STRUCTURE OF INSULIN 



by the osmotic pressure or ultracentrifugal methods insuhn 

 exhibits a molecular weight of about 48,000, but the volume of 

 the unit cell, determined by x-ray measurement, suggested a 

 lower value of 36,000. This discrepancy was resolved by Gut- 

 freund (7), who showed that under certain special conditions 

 insulin dissociated to smaller particles; he suggested that the 

 actual molecular weight was 12,000, and that the higher values 

 were due to aggregation in solution. Other workers have con- 

 firmed this value and agree that it is the minimum molecular 

 weight to which insulin dissociates in aqueous solution. 



Recently, however, Harfenist and Craig (8) have applied 

 an entirely new chemical method to this problem. Insulin was 

 treated with fluorodinitrobenzene under conditions which al- 

 lowed partial but not complete reaction with the amino groups. 

 The partially substituted product was then subjected to counter- 

 current fractionation. Besides unchanged insulin a number of 

 colored compounds were obtained corresponding to the various 

 dinitrophenyl (DNP) derivatives of insulin. By analyses of each 

 fraction it was shown that the one which fractionated nearest to 

 the unchanged insulin and must therefore be the mono DNP 

 derivative contained one DNP group per equivalent weight of 

 6000, the second band two DNP groups per 6000, and so on. 

 There was no evidence of a band containing one DNP group 

 per 12,000 equivalent weight. From this it may be concluded 

 that the minimum particle weight of insulin, representing the 

 smallest unit which is held together by covalent bonds, is 6000. 

 Recent physical studies (4,11) have confirmed this value using 

 certain special conditions for dissociation, but it does not seem 

 to be universally accepted by physical chemists and it would 

 appear that in aqueous solution the 12,000 unit is held very 

 firmly together, presumably by secondary valency forces, since 

 it is very unlikely that the conditions used by Harfenist and 

 Craig would break any of the covalent bonds found in insulin. 

 This value of 6000 will be assumed for the present discussion, 

 though most of the structural work was in fact carried out when 

 it was considered to be 12,000. A more accurate value of 

 5733 may be calculated from the known structure. 



435 



