Proc. Natl. Acad. Sci. USAVol. 90, pp. 5039-5043, June 1993Biochemistry

Sparing effect of hemoglobin F and hemoglobin A2 on thepolymerization of hemoglobin S at physiologic ligand saturations

(sickle cell anemia/polymer fraction/hydroxyurea/fetal hemoglobin)

W. N. POILLON*t, B. C. KIM*, G. P. RODGERS*, C. T. NOGUCHIt, AND A. N. SCHECHTER**Center for Sickle Cell Disease and Department of Pediatrics and Child Health, Howard University, Washington, DC 20059; and tLaboratory of ChemicalBiology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892

Communicated by M. F. Perutz, February 22, 1993

ABSTRACT Recent interest in therapies for sickle cellanemia based on elevating fetal Hb has made accurate estimatesof the sparing effect of fetal Hb (Hb F) and other non-sickle Hbson sickle Hb (Hb S) polymerization essential. We have devel-oped a technique, using HbCO as surrogate for HbO2, thatenables us to assess the solubility ofHb S as a function of ligandsaturation under conditions that mimic those of the sicklingdisorders. Equimolar mixtures of unliganded Hb S with Hb For normal Hb A2 were isosoluble. Solubilities for equimolarmixtures with normal (Hb A) or abnormal (Hb C) Hbs werealso identical but were lower than in the prior case. Thus, thesparing effect of both Hb F and Hb A2 should be considered intherapeutic strategies designed to modify Hb S polymerization.Hemolysates, stripped of 2,3-bisphosphoglycerate, from sicklecell disease patients with Hb (F + A2) levels varying from 6 to25%, as well as from a sickle trait individual, were used toevaluate equilibrium solubiity as a function of ligand satura-tion over the range of pathophysiologic interest (25-70%). Ourresults show that the sparing effect of Hb (F + A2) increasesrelative to that of Hb A as ligand saturation increases, and thatin the absence of ligand, '30% Hb (F + A2) is essentiallyisosoluble with the 60% Hb A of sickle trait. Although detailedknowledge of expected therapeutic benefits is confounded bythe heterogeneity of Hb F distribution and other variables,these data should provide a framework for estimating likelyclinical benefit from pharmacologic efforts to modulate globingene expression.

Sickle cell anemia exists in individuals who are homozygousfor a point mutation (GAG -- GTG) at codon 6 of the 3-globingene, which results in a Glu-6 -* Val substitution in P-globin.This transposition of a hydrophobic for a polar residue on thesurface of sickle hemoglobin (Hb S; a2A32) results in aprofound reduction of its solubility in the deoxy (T) confor-mation. Upon partial deoxygenation of Hb S-containingerythrocytes, such as occurs in the microcirculation, poly-mers of partially liganded Hb S are produced that perturb itsrheological properties and lead to the characteristic vasooc-clusive episodes of this disease. The clinical severity of thesickling disorders is modulated, however, by variations in theHb composition of sickle cells that occur among individualpatients. Therapies based on elevating fetal hemoglobin (HbF; a2y2) have become a major focus of laboratory and clinicalstudies in the last few years. Two recent review articles (1,2) provide an overview of the pathophysiology of sicklingdisorders and the intracellular polymerization process thatunderlies it.Normal [Hb A (a2I3), Hb A2 (a2&2), and Hb F] and

abnormal [e.g., Hb C(a2f3)] hemoglobins that may be pre-sent along with Hb S in human erythrocytes act to elevate itssolubility under conditions of partial or total desaturation and

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

are, thus, said to have a "sparing" effect on intracellular HbS polymerization (3, 4). For Hbs F and A2, this solubilizationcan be described well with a model that assumes completeexclusion of either the homotetramer (a2-y2 or a2&2) or thehybrid heterotetramer (a238y or a2138) from the nascent poly-mer (5-7). For Hbs A and C, the data are compatible withexclusion of only homotetramers (a2I3A and a2K) from thepolymer. Hybrid heterotetrameric forms (a2f3Sf3A anda2138S,c) appear to be incorporated into the polymer, but witha copolymerization probability roughly one-half that of Hb S(8-10). These copolymerization tendencies explain why thesolubility enhancement induced by Hb F or A2 in admixturewith Hb S is greater than that for Hb A or C.The solubilizing properties ofvarious non-S Hbs have been

established by use of a chemical reducing agent (dithionite) tocompletely deoxygenate mixtures of oxy-Hb S and the oxyform of other hemoglobins (Hb X), followed by centrifuga-tion to fractionate the resulting semisolid gel into a tightlypacked polymer pellet and a supernatant consisting of solubleHb S and Hb X. The concentration of all Hb species in thesupernatant is by definition the saturation concentration(csat), or equilibrium solubility, under given in vitro condi-tions (11, 12). This method has been used to assess theequilibrium interaction of various mixtures of totally unli-ganded Hb S and Hb X (8-10) and to screen potentialantigelling agents that might be used to treat sickle celldisease (13-15). However, the 02 tension of the mixedvenous circulation, where vasoocclusive episodes occur, isestimated to range from 20 to 40 torr (1 torr = 133.3 Pa) (16).No equilibrium solubility studies have yet been performed onmixtures of Hb S with non-S Hbs at 02 saturation values(25-70%) corresponding to these physiologic 02 tensions,mainly because of the difficulty in preventing MetHb forma-tion during sedimentation experiments (17).

In contrast to HbO2, however, HbCO is not autoxidizableto MetHb, and its use as a surrogate for the R state wouldobviate this problem. The solubility of Hb S as a function ofligand saturation is virtually identical (up to 85% saturation)for 02 and CO (17, 18), indicating that the two ligands areinterchangeable with respect to stabilization of the oxy (R)conformation of the allosteric equilibrium. Furthermore,sedimentation of mixtures ofHb S partially liganded with COand unliganded Hb S results in a polymer phase from whichHbCO is nearly completely excluded (18-20). Accordingly,we have developed a method to measure csat for mixtures ofHb S and any non-S Hb at physiologically relevant ligandsaturations, in which CO rather than 02 is used to partiallyligate each Hb. Deoxygenation of the residual HbO2 withdithionite then leaves HbCO unchanged in a mixture of oxy(R) and deoxy (T) conformers that is stable indefinitely. By

Abbreviations: c,w, equilibrium solubility;fp, polymer mole fraction;Hb A, adult hemoglobin; Hb S, sickle hemoglobin; Hb F, fetalhemoglobin; MCHC, mean corpuscular Hb concentration.tTo whom reprint requests should be addressed.

5039

Proc. Natl. Acad. Sci. USA 90 (1993)

use of this technique it has been possible to obtain a familyof solubility profiles (Csat vs. fractional saturation with CO)for mixtures of Hb S with levels of Hb F varying from 3.5 to22%. For this purpose, hemolysates, stripped of 2,3-bis-phosphoglycerate, from sickle cell syndrome patients withvariable expression ofHb F, and relatively constant levels ofHb A2 (-3%), were used. Such information will be vital forobtaining reliable estimates of the therapeutic efficacy ex-pected from stimulation of Hb F synthesis by drugs such ashydroxyurea, erythropoietin, and butyrate analogues that arecurrently undergoing clinical trials as treatments for sicklecell disease (21-24).

MATERIALS AND METHODSGeneral Methods. Total Hb was measured spectrophoto-

metrically as cyanmethemoglobin (HiCN) by using the mil-limolar extinction coefficient of 11.0 cm-1 (per heme) at 540nm (25); MetHb was measured (26) from the absorbance ratioA540 (before dithionite)/A555 (after dithionite). Hbs F and A2in hemolysates were quantitated by alkali denaturation (27)and elution from DEAE-cellulose columns (28), respectively.

Purification of Hb S and Non-S Hbs. Unit-size sickle cellblood samples of SF and SC genotypes, as well as anexchange-transfused SS genotype, were obtained from theinventory of the cryopreservation laboratory of the HowardUniversity Comprehensive Sickle Cell Center (29). Informedconsent was obtained before these and other blood collec-tions. Hb S, as well as Hbs A, C, and F, was isolated as HbO2by chromatography of hemolysates on columns of DEAE-Sephacel (Pharmacia) at 4°C (30). Hb fractions were elutedwith a pH gradient of 0.05 M Tris buffer (pH 8.2-7.3 at 25°C);this procedure effectively removes endogenous organophos-phates. Each fraction was concentrated to 38-42 g/dl byvacuum ultrafiltration and dialyzed into 0.15M KPO4 (pH 6.8at 25°C). Hb A2 was obtained by pooling appropriate columneluates from the separation ofHb S from Hb F or A. PurifiedHb C contained <3-4% of Hb A2 as contaminant.

Hemolysates of Hb S Plus Hb X. Three of the four hemoly-sates with various levels ofHb F (Table 1) were also obtainedfrom the cryopreservation laboratory. Packed erythrocytesfrom an apheresis unit were thawed, and glycerol was re-moved. Hemolysates were prepared with toluene by themethod of Drabkin (31). For the hemolysate with the highestlevel of Hb F, whole blood from a patient undergoinghydroxyurea therapy at the National Institutes of Health (21)was used. Sickle trait blood (AS) was used to obtain ahemolysate suitable for evaluating the sparing effect ofHb A.Each hemolysate was dialyzed into isosmotic buffer contain-

Table 1. Blood parameters of sickle cell anemia patients withvarious levels of Hb (F + A2)

Patient

A B1* B2* C Dt E

Genotype SS SS SF S/83P0thal SS AS*Hct, % 24.0 27.9 25.0 30.0 21.2 39.9Hb, g/dl 8.0 8.0 8.8 10.5 7.1 13.2MCHC, g/dl 33.3 28.8 35.1 35.0 33.4 33.1Hb F, % 2.1 2.6 9.9 16.1 22.4 0.5Hb A2, % 4.2 3.5 2.6 2.2 2.6 2.6Hb (F + A2), % 6.3 6.1 12.5 18.3 25.0 3.1

thal, Thalassemia; Hct, hematocrit.*The same patient before (B2) and after (Bi) chronic phlebotomy toinduce iron-limited erythropoiesis.

tElevation of Hb F in this patient was induced by treatment withhydroxyurea at the National Institutes of Health.tHbs quantitated by HPLC were as follows: A, 60.7%; S, 35.8%; andA2, 3.5%.

ing 0.05 M Bistris, 0.1 M KCl, and 0.02 M NaCl (pH 8.0 at250C).

Solubility Measurements. Details of the modification of ouroriginal csat assay method (11) to a microscale have beenpresented elsewhere (13, 14, 32). A description of the pre-cautions necessary to preserve anaerobicity during the han-dling of dithionite, as well as the rationale for the use ofBistris-buffered saline as a near-physiologic medium for thecsat assay, has recently appeared (30). Two sets of csat assaymeasurements were made, one for unliganded mixtures ofHbS plus Hb X in phosphate buffer, pH 6.5 at 30°C, and the otherfor Hb S as a function of ligand saturation at various levelsof Hb (F + A2) in Bistris {2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)-1,3-propanediol} buffer, pH 7.2, at 37°C. Inthe latter case, a portion of the dialyzed HbO2 was convertedto HbCO by equilibration with CO for 30 min in an Instru-mentation Laboratory model 237 tonometer at 250C. Com-plete saturation was verified spectrophotometrically by useofthe ratio A555/A540 after addition ofdithionite (33). SamplesofHbO2 plus HbCO were mixed in various proportions in airto allow for hybridization before deoxygenation. Saturationwith CO was estimated volumetrically from the proportion ofHbCO in the original mixture, after correction for variableMetHb content ('10%). In each case, the total Hb concen-tration, ct, was chosen so that the supersaturation ratio(Ct/Csat) was in the range 1.18 ± 0.05. After equilibration for1.5 hr at 37°C, mixtures were placed in an SW 50.1 rotoradapted to 0.8-ml capacity, equilibrated for an additional 0.5hr in a Beckman L8-55 preparative ultracentrifuge, andcentrifuged for 1 hr at 242,000 x g. After sedimentation, thepH of the supematant was measured at ambient temperaturewith an Ingold combination microelectrode and a RadiometerpH meter (model PHM85). Supernatant pH was corrected to37°C with the factor -0.012 unit per 'C. Measured values ofcsat at 37°C pertain to the pH range 7.10-7.25. [Because oftheprecipitous drop in pH caused by dithionite addition to thepartially liganded mixtures in Bistris buffer (0.7-0.8 unit), thepH of the dialysis buffer was deliberately made higher than7.2 (i.e., pH 8.0) so as to obtain a final pH near 7.2 at 37°C.]These values were normalized to pH 7.20 with the slope ofthecsat vs. pH profile for pH 6.8-7.6 in Bistris buffer (30): 10.8g/dl per pH unit. The total concentration of all soluble Hbspecies (S and non-S, deoxy or carboxy) in the supematantwas measured after conversion to HiCN with Drabkin'ssolution, allowing 1 hr for conversion of HbCO to HiCN.

RESULTSPurified samples of Hbs S, A, C, F, and A2 were used toassess the sparing effect of non-S Hbs on the solubility offully unliganded Hb S in 0.15 M KPO4 buffer, pH 6.5 at 30°C.This buffer was chosen because previous studies addressedto this issue have used it (3, 4, 8, 9) or a variation (6, 7, 10),and any additional data must be compared with those anal-yses. The final pH after addition of dithionite (40-62 mM)falls within the region of csat invariance (6.2-6.8) demon-strated in an earlier study (30). Solubility profiles for eachnon-S Hb species are shown in Fig. 1. Because the data werenearly linear in each case, regression analysis was used toobtain a measure of relative effectiveness (i.e., slope of Csatvs. mol fraction ofHb X for each species). The slopes for HbsF and A2 are nearly identical (21.8 and 20.4 g/dl, respective-ly), as are the slopes for Hbs A and C (13.2 and 14.4 g/dl,respectively). Hb F and Hb A2 are thus equipotent in theirsparing effect on the solubility ofunligandedHb S. To a lesserextent, Hb A and Hb C are also equipotent in their ability toinhibit polymerization.Whole blood hematologic parameters for the four sickle

cell syndrome patients and one sickle trait individual used inthis study are summarized in Table 1. The sum ofHb (F + A2)

5040 Biochemistry: Poillon et al.

Proc. Natl. Acad. Sci. USA 90 (1993) 5041

35

30

D 25

dS 20

15

100 0.2 0.4

Fraction Hb X0.6 0.8

FIG. 1. Solubility profiles for mixtures of unliganded Hb S withvarious non-S Hbs: csat was measured at 30°C in 0.15 M KPO4 buffer,pH 6.5; ,u = 0.38. Total Hb concentrations (ct) were 29.9 g/dl formixtures of Hb S with Hb F or Hb A2 and 26.5 g/dl for those withHb A or Hb C.

varied from 6 to 25% in hemolysates from the four patients.Non-S Hb levels for the AS phenotype were 60.7% (A), 2.6%(A2), and 0.5% (F). Mean corpuscular Hb concentration(MCHC) values were in the normal range (33.1-35.0 g/dl),except for patient Bi, whose MCHC was 28.8 g/dl due to themicrocytosis of iron deficiency (34).Because the phosphate buffer used to obtain the data of

Fig. 1 was nonphysiologic with respect to pH, ionic strength,and temperature, the more physiologic Bistris-based bufferused earlier (30) was chosen for evaluating solubility as afunction of CO saturation at various levels of Hb (F + A2).The plots of measured csat vs. XCO for the six hemolysates ofTable 1 are shown in Fig. 2a. These solubility profiles areapproximately linear in each case (range of correlation co-efficients, r, from regression analyses: 0.974-0.992), witheach curve displaced upward by the difference in solubility(Acsat) at zero CO saturation due to the increment in Hb(F + A2.) Data for the intercepts (c°at) and slopes of theseplots are presented in Table 2 (columns 3 and 4). Because thesolubility experiments reported in Fig. 2a were done at

40

35

- 30

25

20

150 0.2 0.4 0.6 0.8

Fractior

variable total Hb concentrations (ct), chosen such that Ct/Cut1.2, the effect of ct variation on the measured solubilities

must be taken into account; this was done by normalizingeach csat value to a constant ct = 34 g/dl with the two-phasethermodynamic model for the polymerization of partiallyliganded mixture of Hb S with non-S Hbs (35, 36). Plots ofthese normalized csat values vs. CO saturation for each levelof Hb (F + A2) (Fig. 2b) were also subjected to linearregression analysis (r range: 0.942-0.992). Data for the in-tercepts (c..) and slopes of the plots of normalized csat vs.Xco are presented in Table 2 (columns 5 and 6).Comparison of the slopes for the original (Fig. 2a) and

normalized (Fig. 2b) Csat data in Table 2 indicates that as thefraction Hb (F + A2) increases from 0.06 to 0.25 (Table 1,patients A-D), the slopes decrease, but their magnitudes aregreater for the original csat data (cf. the range of 17.9-13.8g/dl vs. 12.6-9.6 g/dl for the slope parameter as X(F+A2)increases from 0.06 to 0.25). This presumably reflects thedependence of the measured csat values on ct and the partic-ular value of ct chosen in each case (Ct/Csat 1.2 vs. ct = 34g/dl). By use of the normalized c?,t values for unligandedmixtures of Hb S + Hb X (i.e., the intercept of each plot atXco = 0), polymer mole fraction (fp) at zero saturation (fMwas calculated from the conservation of mass relationship(37) for each hemolysate (see Table 2 footnote t). Thesevalues off° are collected in Table 2 (column 7) and show areciprocal relationship with fractional Hb X, varying from0.53 to 0.30 as X(F+A2) increases from 0.06 to 0.25.

Values of c?.t (normalized to ct = 34 g/dl) for each level ofHb (F + A2) shown in Table 2 (column 5) were subjected tolinear regression analysis to give intercept (20.2 g/dl) andslope parameters (33.4 g/dl per mol fraction) that pertain tothe solubility of unliganded Hb S in the absence of non-S Hband its dependence on X(F+A2); the value off° correspondingto this c?.t is 0.57. The dependence ofthese two variables (cotandfP) on X(F+A2) shown in Table 2 indicates that the fractionHb (F + A2) that would raise c° to 29.0 g/dl, the solubilityof unliganded Hb (A + S) in the sickle trait hemolysate, is=0.30. Thus, Hb (F + A2) is approximately twice as effectiveas Hb A in its sparing effect on the solubility of unligandedHb S.

Plots of fp vs. fractional CO saturation for each Hb (F +A2) level are presented in Fig. 3. Because the data arecurvilinear in each case, the intercept on the x axis is

b

6.1--6.3

-A - 12.5--8-- 18.3

0 0.- 25-AS

0 0.2 0.4 0.6 0.8n HbCO

FIG. 2. csat as a function of fractional CO saturation for hemolysates from four patients with various levels of Hb (F + A2) plus a sickle traitindividual (see Table 1). (a) csat was measured at 37°C in 0.05M Bistris/0.1 M KCI/0.02M NaCl, pH 7.2; ,u = 0.15 and osmolality = 290 mOsm/kgof H20 as a function of Xco. For the 70 data points, total Hb concentration, ct, was varied such that the supersaturation ratio (c/cjt) had amean value of 1.18 ± 0.05. (b) Measured Csat values were normalized to ct = 34 g/dl with the two-phase model for polymerization of partiallyliganded mixtures of Hb S with non-S Hbs (35, 36) and, thus, pertain to the mean MCHC of unfractionated sickle erythrocytes.

Biochemistry: Poillon et al.

Proc. Natl. Acad. Sci. USA 90 (1993)

Table 2. Solubilities and polymer fraction in the absence ofligand and dependence of solubility on ligation state forvarious levels of Hb (F + A2)

Csat,* Slope,* cO t Slope,tPatient X(F+A2) g/dl g/dl g/dl g/dl f OtA 0.063 20.7 17.9 22.3 12.6 0.51B1 0.061 19.8 19.8 21.6 16.0 0.53B2 0.125 22.5 19.6 24.7 11.7 0.43C 0.183 26.1 15.3 27.2 8.5 0.33D 0.25 27.4 13.8 27.8 9.6 0.30E (AS)§ 0.031 28.9 11.9 29.0 7.6 0.25

*Intercept (c° ) and slope parameters for each measured csat vs. Xcoplot shown in Fig. 2a. Values of ct, the total Hb concentration, werechosen such that the supersaturation ratio (Ct/Csat) had a mean valueof 1.18 ± 0.05.tLinear solubility profiles of Fig. 2b, based on c,,t values normalizedto ct = 34 g/dl (see legend to Fig. 2), were used to obtain theseintercept (coat) and slope parameters.

tfp at zero saturation was calculated from the conservation of massequation (37), with values of 34 and 69.3 g/dl for ct and cp,respectively, and the appropriate normalized values of cst.§A sickle trait hemolysate; see Table 1, footnote t.

uncertain. Hence, no attempt has been made to assign a COsaturation at which fp = 0 for each Hb fraction (F + A2). Ingeneral, the absence of polymer occurs at lower ligandsaturations for the higher fractional Hb (F + A2) values.

DISCUSSIONThe solubility profiles of Fig. 1 show that the sparing effectof either Hb F or Hb A2 on the polymerization of unligandedHb S is substantially larger than that of either Hb A or Hb C.Our findings with Hbs A and C confirm those of an earlier

study (9) in which no significant difference was found in thetendency of Hb A or C to copolymerize with Hb S in thephosphate buffer (,u = 0.38) used here. The solubility of anequimolar mixture of Hb S plus Hb C measured in low-ionic-strength phosphate buffer (,u < 0.2) is, however, lower thanthat of a similar mixture of Hb S plus Hb A (39). Theisosolubility of deoxy-Hb F or deoxy-Hb A2 in a 50:50mixture with deoxy-Hb S confirms an earlier report to thiseffect (4) but is discrepant with others that report Hb A2 tobe either less (7) or more (10) effective than Hb F in enhancingthe solubility of unliganded Hb S. In any case, our results, inwhich buffer composition, pH, and temperature were care-

0.7

c

0

co..-

Cu

a)E

0L

0.6

0.5

0.4

0.3

0.2

0.1

0 L-o

Fraction HbCO

FIG. 3. Polymer fraction (fp) as a function of fractional COsaturation for the six hemolysates depicted in Fig. 2b: fp wascalculated from the conservation of mass equation (38) with valuesof 69.3 and 34 g/dl for polymer (cp) and total Hb (ct) concentration,respectively; values of csat (normalized to ct = 34 g/dl) for eachhemolysate were those used to obtain the intercept (cdat) and slopeparameters of Table 2 (columns 5 and 6).

fully matched for each of the four non-S Hbs, show unequiv-ocally that Hb F and Hb A2 are equipotent in their ability tosolubilize deoxy-Hb S. Thus, in making predictions for theinduction of Hb F synthesis by agents that stimulate eryth-ropoiesis (40, 41) the presence of a fixed amount of Hb A2(=3%) cannot be ignored.The experiments described in Figs. 2 and 3 were designed

to evaluate csat andfp for Hb S as a function ofCO saturationat various Hb levels (F + A2). By using hemolysates, ratherthan mixtures of Hb S with non-S Hbs, and the morephysiologic Bistris buffer (29) at 37°C, the measured solubil-ities more closely mimic those ofthe intact cell. These studiesassess the influence of non-S Hbs on the solubility of Hb Sas a function of ligand saturation under conditions thatapproximate those of the sickling disorders. As such, theyallow more precise predictions to be made of the therapeuticbenefit to be derived from the hydroxyurea-induced elevationof Hb F (21-23, 42).

It should also be noted that with one exception (X(F+A2) =0.063), the linear plots of the normalized csat data (Fig. 2b)tend to converge at higher fractional saturation values. Thisis particularly striking for the plots ofX(F+A2) = 0.25 and thatfor the AS hemolysate, which are displaced by 1.2 g/dl fromeach other at Xco = 0. Because the slope of the dependenceof csat on Xco is greater (9.6 g/dl) for the mixture with Hb(F + A2) than for the AS hemolysate (7.6 g/dl; Table 2), theybecome coincident at a CO saturation near 40%o (see Fig. 2b).The corresponding value offp at this degree ofligation is 0.11.The slope obtained (33.4 g/dl) for the plot of co vs. Xco

agrees well with that extracted from data based on modeling(32 g/dl; ref. 39, figure lc). The value of f° is also inreasonable agreement with that derived from modeling (0.57vs. 0.69), which used a lower value of c° (16 g/dl) than ours.Because polymerization at each 02 saturation in the pre-

capillary arterioles is believed important in the pathophysi-ology of this disease (1, 38), the beneficial effect of Hb Felevation may be somewhat greater than previously esti-mated. This effect, however, is offset by the nonuniformityofthe Hb F response to pharmacologic manipulation becausethe analysis is based on a uniform distribution of Hb F andMCHC, as well as of 2,3-bisphosphoglycerate content andintracellular pH.

Values of polymer fraction as a function of 02 saturationhave been reported for both intact sickle trait erythrocytes(based on NMR spectroscopy) and an AS hemolysate (basedon Csat measurements for gas-deoxygenated samples; refs. 36and 43). The two sets of measurements were in very goodagreement and gave mean values of 0.34 and 0.69 for f°?and Xco (at fp = 0), respectively. These values are compa-rable to those of 0.25 and 0.66 obtained in the present studyfor a sickle trait hemolysate in which csat was measured as afunction of CO saturation. This result attests to the corre-spondence of the solubilities of Hb S partially liganded witheither 02 or CO when measured under comparable condi-tions.Although no csat experiments were done with partially

liganded Hb S in the absence of Hb F or A2, two reportsaddressed to this issue have appeared: one in which ligationwas achieved with CO (19) and one with 02 (17). The one (19)most relevant to our study evaluated csat as a function ofCOsaturation in 0.15 M Bistris buffer at 25°C. The slopesobtained were comparable to ours for the two hemolysateswith the lowest Hb (F + A2) (Table 2) in Bistris-based bufferat 37°C. It appears, therefore, that the slope parameter(Acsat/ACO) for Hb S differs little from that at low levels ofnon-S Hb. A single study exists (17) in which solubility wasevaluated as a function of the physiologic ligand 02. Anapproximately linear solubility profile was observed up to 0.7fractional 02 saturation, with a slope of 14 g/dl per unit molefraction ligation. This slope is comparable to that obtained

5042 Biochemistry: Poillon et al.

Proc. Natl. Acad. Sci. USA 90 (1993) 5043

with CO in Bistris buffer at X(F+A2) = 0.06 (this study, Table2; ref. 19) and shows that the partially ligated R-state speciesproduced by either ligand (02 or CO) are equivalent withrespect to their polymerization tendency.

Sickle erythrocytes have a heterogeneous distribution ofboth MCHC and Hb F between "F" and "non-F" cells.Thus, low-Hb F cells would not achieve as great an inhibitionof polymerization as high-F cells. Conversely, cells of higherMCHC (which decrease during therapy) would require anelevation of Hb F greater than posited above to reduce fp tothat of deoxygenated AS erythrocytes (see Table 2). Al-though these calculations provide only first-order estimatesof the sparing effect of Hb F, they nonetheless represent asignificant advance over those previously available frommodeling (38, 41).An earlier study (30) showed that two other cellular vari-

ables-2,3-bisphosphoglycerate content and pH-also affectpolymerization in interdependent fashion. Because the pro-tons that stabilize the T-state and the polymer are oxylabile,deoxygenation alkalinizes the sickle erythrocyte, elevatingintracellular pH from 7.13 to 7.41 (30). Because SS erythro-cytes are only partialty deoxygenated as they traverse themicrocirculation, this pH increment is not fully realized.Nevertheless, the actual pH of cells in the capillary bedwould probably be >7.2. Values of c4 , therefore, would besomewhat higher than those given in Table 2 because csat isstrongly pH dependent in this region (30). In addition, thebinding of 2,3-bisphosphoglycerate in the ,3 cleft elicits adistortion of the T-state (44) that decreases csat by 1.6 g/dl(30). Because the c?O values in Table 2 pertain to strippeddeoxy-Hb S, the values for 2,3-bisphosphoglycerate-saturated Hb S would be lower by this amount. Althoughexact pH values cannot be assigned for partially deoxygen-ated SS erythrocytes, the expected solubility incrementwould be partially offset by the decrement due to 2,3-bisphosphoglycerate binding. Thus, the final solubilities ofmixtures of Hb S with Hb (F + A2) at the P02 of tissuecapillaries (20-40 torr) should be close to those depicted inTable 2.

We are grateful to Dr. Oswaldo Castro for providing the cryopre-served blood specimens from patients with variable levels of Hb Fthat were used in this study (Table 1) and to Dr. William Winter forthe quantitation of hemoglobins by HPLC for the sickle trait bloodspecimen.

1. Schechter, A. N., Noguchi, C. T. & Rodgers, G. P. (1987) inMolecular Basis ofBlood Diseases, eds. Stamatoyannopoulos,G., Nienhuis, A. W., Leder, P. & Majerus, P. (Saunders,Philadelphia), pp. 179-218.

2. Eaton, W. A. & Hofrichter, J. (1990) Adv. Protein Chem. 40,63-279.

3. Bookchin, R. M. & Nagel, R. L. (1971) J. Mol. Biol. 60,263-270.

4. Nagel, R. L., Bookchin, R. M., Johnson, J., Labie, D., Wajc-man, H., Isaac-Sodeye, W. A., Honig, G. R., Schiliro, G.,Crookston, J. H. & Matsumoto, K. (1979) Proc. Natl. Acad.Sci. USA 76, 670-672.

5. Bertles, J. F., Rabinowitz, R. & Dobler, J. (1970) Science 169,375-377.

6. Goldberg, M. A., Husson, M. A. & Bunn, H. F. (1977) J. Biol.Chem. 252, 3414-3421.

7. Benesch, R. E., Edalji, R., Benesch, R. & Wong, S. (1980)Proc. Natl. Acad. Sci. USA 77, 5130-5134.

8. Sunshine, H. R., Hofrichter, J. & Eaton, W. A. (1979) J. Mol.Biol. 133, 435-467.

9. Bunn, H. F., Noguchi, C. T., Hofrichter, J., Schechter, G. P.,

Schechter, A. N. & Eaton, W. A. (1982) Proc. Natl. Acad. Sci.USA 79, 7527-7531.

10. Cheetham, R. C., Huehns, E. R. & Rosemeyer, M. A. (1979)J. Mol. Biol. 129, 45-61.

11. Magdoff-Fairchild, B., Poillon, W. N., Li, T.-i. & Bertles, J. F.(1976) Proc. Natl. Acad. Sci. USA 73, 990-994.

12. Hofrichter, J., Ross, P. D. & Eaton, W. A, (1976) Proc. Natl.Acad. Sci. USA 73, 3034-3039.

13. Poillon, W. N. (1980) Biochemistry 19, 3194-3199.14. Poillon, W. N. & Kim, B. C. (1982) Biochemistry 21, 1400-

1406.15. Chang, H., Ewert, S. M., Bookchin, R. M. & Nagel, R. L.

(1983) Blood 61, 693-704.16. Klug, P. P., Lessin, L. S. & Radice, B. S. (1974) Arch. Int.

Med. 133, 577-590.17. Sunshine, H. R., Hofrichter, J. & Eaton, W. W. (1982) J. Mol.

Biol. 158, 251-273.18. Hofrichter, J. (1979) J. Mol. Biol. 128, 335-369.19. Chung, L. L. & Magdoff-Fairchild, B. (1978) Arch. Biochem.

Biophys. 189, 535-539.20. Franklin, I. M., Rosemeyer, M. A. & Huehns, E. R. (1983) Br.

J. Haematol. 54, 579-587.21. Rodgers, G. P., Dover, G. J., Noguchi, C. T., Schechter,

A. N. & Nienhuis, A. W. (1990) N. Engl. J. Med. 322, 1037-1045.

22. Charache, S., Dover, G. J., Moore, R. D., Eckert, S., Ballas,S. K., Kosky, M., Milner, P. A., Orringer, E. P., Phillips, G.,Platt, 0. S. & Thomas, G. H. (1992) Blood 79, 2555-2565.

23. Rodgers, G. P., Dover, G. J., Uyesaka, N., Noguchi, C. T.,Schechter, A. N. & Nienhuis, A. W. (1993) N. Engl. J. Med.328, 73-80.

24. Perrine, S. P., Ginder, G. D., Faller, D. V., Dover, G. J.,Ikutu, T., Witowska, H. E., Cai, S.-P., Vichinsky, E. &Oliveri, N. F. (1993) N. Engl. J. Med. 328, 81-86.

25. van Assendelft, 0. W. (1970) Spectrophotometry ofHemoglo-bin Derivatives (Thomas, Springfield, IL).

26. Salvati, A. M. & Tentori, L. (1981) Methods Enzymol. 76,715-731.

27. Betke, K., Marti, H. R. & Schlecht, I. (1959) Nature (London)184, 1877-1888.

28. Huisman, T. H. J., Schroeder, W. A., Brodie, A. N., Mayson,S. M. & Jakway, J. (1975) J. Lab. Clin. Med. 86, 700-702.

29. Castro, O., Hardy, K. P., Winter, W. P., Hornblower, M. &Meryman, H. T. (1981) Am. J. Hematol. 10, 297-304.

30. Poillon, W. N. & Kim, B. K. (1990) Blood 76, 1028-1036.31. Drabkin, D. L. (1946) J. Biol. Chem. 164, 703-723.32. Poillon, W. N. & Bertles, J. F. (1979) J. Biol. Chem. 254,

3462-3467.33. Faulkner, W. P. & Mertes, S., eds. (1982) SelectedMethodsfor

the Small Clinical Laboratory (Am. Assoc. Clin. Chem., Wash-ington, DC), pp. 139-142.

34. Poillon, W. N., Kim, B. C. & Castro, 0. (1991) Blood 78,Suppl. 1, 200a (abstr.).

35. Noguchi, C. T., Torchia, D. A. & Schechter, A. N. (1983) J.Clin. Invest. 72, 846-852.

36. Noguchi, C. T. (1984) Biophys. J. 45, 1153-1158.37. Ross, P. D., Hofrichter, J. & Eaton, W. A. (1977) J. Mol. Biol.

115, 111-134.38. Noguchi, C. T. & Schechter, A. N. (1981) Blood 58,1057-1068.39. Bookchin, R. M. & Balazs, T. (1986) Blood 67, 887-892.40. Noguchi, C. T., Rodgers, G. P., Serjeant, G. R. & Schechter,

A. N. (1988) N. Engl. J. Med. 318, 96-99.41. Noguchi, C. T., Rodgers, G. P. & Schechter, A. N. (1989)

Ann. N.Y. Acad. Sci. 565, 75-82.42. Goldberg, M. A., Brugnara, C., Dover, G. J., Shapira, L.,

Charache, S. & Bunn, H. F. (1990) N. Engl. J. Med. 323,366-372.

43. Noguchi, C. T., Torchia, D. A. & Schechter, A. N. (1981) J.Biol. Chem. 256, 4168-4171.

44. Perutz, M. F. (1987) in MolecularBasis ofBloodDiseases, eds.Stamatoyannopoulos, G., Nienhuis, A. W., Leder, P. & Ma-jerus, P. (Saunders, Philadelphia), pp. 127-178.

Biochemistry: Poillon et al.