Subunit Dissociation of Certain

Abnormal Human Hemoglobins

H. FRANKLIN BUNN

From the Blood Transfusion Division, U. S. Army Medical Research Laboratory,Fort Knox, Kentucky 40121

A B S T R A C T The extent of dissociation of various he-moglobins into subunits was estimated from their elu-tion volumes (V.) on G-100 Sephadex. Under the samecontrolled conditions carboxyhemoglobins A, A3 (A,),F, S, and C all had the same elution volumes. Thecarboxy and cyanmet derivatives of hemoglobin Kansas(a variant with very low oxygen affinity) had a rela-tively high Ve, indicating a decreased mean molecularweight and therefore an increased tendency to formdimers and even monomers. Conversely, the ligandedderivatives of hemoglobin Chesapeake (a variant withhigh oxygen affinity) had a relatively low V., suggestiveof an impaired degree of subunit dissociation. Deoxyhe-moglobin Chesapeake had a V. identical with that ofdeoxyhemoglobin A. Cat hemoglobin, known to have anunusually low oxygen affinity, was found to have ahigher V. than human, dog, rabbit, rat, or guinea pighemoglobins.

Haptoglobin is thought to bind af dimers in preferenceto the a2,82-tetramer. The comparative haptoglobin affini-ties of the human hemoglobins were measured by com-petition between the test hemoglobin and radioactivereference hemoglobin for haptoglobin binding sites.Hemoglobins A, F, S, and C all seemed to bind equallyreadily, but hemoglobin Kansas and cat hemoglobinshowed a higher affinity, and hemoglobin Chesapeake alower affinity.

These results are in accord with recently proposedmodels which predict that hemoglobins which have anincreased degree of subunit dissociation will have a lowoxygen affinity, and vice versa.

INTRODUCTION

Most human hemoglobin variants generally have asingle amino acid substitution in either the a- or pl-poly-peptide chains. Such a structural alteration will com-

Received for publication 25 June 1968 and in revised form20 September 1968.

monly confer a difference in net surface charge allowingthe abnormal protein to be detected and isolated by elec-trophoretic and chromatographic methods. Many ofthese hemoglobins have no unusual chemical or physi-cal properties and appear to be functionally normal.A smaller number of human hemoglobin variants areassociated with clinical features which lead to theirdetection. These hemoglobins almost always have un-usual properties which are readily demonstrable in vitro.

A few human hemoglobin variants are characterizedby marked abnormalities in oxygen affinity (Table I).A mother and son with cyanosis and a normal hematocritwere found by Reissman, Ruth, and Nomura to beheterozygous for a hemoglobin variant, designated asKansas (1). The cyanosis was due to the low oxygenaffinity of the whole blood. Recently Bonaventura andRiggs have determined the structure of isolated hemo-globin Kansas and have confirmed that this hemoglobinhas a markedly reduced oxygen affinity and low heme-heme interaction, but a normal Bohr effect (2). Fivehemoglobin variants: Chesapeake (3), Yakima (4, 5),Kempsey (6), and Rainier (7) and Hiroshima (8)have now been found to be associated with familialerythrocytosis (Table I). In each instance the affectedindividuals have been heterozygous for a hemoglobinwith a markedly high oxygen affinity. In addition, he-molysates containing hemoglobin J-Cape Town havebeen found to have high oxygen affinity (9). Howeveraffected heterozygotes do not have erythrocytosis.

The way in which hemoglobin reversibly binds oxygenhas eluded precise understanding. Most puzzling hasbeen the mechanism by which the oxygen affinity in-creases upon the stepwise addition of ligand: so calledheme-heme interaction. The conformation of hemoglobinis known to be altered considerably upon oxygenation(10). Furthermore, ligand binding enhances the ten-dency of the tetramer to dissociate symmetrically intodimers: aC/t3 z 2a/3 (11, 12). These experimental ob-servations have formed the basis of a model proposed

126 The Journal of Clinical Investigation Volume 48 1969

TABLE IHemoglobin Variants Associated with

Abnormal Oxygen Affinity

Heme-heme

Clinical features Oxygen inter-Hemoglobin variant in heterozygote affinity action

Kansas a202l12 Thr Cyanosis Low LowNormal hct

Chesapeake 2I92 LeuB2Yakima a213299 HisKempsey a212399 Asn Erythro- High LowRainier a2132145 His cytosisHiroshima a2132143 Aspj

J-CapeTown a292 Gln132 Normal hct High ?

Thr, threonine; Leu, leucine; His, histidine; Asn, asparagine;Asp, asparatic acid; Gln, glutamine.

by Benesch, Benesch, and MacDuff for the reaction ofligands with hemoglobin which postulates that the a,8dimer is the primary functioning unit (13). Stepwiseoxygenation is thought to involve rapid exchange be-tween oxygenated and deoxygenated dimers. It is ofgreat interest to examine those hemoglobins with ab-normal ligand affinity in the light of this model. It pre-dicts that if a hemoglobin variant dissociates morereadily than normal from a liganded tetramer to dimerit will have a low oxygen affinity, and vice versa. It isnoteworthy that the amino acid substitution in all of thehemoglobins shown in Table I except Rainier and Hiro-shima is in the nonhelical FG segment, or adjacent Ghelix, regions which are in the interface between ad-joining aft-dimers (14). The substitution in hemoglobinsRainier and Hiroshima is in the H helix which is con-tiguous with the FG segment. A structural alteration inthese areas might well effect a marked change in theequilibrium between dimer and tetramer.

In these experiments the extent of subunit dissociationof various normal and abnormal human and animal he-moglobins was studied by measuring their relative elu-tion volumes on G-100 Sephadex, a cross-linked dextranwhich serves as a molecular sieve. In addition, the rela-tive haptoglobin binding of these hemoglobins wastested.

METHODSPreparation of hemoglobin samples. Blood specimens were

collected in acid citrate dextrose (ACD). Hemolysates wereprepared from blood specimens as previously described (15),and unless stated otherwise, were gassed with carbon monox-ide and stored at 4VC. Hemolysates were dialysed against theappropriate eluting buffer before hemoglobin purification bycolumn chromatography. Isolated hemoglobins were concen-trated by ultrafiltration and again gassed with carbon monox-

ide. Blood specimens from an individual heterozygous forhemoglobin Chesapeake were mailed by Air Express. Hemo-globin Chesapeake was separated from hemoglobin A bychromatography on carboxymethylcellulose as described byCharache, Weatherall, and Clegg (3). Experiments onhemoglobin Chesapeake were performed within 1 wk afterblood samples were obtained. Hemoglobins Kansas and Awere isolated from the hemolysate of a heterozygote bychromatography on carboxymethylcellulose by Drs. JosephBonaventura and Austen Riggs (2) and air-shipped in thecarboxy form packed in ice. In like manner hemoglobinsYakima and A from a heterozygote were isolated by chroma-tography on diethylaminoethyl (DEAE)-Sephadex by Dr.Demetrios Rigas (4) and air-shipped. Hemoglobin F wasisolated from a fresh sample of cord blood on Amberlite(CG-50) by the method of Allen, Schroeder, and Balog withthe use of developer 2 (16). With this procedure hemoglobinA3 (A1)1 was isolated from a hemolysate of a normaladult. Hemoglobins C and S were isolated from the bloodof a patient with hemoglobin SC disease by chromatographyon carboxymethylcellulose (17). Hemoglobin S from thehemolysate of a homozygote was also tested without furtherpurification.

In these experiments hemoglobins-59Fe were used asmarkers. Rabbit hemoglobin was labeled as previously de-scribed (18). Human hemoglobin A-'Fe was prepared bythe exchange of heme between unlabeled ferrihemoglobin Aand 'Fe-labeled methemalbumin. This method, described indetail elsewhere (15), allowed much higher specific activitythan is possible by the incubation of human reticulocyteswith 5'Fe. In brief hemin-55Fe was crystallized from rabbithemoglobin-"5Fe and was then bound to human serumalbumin.2 A mixture containing methemalbumin-'Fe anda heme equivalent amount of unlabeled ferrihemoglobin Awas incubated for 100 min at 37°C and then dialyzed againstdeveloper 2 at 4°C. Under the experimental conditionsemployed there was no net loss of heme from the hemo-globin. The labeled cyanmethemoglobin A was separatedfrom the methemalbumin-5Fe on Amberlite CG-50 (16).

Elution volumes of hemoglobins on G-100 Sephadex. Acolumn of G-100 Sephadex 3 measuring 95.5 X 2.4 cm wasused in all these experiments. When oxy- or carboxy-hemoglobins were studied the column was equilibrated witha solution consisting of two parts isotonic NaCl and onepart isotonic phosphate buffer, pH 7.4 (buffered saline).During the chromatograms of carboxyhemoglobin, it wasnot practical to keep the buffer saturated with carbon monox-ide, so that some of the ligand may have been lost orreplaced by oxygen. When ferrihemoglobin cyanide wastested, the buffer contained cyanide (20 mmoles/liter). Whendeoxyhemoglobin was applied to the column the buffercontained sodium dithionite (1.5 mmoles/liter), and 100%onitrogen was bubbled slowly into the buffer in the supplybottle to exclude as much oxygen as possible from thesystem. Most of the experiments were done at room tem-perature. Before running the column at 4°C, it was repackedto its original dimensions in the cold. Samples of 1.5 ml wereapplied to the column and unless stated otherwise consistedof 8 mg/ml of hemoglobin in buffered saline. In someexperiments human albumin labeled with I 4 served as a

1 A3 or A1 hemoglobin refers to the minor fractions whichare more electronegative than hemoglogin A and compriseabout 8% of the adult human hemolysate.

2 Nutritional Biochemicals Corporation, Cleveland, Ohio.3 Pharmacia Fine Chemicals Inc., Piscataway, N. J.4 Squibb Radio-iodinated ("lI) Human Serum Albumin

USP.

Hemoglobin Subunit Dissociation 127

marker. 1.5 mg of albumin-=I, having an activity of 0.1-0.5,uc, was mixed with the hemoglobin solution immediatelybefore application. A flow rate of 20-30 ml/hr was main-tained. Fractions of equal volume (3.62 ml) were obtainedby a Gilson linear fractionator (Model VL). The hemo-globin concentration of successive fractions was obtainedfrom the absorbance of an appropriate dilution in Drabkinssolution, measured at 540 m,& or, with more dilute solutions,at 415 miu. Radioactivity of an aliquot of successive fractionswas measured in a well scintillation counter.

The relative Sephadex mobility of a given substance canbe expressed by the constant Kav which is independent ofthe column dimensions and the extent of packing (19).

K5V = Ve~-VOVt - VI'The elution volume V. of a substance was determined by thevolume of effluent between its point of application and itspeak concentration during elution. The void volume Vo wasdetermined by the elution volume of Dextran Blue,3 a highmolecular weight polymer which is entirely excluded by thegel matrix and therefore has maximal mobility. Vt was thetotal bed volume of the column.

The determination of V. for hemoglobin was subj ect toexperimental error such as slight variability in the initiationof collection and in the fraction volumes. This uncertaintycould be minimized by the use of albumin-"'1I as a marker.Chromatography of a mixture of hemoglobin and albumin-=Iallowed sharp resolution of the two components despite somedegree of overlap since their respective concentrations weremeasured by different means. In the data reported in TableII and Fig. 1, mixtures containing the test hemoglobin anda small amount of albumin-'I were chromatographed. TheV. of hemoglobin was corrected for slight variations inthe V. of the albumin-MI.

The resolution of hemoglobins on G-100 Sephadex wasgreatly increased by chromatographing a mixture containingequal amounts of the test hemoglobin and a '9Fe-labeledreference hemoglobin. In these experiments the applied

samples contained a total of 8 mg/ml of hemoglobin in eitherthe carboxy or cyanmet form and were chromatographed at250C, since under these conditions subunit dissociation ap-peared to be enhanced (see Results). The specific activityof a given fraction SAf (cpm/mg) was calculated from itsradioactivity Af (cpm/ml) and its hemoglobin concentrationCf (mg/ml).

SAfSAf = CfCf,The concentration of the labeled reference hemoglobin Crand that of the unlabeled test hemoglobin Ct in each fractionwas calculated as follows:

Cr AfSAowhere SAo is the specific activity of the original labeledhemoglobin.

Ct = Cf - Cr.

In these and other experiments, the hemoglobin A isolatedfrom the same hemolysate as the abnormal hemoglobin wasalways tested for direct comparison.

Measurement of haptoglobin binding. The relative bind-ing affinities of various human hemoglobins to haptoglobinwere tested by a competition approach described in detailelsewhere (18). In brief, a mixture of equal amounts ofthe carboxy derivatives of unlabeled test hemoglobin andrabbit hemoglobin-'Fe was added in excess to serum ofknown haptoglobin phenotype. The bound hemoglobin wasthen separated with G-100 Sephadex. The binding of thetest hemoglobin relative to the reference labeled hemoglobinwas calculated as follows:

%Hp bound by test Hb = 100 _ 50 SAHVII,SAmixtureMiscellaneous procedures. Carboxyhemoglobin was con-

verted into oxyhemoglobin as recommended by Riggs and

TABLE I IElution Data for Hemoglobins and Other Substances on G-100 Sephadex

Number of Concentrations of Ve albSubstance determinations applied sample Temperature V. Ksv

mg ml °C

Dextran blue 2 1.0 25 1.42 0.000Human albumin 12 1.0 25 (1.00) 0.224Oxyhemoglobin A 1 1.6 25 0.869 0.359Oxyhemoglobin A 4 8.0 25 0.873 (0.870-0.876)* 0.353Oxyhemoglobin A 1 100 25 0.890 0.339Carboxyhemoglobin A 1 8.0 25 0.874 0.352Cyanmethemoglobin A 1 8.0 25 0.874 0.353Deoxyhemoglobin A 3 8.0 25 0.888 (0.884-0.891)* 0.338Cyanmethemoglobin Chesapeake 1 8.0 25 0.882 0.349Cyanmethemoglobin Kansas 1 8.0 25 0.801 0.436Dextran blue 2 1.0 4 1.43 0.000Humanalbumin 4 1.0 4 (1.00) 0.225Oxyhemoglobin A 3 8.0 4 0.885 (0.883-0.887)* 0.348Oxyhemoglobin A 1 100 4 0.891 0.338

* Mean and range.t Estimated from data in Fig. 2.

128 H. F. Bunn

Herner (20). Cyanmethemoglobin was prepared from oxy-hemoglobin as previously described (15). Methemoglobinwas determined by the method of Evelyn and Malloy (21).Alkali denaturation was tested by the procedure devisedby Huisman and Meyering (17). One part of 12 N NaOHwas added to 100 parts of oxyhemoglobin dissolved in 0.01M phosphate buffer, pH 6.75. The final pH of this mixturewas 11.7. Under these conditions the rate of denaturation ofhemoglobin A was slow enough (half-time 3 min) to detectsubtle changes in other hemoglobins with which hemoglobinA was compared. Auto-oxidation of oxyhemoglobin Chesa-peake and A (3 mg/ml) was measured at 370C in 0.01 Mphosphate buffers at pH 6.6 and 7.4.

RESULTS

Experiments were first done to confirm that the K., ofhemoglobin on G-100 Sephadex was useful as an indi-cator of the extent of subunit dissociation. Conditionswhich enhance the subunit dissociation of hemoglobinwill result in a lower mean statistical molecular-weightand therefore an increase in K.,. Conversely, a decreasedK., would be expected in any hemoglobin which hasan impaired degree of dissociation. Human albumin wasfound to have a considerably lower elution volume (andtherefore a lower K.y) than all the hemoglobins tested(Table II), in agreement with the findings of Andrews(22). Only part of this difference could be due to theslight difference in molecular weight between humanalbumin and hemoglobin (69,000 vs. 64,500). The elu-tion volume of hemoglobin would be further increased

-iI

0.U

l00(

50(

deoxyoxy

albumin--

i .'i

ii

void !volume i

I /

by the extent to which it dissociates into dimers duringthe chromatography. Oxyhemoglobin, cyanmethemo-globin, and carboxyhemoglobin all had very similarKavs (Table II). Guidotti, using osmotic pressure mea-surements, found that the first two forms had identicaldegrees of subunit dissociation, but that of carboxyhe-moglobin was somewhat less (12). In contrast, the elu-tion volume of deoxyhemoglobin was consistently lessthan that of the liganded forms (Fig. 1, Table II).It was not technically possible to run the column underthe strictly anaerobic conditions necessary to maintainhemoglobin in its deoxygenated form. Therefore thedeoxyhemoglobin was chromatographed in the pres-ence of dithionite. There is evidence that this reducingagent may cause certain denaturative effects which couldalter its physical and chemical properties (23). Never-theless, the-relatively low Kav of deoxyhemoglobin sug-gests a lower degree of subunit dissociation in agree-ment with the gel filtration data of Merrett (24), thesedimentation data of Benesch and associates (11), andthe osmotic pressure measurements of Guidotti (12).Elution volumes of oxyhemoglobin were measured overa wide concentration range. Mass action law applied toself-associating systems would dictate that dilute solu-tions have a greater degree of subunit dissociation thanconcentrated ones. The Kav of oxyhemoglobin was foundto vary inversely with its concentration (Table II) in

1ihemoglobin I

zhemoglobin ______ - 0o50 o

0.40z

0.300

.0.20 z

0~00

muz&

50 55

FRACTION -No.

FIGURE 1 Elution profiles of oxyhemoglobin A, deoxyhemoglobin A, and human serum albumin onG-100 Sephadex. Chromatograms of oxyhemoglobin (---) and deoxyhemoglobin (-) were each runin triplicate. The resolution of these hemoglobins was enhanced by chromatographing a mixture con-taining the hemoglobin and a small amount of human albumin-'L. The elution profile of the albumin(- - -) was determined from radioactivity of successive fractions.

Hemoglobin Subunit Dissociation 129

I % A

I %/ jKANSAS

I~~~~~1/~~~~~~~~1

*-.%/ a\\ SPECIFIC ACTIVITY

0.5

z

O *0,4

E - 0,3

09z

Z' E 0,21

.A 0.11

I~ I I I

0

0

0

~~~~~~~~~~~~~~I

I Ir I Itt yI55 60 65 70 75 80

FRACTION No.

agreement with the data of Andrews (22) and Guidotti(25). Concentration dependence of hemoglobin dissocia-tion has also been found from osmotic pressure measure-ments (12) and in recent sedimentation data (26). Theelution volumes of oxyhemoglobin solutions were greaterat 250 than 4VC. This change is unlikely to be due to asignificant alteration in the distribution of water insideand outside the gel matrix, since the mobility of albuminwas unaffected by the temperature change (Table II).Obrink, Laurent, and Rigler have shown that the Kayof Ficoll fractions on G-200 Sephadex varied onlyslightly over a wide temperature range (90-60'C) (27).Our results suggest that the subunit dissociation equi-librium of oxyhemoglobin increases slightly with tem-perature. Benesch and associates have some indirect ex-perimental evidence supporting this conclusion (13).However Kirschner and Tanford found that the sedi-mentation of human carboxyhemoglobin was independentof temperature (28).

The elution volumes of various abnormal human he-moglobins were compared. Chromatography of a mixture

I88 5

FIGURE 2 Elution profile of cyanmethe-moglobin Kansas and cyanmethemoglo-bin A-'Fe. A mixture containing equalamounts of the two hemoglobins was runon the G-100 Sephadex column. From

-60 the optical density and radioactivity mea-surements of successive fractions (lower

panel), the specific activities of these frac-tions were calculated (middle panel). The

i40 top panel shows the resolution of the two'E hemoglobins, as calculated from the ex-

perimental data (see Results). The small-20 E peak under the hemoglobin A

W peak (---) represents unlabeled hemo-globin of unknown identity and question-

_ 0 able significance.90

of labeled reference hemoglobin allowed much greaterresolution than was possible by measurement of the Ka.of the test hemoglobin alone. If the test hemoglobin hasthe same degree of subunit dissociation as the referencehemoglobin, then they should have identical elutionvolumes on G-100 Sephadex, and the specific activitiesof successive fractions should be identical. This wasfound to be true for carboxyhemoglobins A3, F, S, andC. The ion-exchange capacity of G-100 Sephadex ap-peared to be insignificant since these hemoglobins havewidely different surface charges and yet had elutionvolumes identical with carboxyhemoglobin A-2Fe, andtherefore to each other. When carboxyhemoglobin Kan-sas was tested, a marked fall in specific activity of suc-cessive fractions indicated that the unlabeled test hemo-globin had a greater elution volume than the referencehemoglobin, and therefore a lower mean statistical mo-lecular weight indicative of an increased degree of sub-unit dissociation. Identical results were obtained whenthe cyanmet derivatives of hemoglobin Kansas andA-"9Fe were tested (Fig. 2). The K., of Kansas was

130 H. F. Bunn

z 0.40-0

-q 0.30-l E

v E 0,20-z0u 0.10-.a

0 - a * *- 0 m--

200-1-

,, a 150-4(%E

v E 100-

_V

'U 50-

m

0 m Iff --.: I

I I I I

A Kansas

SPECIFIC ACTIVITY

I60 65 70

FRACTION No.

60

40 A

E

E20 ci

0

FIGURE 3 Elution profile of cyanmethe-moglobin A, isolated from the Kansashemolysate, and cyanmethemoglobinA-'Fe.

75

higher than that of any of the other hemoglobins tested(Table II). When the most dilute solution of hemo-globin A was chromatographed, its initial concentra-tion at the point of application was 1.6 mg/ml, and itsmean concentration upon elution was 0.073 mg/ml. Un-der these conditions, during the run most of the hemo-globin would be expected to be in the form of dimers,as estimated from recently derived equilibrium con-stants (12, 28, 29). The fact that the elution volumeof hemoglobin Kansas was far greater suggests thatdissociation had extended to the formation of monomers.These findings are in agreement with the sedimentationvelocity data of Bonaventura and Riggs (2). Theyfound that carboxy- and oxy- but not deoxyhemoglobinKansas had low sedimentation coefficients indicatingthat Kansas dissociated into dimers at concentrationsin which hemoglobin A remained as an intact tetramer.The abnormal Sephadex mobility of hemoglobin Kansaswas not likely to be due to any alterations resultingfrom its purification since the hemoglobin A isolatedfrom the same column showed an elution volume identi-cal with the A-'Fe reference hemoglobin (Fig. 3).

Hemoglobin Chesapeake had a lower elution volumethan the labeled hemoglobin A (Fig. 4). Both the car-boxy- and ferrihemoglobin cyanide derivatives gaveidentical and reproducible results on a total of six chro-matograms prepared from hemoglobins isolated fromtwo separate blood samples. Hemoglobin A isolated onthe same preparative column as the Chesapeake A wasalso tested each time and found to have an elution vol-ume identical with the A-Fe reference hemoglobin(Fig. 5). When cyanmethemoglobin Chesapeake wasrun alone with the albumin-'I marker, its K.y was notvery much less than that of cyanmethemoglobin A ob-tained from the same column (Table II). This smalldifference is probably a reflection of the much higherresolution offered. by the chromatography of hemoglobinmixtures. In contrast to the carboxy and cyanmet deriva-tives, deoxyhemoglobin Chesapeake showed an elutionvolume identical with that of deoxyhemoglobin A-w'Fe.The results from this experiment very closely resembledthose shown in Figs. 3 and 5. Finally when a mixture ofthe cyanmet derivatives of Chesapeake and A-Fe wasrun at 4VC, only a slight increase in specific activity of

Hemoglobin Subunit Dissociation 131

z0

Ua_ 0.30-U "I*-Z E 0.20-

A 0.10z

0M

> 200-1-

@ 150-U E

_ f 100'49I.,v

L" 50.U.$A

oI.

z0

I-.Z-LuE

Z aO E

Ax

m

successive fractions was noted indicating less of a dif-ference in elution volumes than at 250C. Perhaps this isbecause subunit dissociation of hemoglobin A appearsto be reduced at the lower temperature (Table II). Theseresults indicate that the liganded but not the deoxy formof hemoglobin Chesapeake have decreased elution vol-umes and therefore a reduced degree of subunit dissocia-tion when compared to hemoglobin A.

Under the same conditions used for hemoglobinChesapeake (Fig. 4), no difference between the elutionvolume of hemoglobin Yakima (and its companion Ahemoglobin) and hemoglobin A-Fe could be demon-strated.

z0

g 0.30zmu -

UE ]

A /

3-

uE- a

v-

U

3-

uEu _%zC.OE

8

#A

z

200-

I50-

100-

0-loo

The elution volumes of various animal hemoglobinswere measured by chromatographing a mixture of theanimal hemoglobin and human hemoglobin A-mFe, underthe same conditions employed in the previous experiments(Figs. 2-5). The cyanmet and carboxy derivatives gavesimilar results. Dog and guinea pig hemoglobins showedelution volumes identical with human. The results forthese experiments were very similar to the data shownin Figs. 3 and 5. Cat hemoglobin had a larger elution vol-ume than human hemoglobin (Fig. 6). The elution vol-umes of rabbit and rat hemoglobins were slightly greaterthan the human, dog, and guinea pig hemoglobins butless than the cat hemoglobin.

SPECIFIC ACTIVITY

uD - -

0.50'

0.40'

0.30 '

'u t

40 EE

20,20

0

FRACTION No.FIGURE 4 Elution profile of cyanmethemoglobin Chesapeake and cyanmethemoglobin A-"Fe.

132 H. F. Bunn

Chesapeake

SPECIFIC ACTIVITY/

60

40 .:

E

20E

*0

55 60 65 70 75 80FRACTION No.

FIGuRE 5 Elution profile of cyanmethemoglobin A, isolated from the Chesapeakehemolysate, and cyanmethemoglobin A-'Fe.

The relative affinities of the various human hemoglo-bins for haptoglobin are shown in Fig. 7. If the test he-moglobin had precisely the same affinity as the rabbithemoglobin-'Fe, then the specific activity of the isolatedHb-Hp complex would be identical with that of theoriginal hemoglobin mixture. Exactly 50% of the hap-toglobin would be bound by the test hemoglobin. Infact most of the test hemoglobins showed a slightlygreater affinity than the labeled rabbit hemoglobin, form-ing complexes with about 52-55% of the haptoglobin.These results agree well with earlier data on hemo-globin A (18). In contrast to hemoglobins A, F, S, andC, hemoglobin Kansas showed a considerably greateraffinity for haptoglobin, whereas hemoglobins Chesa-

peake and Yakima showed somewhat decreased affinity.Haptoglobin phenotype had no effect on its binding tothe various test hemoglobins. In a parallel experiment,a mixture containing equal amounts of unlabeled cathemoglobin and 'Fe-labeled human hemoglobin wasadded in excess to human serum (type 1-1 haptoglobin).62% of the haptoglobin complexed with the cat hemo-globin, 38% with the human.

Oxyhemoglobin Chesapeake and its companion oxy-hemoglobin A (isolated from the same ion-exchangecolumn) both showed equal rates of alkali denaturation.Furthermore these hemoglobins both auto-oxidized atthe same rate.

Hemoglobin Subunit Dissociation 133

z06-

z -0 Ez CDAd Ez0

-oI

01

> 200-_

Eu 150-

_ WI loo-

X, 50-M-

0.50 -

zO 0.40'6-

Ixa' 1 0.30'Z -Wu E

Z 0.2

O E0.10'

-Ix

I I I I I I

SPECIFIC ACTIVITY

0.50*

0.40 -z

a

Q E0.20z

0.10-

0 s15 5 60 65

FRACTION No.

--I i70 7 5

FIGURE 6 Elution profile of the cyanmet derivative of cat hemoglobinand cyanmethemoglobin A-'Fe.

134 H. F. Bunn

0.30-

0.20-

z0

49

zUA

z0v.az

E

0.101

3o0

300-

I-1-

u4

U

amIL

200.

E0E

100-

.80

*60

i40 E

E.S

20

'0

% OF Hp BOUNDBY TEST HEMOGLOBIN30 50 70

A

A (Kansas Column)

A (Chesapeake Column)

A (Yakima Column)

C

Kansaso

000 0

0 0

30 50

*0Co0

0

0 0

0O

FIGURE 7 Relative haptoglobin binding of varioushuman hemoglobins. To a mixture containing equalamounts of carboxy-9Fe rabbit hemoglobin and thecarboxy hemoglobin to be tested was added a smallamount of normal serum, of haptoglobin phenotype1-1 (0), or 2-2 (0). The per cent of the hapto-globin bound to the test hemoglobin was calculatedfrom the specific activity of the isolated hemo-globin-haptoglobin complex (see Results).

700=1-1 Hp

*=2-2 Hp

DISCUSSION

One of the long-standing interests in hemoglobin chem-istry has been the dissociation of the molecule intosmaller fragments. Subunit formation was readily seenafter subjecting hemoglobin to various stresses such asextremes of pH or very high ionic strength (30). Asthe protein structure was elaborated it appeared likelythat subunit dissociation involved disruption of linkagesbetween the component a and polypeptide chains.Vinograd and Hutchinson presented convincing evidencethat the tetramer split symmetrically (31):

a28t2 2a,8.Experiments along several different lines indicated thatthis equilibrium existed under physiologic conditions ofpH and ionic strength (12, 13, 32) and therefore mightbe an important determinant of hemoglobin function.

Subunit dissociation has been approached by a varietyof methods, all of which involve the measurement ofaverage molecular weight. The observed molecularweight can be related to that of the intact molecule andserves as an index of the extent of dissociation. Osmoticpressure measurements (12) and sedimentation velocitydata (28) have been used to calculate an equilibriumconstant for hemoglobin dissociation under physiologicconditions. Ackers and Thompson (29), and morerecently Chiancone, Gilbert, Gilbert, and Kellett (33),have used G-100 Sephadex to study hemoglobin disso-ciation and have calculated equilibrium dissociationconstants for human oxyhemoglobin. In order to obtainthese quantitative results it was necessary to load the

column with a large enough volume of hemoglobin solu-tion to achieve a plateau type elution profile. In theexperiments reported here and in those of Andrews(22), small volumes were applied, resulting in theformation of sharp peaks. Although this experimentalapproach provided qualitative data, it does not permitthe calculation of equilibrium constants. However, thedata shown in Table II confirm that it can be useful,since conditions known to affect hemoglobin dissociationare reflected in expected alterations in hemoglobinmobility.

In these experiments the elution volume of singlehemoglobin solutions was not sensitive enough to detectsubtle changes in subunit dissociation. When ligandedforms of hemoglobin Chesapeake were run in a mixturewith an equal amount of hemoglobin A-69Fe, a consistentrise in specific activity of successive fractions indicatedthat the Chesapeake had a lower elution volume andtherefore a decreased tendency to dissociate into dimers.Exchange of intact heme groups, a phenomenon whichproceeds quite readily in mixtures containing methemo-globin, was very unlikely to affect the results citedabove. Carboxy and cyanmet derivatives were run at250C, conditions under which heme exchange does notoccur (15). However, another source of error may bemore significant. Guidotti has osmotic pressure data in-dicating that in mixtures of two different hemoglobins, say(a2,2) and (a''a2), the hybrid afa'#' escapes detectionby any method dependent on charge differences for resolu-tion (34). In the chromatographic runs containing mix-tures of hemoglobin A and Chesapeake, the extent that

Hemoglobin Subunit Dissociation 135

Chesapeake

Yakima

I I

-IL

the hybrid aAaCh-fia was present would attenuate theresolution of the two parent molecules into separatepeaks. The extent of this phenomenon cannot be mea-sured directly, but it was unlikely to be a major sourceof error since there was little difference between theSephadex mobility of hemoglobin Chesapeake whenmeasured alone (Table II) as compared to its mobilityin a mixture with the labeled reference hemoglobin(Fig. 4). The failure to observe a difference betweenthe elution volumes of hemoglobin Yakima and thereference hemoglobin does not support the contentionthat Yakima like Chesapeake has a decreased tendencyto dissociate. This possibility is not entirely ruled out,however. The extent of formation of the interspecieshybrid (a2/SAI3Yakima) may have been great enough topreclude resolution of the two hemoglobins.

It was pertinent to compare the affinities of the varioushemoglobins to haptoglobin. This plasma protein bindsspecifically and irreversibly to hemoglobin. 1 mole oftype 1-1 haptoglobin is known to combine stoichiometri-cally with 1 mole of hemoglobin (35). Since type 1-1haptoglobin is a symmetrical molecule (36), it is likelyto have two binding sites. Evidence from several differ-ent experimental approaches indicates that haptoglobinbinds separate a,# dimers rather than the intact hemo-globin tetramer (18, 37-39). If so, a hemoglobin whichhas an increased tendency to dissociate into half mole-cules would have a relatively high affinity for hapto-globin, and vice versa. When viewed in this way, therelatively high haptoglobin affinity of hemoglobin Kansasand cat hemoglobin and the low haptoglobin affinity ofhemoglobin Chesapeake are in good agreement with theSephadex mobility data. However, it is also possiblethat these hemoglobins have alterations in tertiarystructure which affect their affinities for haptoglobin.Comparable data for hemoglobin A treated with varioussulfhydryl agents offer an interesting comparison (18).P-mercuribenzoate-treated hemoglobin, known to havean enhanced degree of dissociation into subunits, hada relatively high haptoglobin affinity. Conversely hemo-globin reacted with bis- (N-maleimidomethyl) ether(BME) has a markedly impaired degree of dissociation(40) and a low affinity for haptoglobin. The G-100Sephadex mobility of BMEhemoglobin was considerablygreater than that of untreated hemoglobin-'9Fe withwhich it had been mixed,5 a finding similar to thatobtained for hemoglobin Chesapeake (Fig. 4). The simi-larities between hemoglobin Chesapeake and BMEhemo-globin extend even further. Of all the sulfhydryl re-agents, BMEappears to effect the greatest increase inoxygen affinity (40-42).

Benesch and associates have proposed a model (13)which is based on two important properties of hemo-

5Unpublished observation.

globin: its change in conformation upon reaction withligand and its reversible dissociation into dimers. Morerecently Guidotti has derived a comprehensive series ofequations which relate subunit dissociation and ligandbinding in quantitative terms (43). When hemoglobinconcentration is large relative to the values of thedissociation equilibrium constants, a condition certainlyexisting in the red cell, the following simplified equationcan be written (43):

p ( Ko \1/4

P0-\K31 Kd,0where Pso is the partial pressure at which half of thehemoglobin is bound by ligand (such as oxygen); Ko andKdeo are the dissociation equilibrium constants for theoxygenated and deoxygenated tetramers respectively, andKs is the association equilibrium constant for the reac-tion of the af dimer with oxygen. It can be readilyappreciated from this equation that if the ratio of thedissociation constant of liganded hemoglobin over thatof unliganded hemoglobin is abnormally high, as is truefor hemoglobin Kansas, there will be an increase in Pzo,or a decrease in oxygen affinity. It is of interest to applythe equilibrium constant derived by Bonaventura andRiggs for oxyhemoglobin Kansas to this equation. Riggs'and Bonaventura's sedimentation data indicate thatKdeoKans = KdeoA. If K3Kans=K3A, an assumption that asyet has no experimental basis, then, from experimentalvalues for KOA (1.0-1.4 X 10-' mole/liter [44, 45]) andKoKans (200 X 10-" [2]), the following ratio can becalculated: P50A/P50K..S = 0.28. The ratio P3oA/P50Kan.derived from the oxygen dissociation curves of Riggsand Bonaventura is 0.30 (2). This good agreement lendssupport for the models proposed by Benesch et al. andby Guidotti. Further support is provided by a compari-son of oxygen affinity and subunit dissociation of certainanimal hemoglobins. Cat hemoglobin, for example, hasa much lower oxygen affinity than that of other mammals(46). Early sedimentation data suggested that cat hemo-globin dissociated readily in dilute solution, althoughthe data are difficult to interpret because of variabilityin experimental conditions (47). Sullivan recently re-ported that cat hemoglobin eluted after human hemo-globin on G-100 Sephadex but presented no data (48).The results shown on Fig. 6 confirm that cat hemoglobinhas a large elution volume and therefore an increaseddegree of subunit dissociation. The adult heterozygoussheep has two hemoglobins, types A and B, which arereadily separated by electrophoresis. The "slow" hemo-globin, type B, has a much lower oxygen affinity than the"fast" type A hemoglobin (49). Gilbert, using a highlysensitive gel filtration approach, similar in principle tothe one used in the present studies, has recently shownthat type B sheep hemoglobin dissociates more readilythan type A (50).

136 H. F. Bunn

The abnormally high oxygen affinity seen in certainhemoglobin variants (Table I) may be due to a lowdegree of subunit dissociation of the liganded form.The data on hemoglobin Chesapeake which have beenpresented here suggest this explanation, but other fac-tors may also be contributory. It may be that the aminoacid substitution in Chesapeake (and similar variants)confers subtle conformational changes which per secould alter ligand affinity. Nagel, Gibson, and Charachehave shown that deoxyhemoglobin Chesapeake differsfrom deoxyhemoglobin A in reactivity to iodoacetamideand the ionization of tyrosine residues (51). In contrastonly the liganded forms of the two hemoglobins differin their Sephadex elution volumes. Furthermore, Nageland colleagues showed that carbon monoxide reactedwith Chesapeake twice as rapidly as A, but this observa-tion does not exclude the postulated importance of sub-unit dissociation as a determinant of ligand affinity.

Hemoglobin Chesapeake had a normal rate of alkalidenaturation. It is noteworthy that hemoglobin Rainier,which resembles Chesapeake in its clinical features andfunctional properties (Table I), is the first adult humanhemoglobin found to be alkali resistant (7). This phe-nomenon which is so striking in hemoglobin F is notlikely to be due to a decreased tendency to form dimers,since hemoglobin F appears to have a normal mobility onG-100 Sephadex. Furthermore, hemoglobin Chesapeakeand BMEhemoglobin,8 both of which appear to disso-ciate less readily into half molecules, have rates of alkalidenaturation similar to hemoglobin A. Finally the veryhigh pH at which denaturation is measured would beexpected to favor dimer formation in all hemoglobins.What seems more plausible is that resistance to alkalidenaturation is due to decreased dissociation of dimerto monomer. Hemoglobin F forms interspecies hybridsmuch less readily than hemoglobin A (52), a phe-nomenon dependent upon the formation of monomer.Furthermore, hemoglobin Kansas has a greater rate ofalkali denaturation than hemoglobin A (2). The datashown in Table II and Fig. 2 suggest that Kansas hasan increased tendency to form monomers as well asdimers.

All the hemoglobin variants depicted in Table I havea low n value, indicative of decreased "heme-heme"interaction. This fact is not predictable from Guidotti'smodel merely on the basis of alterations in the extentof subunit dissociation and point to conformationalchanges in the Oa-dimers themselves, which would affectcooperative interactions. The findings of Nagel andcoworkers (51) indicate that this is probably true forhemoglobin Chesapeake. Furthermore, the apparent tend-ency for hemoglobin Kansas to form monomers mayunderlie a decrease in cooperative interactions.

6 Unpublished observation.

ACKNOWLEDGMENTSDrs. Samuel Charache, Demetrios Rigas, and Austen Riggskindly furnished the blood and hemoglobin specimens whichwere so essential for this study. I am especially indebtedto Dr. Riggs for his constructive comments and for review-ing this manuscript. Dr. Gerald Bloom performed the hapto-globin typing.

REFERENCES

1. Reissmann, K. R., W. E. Ruth, and T. Nomura. 1961.A human hemoglobin with lowered oxygen affinity andimpaired heme-heme interactions. J. Clin. Invest. 40:1826.

2. Bonaventura, J., and A. Riggs. 1968. emoglobin Kan-sas, a human hemoglobin with a neutral amino acid sub-stitution and an abnormal oxygen equilibrium. J. Biol.Chem. 243: 980.

3. Charache, S., D. J. Weatherall, and J. B. Clegg. 1966.Polycythemia associated with a hemoglobinopathy. J.Clin. Invest. 45: 813.

4. Jones, R. R., E. E. Osgood, B. Brimhall, and R. D.Koler. 1967. Hemoglobin Yakima. I. Clinical and bio-chemical studies. J. Clin. Invest. 46: 1840.

5. Novy, M. J., M. J. Edwards, and J. Metcalfe. 1967.Hemoglobin Yakima. II. High blood oxygen affinityassociated with compensatory erythrocytosis and normalhemodynamics. J. Clin. Invest. 46: 1848.

6. Reed, C. S., R. Hampson, S. Gordon, R. T. Jones, M. J.Novy, B. Brimhall, M. J. Edwards, and R. D. Koler.1968. Erythrocytosis secondary to increased oxygenaffinity of a mutant hemoglobin, Hemoglobin Kempsey.Blood. 31: 623.

7. Stamatoyonnopoulos, G., A. Yoshida, J. Adamson, andS. Heinenberg. 1968. Hemoglobin Rainier (145 tyrosine-histidine): alkali resistant hemoglobin with increasedoxygen affinity. Science. 159: 741.

8. Shibata, S., T. Miyaji, I. Iuchi, and H. B. Hamilton.1968. Hemoglobin Hiroshima: a fast moving hemoglobinassociated with polycythemia and increased oxygen af-finity. XII Congress, International Society of Hema-tology. 73. (Abstr.)

9. Lines, J. G., and R. McIntosh. 1967. Oxygen binding byhaemoglobin J-Cape Town. Nature. (London). 215: 297.

10. Muirhead, H., and M. F. Perutz. 1963. A three-dimen-sional Fourier synthesis of reduced human hemoglobinat 5.5 A resolution. Nature. (London). 199: 633.

11. Benesch, R. E., R. Benesch, and M. E. Williamson. 1962.The influence of reversible oxygen binding on the inter-action between hemoglobin subunits. Proc. Nat. Acad.Sci. U. S. A. 48: 2071.

12. Guidotti, G. 1967. Studies on the chemistry of hemoglobin.II. The effect of salts on the dissociation of hemoglobininto subunits. J. Biol. Chem. 242: 3685.

13. Benesch, R. E., R. Benesch, and G. Macduff. 1965. Sub-unit exchange and ligand binding: a new hypothesis forthe mechanism of oxygenation of hemoglobin. Proc. Nat.Acad. Sci. U. S. A. 54: 535.

14. Perutz, M. F. 1965. Structure and function of haemo-globin. I. A tentative atomic model of horse oxyhaemo-globin. J. Mol. Biol. 13: 646.

15. Bunn, H. F., and J. H. Jandl. 1968. Exchange of hemeamong hemoglobins and between hemoglobin and albu-min. J. Biol. Chem. 243: 465.

16. Allen, D. W., W. A. Schroeder, and J. Balog. 1958.Observations on the chromatographic heterogeneity of

Hemoglobin Subunit Dissociation 137

normal adult and fetal human hemoglobin. J. Amer.Chem. Soc. 80: 1628.

17. Huisman, T. H. J., and C. A. Meyering. 1960. Studieson the heterogeneity of hemoglobin. I. The hetero-geneity of different human hemoglobin types in carboxy-methylcellulose and in Amberlite IRC-50 chromatography.Clin. Chem. Acta. 5: 103.

18. Bunn, H. F. 1967. Effect of sulfhydryl reagents on thebinding of human hemoglobin to haptoglobin. J. Lab.Clin. Med. 70: 606.

19. Laurent, T. C., and J. Killander. 1964. A theory of gelfiltration and its experimental verification. J. Chromatog.14: 317.

20. Riggs, A., and A. E. Herner. 1962. The hybridization ofdonkey and mouse hemoglobins. Proc. Nat. Acad. Sci.U. S. A. 48: 1664.

21. Evelyn, K. A., and H. T. Malloy. 1938. Microdetermi-nation of oxyhemoglobin, methemoglobin and sulfhemo-globin in a single sample of blood. J. Biol. Chem. 126:655.

22. Andrews, P. 1964. Estimation of the molecular weightof proteins by Sephadex gel filtration. Biochem. J. 91:222.

23. Dalziel, K., and J. R. O'Brien. 1957. Side reactions inthe deoxygenation of dilute oxyhemoglobin solutions bysodium dithionite. Biochem. J. 67: 119.

24. Merrett, T. 1966. Observations on the dissociation ofoxy-, deoxy-, and ferrihemoglobin in relation to pH andionic strength by the Sephadex method. Biochem. Bio-phys. Acta. 124: 389.

25. Guidotti, G. 1964. Studies on the dissociation of humanhemoglobin. In Structure and Activity of Enzymes. T. W.Goodwin, J. I. Harris, and B. S. Hartley, editors. Aca-demic Press Inc., New York. 157.

26. Schachman, H. K., and S. J. Edelstein. 1966. Ultra-centrifuge studies with absorption optics. IV. Molecularweight determinations at the microgram level. Bio-chemistry. 5: 2681.

27. Obrink, B., T. C. Laurent, and R. Rigler. 1967. Studieson the temperature dependence of chromatography on adextran gel (Sephadex G-200). J. Chromatogr. 31: 48.

28. Kirschner, A. G., and C. Tanford. 1964. The dissociationof hemoglobin by inorganic salts. Biochemistry. 3: 291.

29. Ackers, G. K., and T. E. Thompson. 1965. Determinationof stoichiometry and equilibrium constants for reversiblyassociated systems by molecular sieve chromatography.Proc. Nat. Acad. Sci. U. S. A. 53: 342.

30. Rossi Fanelli, A., E. Antonini, and A. Caputo. 1964.Hemoglobin and myoglobin. Advan. Protein Chem. 19:73.

31. Vinograd, J., and W. D. Hutchinson. 1960. Carbon-14labeled hybrids of haemoglobin. Nature. (London). 187:216.

32. Guidotti, G., W. Konigsberg, and L. C. Craig. 1963.On the dissociation of normal adult human hemoglobin.Proc. Nat. Acad. Sci. U. S. A. 50: 774.

33. Chiancone, E., L. M. Gilbert, G. A. Gilbert, and G. L.Kellett. 1968. Dissociation of hemoglobin into subunits.II. Human oxyhemoglobin: gel filtration studies. J. Biol.Chem. 243: 1212.

34. Guidotti, G. 1967. Studies on the chemistry of hemo-globin. III. The interactions of the af subunits of hemo-globin. J. Biol. Chem. 242: 3694.

35. Jayle, M. F., and J. Moretti. 1962. Haptoglobin. Progr.Hematol. 3: 342.

36. Shim, B. S., and A. G. Bearn. 1964. Immunological andbiochemical studies on serum haptoglobin. J. Exp. Med.120: 611.

37. Laurell, C. B. 1959. Purification and properties of differ-ent haptaglobins. Clin. Chem. Acta. 4: 79.

38. Lionetti, F. J., C. R. Valeri, J. C. Bond, and N. L.Fortier. 1964. Measurement of hemoglobin binding ca-pacity of human serum or plasma by means of dextrangels. J. Lab. Clin. Med. 64: 519.

39. Shim, B. S., T. H. Lee, and Y. S. Kang. 1965. Immuno-logical and biochemical investigations of human serumhaptoglobin: composition of haptoglobin-haemoglobinintermediate, haemoglobin binding sites and presence ofadditional alleles for 8 chain. Nature. (London). 207:1264.

40. Simon, S. R., and W. H. Konigsberg. 1966. Chemicalmodification of hemoglobins: a study of conformationrestraint by internal bridging. Proc. Nat. Acad. Sci.U. S. A. 56: 749.

41. Taylor, J. F., E. Antonini, M. Brunori, and J. Wyman.1966. Studies on human hemoglobin treated with varioussulfhydryl reagents. J. Biol. Chem. 241: 241.

42. Benesch, R., and R. E. Benesch. 1961. The chemistry ofthe Bohr effect. I. The reaction of N-ethyl maleimidewith the oxygen linked acid groups of hemoglobin. J.Biol. Chem. 236: 405.

43. Guidotti, G. 1967. Studies on the chemistry of hemo-globin. IV. The mechanism of reaction with ligands. J.Biol. Chem. 242: 3i04.

44. Kawahara, K., A. G. Kirschner, and C. Tanford. 1965.Dissociation of human CO-hemoglobin by urea, guanidinehydrochloride and other reagents. Biochemistry. 4: 1203.

45. Rosemeyer, M. A., and E. R. Huehns. 1967. On themechanism of dissociation of hemoglobin. J. Mol. Biol.25: 253.

46. Taketa, F., and S. A. Morell. 1966. Oxygen affinity ofcat hemoglobin. Biochemn. Biophys. Res. Commun. 24:705.

47. Svedberg, T., and A. Hedenius, 1934. The sedimentationconstants of the respiratory proteins. Biol. Bull. 66: 191.

48. Sullivan, B. 1967. The effect of dilution on the oxygena-tion properties of cat and human hemoglobins. Biochem.Biophys. Res. Commun. 28: 407.

49. Van Vliet, G., and T. H. J. Huisman. 1964. Changes inthe hemoglobin types of sheep as a response to anemia.Biochem. J. 93: 401.

50. Gilbert, G. A. 1966. Layer technique in gel filtrationfor estimating differences in the degree of dissociationof closely related proteins. Nature. (London.) 212: 296.

51. Nagel, R. H., Q. H. Gibson, and S. Charache. 1967.Relation between structure and function in hemoglobinChesapeake. Biochemistry. 6: 2395.

52. Huehns, E. R., G. H. Beaven, and B. L. Stevens. 1964.Recombination studies on hemoglobins at neutral pH.Biochem. J. 92: 440.

138 H. F. Bunn