By request of the committee which organized the conference on ‘The Cellular Basis of Differentiation’, I undertook to open the meeting with a sort of topical outline that might serve as a guide in the subsequent free and informal discussion. In view of the diverse meanings attached to the term ‘differentiation’ in various quarters, from development in general in the broadest sense to the production of visible specialized cell structures in the narrowest sense, it seemed desirable to circumscribe at least the scope within which the conference was to hold itself. The instructions of the organizing group were to focus the discussion on the cellular transformations which lead to the final functional state of the mature cell. This limitation is taken into account in the following outline. At the same time it is clear that terminal cell specialization is part of a continuous process, hence cannot be neatly separated from antecedent developmental processes leading up to it. Therefore the difference between this presentation and others dealing with early embryonic phases lies in the emphasis rather than in the nature of the objects. In fact, it is becoming ever more evident that the study of special histogenesis and organogenesis may cast as much light on the interpretation of early embryonic processes as the study of early events, e.g. the formation of the fertilization membrane or of the sperm tail, contributes to the understanding of specific terminal cytodifferentiation. Accordingly, the main objective of the following outline is to set the problem of cellular differentiation into proper focus and perspective within the broad problems of ontogeny.

The term ‘differentiation’ has long been used vaguely and loosely; longer than is profitable for either the description and interpretation of facts or the formulation of problems or the design of experiments. The results of a conference such as this should be gauged by its success in reducing this vagueness and looseness. Since science benefits not only from the discovery of new facts but also from the elimination of inconsistencies and misconceptions caused by unfamiliarity with, or disregard of, known facts, a first prerequisite seemed to be for this conference to have before it a complete inventory, as it were, of all the many facts and facets normally subsumed under the single term of ‘differentiation’. Such a table of contents of the problem was therefore prepared and distributed to the members of the conference in advance of the meeting. It is reproduced below. A second task that seemed called for was the following.

As on past occasions (1947,1949, 1950) to which the reader may be referred for further details, I found it necessary in order to deal profitably with the problem of cellular differentiation, to operate with a more realistic and workable conception of ‘cell’ and ‘protoplasm’ than is normally implied in these symbolic terms. I have sketched this conception previously under the name of ‘molecular ecology’. For brevity and pregnancy, I shall condense some of its major aspects into diagrams presented here for the first time. They are admittedly greatly oversimplified models, intended merely to help us to visualize the phenomena and to formulate problems more cogently. They permit us to phrase specific questions which can be answered decisively. As the answers come in, those models will be validated or invalidated or, most likely, appropriately modified. In the meantime they will have served as a temporary scaffolding for the erection of a more tangible and less non-committal notion about ‘differentiation’ than we now possess. Therefore no more than this tentative pragmatic value need be attached to them.

Differentiation is not something that should or could be defined, at least not in our present state of ignorance and conflicting opinions. It is a summary term for a great many diverse experiences and evidently its content will grow and change as more experiences are gathered. Hence whatever definition would be attempted would have no finality. To strip the term of its vagueness would not require definition so much as circumscription, and this is what will be attempted in the following. Instead of engaging in the illusory task of defining the properties of some non-existing entity, called ‘differentiation’, we shall simply list those properties of differentiating living systems that have been observed. This list is a mere sample with no claim to being either categorical or exhaustive. It tries to break down the general problem into a number of specific issues which can serve to bring some preliminary order into the enormous mass of scattered data bearing on the topic.

A. INTRODUCTION TO THE PROBLEM OF DIFFERENTIATION

Progressive changes in cell strains versus the terminal specialization of cell individuals. Changes in composition versus changes in distribution and arrangement. Which parts of the cellular system (a) change with time in all cells equally? (b) change differentially according to cell type? (c) remain unchanged throughout differentiation?

B. CRITERIA OF CYTODIFFERENTIATION

(By what signs can we recognize differentiation?)

  1. Morphological criteria (What signs of differentiation are either immediately visible or can be rendered visible by appropriate tests?)

    • Microscopic characters (intracellular and intercellular structures; their sizes, aggregations, localizations and modifications).

    • Submicroscopic characters (polarization- and electron-optical identification; elemental particulates; orientation; patterned versus irregular deposition; relation of artifacts to reality; microstructural basis of cell shape).

    • Microchemical organization (differential staining reactions, including electron stains; formed secretions; enzyme localization; pH patterns; &c.).

  2. Direct physiological criteria (What signs of differentiation are revealed by changes in the physiological properties of cells and their discharges?)

    • Functional activities (metabolic activity; secretions; phagocytosis; contractility; selective absorption and storage, as of iodine by the thyroid cell).

    • Changes in composition (chromatography; ionic composition; changes of identifiable macromolecular compounds, including proteins, enzymes, antigens, nucleic acids, &c.).

    • Changes in chemical composition and distribution determined by ultraviolet and infra-red absorption.

  3. Indirect physiological criteria (What signs of differentiation not directly observable can be deduced from differential reactions of the living cell?)

    • Selective reaction to drugs, hormones, viruses, &c.

    • Organ-specific serological interactions.

    • Self-sorting of cells according to type (e.g. following dissociation; in wound healing; &c.).

C. MECHANISMS AND CAUSATIVE AGENTS OF CYTODIFFERENTIATION

(The detailed story of just where and how the various criteria of differentiation arise, and of the intracellular and extracellular factors and conditions upon which they depend).

  1. Origin of specialized cell products

    • Sites of specific synthesis and conversion of protoplasmic systems (role of nucleus, genes, cytoplasmic granules, mitochondria, centrosomes, &c., in the manufacture of specialized cell products; continuity of microscopic and sub-microscopic ‘self-duplicatory’ bodies).

    • Secondary rearrangements (coacervation; polymerization; fibril and membrane formation; formation of cilia, sperm tails, insect hairs and scales, keratinization, &c.).

    • Extrusion of cell products (formation of intercellular cements, ground substances, myelinization, chitinization, &c.).

    • Regeneration of cell parts (restoration of intracellular structures in functional cycles and after injury, including merocrine gland cells, protozoans, neurons, muscle-fibres, &c.; fate of differentiated structures during and after cell division).

  2. Metabolism of cytodifferentiation (What are the general metabolic requirements for the functional and structural specialization of the cell?)

    • Thermodynamic considerations and energy requirements for differentiation (relating the metabolic patterns in maintenance, growth, differentiation, and functional operation of the cell).

    • Comparative aspects (relating the metabolic requirements of different cell types to their specialized features, such as cell size, cell life-span, nucleo-plasma ratio, elongation, cell shape, &c.).

  3. Specific accessories to cytodifferentiation (What specific contributions and aids from the environment are prerequisite for the expression of given cytodifferentiations?)

    • Specific nutritive accessories (vitamin A in retinal differentiation; ascorbic acid in connective tissue differentiation; pterins in insect pigments).

    • Hormones (general hormonal requirements for cytodifferentiation; specific hormone-dependence of hormonal end-organs; differentiative interdependence of endocrine glands; neural influences on differentiation).

    • Biochemical genetics of defective cytodifferentiation (e.g. albinism, chondrodystrophia, sickle-cell disease, &c.).

    • Cellular interactions in differentiation (effects of organ-specific discharges other than hormones; contact influences transmitted from cell to cell; homoeogenetic inductions; ‘infective’ propagation of differentiation, &c).

    • Physical factors (orienting and aligning forces in the cellular environment in their relations to cell polarization, locomotion, and structure formation; pressure and tension as factors in muscular, skeletal, and connective tissue differentiation; diffusion and hydrodynamic factors in differentiation; the role of interfaces, &c.).

  4. Pluripotency of differentiation (What is the basis for the fact that most cells prior to their terminal stages are capable of several alternative courses of cytodifferentiation?)

    • Multivalency of cellular equipment (all-or-none principle of differentiation; physical basis of ‘threshold’, ‘dominance’, and ‘pacemaker’ functions, &c.).

    • ‘Position’ effects producing cellular dimorphism (e.g. terminal secretory versus duct cells in glands; endothelial versus blood cells; egg versus nurse cells; ganglion versus satellite cells, &c.).

    • Proliferation and differentiation (antagonistic relations between reproduction and specialization; occurrence and significance of differential cell divisions; endomitosis in relation to cytodifferentiation, &c.).

D. GENERAL PROBLEMS AND CONCLUSIONS

  1. ‘Dedifferentiation’ and ‘redifferentiation’ (To what extent does the loss of certain criteria of differentiation signify (a) loss of type specificity, and (b) true reversion to pluripotent condition, implying the capacity to differentiate in a variety of new directions?)

  2. Genes, nucleus, and cytoplasm in differentiation (it is essential to proceed beyond the trivial assertion that they all are involved in the process of differentiation, and to examine just how and when each component participates; including the hypothetic role of ‘plasmagenes’, somatic mutations, polyploidy, &c.).

  3. Comparative aspects (What lessons regarding cytodifferentiation can be gained from the study of differentiation in bacteria, slime moulds, protozoans, fungi, and higher plants?)

  4. Cytodifferentiation and morphogenesis (How cytodifferentiation determines tissue and organ formation).

  5. Cytodifferentiation and carcinogenesis (Is the primary deviation of the cancer cell a matter of aberrant growth or rather of defective cytodifferentiation? If the latter, which ones among the multiple features of the complex differentiation process enumerated above are predominantly involved?)

With this index as a background, we shall now proceed to single out a number of specific points for discussion.

Ordinarily the term ‘cytodifferentiation’ is used interchangeably in at least four different meanings. It designates:

  1. The variety of terminal characters distinguishing the cells of a developed organism; see, for instance, the frequent statements to the effect that ‘a given organism shows such-and-such differentiations’.

  2. The process by which a cell of more generalized microscopic appearance assumes the more specialized aspect referred to in point 1 by elaborating distinct ‘differentiation products’ characteristic of its kind, declaring colour, as it were; e.g. the differentiation of myofibrils in a muscle cell, of melanin granules in a pigment cell, of hemoglobin in an erythrocyte, of colloid in a thyroid cell, of neurofibrils and Nissl bodies in a nerve cell, of keratin in an epidermal cell, &c.

  3. The process by which the cells expressing themselves visibly in this fashion develop the basic equipment to do so; e.g. the differentiation (or ‘maturation’) of a myoblast into a muscle cell; of a melanoblast into a chromatophore; of a hemo-cytoblast into a blood cell; of a thyroblast into a thyroid cell; of a neuroblast into a neuron. Evidently point 3 grades into point 2, and the empirical distinction between them merely recognizes the fact that even the ‘mature’ muscle-fibre can still form new myofibrils (as in hypertrophy), the ‘mature’ pigment cell can still produce pigment, and so forth.

  4. The process by which each cell type labelled here as ‘-blast’ acquires the peculiar physico-chemical machinery that enables it to manufacture the specialized protoplasm referred to under point 3, and in many cases not in just one single set, but in an unlimited number of identical sets, as myoblasts proliferate myoblasts, melanoblasts more melanoblasts, and so forth. Whether the proliferating cells are strictly localized (‘germinal cords’ or ‘germinal layers’, e.g. in the vertebrate skin, the blood-forming centres, the intestinal mucosa), or whether they are widely scattered, the basic question remains of how they have come by their differential faculties of reproducing each a characteristic type of protoplasmic descendants identified by the specific performances listed (chronologically reversed) in points 1 to 3.

Point 4 thus raises the question of the origin of divergencies among initially identical cell strains. From the testimony of Experimental Embryology and Pathology we know that the same cell can be the progenitor of a variety of derivative types—the more the earlier we test it—hence diversification among descendants of common precursor cells is a fact. The problem of how this diversity comes about is of a different order than that of the subsequent steps 3, 2, and 1, which merely elaborate an already existing diversity. Even so, since the phenomena 4, 3, 2, and 1 are continuous in time, it seems valid for purposes of our discussion, and probably quite generally, to identify differentiation with the whole series of physico-chemical changes described by these four points. Consequently we can state that differentiation (a) is a complex (as opposed to unitary) phenomenon; (b) is a stepwise process, not an abrupt event; and (c) can be meaningfully referred to only in terms of the stage which that series of processes has reached at a given time. To call a cell either ‘undifferentiated’ or ‘differentiated’ without further specification is not only inaccurate but scientifically meaningless.

Ideally one would have to follow a cell and its descendants through development, sample it at various stages along this course, and describe its properties objectively at each sampled stage: the changes of properties registered over the whole course would then add up to a complete record of ‘differentiation’. Regardless of whether or not such objective description is ever attainable, we can at least come closer to it by including in our description as many properties as we can possibly identify, instead of limiting ourselves arbitrarily to those few properties which happen to reveal themselves optically or in some other overt manner. Any test, however indirect, and any reaction, however much delayed, that helps us to distinguish between two cells (or the same cell at different times) is a pertinent index of distinctive cellular ‘properties’. The above Inventory of Cytodifferentiation illustrates the great variety of covert signs of differential cellular constitution that could properly be used for this purpose. The study of differentiation thus resolves itself into (a) the identification of criteria of the state of a cell; (b) the comparison of these criteria, revealing systematic differences between them; and (c) the exploration of the origin of these differences as they arise between (α) a given cell at an earlier stage and the same cell at a later stage, (β) a given cell and its descendant cells, and (γ) different descendant lines stemming from equivalent (or ‘equipotent’) precursor cells.

Differentiation thus connotes the appearance of true differences of constitution —’progressive transformation’—within a protoplasmic continuum extending along the time line, regardless of whether or not this continuum undergoes further subdivision in space by successive cell divisions or remains a continuous mass. Identical courses from identical starts can produce differences only -within, but not between, protoplasmic strains. The emergence of true differences between strains derived from identical sources is due to the fact that their transformations have taken divergent courses.

The term ‘true differences’ implies a distinction between characters inherent in the cell and features reflecting its environment. Now, evidently a cell without an environment is a fiction; hence no property or manifestation of a cell can be divorced from a consideration of the environment in interaction with which it has occurred or been displayed. However, by comparing the same cells in different environments, and different cells in the same environment, their relative contributions to a given reaction, transformation, and eventually form, can be assessed. These behavioural tests of cell character, in contradistinction to environmental expression, are illustrated in the diagram, Fig. 1. It shows in the left third, two cells, A and B, in their regular environments, a and b; the conformance between cell and environment being indicated by identical markings, that is, stripes for a and dotting for b. Cell A is then transplanted into environment b, and cell B reciprocally into environment a. The same operation is carried out in the middle diagram. Yet the subsequent behaviour of the transplanted cells is radically different in the two diagrams. In the left one the transposed cells A and B conform entirely to their new environments, proving that the erstwhile distinctions had not resided in any ingrained differences between the two cells but had been expressions of different behaviour of a single kind of cell in different environments. In the middle diagram, by contrast, the cells after transposition continue to show properties referable to their origin and do not adopt each other’s former properties and appearances. Such behaviour is incontestable proof that some intrinsic differences had existed between A and B prior to their transfer. This does not exclude, of course, that these differences might have been conferred upon them by their previous prolonged residence in those environments. Of this we shall speak later. Nor does this test imply that the transposed cell would not show some response to the new environment. In most cases it will, but the salient feature remains that A in b does not become like B in b, and B in a does not become the like of A in a.

It can readily be seen that this scheme is at the base of most classical experiments on embryonic transplantation, in which a change of behaviour from that of the left to that of the middle diagram has been ascribed to a process of ‘determination’. In the conviction that determination is based on real physico-chemical changes, and since on the other hand we have extended the term of differentiation to include all indices of transformation, and not just the accidentally visible ones, there seems to be no further justification for retaining these two separate categories on the cellular level. Determination then is but the earlier part of the differentiation process, which is less directly discernible. But power of discernment is a property of the observer and his tools, and not of the observed system.

It is evident that these diagrams of behavioural tests apply equally well to the relation between genotype and phenotype. If A and B are taken to be eggs (or seeds) raised in media (or soils) of different compositions, the left diagram evidently represents phenotypic variation, while the middle one reveals genotypic differences. It is important, however, to bear in mind that inherent differences, which in the zygote state would ordinarily indicate differences of genotype, are manifested among somatic cell strains despite their supposedly identical genic equipment (see below).

This latter fact is best demonstrated by the most cogent test of differential constitution, namely, that diagrammed in the right-hand panel of Fig. 1, representing transfer of cells A and B into a common environment, i. As is well known, somatic cell strains transferred to tissue culture will undergo considerable changes, including the resorption of much overt specialized equipment, yet at the same time show the following signs of the preservation of inherent differences. (1) They retain certain gross morphological and physiological distinctions, such as differences of size, of nutrient requirements, of growth rates, of viability, &c. (2) Returned to conditions appropriate for the restoration of specialized equipment, each strain will develop products characteristic of its own original type or a related type. (3) When being transplanted, after prolonged stay in vitro, into a foreign environment in a host organism, they will behave essentially according to the middle rather than the left diagram. (4) While in the in vitro environment, each strain may proliferate and give rise to countless descendant cells, all of which will bear the marks of the parent strain, hence likewise behave according to points (1), (2) and (3); that is to say, while B in small i will behave differently from B in b, and A in i differently from A in a, the differentials between B and A are preserved and passed on to their cellular progeny throughout the subsequent processes of growth and cell division without depletion or attenuation.

Consequently the apparent simplification, or as it is commonly called ‘dedifferentiation’, of cells in tissue culture signifies merely a loss of external criteria such as referred to in section II. 2, without loss of type-specificity. It does not imply reversion to a common type. The wider range of responses of cells kept in different media only proves the plurivalency of cells even in advanced stages of differentiation, rather than a recovery of omnipotency. To determine the breadth of this range, which varies from type to type and from stage to stage, is a purely empirical task. The crucial fact to remember is that this range undergoes progressive narrowing within each sector of the protoplasmic time-continuum. In conclusion, the true, that is inherent, properties which connote cytodifferentiation reside within the cell boundaries and are of such nature that they can be reproduced true to type in unlimited amounts during continued proliferation. No consideration of differentiation that confines itself to the cell individual can thus be complete. A complete account must include the prior history during which equipotential cells have acquired their properties of myoblasts, chondroblasts, nephroblasts, &c., respectively, as well as the subsequent history during which these various cell types can continue to reproduce each its own kind differentially even in an indifferent common medium.

These facts are summarized in the two diagrams, Figs. 2 and 3. In Fig. 2 the segregation of omnipotent cells of the early germ by a series of events into a neuroblastic and a mesoblastic strain is depicted, with the former branching into nerve cells and glia cells, the latter into muscle cells and kidney cells, among other specializations. This diagram takes into account that, even in advanced stages, cells can still assume a variety of expressions; for instance, both muscle and glia cells can appear as either spindle cells or macrophages (see arrows), which are functional conversions, called ‘modulations’, and indicate latitude of expression within a given type rather than instability of the type as such.

In Fig. 3 one particular cell line is singled out to show the transformations its protoplasm continues to undergo even after type specificity has been established. Stage a could represent a medullary plate cell or an epidermal cell maturing (with concomitant growth and division) into stage b, then passing on into stage c, at which there appears for the first time a separation into reproductive and non-reproductive groups. The reproductive ones may remain segregated in special ‘germinal’ layers or cords, such as the neural epithelium lining the brain ventricles or the Malpighian layer of the skin. Knowing that the former can no longer give rise to anything but neural cells and the latter to nothing but epidermal derivatives, it is hardly proper to call them ‘undifferentiated’ as is done in common usage. While the upper cell in stage c keeps turning out more cells of the particular type, the lower one has become a terminal cell individual. As such it produces additional specialized equipment of the sort usually used for its identification.

It has frequently been asserted that there is a certain general inverse or antagonistic relationship between proliferation and terminal specialization. This rule, however, has by no means universal validity and would be weakened even more if instead of cell division the process of protoplasmic reproduction were used as criterion of growth. Neurons, for instance, keep growing throughout life yet hardly ever divide, and from what is known about cellular hypertrophy, one can make a similar case for other cell types. Perhaps the general absence of mitotic activity in terminal stages of differentiation is due to the diversion of materials or energy resources that would be needed for the mitotic apparatus into the building of specialized equipment.

Even the non-reproductive cells of stage c are not necessarily single-tracked. Depending on what type they belong to, they can go on to a variety of functional states. Most common among these are the alternatives of inoperative and operative phases which might be reiterative, either cyclically or aperiodically, or represent a singular event. Familiar examples are the active and inactive states of glands, the castrate or hormonally stimulated states of secondary sex characters, the fixed or mobilized forms of reticulo-endothelial cells, and the like. An operative cell which has become single-tracked may then assume a terminal expression (state e), as in the case of neuron or sensory cell, often eventuating in death, as for instance in keratinized cells of the horny layers, red blood cells, secretory cells or holocrine glands, &c. (stage f).

The course of events here schematized could be compared to the conveyor belt of an assembly line in an industrial plant in which a raw product is gradually transformed into the finished product. The analogy is correct in that, at every step of the process, additional factors must enter to permit it to proceed to the next step towards completion, and also in that the process may be arrested at any one stage. Thus many cells never reach the maximum possible terminal expression. Contrary to a machine assembly, however, the progress of cellular differentiation is marked less by the stepwise addition of new components than by the reorganization and selective rearrangement of existing ones within the system.

Having thus circumscribed the criteria and the nature of differentiation, let us now turn to the question of what parts of the cell take part in it. We may immediately exclude from our consideration elementary molecular constituents that travel freely between the cell and its environment, such as water, electrolytes, and most small organic molecules, and restrict our question to those organic systems which occur solely within the cell space. For simplicity we may lump them into five categories: (1) the genome; (2) the non-genic parts of the nucleus, or the nucleome; (3) the continuous cytoplasm; (4) the cytoplasmic inclusions, among which we may distinguish two kinds, (4a) those in common to a great variety of cell types, e.g. mitochondria, microsomes, the Golgi system; and (4b) specialized products peculiar of a given cell type, e.g. myofibrils, neurofibrils, secretion granules, pigment, &c.

The last class, (4b), obviously differs widely between different cell types. From what we have said before, particularly in section II. 3 and 4, this presupposes the existence of differential production machineries in the respective cytoplasms. Consequently, at least part of system (3) must be assumed to be subject to differentiation. As for the cell organs of class (4a), the answer is uncertain. Except for trivial differences of size, configuration, and density, they are usually considered as equivalent in all cells. It is quite possible, however, that beyond their universal functional similarities (e.g. concentration of respiratory enzymes on mitochondria, lipid character of the Golgi system, ribonucleic acid accumulation in microsomes), they show finer biochemical distinctions corresponding to their respective cell types. The extragenic nucleome (2) must be considered as differentiated, not only from general cytological appearances, but also because of the reported origin of certain specific cell products (e.g. secretions) within the nucleus. The genome (1), on the other hand, is generally assumed to retain its identity in all the various somatic cell types, at least as far as the quality of its composition is concerned. The evidence rests essentially on genetic data. Occasional attempts to connect differentiation with quantitative changes in the genome, such as polyploidy, can be discounted in view of (a) the normal cytological and histological differentiation of animals with haploid as well as polyploid (from triploid to octoploid) chromosome sets; (b) the regular occurrence in some tissues of cells with multiple chromosome sets (e.g. mammalian liver, insect scales) affecting solely cell size, but not basic cell character; and (c) a simple consideration of the very large number of qualitatively different cell types in the higher animals.

Even after narrowing the issue of differentiation to this point, it is still vague and intractable because we have recognized no further inner distinctions within the various subsystems (2) to (4) to which we concede differentiation, treating them as if they were homogenous substances of identical composition and properties throughout. In order to go beyond verbal generalities and to confront the living system with more realistic and analytical questions, we must try to form a more concrete idea of just what the microcosm of a cell really looks like, consists of, and how and by what forces it undergoes its progressive transformations. This calls for replacing the common notion that protoplasm is a ‘substance’ by a more realistic representation which takes into account its character as a ‘system’ composed of populations of molecular species of various properties and groupings, interacting with one another within the ordered facilities, as well as limitations, of the space they occupy. With this in mind, the molecular model of differentiation, presented in the following section, was constructed.

Any attempt to formulate a molecular concept of cytodifferentiation is bound to remain, for the time being, highly speculative. However, the free flight of fancy can be considerably restrained by paying rigorous attention to certain principles of cell behaviour in differentiation that have been derived from countless observations and experiments. In an earlier publication I have labelled three of these basic principles as ‘discreteness’, ‘exclusivity’, and ‘genetic limitation’. Discreteness of cytodifferentiation means that cell types fall into rather sharply delimited classes without intergradations. There are, for instance, no transitions between muscle cells producing actin and myosin, thyroid cells producing thyroxin, Langerhans cells producing insulin, and nerve-fibres producing myelin. This argument is not weakened by the fact that many cell types may contain or produce common components (for instance, collagen or melanin), much as all living cells must be able to reproduce common equipment for their basic physiological functions such as respiratory enzymes. But with regard to other parts of their endowments, different cell types differ radically. Therefore, the principle of discreteness and the absence of a continuous spectrum of transitions point clearly to the fact that differentiation between strains is based on qualitatively different chains of chemical reactions. The second principle, exclusivity, expresses the fact that a cell cannot follow more than one of the several discrete courses originally open to it, at a time. Once it has become definitely engaged in one course, alternative courses are automatically suppressed. Evidently the cell in differentiation behaves as a unit. This principle resembles the principle of complete dominance established in genetics. The third principle, genetic limitation, expresses the empirical fact that the various ontogenetic courses open to a given cell are strictly circumscribed by the genetic endowment of the species to which it belongs. Combined with the principle of discreteness, this means that the finite, although very large, repertory of reaction types of the various descendants of a zygote is strictly limited from the start by the chemical equipment of the genome.

Heeding these clues and from a general critical interpretation of ontogeny, we arrive at the following concept of differentiation. The genome of the zygote endows all descendant cells with a finite repertory of modes of reaction. What is commonly called ‘differentiation potency’ may be interpreted as a finite assortment of chemical entities. These entities, of course, must not be viewed as direct precursors of any final characters, but as a reactive system, the constant interaction of which with systems of the extragenic space will only gradually yield the later specific characteristics of the various cell strains. If we envisage the response repertory of a cell as a system of alternative chains of reactions permitted by the original genic endowment, then differentiation involves the selective triggering off of certain of these chains to the exclusion of others. Divergent differentiation between two cell types thus is due not to differences, of native composition but to the activation of different parts of the common equipment.

The first divergent activations in a germ arise presumably from preformed regional differences in the chemical composition and configuration of the surface layer of the egg; for blastomeres enclosing one particular sector of this surface mosaic will confront their genome with a different reactive background than will those that have received another sector. The ensuing reactions, further diversified by interactions among neighbouring parts, lead to the next steps of activation from the still multivalent, if already somewhat restricted, reaction repertory, and so on down into the late stages pictured in Fig. 3, in a continuous sequence of interactions. Since the extragenic space, i.e. the genic environment, is thus undergoing progressive transformation, it is evident that every new reaction must be viewed in terms of the cellular system in its actual condition at that particular stage, moulded by the whole antecedent history of transformations and modifications, rather than solely in terms of the unaltered genes at the core. Incidentally, keeping this in mind ought to stop the confusing practice of labelling all intrinsic properties of a cell at an advanced stage as ‘genetic’, but those brought out by still later interactions with neighbouring cells or diffusible agents as ‘environmentally’ or ‘hormonally’ induced, forgetting that no cell develops independently, but that all of them have gone through a long chain of similar ‘environmental’ interactions with neighbouring cells and the products of distant ones.

This view of differentiation as a chain of chemical reactions is, of course, not new, but then it is not sufficiently tangible and specific either. It offers no model for the process of selective activation; no explanation of how, despite the great diversity of possible reactions, their systematic order in time and space, underlying functional organization, can be maintained; in general, it is too noncommittal to guide deeper experimental penetration. In an effort to fill it with more specific meaning, I shall sketch in the following a concept the gist of which is incorporated in the diagram, Fig. 4.

The left end of the figure represents a sample of protoplasm (similar models apply to all subsystems listed above in IV. 2–4). We ignore chemical constituents of ubiquitous occurrence and concentrate on those molecular species characteristic of the particular cell. Though their building-blocks are of general occurrence, the peculiar pattern of their assembly into larger systems is unique and apparently reproducible only in the presence and with the aid of pre-existing similar patterns. In order not to encumber our model with unverified assumptions, we shall make no attempt to identify these molecular species chemically as to whether they are proteins, nucleic acids, lipids, polysaccharides, or higher-order combinations of such. Their number may be very large but we shall symbolize them only by four different representatives, depicted as crosses, triangles, crescents, and pins. If the chosen sample were part of the egg cytoplasm, these would be part of the primordial molecular endowment. If the sample represents a cell in a more advanced stage, this would be a derivative population, modified by the past phases of ontogeny.

It is characteristic of living systems that a state of random dispersion like that in the extreme left of the diagram would be unstable and gradually give way to orderly segregations effected and maintained presumably with the aid of metabolic energy and other factors to be mentioned presently. As I have outlined in earlier publications, the mixed molecular populations will sort themselves out according to the specific physical and chemical conditions that prevail in the different regions of the cell space; moreover, as a result of their different localizations, the segregated populations will reciprocally contribute to the establishment of similar ordering conditions for subsequent processes. The resulting organized behaviour of mixed molecular populations in the living cells, their ‘molecular ecology’, contrasts sharply with the behaviour of the same populations in the random dispersed state existing in homogenates. Only those chemical compounds in a cell are of relevance that are given opportunity either to interact or to alter the conditions for the interaction of others. Now it is generally recognized that the more complex a biochemical system is, the more subtle and specific are the prerequisites for its operation and maintenance. That is to say, in order for any one of the symbolic compounds of our model to have enough stability to be demonstrable, the conditions for its existence in that locality must be uniquely favourable. Such conditions may pertain to its synthesis or local accumulation or simply protection against breakdown and dissipation. They will include physical factors as well as chemical requirements such as the proper concentrations of constituent compounds, accessory factors, energy-yielding reactions, &c. Non-random distribution of key compounds thus signals the existence in different parts of the cell space of different sets of conditions favouring different types of reactions. Consequently, just as we have come to recognize in the visible range that morphological differentiation is merely an index of antecedent differentiating processes, so on the molecular scale we may now consider the demonstrable localization of given molecular species as merely an index of underlying physico-chemical conditions favouring either the reproduction or the accumulation of that particular species in that particular sample of protoplasmic space.

Among the conditions favouring selective molecular grouping, interfaces deserve special attention for the following reasons. Adsorption to interfaces can stabilize a molecular array against disruption by thermal agitation or liquid convection. Anisodiametric molecules at the same time will be adsorbed in a definite spatial orientation and may thus be ordered and aligned with the result that (a) if they are enzymes, their activity will be increased because of their closer packing and the common orientation of their active groups; and (b) if they are structural units, their orderly assembly into larger structures will be facilitated. Factors conferring this organizing faculty upon interfaces are not only differences of electric potential and the purely physical conditions favouring the formation of monomolecular films, but in a subtler sense, the chemical bonding between sterically matching or otherwise corresponding chemical compounds to either side of the interface. Compounds may thus be trapped in the interface by their affinity to sterically interlocking compounds already there. Although models of this type of interaction are currently popular as explanations of enzymesubstrate relations and antibody-antigen binding, and although their extension to phenomena of gene reproduction, protoplasmic replication, and surface interactions among cells seems highly suggestive, our model is independent of any such special interpretation. All it assumes is that a condition k, indicated in the upper branch of the diagram by a wavy line, representing a certain physical and chemical constellation along that interface, favours the selective accumulation from the interior of the pin-shaped compounds, whereas the condition l prevailing in another interface, symbolized by the broken line in the lower branch, promotes the adsorption and concentration of the triangular species.

Thus the same protoplasm, faced with two different conditions, will acquire surfaces composed of radically different compounds, hence qualitatively different. Let us call such populations which have assumed singular controlling positions ‘master populations’. To explain subsequent development they must satisfy two demands—first, they must have a governing influence on the further reaction pattern of the cell, and second, they must stimulate the reproduction of their own kind. Qualitative effects on metabolism by border populations are indicated by the fact that (a) they can control selectively the substance traffic between the system and its medium in the way of a ‘living membrane’; (b) if endowed with enzymatic activity, potentiated by their ordered state, they will catalyse characteristic chains of reactions; and (c) as structural elements, they constitute foundations for the anchoring and stacking up of other selected compounds (see later). Moreover, if these master compounds in the surface, either directly or through some of their derivative effects, were to monopolize certain metabolic resources to the exclusion of potentially competing compounds, which do not assume equally favourable positions, the latter would gradually be starved out and disappear irretrievably. This is indicated in the model by the dotting and later omission of triangles in the upper branch, and pins in the lower branch, of the diagram.

According to this model, the earliest steps towards differentiation involve primarily changes in the disposition of existing compounds, some of them being shifted to, and concentrated in, preferred interfacial positions. Only in further consequence, and with the passing of time, will their controlling functions in these preferred positions entail changes in the substantial composition of the systems to which they belong. This distinction between disposition and composition is fundamental, as the former is reversible, whereas the latter is not. So long as their contents in molecular species have not changed, two systems, even though they may have displayed different segments of their molecular populations in master positions, hence manifested different morphological and physiological aspects,can still be returned to a common equivalent state, if the respective key species can be dislodged from their controlling positions. Such reshuffling of the molecular population may come about as the result of solvatization, protoplasmic streaming, or other unstabilization of protoplasm; for instance, following either transfer of a cell into a foreign medium or changes in the inner milieu, as in inflammation or other pathological states.

On the other hand, once a given selected master species of molecules has occupied controlling positions long enough to have caused the competitive depletion of certain other species, then obviously a return to a common condition is no longer possible even though the key compounds may still be displaced from their controlling positions and others be induced to occupy their places. This condition is exemplified by the right-hand part of the diagram. In this the two molecular populations of the upper and lower branches of the left half of the diagram are assumed to have been first thoroughly stirred up by some factor mobilizing the cellular content and then confronted with new conditions, one symbolized as m and the other as n. Surface condition m traps the crescent molecules while the cross molecules aggregate in the n surface. Since both of the original cell types still possess both crosses and crescents, their reactions to conditions m and n are similar, as will be recognized by comparing the two inner branches and the two outer branches in the right half of the figure with each other. It should also be noted, however, that these resemblances do not connote identity and that the erstwhile differences in composition between the strains derived from the upper and the lower branches of our original protoplasmic strain have persisted. The rearrangements indicated in the right half of the diagram are the molecular version of the cellular phenomena previously described as modulations. They represent adaptations of cells to different conditions without change of fundamental equipment, but also without implying that the cells of different strains, which have undergone similar or parallel adaptations in response to identical media, have thereby become constitutionally alike.

According to our model, all irreversible differentiations can be said to arise by way of an initial reversible modulation. The other principles of cellular differentiation listed before can likewise be readily translated into terms of this model. The principle of discreteness is the result of the presence in the molecular repertory of the particular protoplasmic strain of a limited number of discrete key compounds that can assume controlling master positions. The principle of genetic limitation reflects the fact that the number and character of these key species is determined by the genic endowment of the zygote. The principle of exclusivity expresses the complete dominance in the determination of consecutive cell transformations by the molecular master species selected for surface occupancy over other molecular species not so favoured, hence excluded from exercising a determinative role.

Ordering processes of this type, spreading from an interface into the interior, are instruments of progressive organization. They are agents of selective activation and specific communication by which a given surface state can gradually evoke a conforming response from the enclosed parts. It should be borne in mind that in this progressive interaction all other interfaces can act as sites of selective screening and conversion according to their own molecular occupancy, so that one cannot take it for granted that any complex chemical entity can pass through a series of such boundaries (nuclear surface, surface of chromosomes, genes, cytoplasmic particles, &c.) without major modifications. In reverse, the actions thus activated within each enclosed system can alter conditions in the whole hierarchy of systems in ascending order. Accordingly the process we have modelled here crudely for a single protoplasmic fragment must be envisaged as repeating itself in manifold variations as each system interacts with its adjacent space. In this light, one could ask whether what we normally call ‘activation’ of certain components of the genome in cellular differentiation does not likewise consist of the selective segregation into controlling or active master positions of the appropriate fraction of the molecular repertory of the genes.

In conclusion, this model epitomizes the broadest statement that can be made about the biochemistry of the living cell in its organized state, as contrasted with its homogenized or disorganized condition, namely, that what determines the activities of the system is not the totality of chemical compounds it contains but the specifically selected assortment of compounds that have an opportunity to interact or otherwise operate, this being only a relatively small fraction of the total. Thus knowledge of the content of a protoplasmic system is of interest only in that it limits the possibilities of what can happen. However, in order to know just what will happen in a given case requires knowledge of just what part of the content will be placed in the appropriate conditions where it can operate. This is merely another and more explicit description of the property we usually refer to as ‘organization’. To take it into account, biochemistry will have to develop a special field of ‘topochemistry’.

We have assumed in this model that, in order for a protoplasmic strain to undergo divergent differentiation, the two branch lines have to be exposed to specificially different external conditions k and l. Similar branching points will arise later leading to further subspecialization. Some such dichotomies may be merely in the nature of modulations, as in the divergent expressions in media m and n. In modulation, the particular type of organization will last only as long as the respective conditioning environments are actually present, whereas in true differentiation a permanent residue of the response to a particular environment has become fixed in the cell so that it can continue itself even in the absence of the organizing environment. The descendants of a modulating cell may, of course, carry permanent and irreversible characters reflecting the particular state of the mother cell during which they were procreated.

Clearly the tacit assumption underlying this model has been that the original protoplasmic system, unless subjected to either condition k or l, would have remained stationary and unchanged. However, the validity of this assumption is open to question. It may be doubted whether any living system, even when left entirely to its own devices in a stable environment, could remain unaltered over prolonged periods of time. Slow progressive changes are known to occur in unfertilized eggs as well as in ‘ageing’ protozoan and metazoan cells. Moreover, any sequence of protoplasmic transformations that greatly outlasts the duration of the condition that has set it off will give the appearance of autonomous intrinsic change.

This being the case, our model may have to be amended in the sense of Fig. 5. The lower branch shows what would happen to the protoplastic system if subject to no additional differentiating influences. It can be seen that a progressive segregation of molecular species occurs, but in this case ‘autonomously’, that is, by virtue of a course of events initiated much earlier in the cell’s history. This course, however, can now be deflected into a different direction by the appearance of a single differentiating condition k, provided this new condition (a) remobilizes the molecular populations, (b) dislodges the master species of triangles, and (c) replaces it by a master population of pins with greater affinity (meaning perhaps better steric conformance) to the inducing category k.

According to this model, divergent differentiation within a given cell strain would require exposure not to two different sets of conditions but to only one, while the other would simply continue an intrinsic pattern of ‘maturation’. Let us quote some examples. Divergent differentiation of secondary sex characters has often been described as the switching of a neutral cell form into either the male or the female direction by corresponding sex-differentiating factors, including hormones. This implies double-switch action. On the other hand, it has also become clear that in many forms of animals the differentiation of one type of sex character is actually identical with the assumed neutral condition, with the opposite sex development being actively enforced by appropriate hormonal deflexion from the original course. This then is single-switch action. Similarly, when tapetum of the developed urodele eye proves capable of reproducing retinal cells of the optic layer, but not vice versa, one could interpret this to mean that tapetal differentiation represents the autonomous course from which the cells would have to be positively diverted in order to produce retinal derivatives. There are innumerable examples in experimental embryology and pathology amenable to similar interpretations. The two variants of our model, corresponding to Figs. 4 and 5, are summarized in Fig. 6a and b, mainly to show that in practice the decision between them may often be difficult to make, particularly if the branching points, that is, the events causing divergencies, follow each other in quick succession. At any rate, these are empirical questions, and the models are merely intended to help phrasing them in a realistic light.

Our models depict the progressive transformations of a given molecular population in time, but they do not take into account the continued increase of this population, which we call growth. Since the symbolic molecules used in our models are by definition peculiar to the particular type of protoplasm, hence are found only inside the cell, the mechanism of their reproduction must be looked for entirely within the cellular system. Despite the splendid upsurge of work on protein synthesis and particularly on the role of nucleic acid systems in cell growth, we still lack the major keys to the understanding of just what goes on in protoplasmic reproduction. It may be well, therefore, to outline certain basic considerations which any future theory of somatic growth must take into account.

  1. Specialized cell protoplasm of a given cell strain can continue to propagate its own kind, as set forth in section III.

  2. There is no evidence that this reproduction of type-specific protoplasm can be referred to the existence of corresponding type-specific differentiations among the genomes of the various somatic cell strains (see above, section IV).

  3. To reconcile these two points, one might assume that during differentiation the capacity for self-reproduction or self-replication has been conferred upon some of the molecular key species that distinguish different cell strains according to our models.

  4. The autonomy, that is independence from the genome, of such hypothetical self-reproducing cytoplasmic units is contradicted by experimental results in protozoans and yeasts. As for somatic cells, actual observations on suitable objects have made even the very concept of self-reproducibility highly questionable. Our own demonstration of the fact that neurons are in perpetual growth and that this growth proceeds solely from the nuclear territory of the cell space, supported by the recent cytochemical work on the substantial role of the nucleus in cytoplasmic synthesis, indicates strongly that the actual production sites in the process of growth are located in the nuclear territory. Therefore it is not unlikely that the so-called self-reproducing cytoplasmic particles derive the substance for their ‘growth’ essentially preformed from the nuclear (or more restrictedly, genic) space, hence represent stations for the deposition and possible type-specific conversion, rather than for the synthesis, of the basic protoplasmic compounds. Their growth thus would be by accretion.

  5. Considering the fact that the extragenic nuclear space ostensibly undergoes differentiation (see section IV), it would be equally plausible, of course, to assume that the conversion of primordial genic products into type-specific variants occurs already within the nucleus itself.

An effort to bring these various considerations to a common denominator leads to the following integrated concept. The various specialized high-molecular key-compounds in the cytoplasm would not really possess the faculty of catalysing the synthesis of more of their own kind from elementary constituents, as implied in the concept of ‘self-reproduction’, but would only have the role of models in the reshaping or converting of more complex primordial compounds, furnished from the genic space, into conforming patterns. For this reason these model compounds may be given the purely descriptive name of ‘templates’. By imposing their pattern upon other compounds, they would perpetuate their kind without being themselves involved in the process of multiplication. They therefore have the faculty of ‘self-perpetuation’, not self-multiplication. Evidently, if the type-specific master compounds which we have assumed to characterize differentiated strains act in this template fashion, the continued reproduction of the particular type of protoplasm would be ensured. Myoplasm would engender more myoplasm, nephroplasm more nephroplasm, neuroplasm more neuroplasm, and so forth, despite the identity of the genome.

In passing, it may be pointed out that this concept, suitably expanded, also furnishes the clue for the harmonious growth relations between different parts of the same organism and for their automatic regulation upon disturbance. This growth control, explained more fully in previous publications, operates on the basis of the following premises: (a) Some of the master compounds selectively sorted out in different strains act as models for their own multiplication and thus for the perpetuation of the strain, (b) Production of primordial genic (speciesspecific) compounds in each growing cell is superabundant, (c) The rate of their conversion into type-specific protoplasm is proportional to the number of extragenic master compounds free to act as templates, (d) Complementary compounds combining with templates render the latter ineffective, hence inactivate or veritably sterilize their template function, (e) Compounds of such complementary combining power are being thrown off as by-products of the type-specific remodelling process. In a very crude picture we might visualize them as the chips coming off as a primordial compound is whittled down to the shape of the template model in the replication process. Again with a purely descriptive term, we might call these small units of a configuration complementary to the templates ‘antitemplates’, (f) Because of their small size, the antitemplates can diffuse from the cell and pass freely between the cell and its exterior, whereas the templates, because of larger size or conjugation, remain confined to the inside, (g) By virtue of the diffusion gradient between their intracellular production sites and the large extracellular liquid space in blood, lymph, and tissue juices, the antitemplates will leave the cell at a given rate. (h) As their concentration in the outside medium increases and the gradient flattens, the rate of diffusion from the intracellular to the extracellular space will decrease, until finally equilibrium is reached. Correspondingly, as their cellulifugal diffusion falls off, their relative intracellular concentration increases, (i) Since the proportion of free templates to inactivated templates in the intracellular space will decrease as the concentration of antitemplates inactivating them increases, the reduction in the rate of outward diffusion of antitemplates will automatically produce a reduction in the number of free template molecules, hence, according to (c), an automatic retardation in the reproduction of the type-specific protoplasm which we measure as growth. Whether this, in turn, rebounds on the rate of primordial genic synthesis or merely switches the utilization of the primordial products from the reproduction of basic protoplasm to the elaboration of specialized cell products, thus accounting for the observed interference of cell proliferation with specialized cell function, is problematical. Further complications are introduced into this scheme by the consideration that in the intracellular space both templates and antitemplates will be metabolized, whereas in the extracellular space the antitemplates will be katabolized without resynthesis, which presumably would cause a steady cellulifugal drift.

It can be readily seen that this system permits all protoplasm of a given kind, however widely dispersed throughout the body, to retain intercommunication and regulate its total growth. Let us consider, for instance, that part of a given type of growing protoplasm is artificially removed, thus reducing the production of templates and antitemplates of that particular type by a given amount. To the residual part of the body the loss of templates will not be perceptible since, as strictly intracellular entities, they had not been in circulation. The only information of the changed situation will come from the sudden drop of the corresponding antitemplate species in the extracellular pool. As a result of this drop of extracellular concentration, the rate of their diffusion from all residual cells of the same kind will increase, thus leaving uncovered intracellular templates for renewed growth catalysis. This will register as an automatic spurt of growth in all tissues having the same characteristic as the removed one. We have experimental evidence to show that this is at least a major part of the mechanism of so-called compensatory hypertrophy and compensatory hyperplasia.

A second way by which to reduce the effective concentration of a given antitemplate population in the extracellular pool, hence to cause an automatic growth response in the homologous cell types, is to release into the extracellular space free templates from their intracellular confinement. As these combine with their specific antitemplates in the pool, the intra-to extracellular concentration gradient of antitemplates will steepen and their growth-inhibiting effect will be correspondingly diminished. Thus, injury to a tissue, by bringing cell content into circulation, will have the same effect on homologous tissues as has partial removal. On the supposition that the ‘growth-promoting’ effect of embryonic extract in tissue culture is due precisely to this mechanism, we have obtained experimental support of highly suggestive, if not yet fully conclusive, nature. Even stronger support has come from experiments carried out in the embryo itself in which’organ growth can be influenced in the expected direction by spilling cell content into the vascular or extra-embryonic spaces.

Whether the templates and antitemplates are to be conceived in terms of steric fitting, like antibodies and antigens, is wholly conjectural and by no means crucial for the scheme as here presented. The envisaged mechanism views the organism as a vast system of chemically intercommunicating differentiated protoplasts. Communication by special hormones appears as merely a more highly adapted and specialized version of this more general principle, the difference being that hormones are specialized cell products turned out in terminal cell phases (section II. 1 and 2), whereas we are here concerned with the underlying cell substance itself (section II. 3 and 4).

Differentiation in molecular terms implies unscrambling and selective localization of molecular populations, setting the stage for consecutive reaction chains. The result is over-all structural order. We are satisfied that thermodynamically the production and maintenance of a non-random condition requires the constant input of energy but the actual mechanisms by which chemical processes are translated into orderly physical structure are for the most part still obscure. They are ostensibly of many diverse kinds, each to be subject to separate analysis in its own right. A few common examples are presented in the following.

One is based on the ordering effect of interfaces. It starts with the formation of a monomolecular film of oriented molecules, yielding a first-degree planar order, which then constitutes the floor plan, as it were, for a higher degree of order attained by the orderly stacking up of additional layers in the third dimension. The simplest cases are schematized in the upper part of Fig. 7. Let us assume a population of two molecular species in random dispersion (Fig. 7a), the pinshaped kind with a lyophilic group at one end and a hydrophilic group at the other, and a water-soluble kind, e.g. a certain protein. If only the former were present at an oil-water interface, oriented molecular fixation of the kind shown in Fig. 7b would take place. In a more general sense, we may replace the oil-water system by any diphasic system with regard to which different end groups of the same molecule would show differential selective affinities. In the presence of the second species, the mixed population could sort itself out according to the diagram, Fig. 7c. Current concepts of the cell membrane and of the myelin sheath of nerve-fibres support this model. By introducing additional molecular species we can construct more complex systems such as in Fig. 7d, in which the third dimension no longer shows repetitive structure as in b and c, but displays qualitative variety. It is important to keep in mind that the adsorbed surface population maintains its stable arrangement in the face of convections and thermal agitation of an otherwise liquid system. Through the stacking on of additional layers the stable organized crust gains in width and may become microscopically distinct as gelated exoplasm. If this is true of the cell surface, then similar processes must also be conceded to the interfaces along genes, chromosomes, nucleolar and nuclear surfaces, and particulates in general.

In its application to the primordial differentiation of the egg, this scheme implies that if the egg surface contains a mosaic of molecular species of different properties segregated in different sectors, this topographical pattern would retain its stable localization despite the movements of the egg content during cleavage or the experimental reshuffling by centrifugation or stirring. It therefore provides firm bearings for the subsequent changes in individual blastomeres as outlined above in section V.

Differences are thus initiated in different parts of the original protoplasmic mass which may not become effective or manifest until at a much later stage. Yet as soon as differences appear, they present an emergent condition of further differentiation for the various interacting cell strains, for the condition of any one cell is at the same time an environmental factor for its adjacent cells. In consulting our model, Fig. 4, it is clear that if conditions k and l, after having produced differential effects in the respective cells, subside and the two cells come in apposition, each constitutes a new outside condition for the other. Depending on the state of consolidation or responsiveness reached, either cell may now react to the other by a further step of transformation. The interaction is mutual, yet whether or not a response will materialize is a matter not only of the presence or absence of adequately responsive units but also of the degree of mobility and displaceability.

One readily recognizes in this a general model of processes of induction by cellular contact. As I have indicated on earlier occasions, the neuralization of ectoderm by subjacent cell layers, or layers of organic molecules deposited on its surface, might be a case in point, although the evidence that this type of induction is of transmissive rather than transportative nature is by no means conclusive. On the other hand, in the case of lens induction by contact with the retinal layer, signs of cellular orientation signalling molecular regrouping of the requisite kind have been observed.

A second example of typical unrandomization of protoplasmic components is found in the formation of linear complexes with definite orientation in space, best illustrated by the fibrous proteins but presumably applying to all kinds of elongate and anisodiametric molecules. A simple sequence of steps is represented in the lower half of Fig. 7. In e we see constituent molecules with specifically configurated end groups; the addition of molecules of fitting properties (f) will link the original elements into chains (g). Polymerization and coacervation in vitro, blood clotting, protoplasmic coagulation, &c., furnish examples for such linear compounding. The resulting chains (g) are linear but not straight. Save for true crystallization and the formation of tactoids, such filaments require extraneous vectors in order to be straightened out. Such vectors (h) may consist of physical tensions, convection currents, and perhaps strong electrostatic fields. The resulting arrangement (i) leads to the building up of straight fibrils, fibrillar bundles, and fibres by progressive condensation. A still higher degree of order is produced in those cases in which, in addition to the lengthwise alignment, the constituent elements fall in lateral register (j), producing cross-banding, as in chromosomes, collagen fibrils, myofibrils, &c.

These examples may suffice to prove the diversity as well as the intimacy of interrelations between physical and chemical factors underlying the establishment of spatial order which we call structure. They are important not only because of their conspicuous contributions to the microscopic features of terminally differentiated cells, but because of their less overt ordering function at all stages of the differentiation process.

Problems of differentiation are commonly dealt with in terms that tacitly imply identity among the cells within a given cell population. In reality, no two cells are ever exactly alike, and random variations during ontogeny are apt to magnify rather than reduce initial inequalities. This being the case, a given cell population, subject to an inductive or otherwise differentiating influence, will give a uniform response in all its elements only if that influence exceeds a certain critical magnitude; corresponding to what in neurophysiology is known as a supramaximal stimulus, or in nutrition as superabundant food-supply. If, however, the influence is of lower magnitude, intensity, or duration, only a given fraction of the elements of the population will respond, namely, those of sufficiently low thresholds to be affected by the given dose of action. This is explained by the diagram, Fig. 8. In the left portion the frequency distribution of the elements of a cell population according to thresholds of responsiveness to a given agent is plotted, assuming that they vary at random (normal distribution curve). The abscissae represent stimuli of increasing dosage while the ordinates (upper base line) represent percentages of cells beginning to show effective response at the particular dosage level. The sigmoid curve (lower base line) represents the integral of the distribution curve, hence its ordinates give the total number of elements activated at any given dosage level. As the middle portion of the integral curve is nearly a straight line, the number of activated elements increases within the median range in almost linear proportion to the stimulus. Some proportionality between stimulus and response is thus to be expected on purely statistical grounds.

(In terms of the model, Fig. 5, we might view the threshold condition as the one in which the deflecting factor k can attract a sufficient number of the pinshaped molecules into master positions to give them the edge over the competing triangular type.)

Let us assume now that the dosage of a differentiating influence is of the magnitude indicated by the heavy black line. This corresponds to an ordinate value of 25 per cent. That is to say, the probability is that the cell population will contain, on an average, 25 per cent, cells sufficiently sensitive to respond to this given stimulus condition. Evidently this probability will be the same for any size or density of the population. Sample results are depicted in the middle panels of Fig. 8 for three cell populations differing in densities in ratios of 1 : 2 : 4. Although the number of constituent elements varies, the ratio between responding units (black) and refractory units (white) remains the same. In other words, the relative proportions of the two segments of the population are determined although it is impossible to predict for any given unit whether it will respond or not. Statistical regularities of this kind are doubtlessly involved where the stationary composition of a cell population is to be insured throughout growth, multiplication, and repair.

Similar numerical constancy, however, can also be obtained by an altogether different procedure, namely, by precise cell lineage with differentiating divisions. Such precision regulation has been suggested, for instance, for the constant ratio between scale-forming cells and ordinary epithelial cover cells in butterfly wings.

Statistical dichotomy of differentiation is further modified in those instances in which elements which have responded to a given stimulus thereby become the seat of secondary actions which either facilitate or inhibit similar reactions by their neighbours. These effects are illustrated in the right-hand part of Fig. 8. The top panel illustrates the case of a responding element spreading an action which lowers the threshold of other elements. As a result, the responding cells will appear grouped in clusters. Conversely, as is indicated in the right bottom panel, a spreading influence which raises thresholds will inhibit a similar cell response within a given radius, entailing thus a higher degree of regularity of distribution.

These considerations lead over immediately to a section of great morphogenetic significance, namely, ‘field’ responses, in which the fate of a given cell is determined by the position of the element within the group. Geometric position evidently signifies tangible physico-chemical constellations at that particular site.

Examples of this type of interdependence are too common to require listing in this place. As one of the simplest cases, let us take the fate of an embryonic blood island in the chick embryo as indicated in Fig. 9a. A group of equipotential cells (top) undergo divergent differentiation (bottom), the outer ones forming a vascular endothelium, and the inner ones free blood cells. The terms ‘outer’ or ‘inner’ refer not to properties intrinsic to the component cells but to differentials conferred upon them by virtue of their position within the group. If the original cluster (top) were to be rearranged or bi-sected, or if individual cells simply traded places, each cell would then behave according to its actual, rather than its former, position.

Systems of this kind show the simplest form of field behaviour. Their mode of operation can be conceived of as follows (Fig. 9b,c). Let us assume a simple equipotential cell cluster like the one in Fig. 9a (top), bordering on either cells of some other type or on some other medium. As a result of interaction along the free surface, a certain course of reactions will be activated in the cells that share in the surface. Let us assume that this entails the release from those cells of substances or activities that spread inward with a gradient from their source. In addition, we might assume that the whole cell mass is engaged in activities producing metabolites which can diffuse into the outer medium, hence will tend to be most concentrated in the centre. Additional factors of this kind would be crust-core differentials in accessibility to essential food constituents, oxygen supply, and the like. The resultant gradient of these composite conditions is indicated in the top part of Fig. 9b. Let us now assume that the cells of the group can react to a particular constellation or combination of such factors by a differentiating step according to our model, Fig. 5, and that their thresholds for this reaction lie at the level of the broken line t. Then in the central area, within which the mentioned group factors exceed in intensity the requisite threshold minimum, all cells will be switched into the altered course. Whereas in the preceding section we have considered differentials arising within a homogeneous stimulus field by virtue of random variation among cells, the present case demonstrates the response to a graded field of partly self-engendered stimuli. In the transition zone near the threshold level, the prospective response of a given individual cell will again be unpredictable, but the steeper the gradient, the sharper will be the line of demarcation.

A new chain of responses now having been activated in the central cell group (indicated by dots, Fig. 9b, bottom), let us assume that this activity entails two further events, namely, first, a contraction of the central group (from the broken to the solid contour) and second, the release by them of new reaction products into the vicinity. As a result of the former, the outer cells attached to the shrinking core will be subject to centripetal tension, hence become radially elongated, as shown in Fig. 9c (bottom), which in turn may facilitate the outward diffusion or transmission of the newly produced agent. Its emergent diffusion field is represented by the solid curve in Fig. 9c (top). Let us assume further that the earlier substance gradient has persisted (dotted line) and interacts with the new centrifugal factor so as to form an impervious precipitate which will stop further diffusion, and that this happens in the area where the concentrations of the two agents are about equal, as indicated by the intersection of the two curves at the level s (broken line). As a result, a new zone intermediate between the original core and the new barrier will arise with new influences, activating in the constituent cells the next step of their differentiation repertory. In this manner, complexity within the originally equipotential group increases progressively according to a definite orderly pattern of organization in which ‘position effects’ substitute for the various extraneous media k, l, &c., of our models.

It is easy to imagine that by the linking of such spherical fields with linear fields and the further introduction of asymmetries, highly complex products can be obtained. It is also to be noted that by combining such a gradient concept of activating factors with the above molecular concept of discrete differential cell response, the conversion of merely quantitative differentials into qualitatively diverse responses finds a ready formal explanation. On the other hand, too many abstractions of this kind have been accepted in the past as ‘explanations’ prior to and without tangible verifications. They are presented here, therefore, primarily as guides to further analysis and as a hypothetical framework within which proper experiments can be designed. This cautioning note is not to support the scientifically untenable contention, heard on occasions, that a formal operational analysis in dynamic symbols is valueless unless accompanied by a precise identification of the substantial nature of the agents involved. This would be like saying that perfectly valid laws of optics cannot have been developed, as they have, in ignorance of the electromagnetic nature of light. Considering current trends, it would indeed seem more appropriate to caution against the illusion that the mere identification of physico-chemical systems can have much explanatory value unless their formal order of operation in the living system has likewise been revealed. It is of greater pragmatic value to set forth at least formal models of phenomena known to occur in living systems than to ignore or even deny the occurrence of those phenomena just because our present incomplete, oversimplified, and elementary schemes cannot account for them. Field phenomena in supercellular systems are a firm reality to all those observers and analysts of living phenomena who have not deliberately confined themselves to the investigation of elementary and fragmentary processes in which field properties can be legitimately ignored. At the same time, little has been done to instil more concrete content into the various field concepts. The remarks in this last section merely constitute one small and crude effort towards concreteness.

As indicated in the beginning, the study of differentiation stands to make much faster progress if instead of looking for sweeping generalizations or insisting on generalities, which by the very nature of the differentiation process can have little practical meaning, we first disassemble the complex process into its constituent components, many in numbers and of diverse kinds, and then accord each a separate and systematic study in its own right; and if, furthermore, in doing this, we visualize the real objects as concretely as our knowledge and reasoning will permit us, instead of operating solely with formal and symbolic notions as here-tofore. The models introduced here for the purpose of concretization are undoubtedly far too simple and incomplete. Yet they may have the merit of pinpointing the targets at which we shoot our questions and direct our experiments on ‘differentiation’.

The reader will recognize that many of the current discussions on virus reproduction, plasmagenes, adaptive enzymes, and somatic ‘inheritance’ in their relations to the differentiation process can be well accommodated in the terms of the concepts here advanced. However, in view of the narrative character of this article, no literature has been quoted. The main earlier publications of the author referred to in the text as precursors of the present discourse are as follows:

Weiss
,
P.
(
1939
).
Principles of Development
.
New York
:
Henry Holt and Company
.
Weiss
,
P.
(
1940
).
The problem of cell individuality in DEVELOPment
.
Amer. Nat
.
74
,
34
36
.
Weiss
,
P.
(
1947
).
The problem of specificity in growth and DEVELOPment
.
Yale J. Biol. Med
.
19
,
235
78
.
Weiss
,
P.
(
1949
).
Differential growth
.
In: Chemistry and Physiology of Growth
,
135
86
.
Princeton University Press
.
Weiss
,
P.
(
1949
).
The problem of cellular differentiation
.
Proc. 1st Nat. Cancer Conf
.
50
60
.
Weiss
,
P.
(
1950
).
Some perspectives in the field of morphogenesis
.
Quart. Rev. Biol
.
25
,
177
98
.
Weiss
,
P.
(
1952
).
Self-regulation of organ growth by its own products
.
Science
,
115
,
487
8
.