Proceedings of the National Academy of SciencesVol. 66, No. 4, pp. 1213-1219, August 1970

Proinsulin: Crystallization and Preliminary X-RayDiffraction Studies*

W. Wardle Fullerton, Reginald Pottert and Barbara W. LowtCOLLEGE OF PHYSICIANS AND SURGEONS, COLUMBIA UNIVERSITY,

NEW YORK, NEW YORK

Communicatcd by Erwin Chargaff, Mlay 15, 1970

Abstract. Bovine proinsulin has been crystallized under a variety of condi-tions at both neutral and acid pH. Microtechniques were employed with sampleweights of about 200 jig and volumes of 5-20 1AL The crystalline preparations alldiffer from each other morphologically.X-ray photographs of one form, tetragonal bipyramids grown at pH 3 with

added ammonium sulphate solution, established the space group P41212 (or itsenlantiomorph P43212). The cell dimensions are a = 50.8 i 0.2 A, c = 148.0 +0.4 A. The asymmetric unit in this form is a dimer of proinsulin which is alsothe dominant species in solution at this pH.

Introduction. Following the studies of Steiner arid his co-w%-orkersl2 who estab-lished that insulin is synthesized via a single chain precursor protein, proinsulinhas been isolated and characterized from bovine,3-6 porcine,' cod,8 anglerfish,9and human10"1 tissues. The connecting peptide segment of proinsulin whichlinks the carboxyl-terminal group of the insulin B chain to the amino-terminalgroup of the insulin A chain is cleaved in vivo by proteolysis. The lengths andsequences of the C-peptide vary markedly between species. In bovine pro-insulin, where the connecting peptide is 30 amino acid residues long," the pro-insulin molecule with a single 81 residue chain has a molecular weight of 868.5.

Physicochemical studies of porcine proinsulin in solution'2"3 suggest that theaggregation of proinsulin molecules is comparable to that observed with insulin,both as a function of pH and in terms of interaction with zinc. The concentra-tion dependence of the monomer-dimer equilibrium in acid solution has been in-vestigated. At neutral pH in the presence of zinc proinsulin forms hexamers,although in the absence of zinc, the solutions are heterogeneous with respect toparticles in the molecular weight range 22,000-58,000. Jeffrey and Coates"4reported that for insulin in the absence of zinc there is a mixture of even-ag-gregate states. Comparable solution studies of bovine proinsulin have not beenmade; the behavior of bovine proinsulin in solution, however, is probablyanalogous to that of porcine proinsulin.From studies of optical rotatory dispersion and circular dichroism spectra,

Frank and Veros12 have concluded that the connecting peptide of porcine pro-insulin probably exists in a random coil conformation. Their studies are con-sistent with the view that the conformation of the insulin segment in proinsulinis very similar, if not identical, at least so far as the regions of a-helix are con-cerned, to the conformation of insulin as an isolated molecule. Studies in this

1213

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

1214 BIOCHEMISTRY: FULLERTO\N, POTTER, AND LOWV Ptoc. N. A. S.

laboratory,"5 on the other hand, of published sequences for both porcine andbovine proinsulin employing criteria developed here'6 have suggested that thereare probably two short regions of a-helix in both of the connecting peptides.We began crystallization studies with a bovine proinsulin preparation4 given

to us by Prof. Steiner. The material is salt free; if it contained zinc, this wouldimply a strong atypical (unlike insulin) zinc-binding capacity. We were awareof some similarities in solution properties between insulin and proinsulin andsought to exploit them. In particular, salt (sodium chloride) was employed toenhance ionic strength; a phenol' was also used in some studies. Only 8 mg ofproinsulin was initially available. M\licrotechniques were required not, onlyfor the crystallization studies but also for the preliminary solution studies madeto determine proinsulin solubility over a broad range of pH and salt concentra-tion. In order to enhance the possibility that the material could be crystallizedwe chose deliberately to explore several different sets of would-be optimal con-ditions. In all cases, crystals were obtained.Experimental Procedures. Preliminary solution studies suggested that proinsulin

is more soluble than insulin and we were able to prepare solutions at neutral pH con-taining as much as 20-30 mg/ml and at pH 2-3 over 50 mg/ml. Proinsulin is less solubleat 30C than at room temperature and crystallization proceeds more rapidly in the cold.Salt concentrations appropriate to the preparation of proinsulin crystals were determinedby preliminary diffusion experiments (see below). Survey studies were all performed atroom temperature and the crystals which appeared, usually after about 4 weeks, werevery small and ill-formed. In consequence, although some solutions were made up andmixed at room temperature, all subsequent crystallization studies were made in a coldroom at approximately 30C. Mlost of the crystallization studies were conducted in thin-walled (10 Am) glass or quartz capillaries, 1 mm in diameter. These are normally usedto mount protein crystals for x-ray study. The glass capillaries must be washed in acidand subsequently rinsed in distilled water to remove surface alkali. Some of the capil-laries were coated with a bovine serum albumin solution (10 mg/ml in 50% acetic acid)to prevent interaction of proinsulin at the glass wall. The capillaries are approximately60 mm long with a further thicker-walled flared end. The proinsulin preparations wereusually made up at room temperature and allowed to stand in the cold (about 30C).Proinsulin was brought out of aqueous solution by one or more of several effects: tem-perature-dependent solubility, salting out, pH change, or slow evaporation of the solvent.Two variations in crystallization procedure were initially employed: (a) The capillary

was heat-sealed at the thin-walled end. Solutions were added with a syringe, mixedusing a micro-glass stirring rod, and finally sealed with a plug of embedding wax. Here,as elsewhere, the capillaries were maintained vertical at all times. (b) The narrow thin-walled end of the capillary was sealed with acrylamide gel, using the procedure of Zep-pezauer et al.1' The proinsulin solution was introduced into the capillary as before, thenarrow end of the capillary was then immersed in a glass container of precipitating solu-tion which diffuses slowly through the gel. Once some crystalline material had formed,the capillary was removed from the container to halt diffusion and sealed at both endswith embedding wax. Glass apparatus (including glass syringe needles) was employedthroughout except that steel syringe needles were used when a Swinny filter was employed.Sometimes material to be employed in one crystallization (of the order of 200/Ag to be

dissolved in about 5-20 Al) was weighed out separately in the capillary tube. Stock solu-tions prepared with 1 mg of proinsulin in 25-100 Mil were also employed. These solutionswere usually filtered through a Jena sintered-glass crucible (IG4-porosity 5-10 Mm) orsometimes filtered through a Swinny filter.

Crystallization of proinsulin was studied under the following conditions: (1) In phos-phate buffer at pH 7.25: Proinsulin (10-40 mg/ml) was dissolved in 0.0133 M phosphate

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

VOL. 66, 1970 BIOCHEMISTRY: FULLERTON, POTTER, AND LOW12

buffer, which contained 22 mg/ml sodium chloride and also, when a phenol was employed,3 mg/ml of m-cresol. The capillary tubes were filled at room temperature and then re-moved to the cold room.

(2) In citrate buffer at pH 7.25: Proinsulin (10-20 mg/ml) was dissolved in 0.05 Mcitrate buffer containing 22 mg/ml sodium chloride with 0.04% zinc as zinc chloride and,when a phenol was used, 3 mg/ml m-cresol. The capillary tubes were filled at roomtemperature and then removed to the cold room.

(3) In citrate buffer at approximately pH 3: With the exceptions noted, all manipula-tions were conducted in the cold. Proinsulin (30-40 mg/ml) was dissolved in 0.1 -Mcitric acid solution containing 1.75% potassium chloride. Cold 0.08 M di-ammoniumcitrate solution or cold-saturated ammonium sulfate solution was either added to slightturbidity or permitted to dialyze through acrylamide gel until a few crystals appeared.The tubes were then sealed at both ends with embedding wax. A few experiments wereperformed by the initial addition of the ammonium sulfate solution at room temperaturefollowed by removal of the preparation to the cold room.The first x-ray studies of proinsulin crystals were made without removing them from

the thin-walled glass capillaries in which they were grown. In order to cut down excessbackground scattering, most of the mother liquor about the crystals to be photographedwas removed with a microsyringe and only a meniscus of liquid left. Copper radiationfrom a conventional normal-focus GE x-ray source was employed. At a later stage ofx-ray study, crystals were removed from the vessels in which they were grown and mountedin the usual way. X-ray precession photographs were taken with copper radiation froma Jarrell Ash microfocus unit. A pinhole collimator was used to limit the beam to adiameter of about 150 Am. The densities of two crystals which had been used for x-raystudy were measured at 230C in a water-saturated bromobenzene-xylene gradient columnmade up to cover the calculated range of possible densities.

Results. In the studies with phosphate buffer those crystals which appearedafter 1 week in the presence of m-cresol were essentially sphenoids, with somecurved high-order faces still evident; some crystals were truncated (see Fig.IA and B). The crystals did not grow larger on further standing; the largestspecimens observed were about 50 X 30 X 10 ,m. In the absence of m-cresol,crystals appeared after standing for 3 weeks. These were well-formed nonequantthree-sided plates 30 Mm in average diameter and 10 Mm thick and showed obliqueextinction.

Crystals grown from citrate buffer in the presence of m-cresol appeared after1 week. They were thin blades of wedge-shaped cross section, about 10-20Mm long. (Fig. iC). In the absence of m-cresol intergrowths of ill-formedplates, of approximately square (20 Am) cross section. appeared after 3 weeks(Fig. ID).

In the studies at acid pH in citrate buffer, the use of ammonium citrate led tothe separation after about 3 weeks of small thin rectangular plates 10-20 Mmin length, which show parallel extinction. The crystals did not grow on furtherstanding. Those crystals obtained from ammonium sulphate solutions ap-peared in 2 days and reached maximum size after 1 week. They were well-formed, bipyramids, about 90 X 40 X 20 Am, and extinguished parallel to thelong axis (Fig. 1E). X-ray photographs of this form taken with a normal focussource were poor and inadequate to establish unit cell dimensions or space group.

In all the preparations described, the crystals immersed in their motherliquor were stable in the cold for a period of more than 10 months and at roomtemperature for at least several weeks. A second-run crystallization was

12154

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

BIOCHEMISTRY: FULLERTON, POTTER, AND LOW Pitoc. N. A. S.

F J-1 -m, Ia

FIG. 1.-Crystals of bovine proinsulin:(A) and (B) From phosphate buffer at pH 7.25 containing m-cresol.(C) From citrate buffer at pH 7.25 containing m-cresol and zinc.(D) From citrate buffer at pH 7.25 containing zinc.(E) and (F) From citrate buffer at pH 3 containing ammonium sulphate.

attempted on the same scale and under similar conditions with a further 5 mg ofmaterial in order to establish that the crystallization procedures were repro-ducible. There were no differences in time of appearance, growth rate, or finalsize of the crystals obtained in this second study.

Larger scale crystallization was evidently necessary for the preparation of

1216

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

VOL. 66, 1970 BIOCHEMISTRY: FULLERTON, POTTER, AND LOW*

bigger crystals and the m-cresol-free preparations were combined and chromato-graphed on Bio-Gel P-2 with 0.1 'M acetic acid at room temperature according tothe procedure of Steiner.'9 The proinsulin, the first material to come off thecolumn and identified by its absorbance at 275 nm with a Beckman-DB spectro-photometer, was freeze-dried.

Because acid crystallization procedures with ammonium sulphate had yieldedthe largest, well-formed crystals, the whole of the reworked material was initiallyemployed for recrystallization of this acid form on a larger scale. The studyfurther provided an opportunity to investigate solid-state aggregation of pro-insulin at acid pH. These crystals may, in spite of their size and development,be expected to be of poorer quality than those grown at neutral pH if theanalogy between proinsulin and insulin is maintained.The large scale crystallization studies were made in culture tubes (6 X 30

mm) or vials (cut to 9 X 5 mm), each containing about 1 mg of proinsulin. Thecrystallization procedures employed in the culture tubes were completelyanalogous to those in the thin-walled glass capillaries. The vials, however, werenot stoppered but left open in a larger closed vessel to permit slow evaporation.Crystals up to 200 X 100 X 75 ,um were obtained with both procedures withinone week (see lF'ig. IF).X-ray precession photographs were taken on the microfocus unit. Photo-

graphs of hOl, OH, and Iki zones showed systematic absences indicating thespace group P4,212 or its enantimorph P43212. The unit cell dimensions area = 50.S + 0.2 A, c = 148.0 + 0.4 A, with a cell volume of 3.82 X 1&5 A3.The minimum spacing observed for relatively short exposures was 2 A. Thecrystals appear unusually resistant to radiation damage and withstood x-rayexposure for more than 100 hr.The equation

1/P, = Vc = A1VV + (1 - fp)VL (1)

defines the relationship between weight fraction and other variables, where 1V,is the specific volume of the crystal, f, is the weight fraction of the protein, andUL is the partial specific volume of the liquid of crystallization. If it is assumedthat the partial specific volume of the protein in the crystal is that determined indilute solution, and that the liquid of crystallization is identical with the buffersolution used to prepare the crystals, then the equation can be solved for f,.The crystals employed for density determination were small (100 and 120

,um long, respectively). This introduces error-wiping the crystal free of motherliquor rapidly enough to prevent air drying is very difficult. The measureddensity p = 1.20 + 0.01 g/cm3 (estimated error) therefore must be consideredapproximate. The density of the protein-free buffer is 1.028 + 0.002 g/cm3.The partial specific volume of proinsulin has not yet been determined experi-mentallv.

In employing Eq. (1) to solve for fp, we have used two limiting values for thepartial specific volume of insulin up = 0.71 or 0.72.20 This procedure gives valuesfor the weight fraction of protein in the crystal of 0.53 and 0.55. respectively

1217

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

1218 BIOCHEMISTRY: FULLERTON, POTTER, AND LOW Pitoc. N. A. S.

corresponding to weights for the protein component of the asymmetric unit of18,000 and 19,000 daltons.The molecular weight of bovine proinsulin is 8685. If the approximate value

for the protein component of the asymmetric unit corresponds to the two mole-cules of proinsulin, the appropriate crystal density calculated from Eq. (1) is1.19, or 1.185 corresponding to iU, = 0.71 and ip = 0.72 respectively.

Discussion. The close agreement found between the observed and calculateddensity is certainly adequate to establish that a bovine proinsulin dimer is theasymmetric unit in the crystal form studied. The dimer association found insolution at acid pH therefore exists also in the solid state.

Proinsulin crystals grown at acid pH show greater than ideal mosaicity andthe extent of the diffraction pattern observed (2 A) is limiting for structure stud-ies. Before the crystalline phase most appropriate for detailed x-ray study ischosen, other crystal forms should be investigated. A larger scale preparationof the other forms described is now under way. We note that the crystals grownin the presence of added zinc from citrate buffer appear poorer than those grownin the absence of added zinc from phosphate buffer. The relative roles of zincand buffer are now under investigation. Zinc is required for the preparation ofboth rhombohedral21 and monoclinicl insulin crystals at neutral pH. Theproinsulin provided, as we observed earlier, may already have contained tightlybound zinc.As porcine proinsulin differs so markedly from bovine proinsulin in the con-

necting peptide region, crystals of this species may not be formally isomorphouswith those from bovine proinsulin-in contrast to the situation which exists forinsulins from the two species. The crystallization of porcine proinsulin will beinvestigated.

We are happy to acknowledge generous gifts of bovine proinsulin from Prof. D. F. Steinerand thank him for his interest.

* This work was supported by grants from the National Institutes of Health (AMI 01320)and the National Science Foundation (GB 7272).

t On leave from the Department of Physics, University College, Cardiff, Wales.$ Requests for reprints may be addressed to Dr. B. W. Low, Department of Biochemistry,

College of Physicians and Surgeons, Columbia University, 630 W. 168th Street, New York,N.Y. 10032.

1 Steiner, D. F., and P. Oyer, these PROCEEDINGS, 57, 473 (1967).2 Steiner, D. F., D. D. Cunningham, L. Spigelman, and B. Aten, Science, 157, 697 (1967).3 Yip, C. C., and B. J. Lin, Biochem. Biophys. Res. Commun., 29, 382 (1967).4Steiner, D. F., 0. Hallund, A. Rubenstein, S. Cho, and C. Bayliss, Diabetes, 17, 725

(1968).5 Schmidt, D. D., and A. Arens, Hoppe-Seyler's Z. Physiol. Chem., 349, 1157 (1968).6 Nolan, C., and E. Margoliash, cited in Recent Prog. Horm. Res., 25, 250 (1969), by D. F.

Steiner, J. L. Clark, C. Nolan, H. A. Rubenstein, E. Margoliash, B. Aten and P. E. Oyer.7Chance, R. E., R. M. Ellis, and W. W. Bromer, Science, 161, 165 (1968).8 Grant, P. T., and K. B. M. Reid, Biochem. J., 110, 281 (1968).9 Trakatellis, A. C., and G. P. Schwartz, Nature, 225, 548 (1970).

10 Clark, J. L., S. Cho, A. H. Rubenstein, and D. F. Steiner, Biochem. Biophys. Res. Commun.,35, 456 (1969).

Oyer, P. E., S. Cho, and D. F. Steiner, Fed. Proc., 29, 533 (1970).12 Frank, B. H., and A. J. Veros, Biochem. Biophys. Res. Commun., 32, 155 (1968).13 Frank, B. H., and A. J. Veros, Biochem. Biophys. Res. Commun., 38, 284 (1970).14 Jeffrey, P. D., and J. H. Coates, Biochemistry, 5, 489 (1966).

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

VOL. 66, 1970 BIOCHEMISTRY: FULLERTON, POTTER, AND LOW 1219

15 Rudko, A. D., and B. W. Low, unpublished studies.16 Low, B. W., F. M. Lovell, and A. D. Rudko, these PROCEEDINGS, 60, 1519 (1968).17 Schlichtkrull, J., in Insulin Crystals (Copenhagen: Munksgaard, 1958), p. 52.18 Zeppezauer, M., H. Eklund, and E. S. Zeppezauer, Arch. Biochem. Biophys., 126, 564

(1968).19 Steiner, D. F., personal communication.20 Oncley, J. L., E. Ellenbogen, D. Gitlin, and F. R. N. Gurd, J. Phys. Chem., 56, 85 (1952).21 Scott, D. A., Biochein. J., 28, 1.592 (1934).

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020