Purification of Human Erythropoietin*

Human erythropoietin, derived from urine of patients with aplastic anemia, has been purified to apparent homogeneity. The seven-step procedure, which included ion exchange chromatography, ethanol precipitation, gel filtraCon, and adsorption chromatography, yielded a preparation with a potency of 70,400 unitslmg of protein in 21% yield. This represents a purification factor of 930. The purified hormone haa a single electrophoretic component in polyacrylamide gels at pH 9, in the presence of sodium dodecyl sulfate at pH 7, and in the presence of Triton X-100 at pH 6. Two fractions of the same potency and molecular size, by sodium dodecyl sulfate gel electrophoresis, but differing slightly in mobility at pH 9, were obtained at the last step of fractionation. The nature of the difference between these two components is not yet understood.

Human erythropoietin, derived from urine of patients with aplastic anemia, has been purified to apparent homogeneity. The seven-step procedure, which included ion exchange chromatography, ethanol precipitation, gel filtra-Con, and adsorption chromatography, yielded a preparation with a potency of 70,400 unitslmg of protein in 21% yield. This represents a purification factor of 930. The purified hormone haa a single electrophoretic component in polyacrylamide gels at pH 9, in the presence of sodium dodecyl sulfate at pH 7, and in the presence of Triton X-100 at pH 6. Two fractions of the same potency and molecular size, by sodium dodecyl sulfate gel electrophoresis, but differing slightly in mobility at pH 9, were obtained at the last step of fractionation.
The nature of the difference between these two components is not yet understood.
Erythropoietin is an acidic glycoprotein that is present at a very low concentration in plasma under normal conditions. Under anemic or anoxic stress, it is found in relatively large amount in the plasma and is also excreted in the urine. Erythropoietin is the substance that is responsible, in large part, for the regulation of normal red blood cell differentiation. Because of this function, and because it may have a role in replacement therapy of some kinds of anemia, it is important ta have pure erythropoietin in an amount sufficient for chemical characterization.
Reports on the purification of human (1) and sheep (2) erythropoietin have been published.
In the former, the evidence for homogeneity was not convincing, and in the latter, the t&d amount was too low for adequate characterization.
We report in this paper on the preparation of milligram quantities of human urinary erythropoietin in a state of apparent homogeneity.
EXPERIMENTAL PROCEDURES Bioassay-The fasted rat method of bioassay (3), in which the incorporation of labeled iron into circulating red cells is measured, was used routinely to quantitate the amount of erythropoietin activity. Samples for assay were dissolved in 0.1% bovine serum albumin in 0.15 M NaCl, 0.01 M CaCl,. Over the la-month period covered by this work, the In dose-ln response curve obtained when 1, 1.5,2, and 3 units of erythropoietin'lrat were used had the following characteristics: slope, 1.11 + 0.34; intercept, 0.75 * 0.39; correlation coefficient, 0.96 * 0.10. The assay values found for the two final hydroxylapatite fractions were confirmed by the polycythemic mouse method (3) (acrylamide, N,N,N',N'-tetramethylethylenediamine and N,N'methylenebisacrylamide) and  (7).

RESULTS
All of the urine concentrates were treated with phenol paminosalicylate, as described by Chiba et al. (8), so that the loss of activity due to enzymic degradation was reduced. This procedure was carried out on 18 batches which consisted of a total of 7,059,670 units" and a mean potency of 91 ulA (range, 16 to 160). There were 5, 115,110 units recovered, with a mean potency of 109 u/A. In spite of the fact that 28% of the activity was lost and the mean purification factor was only 1.20, it was necessary to use this technique to avoid major losses later in the purification process. The purification method described below was developed as a result of many trials of various standard techniques.
For example, we found that use of gel permeation chromatography early in the procedure did not lead to any significant purification, probably due to the large amount of glycoprotein with similar sizes in the crude urine concentrate; stepwise elution of ion exchange columns was used throughout the procedure since we found that gradient elution decreased the resolution.
Ethanol Precipitation -Sixteen separate batches were precipitated with ethanol by the following procedure. The sample, e.g. 111,600 units at 52 u/A, was dissolved in 50 ml of PBS at 4"; 5 ~1 were removed for assay, and 12.5 ml of 10 M LiC14 were added. Absolute ethanol (62.5 ml) at 4" was added slowly with stirring, which was continued for 30 min after the addition * Potency, or specific activity, is expressed as units of biological activity (u) per absorbance unit (A), measured at 278 nm in l-cm cuvettes.
3 This figure is slightly different (1.2% higher) from that indicated as the amount obtained from the DEAE-cellulose step. This kind of difference is caused by the uncertainty in the bioassay and will also be seen at subsequent steps.
' LiCl was used in the alcohol precipitation procedure in order to increase the solubility of proteins in ethanol (9). Precipitation in the absence of salt resulted in a low potency fraction.
was complete. After the flocculent precipitate had been allowed to settle for 15 min, it was removed by centrifugation at 21,000 x g for 10 min at -15". The pellet was washed three times with 10 ml of 50% ethanol, 1 M LiCl and the supernatants were pooled. The washed precipitate was dissolved in 20 ml of PBS, yielding a turbid solution (50% precipitate).
Sixty-seven milliliters of absolute ethanol were added slowly to the combined supernatants; stirring was continued for 30 min and settling for I5 min. The precipitate was collected as before and washed twice with 10 ml of 65% ethanol, 0.7 M LiCl, and the supernatants were pooled. The washed precipitate was dissolved in 20 ml of PBS (65% precipitate).
To the pooled supernatants, 96 ml of ethanol were added slowly, and stirring was continued for 30 min, after which the precipitate was allowed to settle for 14 h at 4". We have found that this long period in 75% alcohol is required for optimal further fractionation.
The precipitate was washed twice with 10 ml of 75% ethanol, 0.5 M LiCl, the supernatants were pooled, and the precipitate was dissolved in 20 ml of PBS (75% precipitate).
The combined supernatant was brought to 90% ethanol by addition of 540 ml of absolute alcohol, stirred for 30 min, and stored at -20" for 48 h before the precipitate was collected, dissolved in 50 ml of cold water, and immediately frozen (90% precipitate).
The results of one representative ethanol fractionation procedure are given in Table I

DEAE-Agarose Fractionation
-The solution, in water, of a 90% ethanol precipitate was concentrated to about 5 ml on an Amicon UM-10 ultrafllter, then brought to 25 ml with 0.01 M Tris, pH 7.0, and a 50-~1 aliquot was removed. The DEAEagarose, 100 to 200 mesh, was degassed under reduced pressure, suspended in 0.01 M Tris, pH 7.0, and packed into a column 9.2 x 2.5 cm in diameter (bed volume, 45 ml). The gel was washed with 1.5 liters of 0.01 M Tris, pH 6.9; the ratio of absorbance units added to bed volume (ml) was 6.65. The sample was added to the column over a period of 40 min, and 15O-drop fractions were collected. The column was washed with 211 ml of 0.01 M Tris, pH 7, and then eluted with the following buffers: 366 ml of 0.01 M Tris, pH 7.0; 5 mM CaCl*; 270 ml of 0.01 M Tris, pH 7.0; 17 mM CaCl,; 194 ml of 0.01 M Tris, pH 7.0; 30 mM CaCl,; and 65 ml of 0.1 M CaCl,. The elution pattern can be seen in Fig. 1, and the results are given in Table II.
Of the 4,566,240 units of total input, we recovered 4,052,710 (89%) in the 17 mM CaCl, eluate at a mean potency of 1,110 u/ A, representing a mean purification factor of 1.97. From this point on in the fractionation calcium was added to all buffers except those used with hydroxylapatite columns because there were inconsistent results and appreciable losses of activity when buffers without calcium were used. For the next step in purification, we selected three eluates from DEAE-agarose columns, amounting to 2,480,400 units (61% of the total yield) with a mean potency of 1,750 u/A.

Sulfopropyl-Sephadex Chromatogmph
-The three eluates (17 mu CaCl& from DEAE-agarose columns were desalted and concentrated on a UM-10 ultrafllter and then dialyzed against 2 liters of 5 mM CaCl,, pH 7.5, overnight. In the sample run described below, 30 ml of dialyzed solution were brought to pH 4.50 by dropwise addition of 0.1 M HC1; the small amount of precipitate formed was removed by centrifugation and washed with 5 ml of 6 mM CaCl,, pH 4.5. The wash, pooled with the supernatant, was applied to a sulfopropyl-Sephadex column (15.0 x 2.5 cm in diameter; bed volume, 78.3 ml) which had been equilibrated with 5 mM CaCl,, pH 4.50. The absorbance units to bed volume (ml) ratio was 2.47. We found that a low value for this ratio is critical for optimal fractionation on sulfopropyl-Sephadex; for example, if the absorbance unit to bed volume ratio was greater than 10, almost all of the activity was found in the effluent fraction. The following buffers were used in developing the column. Input was: 5 mM calcium acetate, pH 4.50, specific conductivity = 1,075 pmho cm-'. Eluting buffers were: 7.5 mM calcium acetate, pH 4.70, specific conductivity 1,500 pmho cm-'; 12.5 mM calcium acetate, pH 5.25, specific conductivity = 2,100 pmho cm-'; 15 mM calcium acetate, pH 5.5, specific conductivity = 2,400 pmho cm-'; 0.1 M calcium acetate, 0.01 M Tris, pH 7.24, specific conductivity = 11,500 pmho cm-'. The column was run at 0.4 ml/min at 4", and 2OOdrop fractions were collected. After a reading was taken at 278 nm and the appropriate pools were made, the solutions were neutralized (within 1 h a&r elutionl, and aliquots were removed for assay and stored at -20". The elution pattern is presented in Fig. 2 and results of the fractionation are shown in Table III. The overall results of this step in the purification were: 55% recovery (1,352,810 units) in the 12.5 mM calcium acetate, pH 5.55 fraction, at a mean potency of 11,170 u/A, and with a mean purification factor of 6.38.
Gel Filtration -The 12.5 and 15 mM calcium acetate eluates from the sulfopropyl-Sephadex column separations were run in two separate batches on the same gel column. The pools were concentrated on Amicon UM-2 ultrafilters to about 5 ml and equilibrated with 10 mM CaCl,, 10 mM Tris, pH 6.87, before application to the column. The Sephadex G-100 gel was degassed under reduced pressure and equilibrated with the same buffer before the column was poured. The column (100 x 2.5 cm diameter) was calibrated with markers of known molecular size before being used for the erythropoietin fractions. The void volume was 135 ml; bovine serum albumin monomer eluted at 224 ml, ovalbumin at 258 ml, and cytochrome c at 368 ml. The sample was added to the bottom of the column, as was the buffer which was passed through the column at 21 to 22 ml/  h by means of a Mario& bottle with a 42-cm hydrostatic head. Each fraction collected was 4.1 ml (120 drops), and the following pools were made: I, 0 to 131.2 ml; II, 131.2 to 184.5 ml; III,184.5 to 205 ml;IV,205 to 258.3 ml;and V,258.3 to 328 ml (Fig.  3). The first four pools were concentrated by ultrafiltration and aliquots were assayed. In one of the runs, pools I and II contained 17% and 5% of the absorbance units, respectively, but no detectable activity; pool III contained 32% of the absorbance units and 104% of the input activity, yielding a fraction with a potency of 38,850 u/A; and pool IV contained 10% of the absorbance units and 2% of the biological activity. Pool V was not assayed.
For the combined two gel filtration runs, the yield in pool III (184.5 to 205 ml) was lOO%, the mean potency was 39,080 u/A, and the purification factor was 3.04. Hydroxylapatite Chromatography -Hydroxylapatite was packed under unit gravity into a column (6.1 x 1.5 cm diameter) and washed with 500 ml of water and then with 400 ml of 0.5 mM phosphate buffer, pH 7.1, conductivity = 69 pmho cm-' (Buffer I), by use of a peristaltic pump which maintained the flow at 0.3 ml/min. After the buffer wash, the length of the column was 3.4 cm and the bed volume was 6.0 ml. The input sample was concentrated and desalted on an Amicon DM-5 ultrafilter by adding water to the concentrate and reconcentrating three times. The final concentrate and the wash of the filter were centrifuged at 6,000 x g for 20 min at 4". The small insoluble pellet was washed once with 0.5 mM phosphate, pH I. 1, and the wash was added to the supernatant.
An aliquot for assay was removed and the remainder (22 ml) was added to  the column. The ratio of absorbance units added to bed volume (ml) was 1.82. The input buffer was pumped through the column until the eflluent A,,, was less than 0.005 (149 ml), and the following elution schedule was carried out: Buffer II, 1 mM phosphate (pH 7.1, specific conductivity 131 = pmho cm-', 150 ml (Fraction II)); Buffer III, 2 mM phosphate (pH 6.9, specific conductivity = 270 pmho cm +, 220 ml (Fractions IIIA and IIIB)); Buffer IV, 3 mM phosphate (pH 6.9, specific conductivity = 402 pmho cm-', 84 ml (Fraction IV)); Buffer V, 0.1 M phosphate (pH 6.8, specific conductivity = 9.6 mmho cm-l, 134 ml (Fraction VI). The elution pattern is shown in Fig. 4; the results for one such column are listed in Table IV. The total input for the two runs was 1,083,650 units, with a mean potency of 38,770 u/A. The total recovered in Fractions II and IIIA was 721,163 unite (67%) with a mean potency of 82,720 u/A and a mean purification factor of 2.13. Each of the fractions, II and IIIA, from the two experiments was concentrated by means of Amicon DM-5 ultraiilter and stored frozen. When we examined Fractions II and IIIA from the two hydroxylapatite columns by gel electrophoresis in SDS5 (7.5% gels), we found single bands, each with a relative mobility (with reference to the Pyronin Y band) of 0.50. No detectable difference in mobility, in the presence of SDS, between Fractions II and IIIA could be found. Fig. 5 shows the SDS-gel electrophoretic analysis of each of the most active fractions throughout the purification procedure, and Table V summarizes the seven-step method. The overall purification factor was 929 (calculated from initial and final potencies).
Since these fractions further for evidence of heterogeneity. When these fractions were compared by gel electrophoresis at pH 9, it was clear that there was a small, but significant difference in mobility. Fraction II had a mobility relative to the bromphenol blue tracking dye of 0.49, and the value for Fraction IIIA was 0.52. In spite of our finding of similar -potency and molecular size, these two preparations must be considered different. The chemical basis for this difference is now being studied.
Fraction II was run on two gels at pH 9; one was fixed, stained, and scanned, and the other was sliced into 1.1~mm slices which were put into 0.5 ml of 0.10% bovine serum albumin, 10 mM CaCl* in 0.15 M NaCl, and the hormone was allowed to diffuse out of the gel at 4" for 18 h. On assay by the in vitro method, we found the biological activity coincident with the single band of stained protein (Fig. 6).
In view of our previous finding that native sheep erythropoietin was very poorly fixed to polyacrylamide gels and was largely lost during the staining procedure, we adopted the expedient used earlier for the sheep hormone. Fraction II was iodinated with lz51, run on a gel at pH 9 which was then cut into 1. l-mm slices, and counted before and after fixation. The results in Fig. 7 show a single peak of labeled hormone, only a fraction (44%) of which was fixed. The iodinated hormone was then run on a gel at pH 6 in order to confirm the apparent homogeneity. It became clear that there was a large degree of aggregation at the lower pH, since only a small amount of the radioiodine could be found in the gel, with the major fraction remaining at the origin. We then used the observation of Kawasaki and Ashwell (101,who found that aggregation of a liver glycoprotein could be reduced by the use of Triton X-100. When both the native and asialo forms of erythropoietin were run on gels in the presence of 0.05% Triton X-100 (Fig. 81, we found for the former a single symmetrical peak and for the Examination of iodinated Fraction II, both native and asialo, on SDS gels (11) showed single, symmetrical peaks ( Fig.  9) with no evidence of heterogeneity with respect to size.
We measured the absorbance at 278 nm and at 191 nm, using crystalline bovine serum albumin as a standard and correcting for stray light at 191 run, and found that A.:$, for erythropoietin is 8.51. Using this value, the mean potency of homogeneous human erythropoietin can be expressed as 70,400 units/mg of protein.
Espada and Gutnisky (1) isolated a fraction, from urine of patients with anemia due to hookworm, that had a potency of about 8,000 unit.s/mg of protein. They claimed, on the basis of a gel permeation experiment, that this fraction was homogeneous; the poor resolution characteristic of this method of analysis, however, makes it necessary to use additional kinds of information to establish purity. In a subsequent paper, Espada et al. (12) claimed that the same preparation was homogeneous by gel electrophoresis at pH 9, although they pointed out that the stained band was diffuse. In addition, these authors showed an immunodiffision pattern that was inconclusive with respect to immunological homogeneity. Our finding that human erythropoietin has a minimal potency of 70,400 units/mg of protein suggests that Espada's preparation is either about 11% pure or, ifit is homogeneous, is largely in the asialo form that has no activity in Go.
The gel concentration was 10%.
Our previously reported data (2) indicated that the preparation of sheep plasma erythropoietin, with a potency of 9,200 ul A, was free of any contaminant except for a small amount of asialoerythropoietin. If this is truly the case, then human urinary erythropoietin is '7 to 8 times more active than the sheep hormone, when assayed by the same method. This may be due to a greater sensitivity of rata to the human than to the sheep hormone, or it may indicate that human urinary erythropoietin is intrinsically more active than sheep plasma erythropoietin.

4-
The appearance of two fractions with the same potency, as a result of hydroxylapatite fractionation, suggests a degree of heterogeneity which is not, detected upon electrophoresis in SDS, and which might be accounted for by a small difference in the number of terminal sialic acids or of amide groups, or of both. Our findings of single peaks upon electrophoresis at pH 9, pH 6, and pH 7 in SDS constitute reasonable evidence of homogeneity with respect to charge and molecular size for each of the two fractions. At pH 6 in the presence of Triton X-100, the native and asialo forms are clearly separated (Fig. 9), and we could expect to be able to detect an appreciable amount of the latter mixed with the former. With the exception of the Rc. 9. Gel electrophoresis (pH 6, 0.05% Triton X-100) of IzaIlabeled erythropoietin (hydroxylapatite Fraction II). &@ represents native erythropoietin; 0---0 represents asialoerythropoietin. The gel concentration was 10%. 2'. D., tracking dye. small amount of native erythropoietin found in the asialo 1 preparation, both forms appear to be homogeneous. Without added surfactant, there is a considerable tendency for native and asialo erythropoietin to aggregate at pH 6 and for the asialo form to aggregate at pH 9, but at pH 7 in the presence of SDS and dithiothreitol, both forms appear to be monomeric. The human asialo hormone has an apparent molecular weight of 34,000 in SDS, whereas the native form has an apparent molecular weight of 39,000.g These values contrast with the molecular weight of 41,000 found for sheep plasma asialoerythropoietin by the SDS-gel electrophoretic method and the calculated value of 46,000 for the native form latter, one major symmetrical peak with only a trace of the native hormone contaminating the asialo form. At a lower concentration of T&on X-100 (O.Ol%), there was still appreciable aggregation, as detected by label that remained at the OI-igiIl. 6 The molecular weight determined by SDS-gel electrophoresis was the same whether we used 7.5% or 10% gels. of the sheep hormone (13). When we studied the sheep hormone, it was clear that the Weber and Osborn method of molecular weight determination (11) by gel electrophoresis in SDS was not accurate for the fully sialylated hormone, possibly because of a substantial contribution by the sialic acids to the net charge. At present, we cannot estimate the molecular size of the native human hormone from that of the asialo form since we do not yet have an accurate estimate of the sialic acid content.
The method of iodination of erythropoietin deserves comment. We found with the sheep hormone that the unmodified k&nation method (51, in which chloramine-T is used, caused total loss of biological activity. Following the precedent set by Stanley and Metcalf (141, we used a modification of the method of Stagg et al. (6) in which the iodination was carried out in 45% dimethylsulfoxide to protect methionine residues from oxidation. In contrast to the findings of Stagg et al., who found no loss of gastrin activity, and those of Stanley and Metcalf who found no loss of colony-stimulating activity, we found that the iodination method does cause appreciable inactivation of erythropoietin.
For the case of the preparation with 4 iodine atoms/molecule, it can be calculated that less than 2% of the hormone would be noniodinated.
This would suggest that the 25% of the biological activity that remained was due to the iodinated derivative. Until a method can be found for preparation of a fully active, labeled hormone, this less active, labeled erythropoietin may still be useful for the study of a number of biological characteristics of the hormone which, until now, have not been amenable to experiment.