mechanism Kazuhiko Adachi 1 ⁎, Min Ding 1 , Saul Surrey 2 , Maria Rotter 3 Alexey Aprelev 3 , Mikhail Zakharov 3 , Weijun Weng 3 and Frank A. Ferrone 3 1

The Children's Hospital of Philadelphia, Division of Hematology, University of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA 2

Cardeza Foundation for Hematologic Research, Department of Medicine, Jefferson Medical College, Thomas Jefferson University, Philadelphia, PA 19107, USA 3

Department of Physics, Drexel University, Philadelphia, PA 19104, USA

Polymerization of a 1:1 mixture of hemoglobin S (Hb S) and the artificial mutant HbAβ73Leu produces a dramatic morphological change in the polymer domains in 1.0 M phosphate buffer that are a characteristic feature of polymer formation. Instead of feathery domains with quasi 2-fold symmetry that characterize polymerization of Hb S and all previously known mixtures such as Hb A/S and Hb F/S mixtures, these domains are compact structures of quasi-spherical symmetry. Solubility of Hb S/ Aβ73Leu mixtures was similar to that of Hb S/F mixtures. Kinetics of polymerization indicated that homogeneous nucleation rates of Hb S/ Aβ73Leu mixtures were the same as those of Hb S/F mixtures, while exponential polymer growth (B) of Hb S/Aβ73Leu mixtures were about three times slower than those of Hb S/F mixtures. Differential interference contrast (DIC) image analysis also showed that fibers in the mixture appear to elongate between three and five times more slowly than in equivalent Hb S/F mixtures by direct measurements of exponential growth of mass of polymer in a domain. We propose that these results of Hb S/Aβ73Leu mixtures arise from a non-productive binding of the hybrid species of this mixture to the end of the growing polymer. This “cap” prohibits growth of polymers, but by nature is temporary, so that the net effect is a lowered growth rate of polymers. Such a cap is consistent with known features of the structure of the Hb S polymer. Domains would be more spherulitic because slower growth provides more opportunity for fiber bending to spread domains from their initial 2-fold symmetry. Moreover, since monomer depletion proceeds more slowly in this mixture, more homogeneous nucleation events occur, and the resulting gel has a far more granular character than normally seen in mixtures of non-polymerizing hemoglobins with Hb S. This mixture is likely to be less stiff than polymerized mixtures of other hybrids such as Hb S with HbF, potentially providing a novel approach to therapy. © 2006 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: hemoglobin; sickle hemoglobin; HbF; anti-HbS polymerization; fiber formation

Introduction In sickle cell disease the mutation from GTG to GAG in the triplet code at the sixth position from

Abbreviations used: Hb, hemoglobin; DIC, differential interference contrast; LCR, locus control region. E-mail address of the corresponding author: [email protected]

the N terminus of the β-globin chain results in replacement of negatively charged Glu with the uncharged, hydrophobic Val, which is on the molecular surface of the Hb S molecule and decreases its solubility. When deoxy-Hb S concentration exceeds its solubility, hemoglobin aggregates and forms long, multi-stranded polymers, which are comprised of 14 strands. Because one polymer can nucleate others on its surface, the polymers form in attached arrays, called polymer domains. The domains begin with a

0022-2836/$ - see front matter © 2006 Elsevier Ltd. All rights reserved.

Anti-Hb S Polymerization of Hb A β73Leu

single homogeneous nucleus that creates the first polymer, and spread by heterogeneous nucleation of new polymers onto the surface of other growing polymers. Intracellular domains cause a significant reduction in red blood cell deformability (sickling), leading to obstruction of flow in the microcirculation.1,2 Each fiber is composed of seven half-staggered double strands, 3 suggesting that the doublestranded structure of deoxy Hb S crystals is an essential component of fibers within cells. The deoxy-Hb S crystal structure has been extensively refined, permitting a detailed structural comparison with deoxy-Hb A.4,5 The contacts that are made along the axis of the double strand are called axial, while the diagonal contacts between the molecules in different strands (but within the same double strand) are called lateral. The lateral contacts are most critical for Hb S polymerization, since this is the location of the mutant β6Val site, which interacts with a hydrophobic pocket on an adjacent molecule formed by β88Leu, β85Phe and several heme atoms.5 The double-strand structure is such that the lateral contact region occurs between the β-chain of one molecule (the donor) and the other β-chain of the next molecule (the acceptor). In addition, for most of the molecules in the fiber, only one β6Val residue in each Hb S tetramer participates in this intermolecular interaction in the double strand: the β6Val residue that does not form a contact remains exposed to solvent.1,2,6 Thus, a hybrid molecule with one βA chain and one βS chain would have one altered donor, but both acceptor regions remain unchanged. However, some β6 (donor) regions that do not participate in the crystallographic double strands are believed to take on additional roles.7 In some molecules, it appears that both donor and acceptor regions are active in stabilizing the fiber.6,7 There are an additional two such contacts within a 14 molecule slice through the fiber.8 Each molecule thus interacts with one lateral neighbor as a donor, and the other lateral neighbor as an acceptor. In addition, the same residues that make lateral contacts also appear to be the residues involved in heterogeneous nucleation. Computational refinements of X-ray-determined crystal structure clarified the details of many of the axial and lateral contacts. These results show a significant structural change is observed for one of the β73Asp residues in addition to a slight hingelike motion of the A helix including β6Val and β4Thr of the βS -globin subunits.5 The β73Asp residues interacts with β4Thr by a hydrogen bond in α2βS2 tetramers in Hb S crystals, which is more intimately involved in the intermolecular contact than the other β73Asp in Hb S tetramers.9 The nature of the residue at β73 is extremely important in determining how assembly will occur. Hb C-Harlem (β6 (Glu→Val) and β73 (Asp→Asn)) shows dramatically slowed kinetics of assembly albeit with similar solubility to Hb S.10 Hb S (β73 Asp→Leu) has higher solubility and slow kinetics, whereas Hb S (β73 Asp→His) has lower solubility

529 and faster kinetics than Hb S.9 Inhibition of Hb Sβ73Leu polymerization compared to Hb S is most likely caused by eliminating the hydrogen bond interaction of the hydroxyl group of β4Thr with the β73 amino acid in addition to some influence on hydrophobic interactions of β6Val with β88Leu and β85Phe in Hb S polymers. Finding hemoglobin which will inhibit polymerization, especially when hybridized with Hb S, is of great clinical interest. Hybridization occurs because the hemoglobin tetramer can readily split into a pair of αβ dimers, which can then recombine with other dimers from similar or different parent species. Partial replacement of Hb S with Hb F (α2γ2) inhibits polymerization of Hb S in vitro as well as in vivo. Clinically, the presence of high levels of Hb F in patients with sickle cell disease is associated frequently with reduced disease severity.1,2,6,11–13 This inhibitory mechanism depends on Hb F/S hybrids (α2βSγ), which are excluded from polymer formation like non-S hemoglobin, at least when Hb F concentration is less than 20%.12,14,15 In the quest to inhibit Hb S polymer formation, the central theme has been to find hemoglobins that raise solubility by their reluctance to enter the polymer and that also slow the rate of polymerization. The natural limit of such a strategy is reached by hemoglobin that does not polymerize at all. Since it appears that neither Hb F nor its hybrids with Hb S enter the polymer, Hb F and its variants would seem to represent the limit of maximum inhibition. Here, we describe polymerization properties of Hb S/Aβ73Leu mixtures which have superior anti-sickling properties compared with Hb S/F mixtures, and discuss the anti-Hb S polymerization mechanism of Hb A β73Leu which is different from Hb F in an effort to define more effective anti-sickling hemoglobin variants.

Results Solubility of mixtures Hb S and Hb A or Hb Fβ73 variants The solubility of mixtures was measured in 1.8 M phosphate buffer (Figure 1). The solubility of Hb S/ Aβ73Leu mixtures was slightly lower than that of Hb S/F mixtures. Both were substantially greater than that of Hb S/A mixtures. All measured solubilities of mixtures depended on initial concentration, since non-Hb S and some of hybridhemoglobins are solubilized and could not incorporate into polymers. In addition, the solubilties of Hb S/F and Hb S/A mixtures were in quantitative agreement with expected values based on low phosphate results. The lower solubility of the Hb S/Aβ73Leu mixtures is consistent with a small amount of co-polymerization. Solubilities were also measured in 1.0 M phosphate buffers using two methods, with the results summarized in Table 1. In addition to conventional measurement by centrifugation, a new method

Anti-Hb S Polymerization of Hb A β73Leu

530

Figure 1. Solubility of mixtures of Hb S/Hb A and mixtures of Hb S/Hb F-containing 73 His and 73 Leu variants in 1.8 M phosphate buffer. Solubility of the deoxy forms of 1:1 mixtures of Hb S/Hb A β73 His (▵) and Hb S/Hb A β73 Leu (▿), Hb S/Hb F γ73 His (□) or γ73 Leu ( ). Results are compared to Hb S (●), 1:1 AS (▴) and FS (■) mixtures. Solubility was determined in 1.8 M phosphate buffer by centrifugation after completion of polymerization.

using photolysis of CO-Hb S in droplets in oil was also used. At 25 °C the solubility of Hb S/F mixtures by the photolytic method was about 11.3(±0.8) g/dl, and the solubility of Hb S/A β73Leu mixtures was 12.0(±0.7) g/dl, for initial concentrations in the range 14.4 to 15.7 g/dl (see Table 1). At 30 °C, Hb S/F mixtures had solubilities of 9.9(±0.2) g/dl (15 g/ dl initial ) by the photolytic method, and 8.0(±0.7) g/ dl (14 g/dl initial ) by centrifugation. The solubility of Hb S/A β73Leu mixtures was slightly higher, being determined as 11.2(±0.3) (15 g/dl initial) by photolysis and 9.8(±0.9) g/dl (14 g/dl initial) by centrifugation. The composition of the supernatant also was assayed in the centrifugation experiment, and was consistent with the solubility measurements. For Hb S/F mixtures, 75(±3)% non-Hb S hemes were found in the supernatant, while in Hb S/A β73Leu mixtures 65%(±1)% was non-HbS.

Leu had no effect, and the resulting mixtures were indistinguishable from those of S/F (Figure 2(b)). In 1.0 M phosphate buffer, homogeneous and heterogeneous nucleation rates of polymers were determined by laser photolysis at 25 °C. Homogeneous nucleation rates are determined from stochastic distributions of tenth times, while heterogeneous nucleation rates are inferred as the major contribution to the exponential growth rate, B. Homogeneous nucleation rates of Hb S/A β73Leu mixtures were indistinguishable from Hb S/F mixtures as shown in Figure 3(a). The B values did show a systematic difference (Figure 3(b)), which is small given the range of the data, but can be discerned because of the higher precision of the B measurement than the homogeneous nucleation rate fo. The B values for Hb S/A β73Leu mixtures were about three times slower than Hb S/F mixtures. Differential interference contrast (DIC) image analysis of Hb S/Aβ73 Leu mixtures Heterogeneous nucleation allows new polymers to form on the surface of pre-existing ones, at sites believed to be the same contacts as found in the lateral contacts.16 Thus, polymerization naturally and inevitably generates domains. Moreover, the relationship between lateral contacts and heterogeneous nucleation makes it possible that domain structure could change as amino acids in the receptor pocket change. Thus, images of domain structure of Hb S with Hb Aβ73 Leu were observed by DIC microscopy in 1.0 M phosphate buffer at room temperature. We conducted image analysis of the effects of increasing levels of Hb Aβ73Leu in the mixture on the DIC polymer images of the mixtures in 1.0 M phosphate buffer. Whereas DIC images of Hb S β73Leu and Hb C-Harlem showed linear crystals in both 0.1 M and 1.0 M phosphate buffers,9

Table 1. Measured solubilities of 1:1 Hb S/F and Hb S/ Aβ73 Leu mixtures in 1.0 M phosphate buffer (pH 7.35) at 25 °C or 30 °C T(°C) co(g/dl) cs(g/dl)

Kinetics of polymerization of 1:1 mixtures of Hb S/A or Hb F variants We reported previously that Hb S β73His and Sβ73Leu variants polymerized with delay times, requiring, respectively, significantly lower and higher hemoglobin concentrations than Hb S.9 Polymerization of 1:1 mixtures of Hb S/A β73His or β73Leu variants in 1.8 M phosphate buffer was compared to Hb S/A mixtures and Hb S/F mixtures. Polymerization in both mixtures was preceded by a delay time prior to polymerization, like Hb S. Logarithmic plots of delay time versus hemoglobin concentration showed straight lines with similar slopes to Hb S (Figure 2(a)). Hb S/A β73Leu mixtures were comparable to Hb S/F mixtures. Changes in the γ73 position to His or

±

#

e2

25

15.0

12.1

1 0.03

25

15.0

11.8

0.8 2 0.05

25

15.0

12.5

1 0.02

25 25 30

15.7 14.4 15.0

11.4 10.4 11.4

0.7 4 0.08 1 0.10 1 0.07

30 30 30

15.7 14.4 14.0

9.8 10.0 9.8

1 0.17 1 0.13 0.9 3 0.13

30

14.0

8.0

0.7 3 0.30

Hybrid

Method

S/A Photolysis β73Leu S/A Photolysis β73Leu S/A Photolysis β73Leu S/F Photolysis S/F Photolysis S/A Photolysis β73Leu S/F Photolysis S/F Photolysis S/A Sedimentation β73Leu S/F Sedimentation

Error from replication accuracy (±); number of measurements (# ); co-polymerization probability for the hybrid molecules (e2); initial hemoglobin concentration (co); and measured solubility (cs) are indicated.

Anti-Hb S Polymerization of Hb A β73Leu

531 greater than 30%, the mass of domains become obviously smaller than those of less than 20% of Hb Aβ73Leu. When the percentage of Hb Aβ73Leu was 40% and 50%, the size (measured in image pixels) of single domains with fibers which grow in all directions depends on total hemoglobin concentration of Hb Aβ73Leu in the mixture with Hb S; the higher the concentration of Hb Aβ73Leu in the mixture the smaller the mass of domain, which decreased linearly with increasing hemoglobin concentration. In contrast, the mass of the domains of Hb F/S mixtures decreased only slightly with increasing the ratio of Hb F. The change from feathery to spherulitic domains could also be

Figure 2. Relationship between delay time prior to polymerization and hemoglobin concentration for mixtures of Hb S/A or Hb S/F-containing 73His and 73Leu variants. Logarithmic plots of the reciprocal delay time prior to polymerization versus Hb concentration are shown for (a) 1:1 mixtures of Hb S/A β73His (▵) and β73Leu (□), and for (b) HbS/HbF γ73His (▵) and γ73 Leu (□). Results are shown also for Hb S (●), 1:1 Hb AS (▴) and Hb FS (■) mixtures in 1.8 M phosphate buffer (pH 7.4). Experiments were performed using 1.8 M phosphate buffer (pH 7.4), with a temperature jump from 0 °C to 30 °C to initiate polymer formation.

DIC images of domains made from Hb S and Hb S β73His showed the usual domain patterns in 0.1 M and 1.0 M phosphate buffers.9 DIC images of 1:1 mixtures of Hb S/Aβ73Leu were compared to those of Hb S, Hb A/S and Hb F/ S mixtures (Figure 4). The domains of Hb S/Aβ73 Leu mixtures (14 g/dl) were much more compact, dense and spherical than Hb S, Hb A/S and Hb F/S mixtures. In contrast, DIC images of domains of all other mixtures were similar to those of Hb S, and Hb A/S mixtures, even though concentrations required for observation of fibers in Hb A/S mixtures (10 g/ dl) and Hb F/S mixtures (14 g/dl) were much higher than that of Hb S (3.5 g/dl). The difference in the images is not present at low fractions of Hb A β73Leu (Figure 5). As the percentage of Hb Aβ73Leu in the mixture becomes

Figure 3. Homogeneous and heterogeneous nucleation rates for mixtures of Hb S containing 50% Hb F or Hb A β73Leu in 1.0 M phosphate buffer. Laser photolysis of the CO derivative was used to induce polymerization in an array of spots. The distribution yields the homogeneous nucleation rate, while the initial growth of the progress curves is fit to an exponential function of time, with rate B. (a) Top panel shows the homogeneous nucleation rate, fo. Filled squares show results for HbS/F mixtures. Open circles show HbS/A β73Leu mixtures. The one open diamond point is determined by direct measurement of the rate of formation of domains, taken from the data of Figure 8. (b) Bottom panel shows B, the exponential growth rate that is dominated by heterogeneous nucleation, but which also has contributions from the polymer elongation rate. Filled squares show mixtures with 50% Hb F; open circles show mixtures of Hb S with 50% Hb A β73Leu.

532

Anti-Hb S Polymerization of Hb A β73Leu

Figure 4. DIC images of polymerization after various times in Hb S/A β73Leu mixtures in 1.0 M phosphate buffer compared to Hb S, Hb A/S or Hb F/S mixtures. (a– d) DIC images of polymerization of the deoxy forms of Hb S (3.5 g/dl), 1:1 A/S mixtures (10 g/dl), and F/S mixtures (14 g/dl) and 1:1mixtures of S/A β73Leu (14 g/dl), respectively, for different lengths of reaction time and at equilibrium (right panel) are shown at the bottom of each panel. Polymers were prepared after temperature-jump from 0 °C to room temperature (22 °C). The length of the frame (xaxis) in the photographs represents 150 μm.

promoted by increasing total concentration of the mixture even though the percentage of Hb Aβ73Leu in the mixture is kept constant (Figure 6). The growth rates of mass of single domains of 1:1 Hb S/Aβ73Leu and Hb S/F mixtures as assessed by measurements of pixel area of the domain are shown in Figure 7. The mass of a domain in the Hb S/Aβ73Leu mixtures expands much more slowly than Hb S/F mixtures. If the Hb S/

Aβ73Leu mixture's mass is increased by a factor of 23.5, the growth curves overlap for the last 50% of the growth, and are not even very different for the initial 50% (as shown in Figure 7). Observing the area effectively averages the rate of radial growth. Simply taking area as πR2 suggests that Hb S/ Aβ73Leu domains are expanding with linear growth rates which are 4.8 times slower than Hb S/F mixtures.

Anti-Hb S Polymerization of Hb A β73Leu

533

Figure 5. Mass of domains in Hb S/A β73Leu mixtures with different ratios of Hb A β73 Leu in 1.0 M phosphate buffer. The mass of polymers made by single domains and fibers with different ratios of Hb S to Hb A β73Leu was measured by the area of pixel numbers in the DIC images of domains of Hb S/A β73Leu mixtures (●) compared to Hb F/S mixtures (○) at their equilibrium stage. The concentration of Hb S in different percentages of the both Hb A β73Leu and Hb F in the mixtures was kept at 4 g/dl as calculated by the binominal distribution of mixtures.

The number of domains of Hb S/Aβ73Leu mixture increased with time and then plateaued, similar to the behavior of domains of Hb S/F mixtures (Figure 8). However, there were five times as many of Hb S/Aβ73Leu domains as those of Hb S/F mixtures in equal areas. By following the number of Hb S/A β73Leu domains

Figure 6. Mass of domain in Hb S/A β73Leu mixtures with different total concentrations of Hb A β73Leu in 1.0 M phosphate buffer. The mass of polymers made by single domains with different total concentrations and a fixed percentage of Hb Aβ73Leu of 30% was measured by pixel numbers in the area of DIC images of domains in Hb S/A β73Leu mixtures at their equilibrium stage.

(which is equal to the number of homogeneous nuclei) as a function of time, the rate of homogeneous nucleation could be determined as the initial slope of the curve (Figure 8). The slope does not intersect zero because nuclei cannot be observed directly (being smaller than the wavelength of the light). Therefore, a domain generated from a nucleus at time zero cannot be counted until some time later when it reaches some minimum critical size, a limiting case of the familiar instrumental delay time. Domains are no longer formed (nucleation ceases) after the

Figure 7. Growth of domain area in Hb S/A β73Leu mixtures in 1.0 M phosphate buffer compared to those of Hb FS mixtures as a function of time. The size of domains in these mixtures was measured by pixel as a function of time. Open circles are Hb S/F mixtures, while filled circles are Hb S/A β73Leu mixtures. The size of domains in Hb S/A β73Leu mixtures (shown enlarged in the inset for clarity) is much smaller than that of Hb S/F mixtures, but can be overlaid on the Hb S/F domain data by increasing the area for each time point by a factor of 4.8, effectively expanding the y-axis, as shown by the small circles connected by the broken line.

534

Anti-Hb S Polymerization of Hb A β73Leu

show essentially no Hb S/F hybrid in the polymer, and again only a few percent of Hb SA β73Leu hybrids. Kinetics of polymerization

Figure 8. Increase in number of polymer domains in Hb S/A β73Leu mixtures in 1.0 M phosphate buffer compared to those of Hb FS mixtures as a function of time. The numbers of domains in these mixtures per constant area (4 200 μm2) are counted as a function of time. Polymers were prepared after temperature-jump from 0 °C to room temperature (22 °C).

monomer concentration begins to drop as monomers are incorporated into polymers.

Discussion By far the most dramatic result of these studies is the small size and spherulitic shape of the polymer domains that are observed in Hb S/Aβ73Leu mixtures. It is our hypothesis that this results from a transient capping phenomenon, which leads to a lowered polymer elongation rate, which also accounts for the lower exponential growth rate B. To exclude other possible explanations the solubility and kinetics will be examined in turn, before a detailed explanation is presented of the proposed mechanism. Solubility Measurements of solubility provide an overall assessment of the amount of hemoglobin polymerized. For the mixtures prepared here, hybrids will extensively form, and the solubility will therefore determine what fraction of hybrids enter the polymer. At 25 °C, the temperature at which the kinetics and DIC observations are made, there is no significant difference in solubility between Hb S/F mixtures and Hb S/A β73Leu mixtures. In 1.0 M phosphate, the Hb S/F mixture data would suggest about 9% co-polymerization of hybrids, and the Hb S/A β73Leu data about 3% of hybrids. Hence, if hybrids changed the polymer structure, this would be unlikely to affect more than a very small fraction of the observed polymers, and would not be able to create the different domains seen here. The results obtained in 1.8 M phosphate

Polymer domains are the result of heterogeneous nucleation, and so another way in which the domain shape and mass can differ is to change the geometry of the heterogeneous nucleation process. The exponential growth of polymers, which is responsible for the well-known delay time, is highly responsive to homogeneous and heterogeneous nucleation; and, therefore, it is important to consider changes that occur there. A heterogeneous nucleus is a cluster formed on the surface of a polymer, and so it must be determined if the stability of the cluster itself has changed. This is accomplished by measuring the rate of homogeneous nucleation, which depends on the concentration of nuclei (i.e. maximally unstable clusters) for both Hb S/F and Hb S/A β73Leu mixtures. As seen in Figure 3(a), the homogeneous nucleation rates are indistinguishable. From the rate at which nuclei form as seen by DIC microscopy, an entirely independent measurement of the nucleation rate is obtained. This point is shown as the diamond in Figure 3(a), and is completely consistent with data obtained by the stochastic photolytic method. We conclude that homogeneous nucleation is unchanged, and that the Hb S/A β73Leu hybrids do not change stability of the nuclei relative to Hb S/F mixtures. In examining B, which is affected by the delay time to form homogeneous nucleation and is related to heterogeneous nucleation and growth of polymer, a small but definite change is apparent. Can this be due to changing the attachment of the heterogeneous nuclei to the polymer from which they grow? Since only a few percent of the molecules in the polymer can be hybrids, it is hard to see how this change would alter the domain shape and mass beyond adding at most a small percent of fibers that might depart from the original fiber at larger angles to hasten the formation of spherulites. Suppose the mutant provided a new heterogeneous site, which had a higher probability of creating a nucleus than usual. Then it is possible that fibers would grow preferentially from this new site, even if it were only a small percent of the total, since not all possible heterogeneous sites are utilized in making nuclei. However, such a hypothesis would also require the rate of heterogeneous nucleation to have increased. In fact, it decreases, relative to Hb S/F hybrids by about a factor of 3. Thus, an altered heterogeneous nucleation site can be excluded as the source of the domain structure seen here. Thus, the B decrease is not likely to be the result of heterogeneous nucleation per se, and so other contributions to B need to be considered. If J is the rate of polymer elongation, g is the heterogeneous nucleation rate, f is the

Anti-Hb S Polymerization of Hb A β73Leu

homogeneous nucleation rate, it has been shown that:17 B2 ¼ J½g  df =dc It is thus possible to diminish B by decreasing the elongation rate, J, since the heterogeneous nucleation rate is unlikely to be changed, and df/dc, the concentration dependence of f, is also observed not to change. Domain structure of Hb S/F or Hb S/A β73Leu mixtures Very little direct work has been done on the ways in which domains assume the observed structures. With a simple model invoking documented polymer bending as well as heterogeneous nucleation, Dou & Ferrone previously showed that the shape of domains begins with a 2-fold symmetry that fans into a spherical shape, referred to as domain closure, after a period of time.18 If the bending rate is b, then theptime ffiffiffiffiffiffi for domain closure was found to vary as 1= Bb, where B is the exponential growth rate. However, the ultimate shape of the domain depends on whether the closure time would occur before monomers are all consumed (so that the domains become spherical) or after the available monomers are exhausted (in which case the domains are frozen in their 2-fold shapes and closure does not happen). The time for completion of polymerization is proportional to 1/B. So, if B were to decrease by a factor of 4 relative to Hb S/F mixtures, polymerization would be completed four times longer than in Hb S/F mixtures, whereas domain closure would occur only two times later than in Hb S/F mixtures. In other words, domains have the opportunity to become spherulitic before monomers are used up, as observed. Therefore, a decrease in J, leading to a decrease in B, favors the formation of spherulites. This is supported by the measurement of the domain area increase by measurements of exponential growth of the mass of polymer using DIC analysis. It is significant that the shape of the domain approach to equilibrium looks the same for Hb S/F and Hb S/A β73Leu mixtures' domains, for it suggests that the same processes are operative, but just happening with different net rates. The net elongation rate J is given by the equation: J ¼ kþ ðgc  gs cs Þ where γ is the activity coefficient, and the subscript s indicates the quantity which is evaluated at solubility (for 14 g/dl, γ=3, and γc–γscs = 6 mM). The decrease in linear domain growth rate as shown in Figure 7 is a factor of 4.8. This would translate into a change in B of a factor of 2.2 given no change in heterogeneous nucleation rates. It is possible that heterogeneous nucleation rates also get slower because k+ decreases and then B would change by a factor of 4.8. The factor of 3 by which B is observed

535 to decrease is thus well within the range of expected outcomes from a simple slowing of the elongation rates. Since this slowing of linear growth will lead to relatively earlier domain closure times, a decrease in polymer elongation rate can describe all the observed phenomena, and quantitatively unify the different observations. A mechanism for kinetic inhibition of Hb S polymerization by Hb SAβ73Leu hybrids The question is then how such a slowing of growth can occur when so little Hb SAβ73Leu hybrid is incorporated into the polymer? This is possible in a kinetic fashion. Suppose that a hybrid could bind to the end of a polymer in a nonproductive fashion meaning that one side of the hybrid could not support propagation. Then, when that hybrid is bound the polymer cannot grow. If the binding of the favorable side to the polymer were identical to binding of Hb S, then the overall equilibrium constant for binding would be that of binding an Hb S. Due to the dynamic nature of equilibrium, this cap would later fall off spontaneously, and elongation would then resume. But at any given moment, since a fraction of the polymers are capped, growth would be accordingly slower. The sickle polymer situation is more complex because the fiber has a 14 strand cross-section. One unproductive contact alone is unlikely to slow growth, since the remaining 13 strands might be able to grow around this. However, a group of unproductive contacts could be destructive. This type of cooperative destruction is seen in side fiber melting by ligand binding to the heme of hemoglobin. There, a hole of two or more molecules destabilizes the entire fiber.19 A detailed description can be easily formulated. Let the concentration of hybrids be denoted by cSX and the concentration of pure Hb S molecules be cSS. Let K be the equilibrium constant for binding an unproductive hybrid to the end of the fiber, and Kprod be the equilibrium constant for binding an Hb S molecule to the fiber end. cSS and cSX are the concentrations of pure Hb S and hybridized Hb SX molecules, respectively, where HbX stands for any non-Hb S subunit. The fraction of non-productive polymer ends is given by: KcSX =ðKprod cSS þ KcSX Þ in which the activity coefficients have cancelled, since they include all species in the given solution, and are therefore equal. Kprod is simply 1/cs. Assuming a binomial distribution, if X is the fraction of hemes in the initial mixture that do not come from Hb S tetramers, and co is the initial concentration, then cSS = (1–X)2co and cSX = 2X(1–X)co. Thus, the fraction of non-productive ends becomes 2XKcs/ (1-X+2XKcs). If two molecules are required to bind to make the end non-productive, this fraction must now be squared. If instead some other number of molecules (n) were required to make a non-

Anti-Hb S Polymerization of Hb A β73Leu

536 productive end, that would likewise entail raising the fraction to that power, n. This gives the expression for the effective, lowered growth rate, k+eff viz: eff kþ ¼ kþ ð1  ½2XKcs =ð1  X þ 2XKcs Þ2 Þ

For 1:1 mixtures, X = 0.5, and the above simplifies to: eff ¼ kþ ð1  ½Kcs =ð1=2 þ Kcs Þ2 Þ kþ

To reduce k+ by a factor of 4.8 as seen in the DIC measurements requires Kcs = 4. The three fold change in B values requires Kcs = 2.3. This gives the range for K. If K is expressed in terms of a pseudosolubility, c* = 1/K, then c* lies between 0.25 cs and 0.45 cs. This is analogous to a co-polymerization probability of 0.25–0.45, except that, unlike the usual co-polymerization probability which applies to the entire polymer mass, this applies solely to the end of a polymer. In 1.8 M phosphate the capping effects appear to be much smaller (cf Figure 2). Although unanticipated, this result is readily explained within the framework described here by assuming that K and cs respond differently to the increased phosphate. It is known that cs will decrease by about an order of magnitude as the molarity of the phosphate buffer increases from 1.0 M to 1.8 M. If, however, K does not increase equivalently, the overall product Kcs could decrease, thereby diminishing the capping effects. For example, Kcs = 0.32 would give a change of a mere 15% in the apparent on-rate. This capping behavior is easily rationalized in structural terms. A monomer in a polymer makes axial and lateral contacts, with the lateral involving a donor and acceptor region on the beta chains of Hb S molecules. In an Hb SA hybrid, there is only one donor region, viz that on the βS subunit and when that donor is engaged in a contact, the acceptor region on the βA remains. Consequently there is one orientation, in which an Hb SA hybrid can attach, except in those contacts where both donors are required. In the Hb Aβ73Leu, however, the presence of the Leu now substantially weakens the contact in the acceptor region. This is seen in its greater solubility, as reported.9 Thus, an Hb SAβ73Leu hybrid can attach to the end of the fiber using its donor site, but the exposed receptor region will not support further growth. Therefore, of the two possible orientations, there is only one way such molecules can bind to the polymer end. If all end sites were the same, the probability for these would thus be 0.5; however, there are sites where both donors and acceptors are required, and as shown elsewhere this translates into a total probability of 0.375. This is within the experimental uncertainty of the observed 0.25–0.45. The phenomenon is not seen for Hb SF hybrids because even if the βS subunit Val were to dock into a receptor pocket, the γ chain contains groups, which do not make the correct βchain partners, irrespective of the ability of the γ

chain to present an acceptor region. Specifically, the axial contact β22 Glu is changed in the γ chain to an Asp. This failure of axial stabilization offsets the lateral contact that would occur if an Hb S/F hybrid were to polymerize, and so keeps the equivalent kinetic capping effect from happening on Hb S/F hybrids. Implications of Hb A β73Leu There are two major implications of the results presented here. First of all, a new method has been demonstrated by which the overall rate of polymerization can be slowed. This slowing by a kinetic cap may be a helpful addition to the strategies to slow the overall rate of polymer formation. Although not as dramatic as the effects on nucleation rates, the effect adds to what was thought to represent an ultimate limit of inhibition (i.e. no copolymerization), and thus improves upon Hb S/F mixtures as a hybrid system. A second important consequence is that a gel formed by this hybrid may well be naturally “broken up” into separate domains, which would otherwise not form because further nucleation would have been suppressed by the rapidity with which monomers are consumed. This division into multiple domains could have important rheological consequences, since the attachment of polymers to one another is intrinsically weaker than the polymers themselves. It has long been known, for example, that a polymerized sample will flow if stirring has broken the polymers, while the same concentration of polymerized hemoglobin would be solid-like otherwise. That is, re-annealing of the broken fibers does not confer as much mechanical stability as originally long fibers. In addition, a S/ Aβ73Leu mixture is likely to make less stiff gels than polymerized mixtures of other hybrids such as HbS with HbF, potentially providing a novel approach to therapy. Again the “gold standard” of Hb F inhibition is exceeded by this mixture system. Even if Hb F inhibition were only matched rather than exceeded, there may be good reason to consider expression of another type of hemoglobin as a logical target for gene therapy. The efficiency of Hb F in inhibiting Hb S polymerization suggested for some time that transduction of γglobin genes into hematopoietic stem cells might be an effective strategy for sickle cell disease (SCD) gene therapy. 20,21 However, transgenic mouse experiments have demonstrated that locus control region (LCR) γ-globin constructs are expressed at least four times less efficiently than LCR β-globin constructs in adult erythroid tissue.20,22,23 High levels of γ-globin gene expression appear to be difficult to achieve in adult erythroid cells even in the absence of competition between γ-globin with βglobin gene by the LCR interactions in the same viral vector. In order to overcome this deficiency, the βglobin mutant gene has been used to introduce γglobin amino acid substitutions by substitution of Gln for β87Thr.20,21 In fact, several recombinant

Anti-Hb S Polymerization of Hb A β73Leu

hemoglobin variants have been shown to inhibit Hb S polymerization, like Hb F by substitution of Gln for β87Thr, and these Hb A variants were proposed for transplantation of genetically autologous hematopoietic stem cells for curing sickle cell disease.21,24,25 Hb A β73Leu is thus the first Hb variant with antipolymerization properties that exceed those of Hb F. It works by a different mechanism, is due to a single amino acid substitution in Hb A, and may thus be a better candidate for gene therapy of sickle cell disease than Hb F.

Materials and Methods Recombinant Hb A, Hb S or Hb F variants at the 73 po→ His → His sition (α2β73Asp and α2β2β73Asp → Leu, α2γ73Asp 2 2 → Leu and α2γ73Asp ) were expressed in bacteria after sub2 cloning the corresponding cDNAs into pHE2, which was originally constructed to co-express α and β or α and γglobin chains with methionine aminopeptidase (MAP) under transcriptional control of a ptac promoter.26 The β or γ-globin chain variants were constructed by sitespecific mutagenesis of the normal β or γ chain using recombination/PCR as described 9,27 and were coexpressed with α-chains to form tetrameric Hb S variants.9 Recombinant Hb A, Hb S or Hb F variants lack the chaininitiating N-terminal methionine due to cleavage by MAP, which is also co-expressed. Clones of the engineered variants were subjected to DNA sequence analysis of the entire β or γ-globin cDNA region using site-specific primers and fluorescently tagged terminators in a cycle sequencing reaction in which extension products were confirmed using an automated DNA sequencer. The plasmids were transfected into Escherichia coli (JM 109) (Promega Co., Madison, WI), and bacteria were grown at 30 °C with shaking at 225 rpm in one liter of Terrific broth containing 10 μM ampicillin to a density of about 3 × 1010 bacteria/ml. Expression of the variant tetramers was induced for 4 h at 30 °C by addition of 0.2 mM IPTG (Fisher Scientific, Fairlawn, NJ), and cultures then supplemented with 30 μM hemin (Aldrich Chemical Co. Inc., Milwaukee, WI) and 1% (w/v) glucose. Purification of hemoglobin was as described,9 and sample purity was assessed by cellulose acetate electrophoresis on Titan III membranes at pH 8.6 with Super-Heme buffer, by HPLC and by SDS-PAGE. Ratios of mixtures of Hb S or other Hb A and Hb F variants were confirmed using HPLC analysis.9,28 Electrospray ionization mass spectrometry (ES-MS) was performed on the purified recombinant Hb A and Hb F variants using a VG BioQ triple quadrapole mass spectrometer (Micromass, Altrincham, Cheshire, UK). Data analysis employed the MassLynx® software package (Micromass, Altrincham).9 Mass spectral analysis for the β73His-, β73 Leu β-, γ73 His- and γ73 Leu-globin chain variants and α-globin chains was 15,884.7, 15, 864, 16,016.6 and 15,993.6 and 15,125.7 Da, respectively. Hemoglobin concentrations were determined spectrophotometically using a millimolar extinction coefficient of mE 555 = 50 for deoxyhemoglobin and mE 579 = 53.6 for carbonmonoxyhemoglobin (on a tetramer basis). Purified hemoglobins were stored in the CO-liganded form at –80 °C until use. Hemoglobin solutions were concentrated using a Centricon centrifugal concentrator with a membrane cutoff of 30,000 Da (Centriprep, Milipore Co, Bedford, MA). Oxyhemoglobin was prepared by first blowing oxygen across the surface of the CO-Hb solution

537 in a rotary evaporator under a 150 W flood-light bulb in an ice bath for about 1 h. Fiber formation was evaluated by differential interference contrast (DIC) microscopy of Hb A/S or Hb F/S variant mixtures in addition to Hb A/S and Hb F/S mixtures with different ratios, and was initiated on glass slides by first adding 100 mM sodium dithionite to oxyHb S solutions at 4 °C in 1.0 M phosphate buffers (pH 7.0).9 Approximately 1 μl of solution was pipetted to a glass slide which then was sealed with a 18 mm square cover-slip using Mount-Quick solution (Daido Sangyo Co., Ltd, Japan).9,10 Polymer or fiber formation of deoxygenated hemoglobin on the sealed slide glass was induced by the temperature-jump method from 0 °C on ice to room temperature (~22 °C) using a 35 °C water bath for raising the temperature quickly at the initial stage. The total mass of polymer in a domain was analyzed by DIC microscopy by measuring pixel area of a single domain using an Olympus microscope equipped with DIC optics and a 40× oil (1.00 NA) immersion lens. The microscopic images obtained were transferred to a PC via an image grabber board (Universal Image Corporation, Downingtown, PA) and a CCD (chargecoupled devise) camera (Cohu Camera, Cohu Inc, San Diego, CA).10 In order to measure the area of polymer fibers, images were taken of a hemocytometer with 50 μm divisions (American Optical Corp., Buffalo, NY) at 400× magnification. Using the line measurement tool in Molecular Device image analysis system (Universal Image Corporation, Downingtown, PA), we determined the area of polymer domain with fibers by counting pixel numbers. Kinetics of polymerization and solubility of Hb S, the two β73 Hb S variants and mixtures of Hb S with two Hb A and Hb F variants including Hb A/S and Hb F/S mixtures were evaluated in 1.8 M and 1.0 M phosphate buffers (pH 7.3) at 30 °C. Solubility was determined by centrifugation after completion of polymerization.29 Solubility was also measured by a novel method (A. A. et al., unpublished results) in which hemoglobin is mixed with oil to form droplets. Photolysis of part of the droplet allows monomers to move from the unphotolyzed to the photolyzed region as polymerization continues, until solubility is reached, at which time the unphotolyzed region does not decrease its concentration further. Absorption spectra of the unphotolyzed region yield concentration, and thus solubility. Homogeneous and heterogeneous nucleation rates were calculated from kinetics of polymerization of mixtures of Hb S containing of Hb A or Hb F Leu variants at the 73 position and were compared to Hb A/S and Hb F/S mixtures containing similar fractions of Hb A or Hb F with Hb S. The kinetics of polymerization was measured on hemoglobins which were saturated with humidified CO in 1.0 M phosphate buffer (pH 7.3) by laser photolysis methods.10,17 Homogeneous nucleation rate (fo) and exponential growth (B) of Hb S polymerization, which is dominated by heterogeneous nucleation rates, were analyzed at 25 °C using laser photolysis with parallel detection.30

Acknowledgements We are grateful to Drs E. Daikhin and M. Yudkoff for mass analysis of β and γ-globin chain variants

Anti-Hb S Polymerization of Hb A β73Leu

538 performed at the Children's Hospital of Philadelphia Mass Spectrometry Research Core Facilities. This research was supported in part by grants from the National Institutes of Health (HL58879, HL 69256, HL38632, HL57549 and HL58512), and by the Cardeza Foundation for Hematological Research at Jefferson Medical College.

References 1. Eaton, W. A. & Hofrichter, J. (1990). Sickle cell hemoglobin polymerization. Adv. Protein Chem. 40, 63–279. 2. Bunn, H. F. & Forget, B. G. (1996). Hemoglobin: Molecular and Clinical Aspects, Sanders, Philadelphia. 3. Dykes, G. W., Crepeau, R. H. & Edelstein, S. J. (1979). Three-dimensional reconstruction of the 14-filament fibers of hemoglobin S. J. Mol. Biol. 130, 451–472. 4. Padlan, E. A. & Love, W. E. (1985). Refined crystal structure of deoxyhemoglobin S. I. Restrained leastsquares refinement at 3.0-Å resolution. J. Biol. Chem. 260, 8272–8279. 5. Harrington, D. J., Adachi, K. & Royer, W. E., Jr (1997). The high resolution crystal structure of deoxyhemoglobin S. J. Mol. Biol. 272, 398–407. 6. Embury, S. H., Hebbel, R. P., Mohandus, N. & Steinberg, M. H. (1994). SICKLE CELL DISEASE: Basic principles and clinical practice, Raven Press, New York. 7. Roufberg, A. & Ferrone, F. A. (2000). A model for the sickle hemoglobin fiber using both mutation sites. Protein Sci. 9, 1031–1034. 8. Mirchev, R. & Ferrone, F. A. (1997). The structural link between polymerization and sickle cell disease. J. Mol. Biol. 265, 475–479. 9. Adachi, K., Ding, M., Wehrli, S., Reddy, K. S., Surrey, S. & Horiuchi, K. (2003). Effects of different beta73 amino acids on formation of 14-stranded fibers of Hb S versus double-stranded crystals of Hb C-Harlem. Biochemistry, 42, 4476–4484. 10. Ivanova, M., Jasuja, R., Krasnosselskaia, L., Josephs, R., Wang, Z., Ding, M. et al. (2001). Flexibility and nucleation in sickle hemoglobin. J. Mol. Biol. 314, 851–861. 11. Goldberg, M. A., Husson, M. A. & Bunn, H. F. (1977). Participation of hemoglobins A and F in polymerization of sickle hemoglobin. J. Biol. Chem. 252, 3414–3421. 12. Adachi, K., Segal, R. & Asakura, T. (1980). Nucleationcontrolled aggregation of deoxyhemoglobin S. Participation of hemoglobin F in the aggregation of deoxyhemoglobin S in concentrated phosphate buffer. J. Biol. Chem. 255, 7595–7603. 13. Eaton, W. A. & Hofrichter, J. (1995). The biophysics of sickle cell hydroxyurea therapy. Science, 268, 1142–1143. 14. Benesch, R. E., Edalji, R., Benesch, R. & Kwong, S. (1980). Solubilization of hemoglobin S by other hemoglobins. Proc. Natl Acad. Sci. USA, 77, 5130–5134.

15. Adachi, K., Konitzer, P. & Surrey, S. (1994). Role of gamma 87 Gln in the inhibition of hemoglobin S polymerization by hemoglobin F. J. Biol. Chem. 269, 9562–9567. 16. Mirchev, R. & Ferrone, F. A. (1997). The structural origin of heterogeneous nucleation and polymer cross linking in sickle hemoglobin. J. Mol. Biol. 265, 475–479. 17. Ferrone, F. A., Hofrichter, J. & Eaton, W. A. (1985). Kinetics of sickle hemoglobin polymerization. II. A double nucleation mechanism. J. Mol. Biol. 183, 611–631. 18. Dou, Q. & Ferrone, F. A. (1993). Simulated formation of polymer domains in sickle hemoglobin. Biophys. J. 65, 2068–2077. 19. Agarwal, G., Wang, J. C., Kwong, S., Cohen, S. M., Ferrone, F. A., Josephs, R. & Briehl, R. W. (2002). Sickle hemoglobin fibers: mechanisms of depolymerization. J. Mol. Biol. 322, 395–412. 20. McCune, S. L., Reilly, M. P., Chomo, M. J., Asakura, T. & Townes, T. M. (1994). Recombinant human hemoglobins designed for gene therapy of sickle cell disease. Proc. Natl Acad. Sci. USA, 91, 9852–9856. 21. Pawliuk, R., Westerman, K. A., Fabry, M. E., Payen, E., Tighe, R., Bouhassira, E. E. et al. (2001). Correction of sickle cell disease in transgenic mouse models by gene therapy. Science, 294, 2368–2371. 22. Behringer, R. R., Ryan, T. M., Palmiter, R. D., Brinster, R. L. & Townes, T. M. (1990). Human gamma- to betaglobin gene switching in transgenic mice. Genes Dev. 4, 380–389. 23. Dillon, N. & Grosveld, F. (1991). Human gammaglobin genes silenced independently of other genes in the beta-globin locus. Nature, 350, 252–254. 24. Levasseur, D. N., Ryan, T. M., Pawlik, K. M. & Townes, T. M. (2003). Correction of a mouse model of sickle cell disease: lentiviral/antisickling beta-globin gene transduction of unmobilized, purified hematopoietic stem cells. Blood, 102, 4312–4319. 25. Levasseur, D. N., Ryan, T. M., Reilly, M. P., McCune, S. L., Asakura, T. & Townes, T. M. (2004). A recombinant human hemoglobin with anti-sickling properties greater than fetal hemoglobin. J. Biol. Chem. 279, 27518–27524. 26. Shen, T. J., Ho, N. T., Simplaceanu, V., Zou, M., Green, B. N., Tam, M. F. & Ho, C. (1993). Production of unmodified human adult hemoglobin in Escherichia coli. Proc. Natl Acad. Sci. USA, 90, 8108–8112. 27. Adachi, K., Konitzer, P., Kim, J., Welch, N. & Surrey, S. (1993). Effects of beta 6 aromatic amino acids on polymerization and solubility of recombinant hemoglobins made in yeast. J. Biol. Chem. 268, 21650–21656. 28. Haewon, C. K., Adachi, K. & Schwartz, E. (1995). Separation of hemoglobin. In Williams Hematology (Beutler, E., Lichtman, M.A., Coller, B.S. & Kippes, T.J., eds), 5th edit., pp. L35–L42 , McGraw-Hill, New York. 29. Adachi, K. & Asakura, T. (1979). Nucleation-controlled aggregation of deoxyhemoglobin S. Possible difference in the size of nuclei in different phosphate concentrations. J. Biol. Chem. 254, 7765–7771. 30. Cao, Z. & Ferrone, F. A. (1997). Homogeneous nucleation in sickle hemoglobin: stochastic measurements with a parallel method. Biophys. J. 72, 343–352.

Edited by M. Moody (Received 10 May 2006; received in revised form 17 July 2006; accepted 19 July 2006) Available online 28 July 2006