GENETICS OF SOMATIC CELLS 449 



early dependent forms and in late autonomous forms of the same hormone-induced 

 endocrine tumor. 



Another point, already emphasized in previous papers, 719 - 720 is the possibility 

 of analyzing the various phenotypic characteristics entering into the complex pheno- 

 menon of malignancy, by selecting specific pairs of tumor sublines for comparison. The 

 random assortment of various unit characteristics such as growth rate, degree of differ- 

 entiation, invasiveness, ability to spread, dependence on superimposed hormonal and 

 other controls, sensitivity to antimetabolites and other growth-inhibitors during tumor 

 progression 399 - 401 is strongly suggestive of independent determination mechanisms. 

 If they are really partially or completely unrelated, it would follow that these units, 

 rather than malignancy as a whole, should be studied separately in clear-cut systems. 

 It is increasingly realized that conventional biochemical or other comparisons between 

 malignant tissues (often used after long periods of serial transplantation) and their 

 homologous, normal counterparts suffer from serious drawbacks. Representativeness 

 is one of the problems ; a given normal tissue may contain the normal ancestor cell of 

 a given tumor in a low frequency only; pulmonary tumors and normal pulmonary 

 tissue, or hepatic tumors of the cholangioma type and normal liver, are cases in point. 

 Even if this source of error could be minimized, the tumor may still contain considerable 

 amounts of irrelevant material, such as inflammatory infiltrates, fibrous stroma, and areas 

 of necrosis. These may seriously interfere with the meaningfulness of the comparison. 



Even if these risks could be avoided and homologous normal and malignant 

 tissues were obtainable in highly purified cultures and free of necrosis, the stepwise 

 and multiple nature of tumor progression, apparently involving different kinds of 

 cellular changes, would make it difficult to correlate biochemical or morphologic 

 differences with any one of the unit characteristics involved in the development of 

 malignancy. Even such experiments may, therefore, lead to variable results and may 

 fail to demonstrate general differences between normal and malignant cells; but are 

 there any general differences to be expected ? The tacit assumption of a common 

 cellular pathway of carcinogenesis, implicit in many experimental designs, may be 

 nothing but wishful thinking, as pointed out by Berenblum. 80 Perhaps it would be 

 more reasonable to seek for cellular mechanisms that can explain, for example, the 

 change of a conditioned tumor of an endocrine organ, caused by a certain hormonal 

 imbalance of the host rather than by any cellular change, into its autonomous counter- 

 part, no longer dependent on the original hormonal imbalance and caused by some 

 change in the cells themselves. 412, 413, 414 By comparing dependent and autonomous 

 sublines of the same original tumor, if preservable indefinitely by frozen storage, the 

 difficulties discussed above might be largely eliminated. The populations of cells would 

 be more comparable, and differences between them would be more liable to be corre- 

 lated with the biological change than in less related populations. The biological 

 characteristic itself is sharper and better defined than malignancy or normalcy, although 

 clearly related to one of the most essential features of malignant behavior: the changed 

 responsiveness to the superimposed homeostatic mechanisms of the organism. 



