November 25, 1949, Vol. 110 SCIENCE 543 -~ ..-. ~.--.. -. Sickle Cell Anemia, a Molecular Disease1 Linus Pauling, Harvey A. Itano,a S. J. Singer,' and Ibert C. Wells3 Gates and Crellin Laboratories of Chemistry, California Institute of Technology, Pasadena, California4 T HE ERYTHROCPTES of certain individuals possess the capacity to undergo reversible changes in shape in response to changes in the partial pressure of oxygen. When the oxygen pressure is lowered, these cells change their forms from the normal biconcave disk to crescent, holly wreath, and other forms. This process is known as sickling. About 8 percent of Anlerican Negroes possess this characteristic; usually they exhibit no pathological consequences ascribable to' it. These people are said to have sicklemia, or sickle cell trait. However, about 1 in 40 (4) of these individuals whose cells are capable of sickling suffer from a scvcre chronic anemia re- sulting from execssivr destruction of their crythro- cytes; the term sickle cell anelnia is applied to their condition. The main observable difference between the erythro- eytes of sickle cell trait and sickle cell anemia has been that a considerably greater reduction in the partial pressure of oxygen is required for a major fraction of the trait cells to sickle than for the anelllia cells (11). Tests in. viva have demonstrntcd that betweeli 30 and 60 percent of the erythrocytes in tha.venous circulation of sickle cell anemic individuals, but less than 1 percent of those in the venous circulation of sicklemic individuals, are normally sickled. Experi- lnents i+z aitro indicate that under sufficiently low oxy- gen pressure, however, all the cells of both types as- sume the sickled form. The evidence available at the tinre that our investi- gation was begun indicated that the process of sick- ling might be intimately associated with the state and the nature of the hemoglobin within the erythrocyte. Sickle cell erythrocytes in which the hemoglobin is combined with oxygen or carbon monoxide have the biconcave disk contour and are indistinguishable in IThis rrsr:~rrh n-as carried out with the nid of a grant from the United States Public Health Service. The authors are grateful to Professor Ray D. Owen, of the Biology Di- vision of this Institute, for his helpful suggestions. We nre indebted to Dr. Edward R. Evans, of Pasadena. Dr. Travis Winsor, of Los Angeles, nnd Dr. G. E. Burch, of the Tulane University School of Medicine, New Orleans. for their aid in ohtnining the blood used in tlicw expwimmts. 2U. S. Public Health Service postdoctoral fellow of the Nntionnl Institutes of Henlth. x Postdoctoral fellow of the Division of Medical Sciences of the National Research Council. 4 Contribution Ko. 1.133. that form from normal erythrocytes. In this condi- tion they are termed promeniscocytes. The hemo- globin appears to be uniformly distributed and ran- domly oriented within normal cells and promenisco- cytes, and no birefringence is observed. Both types of cells are very flexible. If the oxygen or carbon monoxide is removed, however, transforming the hemo- globin to the uncombined state, the promeniscocytes undergo sickling. The hemoglobin within the sickled cells appears to aggregate into one or more foci, and the cell membranes collapse. The cells become bire- fringent (11) and quite rigid. The addition of oxy- gen or carbon monoxide to these cells reverses these phenomena. Thus the physical effects just described depend on the state of combination of the hemoglobin, and only secondarily, if at all, on the cell membrane. This conclusion is supported by the observation that sickled cells when lysed with water produce discoidal, rather than sickle-shaped, ghosts (JO). It was decided, thrrrfore, to examine the physical and chemical properties of the hemoglobins of indi- viduals with sicklemia and sickle cell anemia, and to compare theIt with the hemoglobin of normal indi- viduals to determine whether any significant differ- ences might be observed. EXPERIMENTA4L METHODS The experimental work reported in this paper deals largely with an elect,rophoretic study of these hemo- globins. In the first phase of the investigation, which concerned the comparison of normal and sickle cell anemia hemoglobins, three types of experiments were performed : 1) with carbonmonoxyhemoglobins; 2) with uncombined ferrohemoglobins in the presence of dithionite ion, to prevent oxidation to methemoglo- bins; and 3) with carbonmonoxyhemoglobins in the presence of dithionite ion. The experiments of type 3 were performed and compared with those of type 1 in order to ascertain whether the dithionite ion itself causes any specific electrophoretic effect. Samples of blood were obtained from sickle cell anemic individuals who had not been transfused within three months prior to the time of sampling. Stroma- free concentrated solutions of human adult hemoglobin were prepared by the method used by Drabkin (3). These solut.ions were diluted just before use with the 544 SCIENCE November 25, 1949, Vol. 110 appropriate buffer until the hemoglobin concentrations were close to 0.5 grams per 100 milliliters, and then were dialyzed against large volumes of these buffers for 12 to 24 hours at 4' C. The buffers for the ex- periments of types 2 and 3 were prepared by adding 300 ml of 0.1 ionic strength sodium dithionite solu- tion to 3.5 liters of 0.1 ionic strength buffer. About 100 ml of 0.1 molar NaOH was then added to bring the pH of the buffer back to its original value. Fer- rohemoglobin solutions were prepared by diluting the FIG. 1. Mohilitr(p)-pH `&rves for carhonmonoxyhemo- glohins in phosphate buffers of 0.1 ionic strength. The black circles and black squares denote the data for experiments performed with buffers containing dithionite ion. The open square designated by the arrow represents an average value of 10 experiments on the hemoglobin of different individuals with sickle cell anemia. The mobilities recorded in this graph are averages of the mobilitles in the ascending and descending limbs. PH FIG. 2. Mobility (IL) -pH curves for ferrohemoglobius in phosphate buffers of 0.1 ionic streIlgth containing dithionite ion. The mohilities recorded in the graph are averages of the mabillties in the ascending and descending limbs. concentrated solutions with this dithionite-containing buffer and dialyzing against it under a nitrogen atmos- phere. The hemoglobin solutions for the experiments of type 3 were made up similarly, except that they were satnrated with carbon monoxide after dilution and were dialyzed under a carbon monoxide atmos- phere. The dialysis hags were kept in continuous motion in the buffers by means of a stirrer with a mercury seal to prevent the escape of the nitr0gc.n and carbon monoxide gases. The experiments were carried out in the modified Tiselius electrophoresis apparatus described by Swingle (14). Potential gradients of 4.8 to 8.4 volts per centi- meter were employed, and the duration of the runs varied from 6 to 20 hours. The pH values of the buffers were measured after dialysis on samples which had come to room temperature. The results indicate that a significant difference es- ists between the electrophoretic mobilities of hemo- globin derived from erythrocytes of normal individuals and from those of sickle cell anemic individuals. The two types of hemoglobin are particularly easily dis- tinguishrd as the carbonmonoxy compounds at pH 6.9 in phosphate buffer of 0.1 ionic strength. In this buffer the sickle cell anemia carbonmonoxyhetnoglobin moves as a positive ion, while the normal compound moves as a negative ion, and there is no detectable amount of one type present in the other.4 The hemo- globin derived from erythrocytes of individuals with sicklemia, however, appears to be a mixture of thr normal hemoglobin and sickle cell anemia hemoglobin in roughly equal proportions. Up to the present time the hemoglobins of 15 persons with sickle cell anemia, 8 persons with sicklemia, and 7 normal adults have been examined. The hemoglobins of normal adult white and negro individuals were found to be indis- tinguishable. The mobility data obtained in phosphate buffers of 0.1 ionic strength and various values of pI1 are sum- marized in Figs. 1 and 2.5 4Occasionally small amounts (less than 5 percent of ihe total protein) of material with mobilities different from that of either kind of hemoglobin were observed in these uacrys- tnllized hemoglobin preparations. According to the ohserrn- tions of Stern, Reiner, and Silher (12) a small amount of a component with a mobility smaller than that of oa~hrmo- glohin is present in human erythrocyte hemolyzates. 5 The results obtained with carhonmonoryhemoglobins with and without dlthionite ion in the buffers indicate that the dithlonite ion plays no significant role in the electrophoretic properties of the proteins. It is therefore of interest that ferrohemoglohin was found to have a lower isoelectric point in phosphate buffer than carbonmonoxyhemoglobin. Titm- tion studies have indicated (5, 6) that oxyhemoglobin (siml- lar in electrophoretic properties to the cnrhonmonoxy corn- poonrl) has a lower isoelectric point than ferrohemoglobin itr November 25, 1949, Vol. 110 SCIENCE 545 The isoelectric points are listed iu Table 1. These rc- snlts prove that the electrophoretic difference between normal hemoglobin and sickle ccl1 anemia hemoglobin TABLE 1 Compound Nortnnl Sickle cell anemia Difference exists in both ferrohemoglobin and carbonmonoxy- hemoglobin. We have also performed several experi- ments in a buffer of 0.1 ionic strength and pH 6.52 containing 0.08 M NaCl, 0.02 M sodium cncodylate, and 0.0063 M cacodylic acid. In this buffer the average mobility of sickle cell anemia earbonmonoxyhemo- globin is 2.63 x lo-", and that of normal carbonmon- oxyhemoglobin is 2.23 x 1OP ctti/scc per T-olt/cm.6 a) Normal c) Sickle Ceil Trait b) Sickle Ceil Anemia d) 50-50 Mixture of a) and b) FIG. 3. Longsworth warning diagrums of carbonmon- oxyhemoglobins in phosphate buffer of 0.1 ionic strength and pH 6.90 taken after 20 hours' electrophoresis at R potential Rradient of 4.73 volts/c~n. These expcrilnents with a buffer quite different from phosphate buffer demonstrate that the difference be- tween the hemoglobins is essentially independent of the buffer ions. Typical Longsworth scanning diagrams of experi- ments with normal, sickle cell anemia, and sicklemia oarbonmonoxyhemoglobins, and with a mixture of the first two compounds, all in phosphate buffer of pH 6.90 and ionic strength 0.1, are reproduced in Fig. 3. It is apparent frotn this figure that the sicklcmia mate- rial contains less than 50 percent of the anemia com- ponent. In orclcr to determine this quantity accu- rately some experiments at a total protein concent,ra- the absrocr of other ions. These results might be reconciled bg assuming that the ferroos iron of ferrohemoglobin forms comple~c~s with phosphate ions which cannot be formed when the iron is combioed aitb oxygen or carbon monoxide. We propose to roni inoe the study of this phenomenon. "The mobility data show that in 0.1 ionic strength cacody- late buffers the isoelectric points of the hemoglobins are increased about 0.5 pH unit over their values in 0.1 ionic strenath phosphate buffers. This effect is similar to that observed by I,onusworth in his study of oralbumin (7). tion of 1 percent were performed with known mixtures of sickle cell anemia and normal carbonmonoxyhemo- globins in the cacodylate-sodium chloride buffer of 0.1 ionic strength and pH 6.52 described above. This buffer was chosen in order to minimize the anomalous electrophoretic effects observed in phosphate buffers (7). Since the two hemoglobins were incompletely resolved after 15 hours of electrophoresis under a potential gradient of 2.79 volts/cm, the method of Tiselius and Kabat (16) was employed to allocate the r-i I I I I J SICKLE CELL INEMIP. C*ReONMONOXYHEMoGLOBIN FIG. 4. The determination of the percent of sickle cell nncmin c;lrbonnlonosyhemoglot)in in known mtitures of the protein with norm:11 cnrbonmonoxShelnoglobin by means of electrophoretic nnnlysis. The esperiments were performed in a cncodylate sodium chloride buffer described in the text. areas under the peaks in the electrophoresis diagrams to the two components. In Fig. 4 there is plotted the percent of the anemia component calculated from the areas so obtained against the percent of that com- ponent in the known mixtures. Similar experiments were performed with a solution in which the hemo- globins of 5 sicklemic individmlls were pooled. The relative concentrations of the two hemoglobins were calculated from the electrophoresis diagrams, and the actual proportio& were then determined from the plot of Fig. 4. A value of 39 percent for the amount of the sickle cell anemia component in the sicklemia hemoglobin was arrived at in this manner. From the experiments we have performed thus far it appears that this value does not vary greatly from one sick- lcmic individual to another, but a more extensive study of this point is required. Up to this stage we hare assumed that one of the two components of sicklemia hemoglobin is identical with sickle cell anemia hemoglobin and the other is identical with the normal compound. Aside from the 546 SCIENCE November 25, 1949, Vol. 110 genetic evidence which makes this assumption very probable (see the discussion section), electrophoresis experiments afford direct evidence that the assumption is valid. The experiments on the pooled sicklemia carbonmonoxyhe~noglobin and the mixture containing 40 percent sickle cell anemia carbonmonoxyhemoglobin and MJ prrpeut norlllal carbonll~onoxyhemoglobin in the cacodylate-sodium chloride buffer described above were compared, and it was found that the mobilities of the respective components were essentially iden- ticaL7 Furthermore, we have performed experiments in which normal hemoglobin was added to a sicklemia preparation and the mixture was then subjected to electrophoretic analysis. Upon examining the Longs- wort,h scanning diagrams we found that the area under the peak corresponding to the normal component had increased by the amount expected, and that no indi- cation of a new component could be discerned. Sim- ilar experiments on mixtures of sickle cell anemia hemoglobin and sicklemia preparations yielded similar results. These sensitive tests reveal that, at least electrophoretically, the tlvo components in sicklemia hemoglobin are identifiable with sickle cell anemia hemoglobin ancl normal hemoglobin. 7) On the Nature of the Diffe+ence betawen Sickle Cell Anemia Hemoglobin and Normal Hemoglobin : Having found that the electrophoretic mobilities of sickle cell anenlia hemoglobin and normal hemoglobin differ, we are left with the considerable problem of locating the cause of the difference. It is impossible to ascribe the difference to dissimilarities in the par- ticle weights or shapes of the two hemoglobins in solu- tion : a purely frictional effect would cause one species to move more slowly than the other throughout the entire pH range and would no& produce a shift in the isoelectric point. Moreover, preliminary velocity ultracentrifuges and free diffusion measurements indi- cate that the two hemoglobins have the same sedimen- tation and diffusion constants.' The most plausible hypothesis is that there is a dif- ference in the number or kind of ionizable groups in the two hemoglobins. Let ts assume that the only groups capable of forming ions which are present in carbonmonoxyhemoglobin are the carboxyl groups in the heme, and the carboxyl, imidnzole, amino, phenolic hydroxyl, and guanidino groups in the globin. The number of ions nonspecifically adsorbed on the two proteins should be the same for the two hemoglobins 7 The patterns were very slightly different in thnt the kt~owtt mixture contained 1 percent more of the sickle cell anemia component than did the sickle cell trait mnterinl. C\Ve itre indebted to Dr. iIf. Moskowita, of the Chemistry Depnrtmrnt, University of California at Berkeley, for pcr- forming the ultracentrifuge experiments for us. under comparable conditions, and they may be neg- lected for our purposes. Our experiments indicate that the net number of positive charges (the total number of cationic groups minus the number of anionic groups) is great,er for sickle cell anemia hemo- globin than for normal hemoglobin in the pH region near their isoelectric points. According to titration data obtained by us, the acid- base titration curve of normal hunIan carbonmonoxy- hemoglobin is nearly linear in the nrighborhood of the isoelectric point of the protein, and a change of one pH unit in the hemoglobin solution in this region is associated with a change in net charge on the hemo- globin molecule of about 13 charges per molecule. The same value was obtained by German and Wyman (ci) with horse oxyhemoglobin. The difference in iso- electric points of the two hemoglobins under the con- ditions of our experiments is 0.23 for ferrohcmoglobin and 0.22 for the carbonmonoxy compound. This dif- ference corresponds to about 3 charges per molecule. With consideration of our experimental error, sickle cell anemia hemoglobin therefore has 24 more net positive charges per molecule than normal hemoglobin. Studies have been initiated to elucidate the nature of-this charge diffrrrncca Inow precisely. Samples of porphyrin dimethyl esters have been prepared from normal hemoglobin and sickle cell anemia hemoglobin. These samples n-erc' shown to be identical by their x-ray powder photographs and by identity of their melt~ing points and mixed melting point. A sample made from sicklemia hemoglobin was also found to have the same melting point. It is accordingly prob- able that normal and sickle cell anemia hemoglobin have different globills. Titration studies and amino acid analyses on the hemoglobins are also in progress. 2) On the Natwe of the Sickling Process: In the introductory paragraphs we outlined the evidence which suggested that the hemoglobins in sickle cell anemia and sicklemia erythrocytes might be respon- sible for the sickling process. The fact that the hemoglobins in these cells have now been found to be different from that present in normal red blood cells makes it appear very probable that this is indeed SO. We can picture the mechanism of the sickling process in the following l\Tay. It is likely `that it is the globins rather than the hemes of the two hemo- globins that are different. Let us propose that there is a surface region on the globin of the sickle cell anemia, hemoglobin molecule which is absent in the normal molecule and which has a configuration eom- plementary to a different region of the surface of the hemoglobin molecule. This situation would be some- what analogous to that which very probably exists in antigen-antibody reactions (9). The fact that siek- November 25, 1949, Vol. 110 SCIENCE 547 ling occurs only when the partial pressures of oxygen and carbon monoxide are low suggests that one of these sites is very nea; to the iron atom of one or more of the hemes, and that when the iron atom is combined with either one of these gases, the comple- mentariness of the two structures is considerably di- minished. Under the appropriate conditions, then, the sickle cell anemia hemoglobin molecules might be capable of interacting with one another at these sites sufficiently to cause at least a partial alignment of the nlolecules within the cell, resulting in the erythrocyte's becoming birefringcnt, and the cell membrane's being distorted to accommodate the now relatively rigid structures within its confines. The addition of oxygen or carbon monoxide to the cell might reverse these effects by disrupting some of the weak bonds between the hemoglobin molecules in favor of the bonds formed betwren gas molecules and iron atoms of the hemes. Since all sicklemia erythrocytes behave more or less similarly, and all sickle at a sufficiently low oxygen pressure (11), it appears quite certain t,hat normal hemoglobin and sickle cell anemia hemoglobin coexist within each sicklemia cell; otherwise there would be a laixture of normal and sickle cell anemia erythrocytes in sicklemia blood. We might expect that the normal hemoglobin molecules, lacking at least one type of complementary site present on the sickle cell anemia molecules, and so being incapable of entering into the chains or three-dimensional frameworks formed by the latter, would interfere with the alignment of these molecules within the sicklemia erythrocyte. Lower oxygen pressures, freeing more of the complementary sites near the hemes, might be required before suffi- ciently large aggregates of sickle cell anemia hemo- globin molecules could form to cause sickling of the erythrocytes. This is in accord with the observations of Sherman (11), which were mentioned in the introduction, that a large proportion of erythrocytes in the venous cir- culation of persons with sickle cell anemia are sickled, but that very few have assumed the sickle forms in the venous circulation of individuals with sicklemia. Presumably, then, the sickled cells in the blood of per- sons with sickle cell anemia cause thromboses, and their increased fragility exposes them to the action of reticula-endothelial cells which break them down, re- sulting in the anemia (1). It appears, therefore, that while some of the details of this picture of the sickling process are as yet eon- jeetural, the proposed mechanism is consistent with experimental observations at hand and offers a chemi- cal and physical basis for many of them. Further- more, if it is correct, it supplies a direct link between the existence of "defective" hemoglobin molecules and the pathological consequences of sickle cell disease. 3) 0% the Genetics of Sickle Cell Disease : A genetic basis for the capacity of erythrocytes to sickle was recognized early in the study of this disease (4). Taliaferro and Huck (15) suggested that a single dominant gene was involved, but the distinction be- tween sicklemia and sickle cell anemia was not clearly understood at the time. The literature contains con- flicting statements concerning the nature of the genetic mechanisms involved, but recently Neel (8) has re- ported an investigation which strongly indicates that the gene responsible for the sickling characteristic is in heterozygous condition in individuals with sicklemia, and homozygous in those with sickle cell anemia. Our results had caused us to draw this inference before Necl's paper was published. The existence of nornlal hemoglobin and sickle cell anemia hemoglobin in roughly equal proportions in sicklemia hemoglobin preparations is obviously in complete accord with this hypothesis. In fact, if the mechanism proposed above to account for the sickling process is correct, we can identify the gene responsible for the sicklin: process with one of an alternative pair of alleles capable through some series of reactions of introducing the modification into the hemoglobin molecule that dis- tinguishes sickle cell anemia hemoglobin from the normal protein. The results of our investigation are compatible with a direct quantitative effect of this gene pair; in the chromosomes of a single nucleus of a normal adult somatic cell there is a complete absence of the sickle cell gene, while two doses of its allele are present; in the sicklemia sotnatic cell there exists one dose of each allele; and in the sickle cell anemia somatic cell there arc two doses of the sickle cell gene, and a complete absence of its normal allele. Correspondingly, the erythrocytes of these individuals contain 100 percent normal hemoglobin, .40 percent sickle cell anemia hemoglobin and 60 percent normal hemoglobin, and 100 percent sickle cell anemia hemoglobin, respec- tively. This investigation reveals, therefore, a clear case of a change produced in a protein molecule by an allelic change in a single gene involved in synthesis. The fact that sicklemia erythrocytes contain the two hemoglobins in the ratio 40: 60 rather than 50: 50 might be accounted for by a number of hypothetical schemes. For example, the two genes might compete for a common substrate in the synthesis of two differ- ent enzymes essential to the production of the two different hemoglobins. In this reaction, the sickle cell gene would be less efficient than its normal allele. Or, competition for a common substrate might occur at some later stage in the series of reactions leading to the synthesis of the two hemoglobins. Mechanisms of this sort are discussed in more elaborate detail by Stern (13). _ 548 SCIENCE November 25, 1949, Vol. 110 The results obtained in the present study suggest anemias be examined for the presence of abnormal that the erythrocytes of other hereditary hemolytic hemoglobins. This we propose to do. 1. 2. 3. 4. 5. 6. 7. 5. Based on a paper presented at the meeting of the National Academy of Sciences in Washington, D. C., in April, 1.949, and at the meeting of the American Society of Biological Chemists in Ddroit in April, 1.949. BON, \V. Textbook of pathology. (::ld Ed.) I'hilx- 9. Pav1,xxir:. L., I'I:I$SSHAK, D., nnd C~ar~`l?i:~.r., I). IX. delphia: Lea and Febiger, 1938. P. stx. Phyeiol. Rev., 1943, 23, 203. Drccs, L. W., AHMXW, C. F., and BIUO, J. Ann. id. 10. PONDER, E. Ann.. N. Y. Acad. Sci., 1947. 48, 579. Med., 1933. 7, 769. 11. SHEKMAN, flopk. HOSP., I. J. RtrZZ. Jol~ns 1940, 67, 309. DRABKIN, D. L. J. biol. Chem., 1946, 164, 703. E~IMEI., v. E. Arch. 6111. Med., 1917. 20. .5S6. 12. STERN, 1;. G., ItEINER, iM. nnd SIr.n%X. R. II. J. biol. GEHYAN, B. nnd WXMAN, J., JR. J. biol. Chem., 1937, Chem, 1345. 161, 731. 117, 533. 13. STeRN, Science, 1948, 108, c. 615. HASTINGS. A. B. et al. J. biol. Clrcm., 1924, 60, 89. 14. SWINGLE, s. M. RFU. sci. Inst., 1947, 18, 1oK Lo~cswo~~a, L. G. Ann. N. Y. Acnd. Sci., 1941. 41, 13. TALI~~ER~O, I\`. 11. and Hucrc, J. G. Genetics, 1923, 267. 8, 59-l. NEEI,, J. V. Science, lW!J. 110. 64. 16. Trss~vs, A. nud HABAT, E. J. cap. Med., 1939, 69, 119.