Enzyme Genetic Programming

Programming is a process of optimization; taking a specification, which tells us what we want, and transforming it into an implementation, a program, which causes the target system to do exactly what we want. Conventionally, this optimization is achieved through manual design. However, manual design can be slow and error-prone, and recently there has been increasing interest in automatic programming; using computers to semiautomate the process of refining a specification into an implementation. Genetic programming is a developing approach to automatic programming, which, rather than treating programming as a design process, treats it as a search process. However, the space of possible programs is infinite, and finding the right program requires a powerful search process. Fortunately for us, we are surrounded by a monotonous search process capable of producing viable systems of great complexity: evolution. Evolution is the inspiration behind genetic programming. Genetic programming copies the process and genetic operators of biological evolution but does not take any inspiration from the biological representations to which they are applied. It can be argued that the program representation that genetic programming does use is not well suited to evolution. Biological representations, by comparison, are a product of evolution and, a fact to which this book is testament, describe computational structures. This chapter is about enzyme genetic programming, a form of genetic programming that mimics biological representations in an attempt to improve the evolvability of programs. Although it would be an advantage to have a familiarity with both genetic programming and biological representations, concise introductions to both these subjects are provided. According to modern biological understanding, evolution is solely responsible for the complexity we see in the structure and behavior of biological organisms. Nevertheless, evolution itself is a simple process that can occur in any population of imperfectly replicating entities where the right to replicate is determined by a process of selection. Consequently, given an appropriate model of such an environment, evolution can also occur within computers.

Encrypted Genes and Their Assembly in Ciliates

Maintenance of normal cell function and structure requires some level of stability of the cell’s DNA—at least the DNA that makes up the genes of the cell. In most eukaryotes most of the DNA in the genome does not encode genes and has no known function beyond forming long spacers between successive genes. For example, the gene density in the germline (micronuclear) genome of stichotrich ciliates (formerly referred to as hypotrich ciliates) is very low; only a few percent of the DNA encodes the approximately 27,000 different genes, and more than 95% is spacer DNA. Powerful DNA repair systems guard the stability both of nongene and gene DNA in contemporary cells, protecting it against mutagenesis. Although species survival depends on DNA stability, cell evolution requires changes in DNA. Presumably, there is a balance between instability of DNA that allows evolution and a stability that protects species from mutational extinction. Could cells evolve strategies that change the balance, allowing a greater rate of DNA change (gene evolution) without jeopardizing species survival? The stichotrichs may, in fact, have evolved such a mechanism, dramatically modifying their germline DNA during evolution to facilitate creation of new genes without reducing the level of cell survival. The modifications of germline DNA in ciliates, in turn, require dramatic DNA processing to convert germline DNA into somatic DNA during the life cycle of the organisms. The ciliate strategy rests on the evolution of nuclear dimorphism: the inclusion both of a germline nucleus (micronucleus) and a somatic nucleus (macronucleus) in the same cell (Figure 9.1; for a general review, see Prescott [6, 7]). Like the example in Figure 9.1, most stichotrich species contain two or more micronuclei and two or more macronuclei per cell. The multiple micronuclei are genetically identical to each other, and the multiple macronuclei are genetically identical; these multiplicities of nuclei have no bearing on the issues addressed in this chapter. The micronucleus is used only in cell mating, and its genes are silent. Hence, micronuclear genes do not support the maintenance, growth, or division of the cell.

The Biology of Integration of Cells into Microscale and Nanoscale Systems

In chapter 5 we focused on the informational interface between cells and synthetic components of systems. This interface is concerned with facilitating and manipulating information transport and processing between and within the synthetic and whole-cell components of these hybrid systems. However, there is also a structural interface between these components that is concerned with the physical placement, entrapment, and maintenance of the cells in a manner that enables the informational interface to operate. In this chapter we focus on this structural interface. Successful integration of whole-cell matrices into microscale and nanoscale elements requires a unique environment that fosters continued cell viability while promoting, or at least not blocking, the information transport and communication pathways described in earlier chapters. A century of cell culture has provided a wealth of insight and specific protocols to maintain the viability and (typically) proliferation of virtually every type of organism that can be propagated. More recently, the demands for more efficient bioreactors, more compatible biomedical implants, and the promise of engineered tissues has driven advances in surface-modification sciences, cellular immobilization, and scaffolding that provide structure and control over cell growth, in addition to their basic metabolic requirements. In turn, hybrid biological and electronic systems have emerged, capable of transducing the often highly sensitive and specific responses of cellular matrices for biosensing in environmental, medical, and industrial applications. The demands of these systems have driven advances in cellular immobilization and encapsulation techniques, enabling improved interaction of the biological matrix with its environment while providing nutrient and respiratory requirements for prolonged viability of the living matrices. Predominantly, such devices feature a single interface between the bulk biomatrix and transducer. However, advances in lithography, micromachining, and micro-/nanoscale synthesis provide broader opportunities for interfacing whole-cell matrices with synthetic elements. Advances in engineered, patterned, or directed cell growth are now providing spatial and temporal control over cellular integration within microscale and nanoscale systems. Perhaps the best defined integration of cellular matrices with electronically active substrates has been accomplished with neuronal patterning. Topographical and physicochemical patterning of surfaces promotes the attachment and directed growth of neurites over electrically active substrates that are used to both stimulate and observe excitable cellular activity.

Genetic Process Engineering

In this chapter we present an engineering discipline to obtain complex, predictable, and reliable cell behaviors by embedding biochemical logic circuits and programmed intercellular communications into cells. To accomplish this goal, we provide a well-characterized component library, a biocircuit design methodology, and software design tools. Using the cellular gates,we introduce genetic process engineering, a methodology for modifying the DNA encoding of existing genetic elements to achieve the desired input/output behavior for constructing reliable circuits of significant complexity.We also describe BioSpice, a prototype software tool for biocircuit design that supports both static and dynamic simulations and analysis of singlecell environments and small cell aggregates. The goal of our research is to lay the foundations of an engineering discipline for building novel living systems with well-defined purposes and behaviors using standardized, well-characterized components. Cells are miniature, energy efficient, self-reproduce, and can manufacture biochemical products. These unique characteristics make cells attractive for many novel applications that require precise programmed control over the behavior of the cells. The applications include nanoscale fabrication, embedded intelligence in materials, sensor/effector arrays, patterned biomaterial manufacturing, improved pharmaceutical synthesis, programmed therapeutics, and as a sophisticated tool for in vivo studies of genetic regulatory networks. These applications require synthesis of sophisticated and reliable cell behaviors that instruct cells to make logic decisions based on factors such as existing environmental conditions and current cell state. For example, a cell may be programmed to secrete particular timed sequences of biochemicals depending on the type of messages sent by its neighbors. The approach proposed here for engineering the requisite precision control is to embed internal computation and programmed intercellular communications into the cells. The challenge is to provide robust computation and communications using a substrate where reliability and reproducible results are difficult to achieve. Biological organisms as an engineering substrate are currently difficult to modify and control because of the poor understanding of their complexity. Genetic modifications to cells often result in unpredictable and unreliable behavior. A single Escherichia coli bacterial cell contains approximately 1010 active molecules, about 107 of which are protein molecules.

Biocomputation in Ciliates

Ciliates are unicellular protists that may have arisen more than a billion years ago. They have since diverged into thousands of species, many uncharacterized, the genetic divergence among ciliates being at least as deep as that between plants and animals. Despite their diversity, ciliates are united by two common features; the presence of short threads called cilia on their surface, whose rhythmic beating causes movement and is also useful for food capture, and the presence of two types of nuclei. The macronucleus contains DNA encoding functional copies of all the genes that regulate vegetative growth and cell proliferation. The micronucleus contains encrypted versions of the macronuclear DNA, is mostly functionally inert, and is only used for sexual exchange of DNA. In this chapter we study the decryption of the macronuclear DNA from a computational perspective. When two cells mate, they exchange micronuclear information. After they separate, the old micronuclei and macronuclei degenerate, while the newly formed micronuclei develop into new macronuclei over hours or days, depending on the species. Few ciliates have so far been studied at the level of molecular genetics: Tetrahymena and Paramecium representing the Oligohymenophorans and Oxytricha (recently renamed Sterkiella), and Stylonichia and Euplotes representing Spirotrichs. The DNA molecule in each of the approximately 120 chromosomes in the micronucleus contains on average approximately 107 basepairs (bp) in Oxytricha species and approximately 18 × 106 bp in Stylonichia lemn. The size of the DNA molecules in the macronucleus is, in contrast, very small. In various Oxytricha species and S. lemnae, macronuclear DNA molecules range in size from 400 to 15,000 bp with most molecules in the 1000–8000 bp range. Macronuclear DNA sequences are derived from the micronuclear sequences through a series of DNA rearrangements as follows. The segments that together constitute a macronuclear sequence (macronuclear destined sequences or MDSs) are present as sub-sequences in the micronuclear DNA. However, in the micronuclear DNA, MDSs are interspersed with long DNA sequences (internal eliminated sequences or IESs) that are excised in the micronucleus to macronucleus differentiation.

The Device Science of Whole Cells as Components in Microscale and Nanoscale Systems

Intact whole cells may be the ultimate functional molecular-scale machines, and our ability to manipulate the genetic mechanisms that control these functions is relatively advanced when compared to our ability to control the synthesis and direct the assembly of man-made materials into systems of comparable complexity and functional density. Although engineered whole cells deployed in biosensor systems provide one of the practical successes of molecular-scale devices, these devices explore only a small portion of the full functionality of the cells. Individual or self-organized groups of cells exhibit extremely complex functionality that includes sensing, communication, navigation, cooperation, and even fabrication of synthetic nanoscopic materials. Adding this functionality to engineered systems provides motivation for deploying whole cells as components in microscale and nanoscale devices. In this chapter we focus on the device science of whole cell components in a way analogous to the device physics of semiconductor components. We consider engineering the information transport within and between cells, communication between cells and synthetic devices, the integration of cells into nanostructured and microstructured substrates to form highly functional systems, and modeling and simulation of information processing in cells. Even a casual examination of the information processing density of prokaryotic cells produces an appreciation for the advanced state of the cell’s capabilities. A bacterial cell such as Escherichia coli ( 2 μm2 cross-sectional area) with a 4.6 million basepair chromosome has the equivalent of a 9.2-megabit memory. This memory codes for as many as 4300 different polypeptides under the inducible control of several hundred different promoters. These polypeptides perform metabolic and regulatory functions that process the energy and information, respectively, made available to the cell. This complexity of functionality allows the cell to interact with, influence, and, to some degree, control its environment. Compare this to the silicon semiconductor situation as described in the International Technology Roadmap for Semiconductors (ITRS). ITRS predicts that by the year 2014, memory density will reach 24.5 Gbits/cm2, and logic transistor density will reach 664 M/cm2. Assuming four transistors per logic function, 2 μm2 of silicon could contain a 490-bit memory or approximately three simple logic gates.

Proteins and Information Processing

This chapter reviews and briefly discusses a set of computational methods that can assist biologists when seeking to model interactions between components in spatially heterogeneous and changing environments. The approach can be applied to many scales of biological organization, and the illustrations we have selected apply to networks of interaction among proteins. Biological populations, whether ecological or molecular, homogeneous or heterogeneous, moving or stationary, can be modeled at different scales of organization. Some models can be constructed that focus on factors or patterns that characterize the population as a whole such as population size, average mass or length, and so forth. Other models focus on values associated with individuals such as age, energy reserve, and spatial association with other individuals. A distinction can be made between population (p-state) and individual (i-state) variables and models. We seek to develop a general approach to modeling biosystems based on individuals. Individual-based models (IBMs) typically consist of an environment or framework in which interactions occur and a number of individuals defined in terms of their behaviors (such as procedural rules) and characteristic parameters. The actions of each individual can be tracked through time. IBMs represent heterogeneous systems as sets of nonidentical, discrete, interacting, autonomous, adaptive agents (e.g., Devine and Paton [5]). They have been used to model the dynamics of population interaction over time in ecological systems, but IBMs can equally be applied to biological systems at other levels of scale. The IBM approach can be used to simulate the emergence of global information processing from individual, local interactions in a population of agents. When it is sensible and appropriate, we seek to incorporate an ecological and social view of inter-agent interactions to all scales of the biological hierarch. In this case we distinguish among individual “devices” (agents), networks (societies or communities), and networks in habitats (ecologies). In that they are able to interact with other molecules in subtle and varied ways, we may say that many proteins have social abilities . This social dimension to protein agency also presupposes that proteins have an underlying ecology in that they interact with other molecules including substrates, products, regulators, cytoskeleton, membranes, water, and local electric fields.

An Introduction to Cellular Computing

The abstract operation of complex natural processes is often expressed in terms of networks of computational components such as Boolean logic gates or artificial neurons. The interaction of biological molecules and the flow of information controlling the development and behavior of organisms is particularly amenable to this approach, and these models are well established in the biological community. However, only relatively recently have papers appeared proposing the use of such systems to perform useful, human-defined tasks. Rather than merely using the network analogy as a convenient technique for clarifying our understanding of complex systems, it is now possible to harness the power of such systems for the purposes of computation. The purpose of this volume is to discuss such work. In this introductory chapter we place this work in historical context and provide an introduction to some of the underlying molecular biology. We then introduce recent developments in the field of cellular computing. Despite the relatively recent emergence of molecular computing as a distinct research area, the link between biology and computer science is not a new one. Of course, for years biologists have used computers to store and analyze experimental data. Indeed, it is widely accepted that the huge advances of the Human Genome Project (as well as other genome projects) were only made possible by the powerful computational tools available to them. Bioinformatics has emerged as the science of the 21st century, requiring the contributions of truly interdisciplinary scientists who are equally at home at the lab bench or writing software at the computer. However, the seeds of the relationship between biology and computer science were sown long ago, when the latter discipline did not even exist. When, in the 17th century, the French mathematician and philosopher René Descartes declared to Queen Christina of Sweden that animals could be considered a class of machines, she challenged him to demonstrate how a clock could reproduce. Three centuries later, with the publication of The General and Logical Theory of Automata [19] John von Neumann showed how a machine could indeed construct a copy of itself.

Cellular Computation and Communication Using Engineered Genetic Regulatory Networks

In this chapter we demonstrate the feasibility of digital computation in cells by building several operational in vivo digital logic circuits, each composed of three gates that have been optimized by genetic process engineering. We have built and characterized an initial cellular gate library with biochemical gates that implement the NOT, IMPLIES, andANDlogic functions in E. coli cells. The logic gates perform computation using DNA-binding proteins, small molecules that interact with these proteins, and segments of DNA that regulate the expression of the proteins. We also demonstrate engineered intercellular communications with programmed enzymatic activity and chemical diffusions to carry messages, using DNA from the Vibrio fischeri lux operon. The programmed communications is essential for obtaining coordinated behavior from cell aggregates. This chapter is structured as follows: the first section describes experimental measurements of the device physics of in vivo logic gates, as well as genetic process engineering to modify gates until they have the desired behavior. The second section presents experimental results of programmed intercellular communications, including time–response measurements and sensitivity to variations in message concentrations. Potentially the most important element of biocircuit design is matching gate characteristics. Experimental results in this section demonstrate that circuits with mismatched gates are likely to malfunction. In generating biology’s complex genetic regulatory networks, natural forces of selection have resulted in finely tuned interconnections between the different regulatory components. Nature has optimized and matched the kinetic characteristics of these elements so that they cooperatively achieve the desired regulatory behavior. In building de novo biocircuits, we frequently combine regulatory elements that do not interact in their wild-type settings. Therefore, naive coupling of these elements will likely produce systems that do not have the desired behavior. In genetic process engineering, the biocircuit designer first determines the behavioral characteristics of the regulatory components and then modifies the elements until the desired behavior is attained. Below, we show experimental results of using this process to convert a nonfunctional circuit with mismatched gates into a circuit that achieves the correct response.

The Enterococcus faecalis Information Gate

Information exchange between cellular compartments allows us to engineer systems based around cooperative principles. In this chapter we consider a unique bacterial communication system, the conjugative plasmid transfer of Enterococcus faecalis. Using these bacteria, we describe how to engineer a logically controlled information gate and build a logical inverter based upon it. Cellular computing is an alternative computing paradigm based on living cells. Microscale organisms, especially bacteria, are well suited for computing for several reasons. A small culture provides an almost limitless supply of bacterial “hardware.” Bacteria can be stored and easily modified by gene recombination. In addition, and important for our purposes, bacteria can produce various signal molecules that are useful for computation. DNA-binding proteins recognize specific regulatory regions of DNA, bind them, and regulate their genetic expression. These proteins are available for use as computing signals inside the cell. Weiss et al. have shown, for example, how to construct logic circuits based on gene expression regulated by DNA-binding proteins. Some signal molecules are associated with intercellular communications between individuals. Intercellular communication is one of the fundamental characteristics of multicellular organisms, but it is also found in single-celled microorganisms, including bacteria. Communication mediated by homoserine lactones can widely be seen in various Gram-negative bacteria. The mechanism of this behavior was well characterized in Vibrio fischeri, due to their bioluminescent activity mediated by homoserine lactones. It has been shown that bacterial information transfer can be engineered as an extension of Escherichia coli into which the lux genes of Vibrio fischeri are transformed. The communication abilities of bacteria therefore allow us to build microbial information processors for cellular computing. Communication mechanisms in Gram-positive bacteria are not yet well understood. One of the exceptions to this is the conjugative plasmid transfer system in Enterococcus faecalis. E. faecalis conjugate in response to a pheromone is released by other cells. Pheromones are seven- or eight-residue amino peptides produced in E. faecalis. In the case of cPD1, the pheromone is produced by truncation of a 22-residue precursor that is the signal peptide of a lipoprotein.