Synthetic forms most beautiful: engineering insights into self-organization

Abstract

Reflecting on the diversity of the natural world, Darwin famously observed that “from so simple a beginning endless forms most beautiful and most wonderful have been, and are being evolved”. However, the examples that we are able to observe in nature are a consequence of chance, constrained by selection, drift and epistasis. Here we explore how the efforts of synthetic biology to build new living systems can expand our understanding of the fundamental design principles that allow life to self-organize biological form, from cellular to organismal levels. We suggest that the ability to impose a length or timescale onto a biological activity is an essential strategy for self-organization in evolved systems and a key design target that is now being realized synthetically at all scales. By learning to integrate these strategies together, we are poised to expand on evolution’s success and realize a space of synthetic forms not only beautiful but with diverse applications and transformative potential.

Introduction.

Evolution has built a biological world that is rich with diverse forms and functions that arise from self-organization at all scales, from complex single cells to multicellular plants and animals^1–5. Each extant system we find in nature provides an example that can, in principle, be deconstructed to reveal how self-organizing mechanisms can craft and sculpt biology. Comparison of multiple examples, especially convergent ones, can suggest higher-order themes and theories that provide over-arching principles that may govern biological self-organization more generally^6–8. However, the forms and functions realized by evolution are also constrained by selection, chance, and epistasis^9–12. These constraints hinder our ability to test theoretical predictions or models systematically and satisfactorily. As a result, applying purely deconstructive approaches has limitations that impact our understanding of biological self-organization.

Synthetic biology provides a complimentary approach to understanding self-organization that can potentially overcome these limitations. A major goal of the synthetic approach is to reconstitute the complexity of living systems from the bottom up by creating regulatory relationships between parts or modules that often have not previously served in or acted in a way at all related to the specific target of design^13–16. Such approaches are constrained not by selective forces and evolutionary history, but by our theoretical understanding and hypotheses about the mechanisms that underpin the phenomena of interest¹⁷.

While there has been a great deal of historical focus in synthetic biology on controlling the input/output capabilities of cells^18,19, there has been increasing emphasis and renewed focus on synthetic approaches that can encode new patterns of self-organization, altering the core ground state behavior of living systems and providing a toehold for the design of new cellular and multicellular forms and functions²⁰. By challenging ourselves to use imagination to identify mechanisms that can encode self-organization using new components and imposing those patterns onto living systems, synthetic biology can potentially push living systems into regimes hitherto unrealized by evolution but completely within the bounds of biology’s untapped potential (Fig. 1).

Figure 1. Synthetic biology can explore biological forms and functions inaccessible to evolution.

Venn diagram illustrating the space of biological forms and functions. Metazoan systems, frequently the focus of biological research, occupy only a small subset within the space of biological forms and functions that have been accessed by evolution. However, evolutionary processes cannot sample the entirety of what is possible for biological components to achieve, as they are limited by selection, historical epistasis, and chance processes. Synthetic biology provides an alternative path for exploring this space.

Here, we review synthetic biology approaches that can approximate and reconstitute the self-organizing phenomena of highly evolved systems across different length and timescales. We first introduce and recontextualize landmark synthetic genetic circuit designs as systems that self-organize gene expression dynamics. This identifies the notion of timescales as a key design module that can be emulated and encoded synthetically. We then explore recent developments that allow self-organization of spatiotemporal behavior at subcellular and multicellular levels, identifying both length scales and timescales in evolved systems as synthetic targets for coupling to other activities to carve out new emergent form and function. We suggest that by emulating the ability of living systems to self-organize vastly different length and timescales and efficiently integrate them together, synthetic biology is poised to realize its full potential to not just imitate the splendor of evolution’s diversity, but to expand beyond it.

Genetic circuits self-organize gene expression.

A major focus of synthetic biology has historically been in the development of gene circuits that perform specific functions^{14,16,20–22}. Evolved gene circuits play many critical roles in biology, from the coordination of stress responses in individual cells to the establishment of different cell-types across an organism^23–25. The networks of protein/protein and protein/DNA interactions at play to implement an evolved circuit can be extraordinarily complex, producing “hairball” plots of connectivity with redundancy and cross-reactivity that can be difficult to interpret or understand²⁶. However, the core logic that powers these networks can be simplified and reduced to modules—transcriptional regulators and their associated DNA binding sites—that are connected in different ways to encode specific functions (Fig. 2A)^14,20,21. Indeed, one of the earliest achievements in synthetic biology was in its ability to demonstrate that the function of a genetic circuit could largely be explained by the regulatory interactions between its modules. This approach is exemplified by the design of the repressilator, an E. coli genetic circuit that connects three completely arbitrary transcriptional repressors in a loop to generate oscillations in the expression levels of a fluorescent protein²⁷. Thus, although the specific repressors used—TetR, λcI and LacI—have no roles in timekeeping or biological rhythms, they can be made to encode a self-organizing dynamic function that mimics the overall behavior of highly evolved timekeeping circuits, like the CLOCK/per2 system that drives circadian rhythms (Fig. 2B)²⁸.

Figure 2. Genetic circuits self-organize gene expression dynamics.

(a) Genetic circuits can be decomposed into modules (transcriptional activators and repressors) and controllers (regulatory links that organize expression).

(b) Example of a natural genetic circuit (the mammalian circadian clock) that encodes a specific timescale: the 24-hour day/night cycle.

(c) Example of a synthetic genetic circuit (the repressilator) which generates a synthetic timescale. This timescale is, in principle arbitrary, and can be tuned by the user through engineering of the regulatory links in the circuit.

Since the initial development of the repressilator, numerous synthetic genetic circuits for timekeeping have been developed using different regulatory strategies and architectures^29–31. However, in contrast to the circadian clock, which has undergone selective pressures to establish a roughly 24-hour cycle that matches the day/night cycles of our planet, the timescale of these timekeepers is synthetic and tunable by the experimentalist. As a result, a key self-organizing capability that can be synthetically instantiated by gene circuits is a new timescale for the cell. In the case of oscillations, this timescale can be used to set a periodicity that self-organizes sequences or patterns of gene expression. Similarly, other non-oscillating synthetic gene circuits such as toggle switches^32,33 or pulse generators^21,34,35, provide opportunities to self-organize a synthetic timescale in their responses that can be imposed on to biological targets. However, it is critical to note that the timescales available to genetic circuits are inherently constrained by the timescales of gene expression³⁶. Nonetheless, we argue that synthetic genetic circuits can be viewed as self-organization systems that can project periodic or exponential timescales onto the expression of a cell’s genetic material.

Spatiotemporal signaling self-organizes cellular form and function.

The evolved and synthetic genetic circuits we discussed above operate largely agnostic of any particular cellular structure and form. However, evolution has produced a remarkable array of diverse cells of all shapes and sizes for which structure is critical to function. These range from astonishingly elegant protists, who exhibit dynamic animal-like shapes and behaviors at the single-cell level^5,37–40; to the myriad morphologically distinct cell-types—neuronal, glial, muscle, cochlear, etc.—that cooperate and coordinate to allow for multicellular life^41,42. Such cellular forms arise through self-organization of their protein hardware in space and time.

To begin to approach self-organization of cellular biology synthetically, it is useful to extend the concepts we derived from genetic circuits by defining the modules and controllers that can be connected together in different ways. However, unlike genetic circuits, where it is straightforward to abstract and compose modular DNA binding proteins and their targets together, the structure of cells depends on a motley crew of diverse and often structurally unrelated components.

For example, the cytoskeleton of Eukaryotic cells consists of multiple distinct filament systems, each with their own unique dynamics, regulators, and mechanical properties (Fig. 3A). In a typical human fibroblast, microtubules (25 nm) radiate outwards from the centriole to the cell cortex⁴³; actin filaments (7 nm) assemble into stress fibers and barbed networks at migrating edges⁴⁴; and diverse intermediate filament systems (10 nm) fill the cell interior to hold together cell shape⁴⁵. Yet, these same cytoskeletal modules can be self-organized in completely different ways to encode completely different cellular shapes, such as the stable arrays of cortical microtubules in protists and apicomplexan parasites^{38–40,46–51}, or the beautiful bundles of actin in cochlear ear cells⁵².

Figure 3. Spatiotemporal signaling self-organizes cellular form and function.

(a) Cell morphology can be decomposed into an assortment of modules that facilitate the generation of structure. These include cytoskeletal filaments, biological condensates and sub-cellular compartments, as well as force generating and membrane curvature-inducing factors.

(b) These structuring elements can be organized at a sub-cellular level through the action of spatiotemporal signaling controllers that regulate their assembly and disassembly. These spatiotemporal controllers may use relocalization, polarization-mechanisms, or reaction-diffusion patterning to facilitate this organization.

(c) An example of an evolved circuit that generates single-cell polarization in the mammalian epithelium; and a designed circuit that creates a polarized structure of synthetic PIP3 in budding yeast based on a similar network architecture.

(d) An example of an extraordinarily complex cell geometry seen in the protist Tetrahymena thermophila, in which basal bodies are organized by two spatial wavelengths (λ₁ and λ₂) along a polarized anterior-posterior axis. This is juxtaposed with an example of how multiple spatial wavelengths (λ₁ and λ₂) can be generated by combining synthetic reaction-diffusion patterning with droplet-forming protein sequences. Synthetic reaction-diffusion patterning has also been used to position actin polymerization, suggesting a path towards hierarchical self-organization of cytoskeletal structures akin to those seen in elaborate protist geometries.

In addition, cells have many other structuring modules at their disposal. Biological condensates, formed through weak multivalent interactions between intrinsically disordered protein sequences, have recently emerged as a powerful tool that can organize molecules into discrete micron-scale phase-separated structures with unique exchange rates, concentrations, and chemical environments^53–55. Similarly, Eukaryotic cells have membrane-bound subcellular compartments which provide additional spaces with which to partition the distribution of proteins⁵⁶. Finally, cells can use mechanical mechanisms such as force generation or curvature-inducing proteins like BAR proteins to manipulate or deform their structures^57,58.

Clearly, there is a rich array of subcellular structuring modules that exists, so what mechanisms does biology have at its disposal to self-organize them? A general theme that has emerged through decades of cell biology is that patterns of spatiotemporal signaling, operating at a subcellular level, provide a control-logic for self-organizing these diverse systems (Fig. 3B)^59–61. This is because most micron-scale structures within the cell can be assembled or disassembled locally by the presence or absence of other specific regulatory factors^44,62,63. Indeed, the critical role for signaling as a primary controller for the organization these structures is exemplified by the extent to which exogenous control can be imposed on our cells by a wide array of bacterial effectors, which hijack our structuring machinery to create new structures or motility that support invasive functions^64–67, or alternatively through optogenetic restriction of such factors to specific subcellular locations^68,69. Thus, the design challenge can be refocused around the question: how can synthetic biology construct new patterns of spatiotemporal activity within the cell?

Polarization mechanisms: establishing a landmark and axis for cellular organization.

Polarization is a powerful organizing mechanism in cell biology^70,71. For example, actin polymerization at the leading edge of a migrating cell can arise through localized stimuli triggering activation of small GTPases which recruit factors like Arp2/3 which stimulate nucleation of actin filaments⁷². Importantly, the establishment of polarized distributions of regulators need not be organized by stimuli, as they can also self-organize through specific regulatory relationships between signaling molecules^73,74. For example, local activation/global inhibition signaling networks whose core connectivity contains a combination of positive feedback (at the pole) and mutual inhibition (between the cell front and back) are thought to underlie the self-organized polarization of specific molecules in systems ranging from the C. elegans embryo to the mammalian epithelium (Fig. 3C)^75,76.

These evolved polarization networks typically contain a complex tapestry of regulatory links between multiple components, with additional redundancy and feedback loops layered on top of the core architecture. Remarkably, much in the same way that oscillating gene expression can be reconstituted from arbitrary parts, self-organizing polarity can be reconstituted synthetically as well (Fig. 3C). Taking advantage of the fact that budding yeast cannot synthesize the phosphoinositide PIP3, Chau and Walter installed the key machinery needed to read, write, and erase this lipid into yeast as an orthogonal signaling currency⁷⁷. Then, by synthetically connecting these activities together with regulatory relationships that resemble natural polarization systems, they were able to generate synthetic distributions of polarized PIP3 in yeast.

Critically, because these poles are orthogonal to yeast biology, they can potentially be used as a new landmark in the cell to recruit or control other activities. A synthetic polarized structure on the cell membrane could be used to direct the formation of new cytoskeletal structures or deliver specific payloads to the cell membrane. Alternatively, these poles could serve as an anchor point for concentrating factors that generate gradients emanating from the pole, with a resulting spatial wavelength set by the balance between localized activation and global turnover. While this PIP3 strategy is not generally extendable to most eukaryotic systems (which have endogenous PIP3), similar approaches could likely be implemented using other synthetic signaling currencies and systems that have recently been developed^78–81.

Carving out subcellular structure by wavelength using synthetic reaction-diffusion circuits.

While the ability to self-organize the front and back of a cell is important, it represents only one possible partitioning of cellular structure and one whose length scale is set by the size of the cell. For example, many cells must organize structured arrays of cytoskeletal structures into specific spacings for optimal performance or function^52,82,83. In ciliated protists like Tetrahymena geometric arrays of basal bodies and their associated motile cilia are typically organized along the cell-scale anterior-posterior axis at two distinct spatial wavelengths: a short wavelength spacing between adjacent basal bodies on the same track (λ₂: called a kinety); and a long wavelength spacing between the tracks themselves (λ₁: inter-kinety spacing) (Fig. 3D)^46,47.

The level of complexity in the organization of these single-cell geometries is simply remarkable, but at present so poorly understood that any hope for synthetic approximation might at first seem insurmountable. However, by recognizing that this complexity to a first approximation can be reduced to a few specific wavelengths that couple to the cytoskeleton, we can reframe the design challenge to one that asks how one can synthetically generate these spatial wavelengths within the cell. These design targets echo the temporal period of the oscillating gene expression circuits we considered earlier, but now operating in space.

Reaction-diffusion mechanisms provide one strategy for generating periodic spatial organization into cells⁸⁴. Originally conceptualized by Alan Turing, these systems describe the spatiotemporal behavior of activator and inhibitor molecules whose reactivity with one another leads to self-organizing static or dynamic pattern formation in the presence of differing diffusion rates. In the simplest example, an activator molecule promotes production of itself and its inhibitor, while the inhibitor negatively regulates activator production^85,86. Depending on the system parameters, a wide range of self-organizing concentration profiles can be generated, ranging from dynamic oscillations, traveling or spiral waves, and stationary patterns that display a consistent wavelength.

Reaction-diffusion waves have been observed to operate natively within individual eukaryotic cells, such as rho-actin waves in the cell-cortex of marine embryos, basophils and neutrophils^87–89. It also is a mechanism widely employed by many bacterial cells to organize the position and distribution of their organelles, flagella, chromosomes, or divisomes^90,91. In the best understood example of a bacterial positioning system, the E. coli MinDE system uses an ATPase, MinD, and its ATPase activating protein, MinE, to generate fast reaction-diffusion pole-to-pole oscillations on the membrane that control the spatial distribution of MinC, an inhibitor of divisome formation^90,92,93. In this way, these waves generate a nodal structure that provides the geometric information for the cell to define the midpoint for cell division.

Recently, the bacterial MinDE system has been repurposed as an orthogonal spatiotemporal signaling node to synthetically self-organize subcellular activities in human cells (Fig. 3D)⁹⁴. When MinD and MinE are expressed in mammalian cells, the two components self-organize into a dazzling array of waves, oscillations, and static patterns that fill the cell interior and are programmed by their relative and absolute expression levels. At low expression levels, fast (seconds-to-minutes timescale) waves of MinD and MinE proteins with persistent wavelength propagate throughout the cell, generating local oscillations with a genetically encodable frequency. At higher expression levels, stationary domains of MinD and MinE self-organize into geometric patterns with consistent spacing. Thus, surprisingly, completely new periodic spatial or temporal wavelengths are synthetically accessible with just the expression of two new proteins in a human cell.

These synthetic reaction-diffusion wavelengths are initially inert but can be further engineered to pattern other structuring elements in the cell. For example, when ActA, a bacterial effector that stimulates actin polymerization⁶⁶, is recruited to MinDE, it leads to localized production of filamentous actin controlled by the spatiotemporal behavior of the synthetic pattern⁹⁴. For traveling MinDE waves, this reconstitutes synthetic actin waves that resemble the rho-actin waves of natural evolved systems; and for stationary patterns it results in the formation of discrete, periodically spaced zones of actin. Thus, reaction-diffusion systems can synthetically project completely new spatiotemporal wavelengths onto the cell’s existing structure-building modules.

Combining these strategies with other organizing systems can enable multiple spatial wavelengths to be synthetically organized in parallel (Fig. 3D). For example, when droplet-forming intrinsically disordered protein sequences are recruited to MinDE waves in human cells, it triggers protein condensation⁹⁴. The resulting condensates are organized by the underlying wavelength of the reaction-diffusion system, as well as interactions between molecules both in droplets and the dilute phase. As a result, this can result in subcellular self-organization of droplets along two distinct wavelengths: a short wavelength (λ₁) controlled by interaction between adjacent droplets; and the long wavelength of the reaction-diffusion patterning system (λ₂).

Using synthetic approaches, the elegant organization of Tetrahymena basal body arrays that at first seemed impossible to reconstitute is becoming increasingly more accessible to engineer (Fig. 3D). A surprisingly simple patterning circuit that connects reaction-diffusion and droplet formation together can create a lattice with two distinct wavelengths, providing a coordinate system for recruiting factors that pattern the assembly of cytoskeletal structures using other emerging synthetic strategies^95–97. Layering in synthetic polarization systems could further orient these arrays along a new defined axis or restrict their formation to a specific length scale imposed by diffusion from a polarized landmark. Together, encoding a hierarchy of subcellular length scales provides a viable starting point for understanding and engineering complex cellular geometries.

Cell-cell communication self-organizes multicellular form and function.

The self-organization strategies and mechanisms that we described above act to partition and carve up subcellular geometry and structure. Those same concepts can be naturally extended to multicellular biological systems by recognizing the appropriate modules and controllers at play. For a multicellular system, the modules are played by the cells themselves, whose gene expression states, structures, and functions can be organized by the principles described in the previous sections. Consequently, the spatiotemporal controllers organizing cells as modules will take the form of cell-cell communication⁹⁸.

This cell-cell communication can be realized through a number of mechanisms that differ in the range by which they act. Secreted molecules like morphogens, hormones, or growth factors, enable cell-cell communication to occur over long distances⁹⁹. The specific distances that can be communicated in this way depend on the solubility and stability of the secreted factor, as well as the density of cells nearby to intercept these signals^98,100. As cells communicate with their neighbors, they can also communicate with themselves through autocrine signaling¹⁰¹. For example, when a T-cell activates and begins secreting interleukin-2 (IL-2), these signals not only act to stimulate the activity of nearby T-cells, but can also bind to IL-2 receptors on the secreting cell itself¹⁰². This can create strong positive feedback wherein the sender cell amplifies its own activity as it communicates with others^101,103.

Cells can also communicate with one another locally in a manner that depends on direct physical contact¹⁰⁴. For example, Notch/Delta signaling, T-cell receptor signaling, and Cadherin signaling all require receptor/ligand engagement to occur between two adjacent cells⁹⁸. This requirement for physical contact between cells can be rooted in mechanics, such as the force-mediated activation of Notch receptors¹⁰⁵; or through enhanced clustering of receptor/ligand between surface-presented molecules¹⁰⁶. Cells may also directly use physical forces to communicate with one another, using the mechanics of pulling and stretching to trigger signaling activities in nearby targets^107,108.

Because multicellular organization depends so strongly on cell-cell communication, many strategies have been developed to create new channels that allow for synthetically controlled communication between cells¹⁰⁹. Generally, these strategies all take the form of an engineered signaling receptor that either 1) changes the ligand that is recognized by the receptor to something synthetic; 2) redirects the signaling output of a receptor to a new kind of downstream target; or 3) a combination of 1 and 2.

For example, Chimeric Antigen Receptors (CARs) are synthetic signaling receptors that fall into category 1: they activate T-cell signaling by coupling synthetic antigen recognition (through a single-chain antibody) to “signal 1” (CD3ζ-chain) and “signal 2” (co-stimulation; typically 41BB or CD28) intracellular domains¹¹⁰. Such receptors have shown tremendous clinical promise in adoptive immunotherapy, where they can be programmed to target tumor cells for destruction¹¹¹. In contrast, synthetic Notch receptors¹¹² and intramembrane proteolysis receptors (SNIPRs)¹¹³ fall into category 3: like CARs, they use an extracellular ScFv to recognize an antigen, but now couple this to release of a synthetic transcription factor which can be used to generate any downstream transcriptional output of interest. The space of synthetic receptors available now is vast, from synthetic adhesion systems¹¹⁴ to synthetic GPCRs¹¹⁵, and more. We direct the reader to Manhas et al. for a more detailed review on synthetic signaling receptors¹⁰⁹.

Synthetic cell-cell communication can organize multicellular patterning.

Armed with enabling tools that provide synthetic biology new ways to engineer and manipulate cell-cell communication, what types of evolved multicellular morphologies can be emulated? We begin by considering synthetic analogues to one of the well-understood organizing mechanisms in multicellular development: gradients of the bicoid morphogen during drosophila embryo^116,117 (Fig. 4B). This gradient is generated through the localized maternal deposition of bicoid RNAs at the cell anterior, whose highly polarized distribution self-organizes a smooth gradient of bicoid protein running along the anterior posterior axis. Because bicoid is a transcriptional activator, the length scale of this concentration gradient can be exploited by genetic circuits to self-organize different spatially restricted domains of gene expression, such as hunchback in the anterior or bands of krüppel in the midzone⁹⁹.

Figure 4. Cell-cell communication self-organizes multicellular form and function.

(a) Self-organizing multicellular spatial patterning can be decomposed by treating individual cells as modules organized by cell-cell communication. Examples of different cell-cell communication strategies and emergent features are highlighted on the diagram.

(b) Bicoid gradients in the developing Drosophila embryo provide an example of an evolved morphogen gradient that establishes a single spatial wavelength operating at a multicellular level. Genetic circuits can decode this gradient into different patterns of gene expression.

(c) Morphogen gradients with synthetically tunable length scales have been generated through reconstitution of natural Hedgehog signaling in NIH 3T3 cells.

(d) Fully synthetic morphogen systems have been developed using engineered synthetic notch receptors that respond to secreted fluorescent proteins. This allows more complex gradients to be self-organized through the action of secreted activator and inhibitor molecules into higher-order circuits, leading to multi-band patterns of gene expression resembling those seen in the evolved Drosophila embryo.

(e) An example of an evolved multicellular reaction-diffusion patterning system. Wnt and Dkk establish an activator-inhibitor architecture that establishes a spatial wavelength for Wnt signaling that sets the spacing of hair follicles in a mouse.

(f) An example of a synthetic multicellular reaction-diffusion patterning system instantiated in E. coli using quorum sensing. By varying the strengths of the different regulatory nodes within the circuit, different spatial wavelengths could be encoded in the growing colony morphology.

The key features of these self-organizing gradients and bands of gene expression have recently been implemented synthetically. In the simplest example, a synthetically controllable Hedgehog (HH) signaling pathway was reconstituted in NIH 3T3 cells (Fig. 4C)¹¹⁸. For this system, a “sender” population of cells was generated by placing HH secretion under inducible control. An associated “receiver” population of cells was engineered such that HH activation of their target PTCH receptors drove expression of a fluorescent marker. Using a specialized cell-culture system, a local sender population could be pre-organized at one end of a linear geometry, with the receiver population extending from there. In this setting, sender cells generate gradients of HH activation whose spatial length scale can be synthetically controlled by either the secretion rate or overall density of the sender population.

The HH example explores patterning using natural morphogens and receptor systems, which limits the amount of control over how additional complexity can be introduced into the gradient formation process. In a second example, a fully synthetic morphogen system was developed using synthetic Notch receptors (Fig. 4D)¹¹⁹. Here, arbitrary secreted molecules (GFP and mCherry) were engineered to act as synthetic morphogens for cognate α-GFP and α-mCherry synNotch receptors that can activate downstream genetic circuits to introduce feedback into the system. Additionally, soluble antibodies could be secreted to act as morphogen inhibitors, sequestering them away from their cognate synNotch receptors. Connecting these morphogen possibilities together in different ways, the authors could create a number of different spatial patterns with tunable wavelengths of activity, including a two-zone expression pattern that resembles the pattern of hunchback and krüppel expression seen in the drosophila embryo.

The demonstration that cell populations can self-organize complex synthetic gradients using activating and inhibitory morphogens is quite intriguing, as it suggests that synthetic analogues of reaction-diffusion patterning should be possible. Such systems would emulate the function of evolved reaction-diffusion mechanisms, such as the Wnt/Dkk driven periodic spacing of hair follicles in mice (Fig. 4E)¹²⁰ and shh/bmp driven positioning of shark denticles¹²¹, among others. However, as synthetic morphogen systems can theoretically be fully orthogonal and arbitrarily tunable, these systems would likely be able to project any periodic spatial wavelength of interest onto cellular states within the population.

Despite experimental and theoretical progress^122,123, a multicellular reaction-diffusion patterning circuit has not been fully implemented with mammalian cells. However, such systems have been constructed synthetically in bacteria to program tunable periodic wavelengths of expression within the growing colony (Fig. 4F)¹²⁴. To achieve this, researchers took advantage of natural bacterial quorum sensing systems that enable cell-cell communication and connected these together synthetically to achieve an activator-inhibitor architecture. Because the quantitative behavior of reaction-diffusion systems depends strongly on both the strength of these regulatory relationships as well as the diffusivity of the quorum sensing molecules, they added in a layer of small molecule control over each of these key nodes. By varying the small molecule concentration, they could alter the overall system parameterization to control the emergent patterning wavelength that develops within the colony.

Insights from these engineered bacterial reaction-diffusion systems suggest paths forward for recapitulating these systems in mammalian cell settings^122,123. Greater control and flexibility over how different morphogens cross-regulate secretion rates and activation will likely be needed to put cells in the right parameterization regime for bona fide patterning. In addition, strategies for manipulating the relative diffusivity of different morphogens will be critical for tuning the resulting spatial wavelengths that emerge. Because natural morphogens frequently appear to travel not by free diffusion in solution but along cell layers through binding and release to receptors or extracellular matrix anchors¹¹⁸, programmability of a synthetic morphogen’s effective diffusivity might in fact be straightforward to achieve using protein engineering strategies that control the lifetimes of diffusing morphogens bound to their extracellular anchors.

Finally, all the multicellular self-organizing examples so far have been implemented using pre-organized landmarks. That is, sub-populations of senders or inhibitors are pre-positioned within the multicellular structure to establish the resulting gradients. A future challenge for synthetic biology to overcome is to develop strategies to self-organize these anchor points de novo from within an expanding clonal population. This may require developing synthetic cell motility or multicellular edge-detection circuits that interface with fate-determining genetic circuits to establish a niche population around which the rest of the spatial organization is implemented. This is analogous to the generating a self-organized polarized landmark at the single-cell level discussed in the previous section, and will likely benefit from abstracting the architectures and principles of those single-cell systems to a multicellular context.

Integrating different length and timescales: a path towards synthetic forms most beautiful.

In this review, we have identified how many evolved aspects of biological form and function require the ability to self-organize modules and activities around specific length or time scales and have highlighted progress in engineering synthetic systems that can instantiate these features operating at different operational scales (Fig. 5). We recontextualized genetic circuits as self-organizing systems that can specify either a single or periodic timescale in the expression dynamics they generate. At the single-cell level, we discussed how synthetic polarization and reaction-diffusion circuits can specify analogous spatial wavelengths that operate at a subcellular level. Finally, we considered how those kinds of spatial wavelengths can also be generated at much longer length scales in multicellular systems using engineered cell-cell communication, synthetic morphogen gradients, and analogous reaction-diffusion mechanisms.

Figure 5. Synthetic and evolved forms as hierarchies of linked self-organizing mechanisms.

Schematic depicting how different biological modules and activities can be self-organized by different length scales and timescales. By encoding new spatial and temporal scales that be integrated together and project onto different cellular and biological activities, we are poised to begin engineering constructing and endless array of synthetic forms most beautiful.

Given the diversity of self-organizing length and timescales that can now be engineered, synthetic biology is well-positioned to integrate these different hierarchies together and begin the challenge of building self-organizing lifeforms. Such systems could contain novel cell-types with structures and dynamics that are synthetic in nature but achieve specific, engineered objectives. Those cell-types in turn could cooperate in multicellular compositions to create synthetic tissues that divide up labor and behavior in new ways. If the cell-types and behaviors are narrowly altered, such self-organizing biological composites could be imagined as serving therapeutic ends, interfacing with the human body to reinvigorate damaged tissues or interface with and remodel tumor microenvironments recalcitrant to existing therapies.

On the other hand, as the scope of synthetic cell-types and multicellular organizations grows, these synthetic forms may expand to become new lifeforms all their own, with capabilities and utilities that extend beyond mere augmentation of what evolved life already does. Such self-organizing systems could become a designable platform for soft robotics, energy production, environmental remediation, or parallelized chemical computing. Unconstrained by selective pressure and epistasis, the space of synthetic forms and functions we might create is likely to be limited only by our imagination and determination.