THE SOLUBILITY CURVE AND THE PURITY OF INSULIN

THE SOLUBILITY CURVE AND THE PURITY OF INSULIN BY J. LENS (From the Organon Laboratories, Oss, Holland) (Received for publication, December 29, 1945) A method suitable for determining the degree of purity of insulin, which avoids the use of animals, would be a great help both for research and in the large scale production of the hormone. Of the many methods proposed not one has been universally recognized as equal in value to the biological assay, but the latter is so cumbersome, time-consuming, and inaccurate that the advantage of a quicker and still reliable method is beyond question. Kunitz and Northrop (1) were the first to adopt the extensive use of the solubility curve as a criterion of purity for proteins, a method originally used by Sgrensen and H$yrup (2), Landsteiner and Heidelberger (3), and Bonot (4). The demonstration of the purity of the protein by its solubility curve is based on Gibbs phase rule. With increasing quantities of solid phase present a pure protein will show a constant solubility, while a solubility varying with the amount of solid phase present is a clear indication of impurity. Kunitz and Northrop (1) give a method of calculating the composition of the system if the protein is not pure. This method has not yet been applied to the determination of the purity of insulin and it appeared valuable to us to see whether it might be of any use for this purpose. As is shown in this paper, the method is unreliable and cannot replace t.he biological assay in the case of insulin. EXPERIMENTAL All insulin used in this investigation was crystalline and prepared from beef pancreas in our own laboratories. The biological standardizations were carried out on rabbits. 60 to 80 animals were used for the standardization of each sample. The conditions for determining the solubility curve are as follows: The solubility of the protein under investigation should not be too high nor too low in the given medium, since an unduly high solubility requires large quantities of material and makes the test rather expensive, while, if the solubility is too low, the determination of the amount dissolved may be difficult. Preferably, the medium should be such that it does not interfere with the analysis of the dissolved fraction. The ph of the solution should be easily 223

221 SOLURILITY CURVE FOR INSULIX reproducible and quite constant over.a range of dilutions, since the solubility depends on it. A medium should be chosen so that the effect of temperature on the solubility is at a minimum. Equilibrium should be reached within a reasonably short time to minimize appreciable deterioration of the protein during the experiment. We found all of these conditions fulfilled to a large extent with insulin at its isoelectric point in a sodium acetate-acetic acid buffer solution. This buffer not only fixes the ph at the desired value but also makes the results independent of the small amounts of electrolyte possibly present as impurities in the insulin samples under investigation, which would otherwise interfere by increasing the solubility. The determination of the amount of protein in solution can be made by a simple micro-kjeldahl analysis of the filtrate. We found the following technique convenient and suitable for routine work. 75 mg. of the insulin are dissolved in 10 ml. of 0.1 N acetic acid, and to this solution are added 20 ml. of 0.1 N sodium acetate, giving a buffer solution of ph 4.95. Varying amounts of the suspension thus obtained are added to a series of 50 ml. Erlenmeyer flasks, each containing 25 ml. of a buffer solution of the same composition. The amounts of the suspension added were usually 0.5, 1, 2, 4, and 8 ml. After an hour to attain equilibrium the suspensions are filtered through an ashless filter paper of G cm. diameter, the filtrate being filtered through the same paper repeatedly until it becomes clear. Micro-Kjeldahl anal- yses are then made of the filtrates and of the original suspension. At first sight the time to establish equilibrium seems to be surprisingly short, but experimental evidence shows that it is sufficient. A run was made with five solutions of exactly the same composition, each one containing 25 ml. of buffer and 4 ml. of the suspension of a crystalline insulin powder. They were filtered and analyzed after equilibration times of 1, 3, 7,23, and 31 hours; the filtrates contained in all but the last instance 0.019 mg. of N per ml. After 31 hours we found 0.020 mg. of N per ml., an insignificant difference. The temperature has not much influence on the solubility, since a clear filtrate does not become cloudy when cooled from room temperature to the freezing point. Since it was not our intention to measure absolute values, we did not particularly fix the temperature by means of a t,hermostat. As all the solutions of one run were made simultaneously, they always were all at room temperature and a comparison within one run is justified. The choice of the pi-1 is of great importance. Xot only the level of the 1 Crystalline insulin usually does not dissolve quickly in this solution; sometimes it takes as much as 2 hours.

J. LENS 225 I curve is dependent on the ph but also its slope, the latter being essential for the determination of the percentage composition. WC chose ph 4.95 arbitrarily, because the solubilities at this ph have about the desired value and because a suspension of insulin crystals in water gives roughly the same ph. In Fig. 1, a and 6, the curves arc given for the solubilities of two samples of the same insulin estimated at ~1-14.95 and in a phosphate buffer of ph (3.0. In a phosphate buffer of ph 7.4 the insulin is completely in solution at all concentrations investigated. The calculation of the percentage composition depends on the type of the precipitate. If the components form a solid solution, the calculation of the 20 IO micrograms 50 N/ml. SUSPENSION lh. I, a - ph 4.Q5. PH 6.0 20 10 microgram5 hg. 1, 6 I So N/ml. SUSPENSION FIG. 1. Solubility curves of insulin nt. pi-14.95 and pii G.0: (a) Bntch 148, (6) Bate h,,,, omposition becomes impossible, since then the solubility of a.11 of the omponents is involved in the formula. Only if the components are a mixture is it possible to calculate the purity of the insulin preparations. Fig. 2 represents a theoretical solubility curve in the latter case with only one impurity present. Assuming insulin to be the less soluble component of the two, at B the solution is saturated with insulin. A suspension with the quantity AF per ml. contains accordingly the quantity EC impurities and AF - EC of insulin. Assuming that 1 I.U. equals 45 y of pure insulin, the activity of the insulin sample can be expressed as 45AF/(AF - EC) y per I.U. if the preparation is completely dry (1 y = 0.001 mg.). If it contains a per cent moisture, the result becomes 1 I.U. = 4500Ak / (100 - a)(& - EC) y. The value of 45 y for 1 I.U. of pure insulin will not be far from the truth and as a first approximation it will be sufficiently reliable, We have the impression, however, that the activity of absolutely pure insulin might be slightly greater, 1 I.U. in about 40 y.

226 SOLUBILITY CURVE FOR INSULIN Results An absolutely pure sample of insulin should give a completely horizontal solubility curve. Of the scores of samples investigated we found only one, that obtained without any special precautions by the method of Roman, Scott, and Fisher (5), to be absolutely pure. The solubility curve was at a very low level, the filtrates containing as little as 0.005 mg. of N per ml., and the suspension up to 0.08 mg. of N per ml. The biological act.ivity of this particular batch was about 40 y per unit. Owing to lack of material, we were able to standardize it on only twenty rabbits. We went to considerable trouble to reproduce this result but without success. We recrystallized some batches of rather well shaped crystals seven to ten times but, as judged by the slope of the solubility curves, they did not much improve by this treatment. Anyhow this one A FIG. 2. Theoretical solubility curve for insulin in the presence of a single impurity. The abscissa represents the amount of insulin per ml. added; the ordinate, the amount per ml. dissolved; solid line, curve for the impure batch; ABE, curve for pure insulin; AD, solubility of pure insulin. favorable result demonstrates that the possibility of obtaining a horizontal solubility curve is not a purely theoretical one. With impure samples the question of solid solution or mixture must be solved first. If it is a mixture, the concentration of insulin in the filtrate should be independent of the total quantity present; if it is a solid solution, the concentration varies according to Raoult s law. In two instances we found by standardization of the filtrates on rabbits that the concentration of insulin dissolved in the filtrates is roughly proportional to their nitrogen content. The ratio of the N content of the filtrates of two different suspensions of the same insulin was 1: 1.67; the ratio of their biological activities was 1: 1.7. In another instance we found for the N content a ratio of 1:2.0 and for the activities 1:2.5. The standardizations were performed on twenty rabbits each. These results therefore prove that insulin and the impurities form a solid solution. It is thus impossible to determine the percentage composition by means of the solubility curve without knowledge of the solubilities of each of the components. F

J. LENS 227 However, we observed the remarkable fact that in some instances, particularly for insulin prepared by the method of Gerlough and Bates (6) and crystallized from a phosphate buffer according to Scott (7), the system TABLE Activity of Insulin Preparations Results obtained by biological assay and by calculation from the solubility curve. Insulin obtained by the method of Gerlough and Bates and crystallized from phosphate-buffer solution, or by the method of Roman, Scott, and Fisher. I Batch No. I Biological test Chemical test Ratio of chemical to biological value 139 140 142 146 147 148 150 154 155 157 158 241 329 4B Average... Standard error. - Gerlough and Bates y per unit y fler unit 52 60 63 63 64 68 64 68 60 58 60 60 84 78 82 73 75 74 75 66 69 73 58 55 59 52 45 50...................................... Roman, Scott, and Fisher 1.15 1.06 1.06 1.06 0.97 1.06 0.93 0.89 0.99 0.88 1.06 0.95 0.88 1.11 0.99 0.09 235 40 45 240 60 70 244 43 60 251 65 70 256 61 70 266 50 I 73 Average... Standard error... - 1.11 1.16 1.40 1.08 1.15 1.46 1.23 0.16 can be treated as if the solid phase was a mixture and not a solid solution. The solubility curve is practically straight in the investigated range; i.e., up to 0.09 mg. of N per ml. The activity was calculated by the above formula with the quantity EC (Fig. 2) obtained by graphical means. The

228 SOLUBILITY CURVE FOR INSULIN ca.lculation has, of course, no theoretical foundation whatsoever but in practice it gives a good approximation of the activity. The results, presented in Table I, are almost convincing of the value of the methods as a first approach for estimating the activity. However, for insulin prepa.red by another method this rule of thumb failed completely and accordingly the usefulness becomes very doubtful for samples of unknown origin. In Table I are also summarized the results of our experiments with samples obtained by the method of preparation of Roman, Scott, and Fisher. Apparently the nature of the impurities is somewhat different from that of Gerlough and Bates preparations and this has a profound influence on the slope of the solubility curve which, by the way, again approaches a straight 1 50 micrograms N/ml. SUSPENSION FIG. 3. Solubility curves of Batches 142 and 154 of insulin before (solid line) and after (broken line) alkali inactivation. We can add another example of the limited value of the method even with crystalline insulin by the process of Gerlough and Bates. According to Freudenberg et al. (8) insulin can be completely inactivated by keeping it for 5 hours at 37 in l/30 N NaOII. Considerable quantities of hydrogen sulfide and ammonia are liberated during this treatment, but the slope of the solubility curve of the inactivated product is not appreciably changed. This is demonstrated in Fig. 3 for two different insulin samples. It is very unlikely that by treatment with alkali a single chemical entity results; it is much more probable that there will be more than one decomposition product. Still, so long as this has not been proved, there is a valid objection to any conclusion drawn from these experiments. If WC suppose that all of the insulin has been decomposed to one and only one other substance, the percentage purity of the decomposition product will be equal to the percentage purit,y of the insulin in the original sample.

J. LENS 229 Therefore samples which are only partly decomposed are of particular interest. We obtained these samples by mixing, in different proportions, an original sample with a fully decomposed one. The suspension of the decomposed insulin contained 0.047 mg. of N per ml., of the original 0.077 mg. of N per ml. With the acetate buffer of ph 4.95, mixtures of these two suspensions were made and the filtrates analyzed for?j content. The data obtained show that. substitution of insulin by its alkali decomposition product does not alter the total solubility appreciably. This (see Table II) is another indication that we are dealing with a solid solution. Thus TABLE Solubility of Mixtures sf Crystalline Insulin and Alkali-Inactivated Insulin at ph 4.95 Insulin per cewt 0 35 62 82 100 Decomposition product Suspension per celtt mg. N per ml. mg. N per ml. 100 0.047 0.015 65 0.050 0.019 38 0.060 0.020 18 0.068 0.021 0 0.077 0.024 II - Found Filtrate - Calculated mg. N per ml. 0.018 0.020 0.022 admixture of alkali-inactivated insulin with an active insulin sample is accordingly not demonstrable in this may. Denatured insulin is much less soluble than native insulin and, since it is insoluble in 0.1. N acetic acid, its presence is readily detected. The solubility in the buffer of ph 4.95 is very low indeed. We found values between 0.002 and 0.004 mg. of N per ml. in the filtrate. This is not due to slow equilibration, since the values after 1 and 24 hours, 0.003 and 0.002 mg. per ml. respectively, were identical within experiment,al error. The stability of the denatured insulin under these circumstances is remarkable. We expected to see an increase of the solubility after 24 hours, owing to a shift in the equilibrium to the native product. DISCUSSION Provided a specimen of insulin is pure, the solubility curve is a valuable criterion for demonstrating its absolute purity and in fact the only proof at present available. It is apparently very sensitive. For an impure batch the solubility curve gives only a clear proof that it is not absolutely pure, but the amount of impurity cannot be derived from the solubility curve with a sufficient degree of certainty. This is due to the formation of solid solutions of the insulin and t.he impurities, which is more or less

230 SOLUBILITY CURVE FOR INSULIN to be expected here, since the main impurities of insulin crystals will consist of closely related proteins with properties so similar to insulin that removal during the purification process has not been effected, while other impurities may also consist of decomposition products of insulin. There is still a remote possibility of obtaining a solubility curve of the mixed component type by altering the conditions for precipitation. Salting-out or precipitation with alcohol seemed a likely method, but experiments carried out with sodium sulfate or ethyl alcohol gave unsatisfactory results. As a routine test, instead of the ordinary biological assay for estimating the activity of insulin, the solubility curve seems unsuitable, but it remains of great value as a criterion for absolutely pure material. It might be desirable to have an absolutely pure international sta.ndard, though possibly the difficulty of obtaining sufficiently large quantities outweighs the advantage of reproducibility. The solubility curve would be a useful guide in the selection of batches suitable for this purpose. SUMMARY The solubility curve of insulin samples in a sodium acetate-acetic acid buffer of ph 4.95 has been determined. This curve is of the solid solution type and it is impossible to calculate exactly the purity of the samples from the results, since the solubility of the impurities is unknown. We succeeded once in obtaining an absolutely pure sample of crystalline insulin with a solubility independent of the amount undissolved, Occasionally the treatment of the system as being of the mixed crystal type gives a fair approximation to the percentage insulin of the sample but the method is unreliable and cannot be generally recommended. The solubility curve is not appreciably altered by alkali inactivation of the insulin. Thanks are due to Dr. C. L. Hewett for revision of the manuscript and to the Board of Directors of the PIT. V. Organon, Oss, for permission to publish these data. BIBLIOGRAPHY 1. Kunitz, M., and Northrop, J. H., J. Gen. Physiol., 13, 781 (1930); also in Cold Spring Harbor symposia on quantitative biology, Cold Spring Harbor, 6, 325 (1938). 2. Splrensen, S. P. L., and H$yrup, M., Compt.-rend. trav. Lab. Curlsberg, 12, 213 (1927). S@rensen, S. P. L., Compt.-rend. trav. Lab. Carkberg, 16, 1 (1925); 18, 1 (1930).

J. LENS 231 3. Landsteiner, K., and Heidelberger, M., J. Gen. Physiol., 6, 131 (1923). 4. Bonot, A., Ann. chim., 8,425 (1937). 5. Roman, R. G., Scott, D. A., and Fisher, A. M., Ind. and Eng. Chem., 32,508 (1940). 6. Gerlough, T. D., and Bates, R. W., J. Pharmacol. and Exp. Therap., 45, 19 (1932). 7. Scott, D. A., Biochem. J., 28, 1592 (1934). 8. Freudenberg, K., Dirscherl, W., and Eyer, H., Z. physiol. Chem., 187, 89 (1930).

THE SOLUBILITY CURVE AND THE PURITY OF INSULIN J. Lens J. Biol. Chem. 1946, 164:223-231. Access the most updated version of this article at http://www.jbc.org/content/164/1/223.citation Alerts: When this article is cited When a correction for this article is posted Click here to choose from all of JBC's e-mail alerts This article cites 0 references, 0 of which can be accessed free at http://www.jbc.org/content/164/1/223.citation.full.h tml#ref-list-1