ON T H E D E F I N I T I O N OF L I F E
ABEL SCHEJTER AND JOSEPH AGASSI
SUMMARY. Schr6dinger's definition of life needs a slight modification to absorb the criticism of it. It is the comparison of the entropy level of a system before and after a process which makes one view it as living: we consider the stability of the deviation from the probable a sign of life. This explains why we do not hesitate to consider as remnants of living systems skeletons and fossils anywhere and physical culture on any archeological site.
Key words: definition of life, Schr6dinger.
The need to give a definition of life is questioned by quite a few writers; much of their objection may be put aside by the formulation of such a definition as a hypothesis to be critically examined. Some objections to such hypotheses are based on the view that any such hypothesis will be reductionist, i.e., will make biology part-and-parcel of chemistry and physics. This need not be so: E. Schr6dinger, in 1946, attempted a definition aimed at stressing the non-reductive part of the definition (Schr6dinger, 1946). As we shall soon see, his attempt was not a complete success. Yet some improvement will be offered in the following pages which m a y do just that: define life as non-reducible to physics and chemistry. What is required of an adequate definition? Apart from its not being trite and uninformative (circular, to use a traditional term), it should be neither too wide nor too narrow; it should not exclude living things and it should not include dead ones. Somehow, this is highly troublesome since it looks quite unenlightening to tell us that what we know to be living is living and what we know to be inanimate is inanimate. Attempts to get over this problem were made in the direction of narrowing the borderlines; viruses are notoriously borderline cases, perhaps the only ones; perhaps computers are borderline cases. Yet borderline cases are usually disputed before the definition is offered and so judgment concerning them may be biased. This is not to say that a definition cannot be criticized. Any theory whose description is too wide m a y be criticized, and the conditions it puts may be deemed necessary but not sufficient. The classical example is Samuel Butler's successful application of the theory of natural selection to machines, which proves Darwin's theory at best a necessary, but not a sufficient, description of the evolution of the living species (Butler, Erewhon, 1872). Nevertheless, a definition may be too narrow or too wide and in a challenging way. Even if we had - and we do not have - a definition which is adequate in all respects, declaring certain servo-mechanisms as Journal for General Philosophy of Science 25: 97-106, 1994. © 1994 Kluwer Academic Publishers. Printed in the Netherlands.
98
ABEL SCHEJTER AND JOSEPH AGASSI
alive might be a challenge. This, indeed, is what general-systems theorists claim: a self corrective complex system is alive. There is a series of refutations of this hypothesis; yet a modification of it, not open to standard criticisms of general-systems theory, will be offered below. With these preparatory discussions, we are now ready to discuss Schr6dinger's definition, its criticism, and its modification. In what sense are biological phenomena different from physical phenomena? We may analyze biological phenomena in as much detail as we are able to - and here the limiting factor is a technological one - and explain each and every one of them in terms of physics and chemistry. Therefore, when we speak about the distinction between physical and biological phenomena, we are assuming that the latter include the former, do not escape their laws, but cannot be explained completely unless we add new laws. What may these laws be about? To answer this question, we start with Schr6dinger's answer: living organisms stay alive by virtue of their ability to get rid of the entropy that is created by the processes by which the organisms live. Now, let us examine the content and the consequences of this hypothesis; and let us do it by performing a thought experiment, a modification of an idea originally suggested by Bridgman (1961). Assume a system that is chemically isolated and at thermal equilibrium. By making the system sufficiently large, we may consider the change in its temperature negligible. In the initial state, the bottom of our system contains a pool of water, and dissolved in it are amino acids, nucleotides, glucose, other sugars, fatty acids, phosphate, vitamins, inorganic salts, etc. In addition to this, the system has, in the initial state, a fertilized egg of a cat. We shall dispense with the problems of embryological development and assume that ontogenesis can also occur in this test tube situation: that is, given the information contained in the fertilized egg, it will develop and eventually become a cat. The mass of the growing cat will increase by incorporation of the metabolites present in the solution. By making the system sufficiently large, we may simplify matters. Let us assume that our cat is an anaerobic organisms and that it is unable to synthesize its own amino acids, purines, pyrimidines, etc., but it is able to absorb them directly from the aqueous solution and build from them nucleic acids and proteins. Any energetic costs incurred by these processes can be paid in some unspecified way by the commonly used chemical currency, ATP hydrolysis. Thus, essentially, the chemical reactions up to this point are: glucose amino acids nucleotides sugars
, , , ,
2 lactate proteins + H20 nucleic acids + H20 polysaccharides + H20, etc.
(1)
Hence, some structures have been formed, some polymers synthesized, some small molecules consumed and some water produced. Our cat will remain
THE DEFINITION OF LIFE
99
alive for as long as there is food - namely, organic molecules, vitamins, salts, etc. - in the pool where it lives. Now, at some point the food will be entirely consumed, and the cat will eventually die of hunger. The polymeric structures formed in equations (1) will be dissociated into monomers, since this is the spontaneous direction of the reactions described by equations (1). These processes may be extremely slow; but what matters to our experiment is not the rate, but the thermodynamic direction of the events. Eventually, we shall be left with a system containing only small molecules. This is our final state, at the same temperature as the initial state. Hence, the overall process can be divided into two: (I) (II) (II)
glucose cell + monomers water + cat
, , ,
2 lactate + heat water + cat monomers
,
lactate + heat (3)
(2)
which, when added up, give: glucose + cell
If the mass of the system is large enough, we can neglect the contribution of the original 'cell' to the overall entropy and enthalpy of the system. It is clear that the total energy of the system plus environment has decreased, while the total entropy has increased. Bridgman (1961) proposed this experiment as a test of the applicability of the second law of thermodynamics to living organisms. He had few doubts regarding the outcome of the experiment, although he was careful to point out that the operations necessary to assign a certain entropy to a living object, and thus to verify qualitatively the second law, were equivalent to the laboratory synthesis of a living system. Bridgman (1961) was highly critical of the fuzziness of thought prevailing at the time of the writing of his book, in 1941, around the subject of order, entropy and life. Much of this fuzziness was dispelled a few years later by Schr6dinger's (1946) answer to the question: what is life? Interestingly, in the preface to a later edition of his book, Bridgman (1961) did not mention Schr6dinger's result; instead he pointed to the progress made in irreversible thermodynamics by de Groot, Prigogine and Onsager. The theoretical work of these authors led to our present understanding of fluctuation phenomena, and its applicability to the problem of life, as we shall soon see. Let us return now to the experiment. At a certain time during the process, the system contained a live cat. We might have said, then, about the system, that it contained some highly organized structures. Had we wished to speak the language of statistical thermodynamics, we might have said that these structures were improbable and, in terms of Boltzmann's relationship, that we were observing a local decrease in the entropy of the system. Thus, in the history of the system and its environment, we have described an intermediate state, characterized
100
ABEL SCHEJTER AND JOSEPH AGASSI
by a local decrease in entropy, and maintained for as long as a source of energy was available. Once the energy supply ran out, the further evolution of the system led to a final loss of its characteristic region of lower entropy. Now, it is well known that we can describe chemical systems that are amenable to a parallel description of their thermodynamic history. The classical exemple of this is the Zhabotinsky reaction (Zhabotinsky, 1964). Imagine that in the initial state, a system contains bromate, malonic acid and salts of cerium in aqueous solution. After some time, we shall be able to observe the existence of colored concentric fringes, indicating the formation of a structure in which Ce 3+ and Ce 4+ ions have become separated. The chemical reactions describing the system are now, e.g. (Degn, 1967): (CHz(COOH)2 + 6 Ce 4+ + 2 H 2 0 10 Ce 3+ + 2 HBrO3 + 10 H ÷
, 2 CO2 + H C O O H + 6 Ce 3+ + 6 H + , 10 Ce 4+ + Br2 + 6 H20 (4)
When either the bromate or the malonic acid run out, these reactions will cease to occur; the final products of the reactions, including the ions Ce 3÷ and Ce 4+, will now distribute themselves equally over the entire volume and the structure will disappear. Historically, the system has developed from non-structure to structure and back to non-structure. We can say that the fringed structure - or, what is equivalent, the presence of regions of low entropy - was maintained at the expense of the energy liberated by reactions (4). These equations describe the overall chemical process. The entropy of the ensemble of system and environment has increased, since heat was produced by the chemical reactions. We observe, then, the parallelism between the description of the cat system and the Zhabotinsky reaction system. Of course, the chemical processes are very different in nature; each of these can be analyzed in as much detail as our technology allows. What is more significant, however, is that the structural history of one - e.g. the cat system - can be described in the same terms as that of the other - the chemical system - since, essentially, all that occurs structurally is that certain molecules or ions become distributed in certain particular ways. For example, for a portion of the time along the history of the system, the ceric and cerous ions are distributed in a particular structure of concentric fringes; while in the cat, the amino acids and nucleotides are, for a portion of the time, distributed in a particular structure of ordered polymers. It is also interesting that the structured ceric and cerous ions play the role of catalysts in the chemical system; while the structured amino acids, the proteins, play the role of catalysts in the cat system. In summary, from the point of view of their thermodynamic history, the eat system, which is biological, and the Zhabotinsky system, which is chemical, can be described in identical terms. In other words, up to this point life can be reduced to chemistry and both can be explained in the language of physics. We could even go one step further and get rid of the chemistry too.
T H E D E F I N I T I O N O F LIFE
I01
For example, in Benard's famous experiment (Chandrasekhar, 1961), a beaker containing water is heated from the bottom and after a certain time, the lines perceived at the borders between different layers coalesce into a definite cellular-like structure that persists for as long as the temperature gradient across the beaker is maintained. That is, we can create a structure - a localized region of lowered entropy - at the cost of energy, in a system where there is no chemical change, in the sense that no (covalent) bonds are broken or formed. The thermodynamic history of the Benard system is identical to that of the cat system or the Zhabotinsky system, and from this thermodynamic standpoint biology can be reduced directly to physics. Clearly, the systems we have described correspond to the well-known dissipative structures of irreversible thermodynamics, which have been analyzed in much detail by Prigogine and others (Prigogine and Nicolis, 1971). These phenomena, beautifully described in the expression "order through fluctuations", are so striking that Prigogine referred to them as a "new state of matter". If we went one step forward, we could identify this "new state of matter" with "living matter". The preceding discussion is very comfortable to the reductionist: indeed, it seems to say that cat-Zhabotinsky-Benard structures are merely a somewhat esoteric property common to all types of matter, under some particular - far from equilibrium - conditions. The major problem that the reductionist may now face is that of explaining how the experiment - cat, Zhabotinsky or Benard - was set up spontaneously. We understand that the spontaneous setting up of each of these experiments by pure chance is highly improbable; but this is a matter that does not concern us at this point. All we wanted to show in our reductionist exercise is that, if life is taken literally as defined by Schr6dinger in thermodynamic terms, then we can also consider as living, systems that from these considerations we usually define as non-living. We conclude that Schr6dinger's definition expresses conditions that are necessary to distinguish between life and certain types of purely physical or chemical systems: structures arising as "order through fluctuations" are clearly comprised in SchrSdinger's definition. To see how we can modify it, we must turn back to Bridgman's thought experiment. Bridgman wrote of a spore germinating in a system like the one we described before, but he added to it a specific quality; Bridgman's organism, being "geotropically negative", climbed to a shelf located at a certain height, where, eventually, it died. We know that the final fate of the organism's remains must be the total hydrolysis of its polymeric structures, since there is enough water in the system to accomplish the degrading process. In this final state, we shall find a solution of amino acids, etc., in a pool that is now at a higher level than it was in the initial state. Thus, the final outcome of life is now that a certain mass of matter has been transported to a higher level. Naturally, this has been paid for by the metabolic energy.
102
ABEL SCHEJTER AND JOSEPH AGASSI
The fact remains, however, that also in the final state, we can distinguish a local region whose entropy has been permanently lowered, since raising a mass to a certain height is the opposite of a spontaneous process. We now face a number of questions. First, is the thermodynamic history of this system in any way distinguishable from analogous, non-biological systems, such as Zhabotinskylike reactions? The answer to this is, yes. Thinking about a plant growing upwards, or a cat jumping to an upper level, does not overtax our imagination; quite to the contrary, this represents our c o m m o n experience of observing living beings moving against gravity, or against gradients, above the limits of the systems that contain them. However, the image of a mass of non-living matter moving to an upper level spontaneously goes directly against our experience, even if we specify a source of energy that can be spent to compensate for the change. In a classic exemple, we are prepared to see a stone falling spontaneously on a piece of ice, and some of the ice melting into water; but we never expect the water to freeze and the stone to rise back spontaneously to its original level. We don't expect to observe the spontaneous upward movement of a non-biological system, even when a Zhabotinsky reaction or a Benard gradient are taking place in it; but the observation of a living animal jumping upwards spontaneously is not contrary to our experience nor to our expectations. Since the animal is carrying with it local regions of lowered entropy made up from matter from its environment, its movement against gradients of gravitational, chemical or other forces, will eventually leave a permanent trace, even after the transported matter is chemically decomposed and finally returned to its original high entropy state. We conclude then, that the thermodynamic history of a living system can be distinguished from that of all possible non-living systems, including non-living local instabilities. The distinguishing factor is the appearance of a permanent structural modification, that is, a difference between final state and initial state, through which the entropy of the system has decreased. This is because the temporary (dynamic) local instability - the living organism - has originated a permanent (static) modification in the environment. Only living systems, we say, are able to produce such a sequence of events: this is the modification that we propose to add to Schr6dinger's definition of life, so that it includes only life and excludes all other "order through fluctuation" systems imaginable. The amazing fact about the argument from this thought experiment is its triviality. After all, it is a commonly known fact that we never doubt the origins of fossil skeletons and shells and trees to be living things. We likewise never question that a find of an archaeological site is part-andparcel of a material culture. And when we find a clock, we surmise that it did, indeed, have a maker. Now, this is Paley's celebrated argument. (He took it from Robert Boyle who claimed no originality for it yet christened it "physico-theological".) Yet Paley was evidently in error: when we see
THE DEFINITION OF LIFE
103
a watch, we surmise its maker, but not so when we see the solar system. Boyle was in error when he said it was a clock. Of course, it does serve us as a clock, yet we deny that it is designed as a clock and so deny the physico-theological argument (also aptly known as the argument from design). We see here that living systems are taken to be those which render improbable processes probable - whether by raising stones, by developing skeletons or by designing and making clocks. Yet the design of the solar system is a different matter, it seems, since it creates nothing improbable, and since with the demise of the solar system, nothing improbable will result, we assume. H a d we assumed the contrary, then perhaps we would thereby grant 'old Sol' a soul! The stability of the solar system is obviously a necessary condition for life on earth as we know it, yet the stability is assumed to depend on certain initial conditions and thus to be t e m p o r a r y on a cosmic scale. On the solar scale, we see systems as dissipating to the full, or not to the full, and the latter we do in fact deem living: not the initial conditions but the final conditions force us to do so, or rather the final conditions as c o m p a r e d with the initial conditions. Thus, a skeleton, though improbable, is dead; yet the system ending up with a skeleton is alive, just because it built itself from scratch. Of course, the skeleton need not be presevred. It is the ability to create it and for it to stay preserved that we usually do take as a sign of life. Why then was this fact not introduced into the centuries-old debate about life? Perhaps because physicists paid attention to initial conditions and biologists to final ones - the cannon ball and the jumping cat - they both failed to see that contrasting initial and final conditions is what we do when declaring a fossil or an archeological site remains - whether the remains of an animal ot the remains of h u m a n material culture. If so, then here is a whole new area of investigation opening up. An objection to the view presented here m a y be the argument that a volcano is as improbable as a fossil and as stable an outcome of a process of energy dissipation. The answer to this argument is that the facts are not described properly in the argument. We all assume - rightly or wrongly - that an eruption leads to diverse topological remains, randomly distributed; we all assume each snail to end up with a dead shell. H a d all volcanic eruptions ended systematically in an improbable result, we would, it should be assumed, consider earthquakes signs of life. What makes a snail different from the volcano is its ability to channel energy, not contrary to laws of thermodynamics, yet with systematic improbable outcome. The second objection rests on the ambiguity of our answer to an obvious question. It is the question that we face in the application of the modified definition to different systems: which of them are alive? Obviously, plants, animals and bacteria are alive. Are viruses living systems? The answer is, unambiguously, yes. Assume that our original pool contains, not a fertilized
104
ABEL SCHEJTER AND JOSEPH AGASSI
egg, but a virus. A virus is unable to carry out a biosynthetic function independently, except for those reactions by which it builds copies of itself. In this sense, however, the virus is not so different from the cat, whose metabolism and biosynthetic processes were qualitatively limited by the variety of small molecules originally present in the pool. It will continue building more copies of itself for as long as it can be able to find the building blocks of its D N A or RNA and proteins, and for as long as there are energy-rich molecules available for this purpose. Furthermore, viruses are sometime transferred from infected to non-infected cells. Consequently, we may conceive also their possible transfer from an infected to a noninfected pool. Assume that the system contains two pools, with identical concentrations of the necessary compounds, and a virus in one of them. The virus will start and continue building copies, until one of the essential chemicals will be entirely consumed. Some members of the population may then be transferred to the other pool, and continue their replication. Eventually, the replication will have to stop here also. The net final result will be the (entropically unfavorable) permanent transfer of matter against a gradient. This, we argue, is the unique characteristic of life. What about a computer? Assume that there is a computer that extracts energy from a solar battery, and that has been programmed to perform certain logical operations and store their results. Physically, this is done by performing alterations in, say, a magnetic tape. The outcome of this physical change is a local decrease in the entropy of the tape. Furthermore, there is a limit to the storage capability of the magnetic devices so that if we want the computer to continue performing and storing its calculations, we must provide it with more blank tape. Now, we may imagine a robotic computer that is able to feed itself with blank tapes that it finds in its environment, and is able, at the same time, to move and search for stores of blank tape. The robot will follow a certain path, picking up tapes and filling them with information, and will stop only when there are no more blank tapes left around. However, the filled tapes themselves are not essential for the activities of the robot; they may as well be discarded, and there is nothing that demands that they should be discarded in a way which is different - more organized - from the way in which they were originally spread about. Thus, given our ability to observe the initial and final states of this system, we shall not be able to recognize, at the end of time, that life was there. This is simply because, though the computer was able to move in search of matter, as the plant moved its tissues upwards, or the cat jumped to a higher pool, it did not incorporate the converted matter - the temporary record of life processes - as an integral part of itself. Now, we may obviously imagine a robot programmed so that it feeds on magnetic tape, builds copies of itself out of the material of the tape, and searches for tape in the environment. We could hardly distinguish its nature from that of a virus. What is more difficult to imagine is what purpose whould serve to such a robot the ability to perform logical operations
THE DEFINITION OF LIFE
105
and print their results on the tape. Hence, a robot p r o g r a m m e d to make copies of itself we deem 'alive' within the context of our definition, but the added computer capability is not essential to robotic life. True, one may say that the computer capability may be useful for the robot to make decisions as to what direction is better to follow in the search for tape; that one species of robot will reproduce faster than another with less computer capability, or with none. But this is another story: life as we define it is already inscribed in the original program, and the computer capability is not essential to the performance of the program. The third question that we face is: what are the properties of non-living systems? Schr/Sdinger himself hinted at the existence of such properties in the last chapter of his book. He believed that the biological principle or law was not alien to physics, but was to be found as a consequence of quantum theory. We deal now, not with the microscopic fundament of the law, but with its macroscopic consequences. We assert that the macroscopic thermodynamic property that characterizes living systems is the ability of their transformed matter to move in search of sources of matter and energy. A Zhabotinsky reactor is limited to the matter from which it is built; once one of its chemical components is consumed, it ceases to react and its structure is destined to dissipate within the same set of physical coordinates where it was originally formed. Were we able to add to it the ability to translate itself to a different set of coordinates where more malonic acid or bromate are available, we could say that we had created a living system. The same goes for a Benard system in which the source of heat runs out: the structure will disappear, and the system will be unable to move by itself in search of another source of heat. We agree with Bridgman's (1961) criticism of animistic thought regarding living systems, and the attribution of 'volition' to them as their specific quality: yet we must distinguish clearly between this type of vitalism and our definition of living systems. Indeed, our definition itself is hidden in Bridgman's attribution of 'negative geotropism' to his organism. Better, then, to recognize it for what it is: a property that could never be attributed to a growing crystal or to a stalagmite. The spontaneous ability to utilize energy in order to maintain structure is not enough to characterize a living system: we must add to it the ability to add new matter to the structure and the ability to move in search of appropriate sources of matter and energy. This ability may be, but is not necessarily, equated to the ability of an organism to change its physical coordinates: plants or sessile animals able to attract microorganisms or molecules of food by operating on the surrounding fluid, are certainly alive. When we refer to their movement, we consider living beings as placed in a special coordinate system: a system of ecological coordinates or a biosphere. We state that the characteristic of living beings is that they are able to transform their ecological coordinates in the biosphere, so that they exist at all times in a viable ecosystem: one where energy and matter
106
ABEL SCHEJTER AND JOSEPH AGASSI
appropriate for the maintenance of a particular kind of life are available. In this context, ecology becomes not just a series of descriptions of particular environments, but an essential element of the definition of life. Let us summarize now the properties of the different types of living organisms included in our definition. The property c o m m o n to all of them is their ability to incorporate matter, convert it into organized structures, and move at random in search of sources of matter and energy. A 'borderline' group of organisms that fits in this category is that of viruses. Another is that of p r o g r a m m e d robots described above. Another, more developed group, is that of organisms endowed with systems that are able to recognize information, process it, and make decisions regarding the direction of their movements within their own environment. These are the characteristics of servomechanisms. Thus, while this higher hierarchy of living organisms - bacteria, plants, animals - must possess servomechanisms of some kind, we are forced to conclude that the servomechanisms themselves are not alive. Furthermore, as long as we include viruses in the category of living beings, the possession of servomechanisms ceases to be a necessary condition for life. It is essential only to a certain type of life - that of bacteria, plants, animals and programed robots that use servomechanisms to probe, explore and exploit their environment. Finally, there is a group of organisms that use special devices for the purpose of solving logical problems and thus facilitating the performance of their servomechanisms, e.g., by modifying either themselves or the environment. This group includes humans, with minds, and p r e - p r o g r a m m e d robots provided with servomechanisms and computers. Clearly, however, computers, by themselves, are not alive, nor is their possession a necessary condition for life. REFERENCES Schr6dinger, E.: 1946, What is Life? Cambridge: Cambridge University Press. Butler, S.: 1872, Erewhon, New York: Collier Books (1961 edition). Bridgman, E W.: 1961, The Nature of Thermodynamics. New York: Harper & Brothers. Zhabotinsky, A. M.: 1964, 'Periodic course of oxidation of malonic acid in solution (Investigation of the kinetics of the reactions of Belousov)',Biophysics 9, 329-335. Degn, H.: 1967, 'Effects of bromine derivatives of malonic acid, and the oscillating reaction of malonic acid, cerium ions and bromate', Nature 213, 589-590. Chandrasekhar, S.: 1961,Hydrodynamics andHydromagnetic Stability, Oxford Clarendon Press. Prigogine, I. and Nicolis, G. 1971, 'Biological order, structure and instabilities', Quart. Revs. of Biophys. 4, 107-148. Sackler Institute of Molecular Medicine, Sackler Faculty of Medicine, and Department of Philosophy, Faculty of Humanities, Tel-AvivUniversity, Ramat Aviv, Israel.