Life's Demons: Information and Order in Biology

outlookoutlook Life’s demons: information and order in biology

What subcellular machines gather and process the information necessary to sustain life? Philippe M. Binder & Antoine Danchin

n his seventeenth-century classic, literature, in which it is frequently used to …biological Demons use Novum Organum, Francis Bacon wrote, refer to stored data, or to some mysterious information from within the I“we cannot command nature except quality carried by matter or energy. Instead, cell to counteract the inexorable by obeying her” (Bacon, 2010). Although we define information as a quantity that our knowledge of living systems is much is transferred—without necessarily being effects of ageing, as well as improved since Bacon’s time, we are still conserved—during a measurement pro permitting the creation of a far from understanding—or commanding— cess, as correlations are created between young progeny from old cells all the complex mechanisms of life. To take the states of the measured and measuring full advantage of living organisms for the systems. Such a view of information has interacting complex networks of genes and benefit of mankind, we will need to under- increasingly been considered by physicists proteins corresponding to functional mod- stand those mechanisms to the furthest pos- as an extension of and alternative to the ules (Oltvai & Barabási, 2002); or included sible extent. To do so will require that the more-traditional concept of entropy (Caves, the concept of a set of digital-like con- concept of information and the theories of 1993; Landauer, 1991; Zurek, 1989). In nected modules that represent biochemical information science take a more-prominent support of Landauer’s idea, Toyabe and col- functions, within which information—in the role in the understanding of living systems. leagues recently claimed to have converted informal sense—flows between modules Two decades ago, Rolf Landauer information directly into energy (Toyabe (Nurse, 2008). These concepts, although (1991) argued that “information is physi- et al, 2010). This definition contrasts with, useful, are overly mechanistic and fall short cal” and ought to have a role in the scien- for example, the measure of how many of capturing life’s unbounded spontaneity tific analysis of reality comparable to that bits can be carried by a memory device and adaptability. of matter, energy, space and time. This such as DNA—also called information would also help to bridge the gap between capacity—and with the most-compressed e therefore summon Maxwell’s biology and mathematics and physics. description of a specific arrangement of Demon (MxD; Leff & Rex, Although it can be argued that we are liv- bits, such as a genome, which is also called W2002) and its distant relative ing in the ‘golden age’ of biology, both algorithmic information. the ratchet (Serreli et al, 2007), and apply because of the great challenges posed by these to biology. The original Demon medicine and the environment and the …we feel that information was the brain-child of Scottish physicist significant advances that have been made— as an essential aspect of life James Clerk Maxwell. In 1867, in a let- especially in genetics and molecular and ter to Tait, Maxwell proposed a thought cell biology—we feel that information as an has been neglected, or at least experiment to “show that the Second Law essential aspect of life has been neglected, misunderstood of Thermodynamics has only a statisti- or at least misunderstood. cal certainty” (Leff & Rex, 2002). Maxwell Theoretical biology has not yet em- conceived of a device to identify and sepa- ere we propose to look at living braced such a notion of information, rate fast and slow particles in order to set organisms as information gather- although the field has advanced beyond up a temperature or pressure gradient, thus Hing and utilizing systems (IGUSs; the idea of open, non-equilibrium systems producing an asymmetrical gas configura- Gell-Mann, 1994) that process information that sustain structures and functions as mat- tion to yield useful work with an apparent about their environment to make decisions ter and energy flow through them. Recent decrease in the entropy of the Universe. about their actions. To do so, we require a attempts to describe life have used large sets Although imaginary, MxD has become definition of ‘information’ that transcends of coupled, non-linear differential equa- a crucial tool for physicists who seek to the informal use of the term in the biology tions (Tomita et al, 1999); considered large, understand information and its processing.

so complete bacterial genomes (Lagesen et al, 2010) that have been sequenced so far, let alone all the other genomes of eukaryotic organisms. Fortunately, analysis of gene persist- ence—the tendency of genes to be present in a quorum of genomes—has allowed us to identify the paleome, a common set of genes that code for the essential functions that maintain life (Danchin et al, 2007). Most of these functions are easily accounted for as they are involved in the construction, reproduction and maintenance of cellular structures and the replication of genomes. Yet, some functions, thought to be involved in maintenance or degradation, are found to use energy in a way that is not obviously Fig 1 | A schematic of Maxwell’s Demon operating within a cell. MxD molecule (with a trident), coded linked to their role. Hence, the unexpected by DNA (helix) along with other proteins, distinguishes functional and damaged proteins. The latter are use of apparently wasted energy can be contained or discarded. MxD, Maxwell’s Demon. seen as an evidential ‘smoking gun’ for MxDs of the sort we describe. One example of such energy ‘waste’ Our demon is somewhat different: it is variety of reasons: repeated transitions might be a kinetic proofreading mechanism an existing biological mechanism that uses between metabolic states, hot or cold during translation (Gueron, 1978; Hopfield, information from its surroundings to create a temperatures, reactive oxygen species, 1974), which involves information transfer, locally ordered environment, while obeying glycation, or other aggressive chemical memory and reset. Reading a codon of the the second law of thermodynamics. Thus, reactions. Perhaps as importantly, some genetic message requires interaction with a by definition, our MxD is an IGUS. As infor- amino-acid residues undergo spontaneous cognate-adaptor molecule—a transfer RNA mation transfer implies a change in the state isomerization that causes functional decay loaded with the amino acid corresponding of the MxD device itself, it turns out that in in a way that is conceptually reminiscent to the codon—on the ribosome, a nano order to operate again, the MxD first has to of the half-life of radioactive isotopes. In machine that translates messenger RNA be reset to its original state; this step carries order for life to go on, these aged and dys codon after codon, in a ratchet-like man- an entropy cost, usually with an associated functional metabolites must be identified ner. There is enough energy in the chemi- energy input, a fact that has an important and either discarded, degraded or repaired cal bond that links the amino acid to the role, as discussed here. during normal cell operation, and even transfer RNA to permit polymerization into more so during cell reproduction, in which a protein. Yet, a further step is required, …the unexpected use of the segregation of new and old molecules with an apparently useless consumption of apparently wasted energy can guarantees the survival of new daughter energy. The loaded transfer RNA is bound be seen as an evidential ‘smoking cells. Enter Maxwell’s Demon: a mecha- to a factor (EFTu) that tests the accuracy of nism that differentiates functional and dys- the decoding process; there is competi- gun’ for Maxwell’s Demons of the functional proteins, perhaps by measuring tion between transfer RNAs at the entry sort we describe their shapes and vibration spectra (Schwartz site in the ribosome, and incorrect transfer & Schramm, 2009), and segregates normal RNAs are more common than correct ones. In a biological context, we postulate proteins from defective ones using exist- When the correct one is bound, free EFTu the existence of MxDs in every cell, but ing cellular structures, or by generating is released at the expense of a molecule of with a different role than their counterparts new ones (Fig 1). energy-rich GTP. Energy consumption is from physics: biological Demons use infor only used at this stage to achieve a more- mation from within the cell to counteract ow can we identify the entities, ‘ordered’ state of genetic-code use during the inexorable effects of ageing, as well as proteins or complexes that play translation. At the least, this is the work of permitting the creation of young progeny Hthe role of MxDs? If they exist, an IGUS, and at best it could be the earli- from old cells. This is based on a view of they must be in genomes. They must also be est MxD (unknowingly) discovered in a cells as machines that make copies of them- essential for life to perpetuate in a sustain- cellular process. selves through organized coupling of the able way. Hence, the corresponding func- processes of replication—copying the DNA tions should be encoded in all genomes. If ow then might other biological ‘program’—and reproduction: a slightly it was possible, an easy way to find them MxDs manifest themselves? All imperfect duplication of the metabolic would be to search for genes that are com- Hcells, including bacteria, age and machinery and cellular structures. mon to all genomes. Unfortunately, because eventually deteriorate in unfavourable con- The proteins that implement metabo- life conserves functions but not structures, ditions. As this happens, proteins aggregate lism and replication degrade with time for a there are no genes common to the 1,000 or into potentially harmful arrangements.

Remarkably, these protein aggregates do Another candidate that deserves further Our proposal is for truly not distribute randomly within the cell. study is the stabilization of neural synapses. interdisciplinary, long-term In some cases, they are transported by an It is generally accepted that learning and collaborations in which energy-dependent process to the poles of memory are based on a selective process Escherichia coli cells, where they can either that retains functioning synapses and dis- biological experiments are be stored and discarded when the cell dies, cards the others. A variety of processes have inspired by cutting-edge research or be more easily degraded by energy- been proposed to explain this, but the most in theoretical physics dependent proteases (Rokney et al, 2009). obvious one is an energy-dependent sys- In other cells—budding yeast, for exam- tem that would tag receptors of functional immediate action, the device keeps modi- ple—old proteins are not included in newly neurotransmitters, leaving the remainder fying its state and becomes a de facto formed buds, but remain within the mother vulnerable to degradation (Milholland & record of a sequence of measurements. cell, which acts as a bin. A family of GTP- Gordon, 2007). This is probably what happens in natural dependent proteins—septins—seems to be selection. In contrast to the intracellular involved in the process; in the absence of he typical modus operandi of an situation, organisms repeatedly ‘measure’ septins, old proteins do not remain in the MxD is to quickly use any infor their environment over generations, lead- mother cell (Barral, 2010; Hu et al 2010; Tmation acquired—resulting in a cor- ing to a cumulative record of mutations and Kim et al 2010). A similar process has been related change of its own state—to produce recombinations in the genome of surviving demonstrated in Caenorhabditis elegans a desirable action—for example, to gener- members of a species. The negative version cells, in which a mother cell divides into a ate work, or to restore the functionality of of this effect is known as Muller’s ratchet. cell carrying old protein and a young egg a degraded protein—and then to be reset To further test these ideas, one can cell (Goudeau & Aguilaniu, 2010). We to a neutral state, in which it can perform explore the information-processing behav- believe that these are legitimate examples another measurement. In other cases, iour of the paleomic energy-dependent of non-trivial, Demon-like information measurements might accumulate through gene products under a variety of conditions. processing in biology (Fig 2). a ratchet-like mechanism. Instead of taking A practical approach could investigate the

A humans. If this is the case, the scaling up of synthetic-biology processes might hit snags. The MxDs we have seen operate within a narrow cellular context; this highlights the need for a theory of semantic information (for example, Floridi, 2004) as part of the proposed programme to achieve greater command of biological entities. In particu- lar, it will be essential to understand which measurements are required to trigger specific biochemical behaviours, especially in cases in which several functions overlap. B Newly Creators of synthetic constructs will need synthesized proteins to harness the function of specific MxDs for their own goals. This will require that Proteins biology also informs those investigating the foundations of information theory, so that they make the best educated guesses to enhance their chances of success. We regard the acceptance and adoption of Old proteins information as the focus of a fascinating and challenging collaboration between biologists, mathematicians and physicists in the twenty-first century.

CONFLICT OF INTEREST Fig 2 | Examples of biological Maxwell’s Demons. (A) Protein aggregates are transported to Escherichia coli The authors declare that they have no conflict of interest. cell poles for easy disposal. (B) Septin rings prevent old cell products from entering young yeast buds. ACKNOWLEDGEMENTS P.M.B. acknowledges the support of Research evolution of synthetic-biology constructs. interdisciplinary, long-term collaborations Corporation and A.D. aknowledges the support Bacteria engineered to produce new in which biological experiments are of the Microme FP7 KBBE‑2007‑3-2‑08‑222886‑2 metabolic pathways would be grown on inspired by cutting-edge research in theo- grant. media with just enough nutrients to main- retical physics. The potential benefit of such REFERENCES tain colonies, together with a metabolite a programme will be an understanding of Bacon F (2010) Novum Organum (reprint that is chemically related to some step in living systems as active IGUSs, rather than and translation of the 1620 Latin original). the pathway, but which they cannot use as mere autonomous mechanisms. Not pur- Charleston, SC, USA: Forgotten books directly. As the genome of the cell ages, suing these ideas might considerably slow Barral Y (2010) Septins at the nexus. Science 329: 1289–1290 we expect that continuous turnover of old the advance of biology, especially if we fail Caves CM (1993) Information and entropy. gene products will occasionally replace to recognize apparently unrelated obser- Phys Rev E 47: 4010–4017 some proteins with counterparts that per- vations as manifestations of a single, fun- Danchin A, Fang G, Noria S (2007) The extant core mit the corresponding colonies to resume damental principle (Danchin et al, 2007; bacterial proteome is an archive of the origin of growth, using the new metabolite. This can Rokney et al, 2009; Barral, 2010; Goudeau life. Proteomics 7: 875–889 Floridi L (2004) Outline of a theory of strongly be helped by transcription, which consider & Aguilaniu, 2010). semantic information. Mind Mach 14: ably increases the local mutation rate of 197–222 expressed genes (Kim & Jinks-Robertson, Creators of synthetic constructs Gell-Mann M (1994) The Quark and the Jaguar. 2009). Manipulating MxD genes in this San Francisco, CA, USA: WH Freeman will need to harness the Goudeau J, Aguilaniu H (2010) Carbonylated experiment will modulate innovation, lead- function of specific Maxwell’s proteins are eliminated during reproduction ing to either colonies that progressively in C. elegans. Aging Cell 9: 991–1003 die out if they fail to produce progeny with Demons for their own goals Gueron M (1978) Enhanced selectivity of adaptive mutations, or that grow faster in enzymes by kinetic proofreading. Am Sci 66: the opposite case. The view presented here has implica- 202–208 Hopfield JJ (1974) Kinetic proofreading: a new tions for the ‘engineering’ branch of biol- mechanism for reducing errors in biosynthetic his would be only the first step of ogy: synthetic biology. If, as we surmise, processes requiring high specificity. Proc Nat an ambitious research programme biological systems accumulate informa- Acad Sci USA 71: 4135–4139 in which biologists and physicists tion in a myopic way—MxDs are local Hu Q, Milenkovic L, Jin H, Scott MP, Nachury MV, T Spiliotis ET, Nelson WJ (2010) A septin naturally collaborate, the latter provid- devices without grand designs—then they diffusion barrier at the base of the primary ing knowledge of information theory and will evolve by tinkering in their own way, cilium maintains ciliary membrane protein related subjects. Our proposal is for truly rather than by fulfilling the design goals of distribution. Science 329: 436–439

Kim N, Jinks-Robertson S (2009) dUTP Rokney A, Shagan M, Kessel M, Smith Y, incorporation into genomic DNA is linked to Rosenshine I, Oppenheim AB (2009) E. coli transcription in yeast. Nature 459: 1150–1153 transports aggregated proteins to the poles by a Kim SK et al (2010) Planar cell polarity acts specific energy-dependent process. J Mol Biol through septins to control collective cell 392: 589–601 movement and ciliogenesis. Science 329: Schwartz SD, Schramm VL (2009) Enzymatic 1337–1340 catalysis and the nature of transition state barrier Lagesen K, Ussery DW, Wassenaar TM (2010) crossing. Nat Chem Biol 5: 551–558 Genome update: the 1000th genome—a Serreli V, Lee C‑F, Kay ER, Leigh DA (2007) cautionary tale. Microbiology 156: 603–608 A molecular information ratchet. Nature 445: Landauer R (1991) Information is physical. 523–527 Philippe M. Binder (left) is professor and Phys Today 44: 23–29 Tomita M et al (1999) E‑CELL: Software Leff H, Rex AF (2002) Maxwell’s Demon 2. environment for whole-cell simulation. chair at the Department of Physics and London, UK: Taylor and Francis Bioinformatics 15: 72–84 Astronomy, University of Hawaii, Hilo, USA. Milholland RB, Gordon H (2007) A role Toyabe S, Sagawa T, Ueda M, Muneyuki E, Antoine Danchin is founder and CEO for acetylcholine receptors in their own Sano M (2010) Experimental demonstration of AMAbiotics, Evry, France. aggregation on muscle cells. Devel Neurobiol of information‑to‑energy conversion and E‑mail: [email protected] 67: 999–1008 validation of the generalized Jarzynski equality. Nurse P (2008) Life, logic and information. Nat Phys 6: 988–992 Received 8 February 2011; accepted 15 April 2011; Nature 454: 424–426 Zurek WH (1989) Thermodynamic cost of published online 6 May 2011 Oltvai ZN, Barabási AL (2002) Life’s complexity computation, algorithmic complexity and the EMBO reports (2011) 12, 495–499. pyramid. Science 298: 763–764 information metric. Nature 341: 119–124 doi:10.1038/embor.2011.83