Concerning the Origin of Malignant Tumours by Theodor Boveri. Translated and annotated by Henry Harris

In the year 1902, I tacked onto the results of my experiments on the development of doubly fertilised sea urchin eggs the speculation that malignant tumours might be the consequence of a certain abnormal chromosome constitution, which in some circumstances can be generated by multipolar mitoses (Boveri, 1902). I had intended, even at that time, to put that assumption on a firmer footing in a separate article. But the scepticism with which my ideas were met when I discussed them with investigators who act as judges in this area induced me to abandon the project. I had to admit that a hypothesis in this field can be of value only if it leads to targeted new research and, above all, to new experimental investigations. And who ought it to be to decide to undertake such investigations if not the originator of the hypothesis himself? So, for some time now, I have carried out experiments that seemed to me to be relevant, admittedly without success so far, but my convictions on this score have nonetheless not been shattered.It was the appearance of an article by O. Aichel (Aichel, 1911) that first prompted me to take the decision to communicate the hypothesis, and the evidence that supported it, in greater detail, even though I had essentially nothing more to offer than I had done ten years previously. To explain the origin of tumours, Aichel's article relies on facts that I had established in sea urchins and that gave rise to my own conception of tumours. But the hypothesis that multipolar mitoses might generate tumours is amalgamated by Aichel with his own view, published earlier, that the initial event in the generation of a malignant tumour is the fusion of a tissue cell with a leucocyte(1). This notion casts the key elements of the hypothesis for which I am responsible in an essentially different light, so much so that it has become a matter of importance to me, now that my views have again become the subject of discussion, to set them out in their original form for readers familiar with the subject. But, in the end, I had another motive. The idea that there might be a connection between abnormal mitoses and malignant tumours has certainly cropped up often enough, but it has always been rejected, indeed so completely that in the recent literature it is curtly dismissed, if mentioned at all. The arguments that contradict the idea are easy to see. But one can nonetheless ask whether these objections are merely apparent and whether the more complete information that we now have about these chromosomal abnormalities might warrant, indeed necessitate, a reassessment of their connection with malignant tumours. What follows might serve as a stimulus to that end.1This idea has recently been resurrected. It has been proposed that some of the characteristics of malignant cells such as invasiveness might be conferred on normally sessile tissue cells by their fusion in vivo with leucocytes. Fusion of tumour cells with polymorphonuclear leucocytes has not yet been described, but fusion with macrophages and lymphocytes has. It is commonly found that malignant tumours contain hybrids that have been formed by the fusion in vivo of the tumour cells with tissue cells of the host [Wiener, F. et al. (1972) Nature New Biol. 238, 155], but whether this phenomenon has any important role in determining the character of the tumour is not at all clear. Lines of mononucleate hybrids between macrophages and tumour cells have been established, and these continue to express macrophage markers in vivo. However, the hybrids isolated so far grow more slowly than the tumour cells themselves and are thus unlikely to be responsible for the progressive or metastatic character of the tumour formed.I write about this problem as a zoologist. I have no personal experience worth mentioning in any of the numerous specialised fields of tumour research. My knowledge comes almost exclusively from books. Given this, it is inevitable that I am unaware of many reports in the literature, that I overestimate the significance of many known facts and that I do not set enough store by others. But this article will doubtless contain even more serious defects, as is so often the case when an author makes an incursion into a field with which he is unfamiliar. You may well ask how anyone who has this to say about himself can hope to offer something worthwhile to investigators who have devoted years and even decades of work and thought to the tumour enigma. But there is one thing that has to be taken into account. The tumour problem is a cell problem (2) and, at the least, it is not impossible that a biologist who seeks to fathom certain phenomena in the living cell might be led to consider properties that cannot emerge from the study of tumours themselves but that nonetheless determine their essential nature (3). I ask that what follows be received with these considerations in mind.2There are still some who contest this and in various forms advocate the view that the tumour problem is a `tissue problem', not a `cell problem'. By this they mean that the fundamental abnormality underlying the formation of a malignant tumour is to be found not in the genetic constitution of the malignant cells, but in the disruption of the organisation of the tissue in which the tumour arises.3Boveri, not having received a medical training, is very conscious of his position as an outsider. One cannot escape the impression that he must have met with a good deal of incomprehension from the medical fraternity. It is remarkable that his views on the origin of malignant tumours are guided almost entirely by his work on sea urchins and intestinal round worms.I shall try to expound my thoughts in the briefest possible manner. I do not regard it as incumbent on me to measure my views against those of other authors. If my comments contain anything useful, it will be obvious without detailed discussion. There is only one author whom I need mention and that is David Hansemann(4), who long ago and on many occasions put forward ideas that are closely related to my own. Since his views on the significance of abnormal chromosome constitutions in the cells of malignant tumours have hardly been taken seriously, as far as I can see, what I have to say might seem doomed at the outset. Nonetheless, for all their similarity, there are such great differences, both in conception and in argumentation, between Hansemann's ideas and my own that the hypothesis I am now putting forward might perhaps contain precisely those elements that are missing in Hansemann's own.4David Hansemann (1858–1920), later ennobled as von Hansemann, is generally regarded as the originator of human cancer cytogenetics, although Hansemann himself acknowledges Hauser as a precursor. The chromosome counts that Hansemann made on mitotic figures in human cells fell far short of the actual number of chromosomes present, but he freely admitted the inadequacy of the techniques then available. He confirmed and extended the observations of several earlier workers who had noted the presence of abnormal mitotic figures in tumours, and he appears to have been the first to argue that the abnormal chromosome constitution produced by aberrant mitoses was an essential determinant of malignant tumorigenesis.Since no one will regard this essay as a source of information about the present state of tumour research, I shall cite only those contributions that workers in this field might not know about.At this point I should like to express my warmest thanks to my former colleagues in Würzburg, Professor M. Borst in Munich, Professor R. Kretz in Vienna, and also to the present holder of the chair of pathological anatomy of our university, Professor M.B. Schmidt, for their many valuable suggestions.The following observations on the essential nature of tumours seem to me to be best supported by the evidence. The cells of even the most malignant tumours can be formed from normal tissue cells. The determinants of this abnormal behaviour are to be found in the tumour cells themselves, not in their surroundings(5). Although benign and malignant tumours have many properties in common, I have to agree with those authors who draw a sharp line between the two. In my view, there must be a fundamental difference between a tumour that grows in the same way as the tissue from which it is derived and one that does not. It seems to me that the transformation of a benign tumour into a malignant one, so often described, is a phenomenon of the same kind as the appearance of a malignant tumour at some site in normal tissue (6).5This is a further declaration of Boveri's position with respect to the two views discussed in Note 2. Theories that argue that malignancy arises not from an event or events taking place within the cell, but from disruption of tissue architecture, are sometimes referred to as `field' theories [see, for example, Willis, R.A. (1948) Pathology of Tumours, Butterworth]. They have a long history, but have not found general acceptance in modern times because they do not account for the demonstrable clonality of many malignant tumours. Boveri, immersed in the study of chromosomes and guided by the belated flowering of Mendelian genetics, naturally assumes that malignancy, which is a heritable condition at the level of the cell, is determined by events that involve chromosomes, the carriers of the cell's heredity.6Here, Boveri confronts another problem for `field' theorists: the stochastic nature of tumour formation. Boveri's own work on chromosomes, which is discussed in some detail later, leads him to conclude that the genesis of a malignant tumour is a rare cellular event. He is arguing here that when a benign tumour becomes malignant that change is again initiated by a rare cellular event within the benign tumour and not by wholesale transformation of a large number of benign cells.This statement does not cast doubt on the possibility that malignant tumours might sometimes arise from cells that are retarded in their histological differentiation (so-called `immature' cells). But such cases seem to me to be exceptions compared with those in which the more pronounced independence of malignant cells is a secondary phenomenon determined by the loss of properties that were present at an earlier stage (7).7Boveri is not contending that differentiation and cell multiplication are mutually exclusive states, a primitive notion that sometimes sees the light of day even now. He is drawing a distinction between the possibility that a malignant tumour might grow out of a cell that is `immature' for some other reason, and the idea that malignancy, caused by the deletion of some normal cellular component, might later incur a loss of specialised functions.The essential elements of my point of view may therefore be summarised as follows. A malignant tumour cell is a cell with a specific defect; it has lost properties that a normal tissue cell retains (8). In this respect, I am in complete agreement with the concept to which Hansemann has given the name `anaplasia'. A cell in this drastically altered state reacts differently to its environment, and it is possible that this alone might account for its tendency to multiply without restraint. Such unrestrained proliferation is no doubt a very primitive property of cells (9). Woodruff (Woodruff, 1913) has recently reported that, over a period of five years, he grew 3340 generations out of a single Paramecium under conditions where conjugation could not have taken place and without any special artificial stimulation. Woodruff has calculated that if all the individuals in this culture had been kept alive, the amount of protoplasm generated from this one organism would have exceeded the mass of the planet earth by a factor of 101000.8At this point, Boveri nails his colours to the mast: malignancy represents a loss of cell function, not the gain of a new function. The growth of the oncogene industry initiated more than half a century after the publication of Boveri's monograph was based on the assumption that oncogenes induce the gain of a new function that acts in a genetically dominant fashion. This view was strongly reinforced by the findings of tumour virology. Here, the assumption was that the virus, by introducing new genetic information into the cell, induces a new function that would also be expected to behave in a genetically dominant fashion. These views were, of course, incompatible with Boveri's insistence that malignancy was due to a loss of normal function. Experimental evidence in support of Boveri's position had to await the introduction in 1965 of cell fusion as a technique for studying the genetics of somatic cells. By means of this technique, it could be shown that normal cells contain genes that are able to suppress the malignant phenotype. Malignancy was thus seen to be a recessive and not a dominant character in the classical Mendelian sense and patently represented a loss of normal cellular function, not the gain of a new function. The discovery of genes that can suppress the growth of malignant tumours led eventually to the displacement of a sea of oncogenes by a tidal wave of tumour suppressor genes.9One way of looking at the multiplication of cells is that exponential multiplication and not `rest' is the inherent steady state of all cells. As a consequence of evolution, all cells are so constituted that, given adequate nutrients and a clement environment, they will multiply exponentially without further stimulation. This idea evidently appeals to Boveri, who uses the data provided by Woodruff for dramatic effect. The key words are `without artificial stimulation'. Boveri obviously believes, and so do I, that cells will go on multiplying of their own accord unless they are in some way restrained. There has been some controversy about the origin of this idea. It seems to have been first set down in print in the English-language literature in the 1950s, but it was apparently familiar to German biologists before the First World War.It is only at the stage when the cells of solid tissues can be divided into those that merely propagate themselves and those that assume special functions that the latter stop multiplying exponentially. The cell ceases to be an egoistical entity and becomes an altruistic one, in the sense that it does not multiply except when the needs of the whole organism require it(10).10This colourful anthropomorphic metaphor demonstrates once again that Boveri is not asserting that differentiated cells stop multiplying in metazoa, but that multiplication in metazoan cells is normally governed by the localised requirements of the body whereas malignant cells no longer recognise this form of control.If now we regard the malignant cell as one that has lost certain properties and hence its normal reactivity to the rest of the body, then this change may well be enough to induce an altruistic cell to revert to its egoistical mode and thus release its multiplication from restraint. (Relapse of `organotypic growth' to `cytotypic growth', to use the picturesque terminology of R. Hertwig.) But it is also possible that, in the tissue cells of metazoa, special inhibitory mechanisms have developed that have to be eradicated before unrestrained multiplication can take place(11). Be this as it may, both possibilities assume that elements present in the normal cell are missing in the malignant tumour cell. The following observations are an elaboration of this concept.11This formulation of the central idea gets even closer to the concept of tumour suppressor genes. Boveri envisages the existence of cellular mechanisms that normally exert a facultative control over cell multiplication and that are eliminated or impaired in malignant tumour cells.The thoughts briefly outlined in the preceding section give rise to the question: how can something be removed from a cell, and what are the consequences of such a deficit? Chemical and physical interventions may perhaps destroy certain cellular components without impairing the viability of the cell. I shall have something to say about such possibilities at a later stage. At this point, I am going to talk only about deficits created by the mechanical removal of parts of the cell. Fragments of cytoplasm can easily be removed from many kinds of cell, especially protists and egg cells. The numerous experiments that have been done to investigate this problem have, in general, reached the conclusion that in every bit of cytoplasm all the properties of the cytoplasm as a whole are latent, or, at least, that under the influence of the nucleus they can be regained. Any fragments taken from a protozoon, so long as they contain a nucleus, will regenerate complete animals, and nucleated fragments taken from eggs will produce normal embryos.12This section is a dramatic illustration of just how much of an outsider Boveri was in the field of cancer research. The observations described in it were made entirely on sea urchin eggs, experimental material introduced by Oskar Hertwig (1849–1922) and exploited by Boveri with great virtuosity. It was on the analysis of sea urchin eggs that Boveri based his three great principles of chromosome behaviour: the individuality of chromosomes, their continuity from one cell generation to the next, and the different capacities of individual chromosomes to transmit heritable traits.However, there are exceptions to this rule. In many kinds of egg, the process of differentiation is such that fragments taken from them generate fragments of embryos. Nonetheless, it is notable that, even in such cases, the cells that grow out of the egg fragment are not abnormal or sick. What happens is that the fragment can generate only certain sorts of cells. And even with eggs that behave in this way, the oocyte from which the egg is derived turns out to be totipotent. Nucleated oocyte fragments give rise to normal dwarf embryos.There is no reason to believe that tissue cells behave differently. Here too, a fragment containing a nucleus will retain the ability to regenerate the whole cell so that, in all probability, it is impossible – provided the cell survives the operation – to produce a permanent deficit in the cell by removing a bit of its cytoplasm.In the case of the nucleus, similar experiments yield an entirely different result.It is perhaps not inappropriate if, before discussing these experiments, I set out briefly the more recent findings concerning the structure of the resting nucleus and the contribution of the chromatin in somatic cells.At fertilisation, two nuclei are brought together and (except in one particular respect that is of no interest to us in the present context) they are alike in the make-up of their chromatin. Not only do each of the two nuclei contain the same number of chromosomes but also, in favourable experimental material, it can be shown that, where a sperm contains chromosomes that have a characteristic size and shape, corresponding chromosomes of the same size and shape will be present in the egg. So one can draw the quite general conclusion that every chromosome in the sperm nucleus has its homologue in the egg nucleus.Let us label the chromosomes in the one nucleus a, b, c and d. The other nucleus will also contain the same set, a, b, c and d. At fertilisation, the two haploid nuclei are amalgamated into one diploid nucleus, which now contains 2a, 2b, etc. Each chromosome in this duplicate set is then split into two, and a tightly controlled karyokinetic separation(13) of the daughter chromosomes ensures that a complete duplicate set is inherited by each of the two cells produced by the cleavage of the fertilised ovum. In the resultant resting nuclei, the individual chromosomes seem to break down. But we have every reason to believe that, within the stroma of the resting nucleus, every chromosome that contributes to the makeup of that nucleus continues to exist as a discrete region that reappears as the same `chromosome' when the cell prepares to divide again. (Theory of the individuality of chromosomes.) In this way, the two sets of chromosomes amalgamated at fertilisation are inherited by all the cells of the individual. It is only in the germ cells that the so-called reduction division converts the duplicate set once more into a single set.13Longitudinal splitting of the chromosomes and the details of karyokinesis were described by Walther Flemming in 1879 [Flemming, W. (1879) Arch. f. mikr. Anat. 16, 302]. The didactic exposition given here by Boveri would have been old hat for most biologists in 1914, but apparently doctors deeply immersed in the study of cancer needed it.This symmetrical transmission of the chromosome constitution of the one-celled embryo to all the cells of the body is only possible if the mitotic figures are bipolar. If three or more poles take part in the mitosis – in the observations that follow we are going to assume that there are four poles – then the daughter cells will inherit an abnormal, and in its details an extremely variable, combination of chromosomes. The principal reason for this is that each chromosome can only split into two identical halves. Thus, if four poles are present, only two of the four daughter cells produced at the one time can receive the split product of any particular chromosome; the other daughter cells get nothing from that chromosome.In addition to this, there is another important issue that must be taken into account. It is a matter of chance (14) which two of the four poles make contact with a particular chromosome, so that the four daughter cells not only have different chromosome numbers, but also have different chromosome combinations.14A further affirmation of Boveri's conviction that cancer is initiated by a stochastic event.Let us assume, for diagrammatic purposes, that a cell contains eight chromosomes – 2a, 2b, 2c and 2d – and that it forms four poles instead of the normal two. One of the possible outcomes is shown in Fig. A. If we examine pole 1, we see that it makes contact with both chromosomes a, one chromosome b, one chromosome c and one chromosome d. From each of these chromosomes, pole 1 captures one of the two daughter chromosomes produced. The daughter cell formed around pole 1 thus receives the chromosome constitution 2a, b, c, d, as shown in Fig. B. In the same way, out of the configuration shown in Fig. A, each of the three other daughter cells will acquire the chromosome constitution shown in Fig. B.It follows, first, that out of the same chromosome complement, a tetrapolar mitosis delivers a smaller number of chromosomes to the daughter cells than a bipolar mitosis: in our case, an average of four instead of eight. Second, it rarely happens that a daughter cell actually receives this average number; in general, some daughter cells get more than that and others get less. Third, it must happen that one or other of the daughter cells, and under some circumstances all four, fail to receive certain chromosomes. Thus, in our case, only daughter cell 1 contains a copy of each kind of chromosome: in cell 2, a and b are missing; in cell 3, a is missing; and, in cell 4, c is missing.In the foregoing case, the cell has a typical normal chromosome number but nonetheless has four poles. This does actually occur in nature and is probably a result of abnormal division of the centrosome.The simplest way to produce a tetraploid mitosis is double fertilisation, which can be achieved in sea urchin eggs, for example, by adding very large amounts of sperm. Because each of the spermatozoa introduces one centrosphere that divides into two, four poles are formed. But these four poles are faced, not with a duplicate set of chromosomes, but with a triplicate set: one in the egg nucleus and one in each of the sperm nuclei. In this case, the chances that one of the four daughter cells might contain a copy of each kind of chromosome are much better.Another way to generate tetrapolar mitosis is to suppress cell division in midstream. This can be done with eggs and blastomeres by pressing them together and shaking them. In this case, the two centrosomes and both groups of daughter chromosomes, instead of being allocated to two cells, are held together in the undivided mother cell and then proceed to form the new resting nucleus just as if cell division had been completed. When this composite double cell gets ready to divide, each of its duplicated pairs of chromosomes divides into two, and four poles are formed; but the nucleus now gives rise to four sets of chromosomes. As a result, each of the four daughter cells, formed at the one time, will receive, on average, the normal number of chromosomes, and the chances that at least one copy of each kind of chromosome will be present are again much better.These observations are meant to show that, in multipolar mitoses, we have a means of achieving something that can otherwise hardly be done without collateral damage to the cell, namely the production of nuclei in which some parts are missing. And, in this way, we can answer the question that we posed in connection with the cytoplasm. What are the consequences of such a deficit in the case of the nucleus? To put the question more precisely: can a nucleus with a deficit of this kind regenerate what is missing and, if it cannot, can it remain viable without it?The answer to the first question is that, as far as we know, even bits of chromosomes cannot be replaced. An abnormal chromosome number is inherited by all daughter cells provided that all subsequent mitoses are bipolar. The answer to the second question is that the overwhelming majority of cells with nuclei produced by multipolar mitosis are sick and perish(15).15This indicates that Boveri was well aware that aneuploidy was not in itself the precipitating cause of cancer, a confused view that one still sometimes hears today. Boveri regarded the chromosomal variability produced by aneuploidy as a mechanism that greatly increased the probability that cancer, a rare and stochastic event, might occur.To determine whether this ruinous effect that multipolar mitoses have is really a result of an abnormal combination of chromosomes, I have carried out a large number of experiments with doubly fertilised sea urchin eggs. I gave a short account of this work in 1902 and described it in detail in 1907 (Boveri, 1907)(16). Here, my account of this work must be quite brief.16The paper referred to here is the last of the classical series begun in 1887. This work laid the foundation of our modern understanding of chromosome cytology.Three variants of dispermy can be distinguished in sea urchin eggs. We can call them the tetraaster type, the triaster type and the double spindle type.The developmental prospects of these three types of dispermy are very different. A dispermic egg with a double spindle, assuming that the egg divides into four, will generate one normal larva showing, at worst, a certain asymmetry. A dispermic egg of the triaster type generates, in addition to many completely pathological products, a number of partially or wholly normal embryos. Dispermic eggs of the tetraaster type, almost without exception, develop in a completely pathological manner; only a few partially normal larvae are found.The first conclusion to be drawn from these differences in behaviour is that the three- or fourfold partitioning of the cytoplasm caused by dispermy cannot be responsible for the damage; for, if it were, no normal larvae at all would be produced by the dispermic egg. Nor can the differences we have described be explained by the abnormal chromosome numbers that the individual cells receive as a consequence of the multipolar mitosis. Furthermore, we know from experiments on merogony and partial fertilisation that eggs or parts of eggs containing as few as half the regular chromosome number can develop normally.So, to explain the extremely variable outcome of double fertilisation, there is only one assumption left, namely that it is the wrong combination of chromosomes that makes dispermy so ruinous for the embryo. Put simply, the individual chromosomes must possess different properties such that only certain combinations permit the cell to function normally or, at least, keep it alive.Both the egg nucleus and the sperm nucleus contain the right combination of chromosomes, which we have labelled a, b, c and d. This at once explains why, with a double spindle, the egg develops normally. For, in this case, two of the four primary blastomeres receive the derivatives of the normal nucleus of a fertilised egg; the other two receive the derivatives of a sperm nucleus.The reason why, almost without exception, dispermic eggs of the tetraaster type develop into pathological forms is, in our view, because it is extremely unlikely that a tetrapolar mitosis would deliver at least one copy of all the different kinds of chromosome to each of the four primary blastomeres. With the same chromosome constitution and only three poles (the triaster type), the chances of this happening are appreciably better, which explains why one gets quite a few normal larvae from dispermic eggs of this type.For further details and more substantial evidence, I must refer the reader to my specialist publications. In the last of these, in 1907, I remarked how closely the different ratio of normal to abnormal outcomes accords with the probability that the chromosomes are appropriately or inappropriately distributed(17). I showed, moreover, that in many dispermic embryos only a proportion of the primary blastomeres generate pathological derivatives and the rest do not; this, on our assumption, is to be expected. Finally, I tried to justify the conclusion that, besides the chromosomes that are indispensable for the life of the cell, there appear to be some whose absence does not limit the viabilit