Structural organization is one of the most conspicuous features of cells, and possibly the most elusive. No one really doubts that cell functions commonly require that the right molecules be in the right place at the right time, or that spatial organization is what distinguishes a living cell from a soup of its molecular constituents. But the tradition that has dominated biological research for the past century mandates a focus on the molecules, and so our first step is commonly to grind the exquisite architecture of the living cell into a pulp. Few molecular scientists have asked whether anything irretrievable is lost by this brutal routine. Such questions as how molecules find their proper place in a framework of orders of greater magnitude, or how spatial order is transmitted from one generation to the next, have been largely neglected until recently.

Two current and quite excellent short reviews afford an entry into the wilderness. Eric Karsenti takes an historical approach to the role of self-organization in creating order on the cellular scale. The physical principles are arcane, but some aspects are actually quite familiar. We have known for half a century that supra-molecular complexes often arise by self-assembly, without any input of either information or energy; examples include lipid bilayer membranes, ribosomes, microtubules, S-layers, and virus particles. But the scope of self-organization has been greatly enlarged in recent years by the discovery that an array of dynamic structures can be generated in the presence of an energy source, usually ATP or GTP. The mitotic spindle of eukaryotes has been identified as a self-organizing machine; the endomembrane system may be another. Like self-assembly, self-construction (my term) requires no external source of information, but it does entail continual energy consumption. In a complementary article, Allen Liu and Daniel Fletcher survey a selection of efforts to reconstitute cellular functions in simplified systems, starting with cell-free extracts or purified proteins. Ingenious experimenters have managed to reconstitute the essentials of actin-based motility, membrane protrusion, the oscillatory system that localizes the midpoint of bacterial cells, and now also the contraction of the Z-ring. Though much remains to be learned, it is safe to conclude that the lower levels of cellular order, at least, are products of pure chemistry: they arise by interactions among the molecular constituents in ways that require the cell as a whole to supply energy and a permissive environment, but no spatial instructions.

      This is excellent science, which takes us some way towards bridging the gulf between nanometer-sized molecules and cells in the range from micrometers to millimeters. It also extends the genome’s reach deep into cellular structure. In a self-organizing system, the “instructions” must be wholly inherent in the molecular parts, and ultimately derive from the corresponding genes. It is the genome that specifies the architecture of the mitotic spindle, not explicitly but indirectly: the form and even functions of the spindle are implied in the structure of the spindle proteins, and in their interactions. And if the spindle can be envisaged as a creature of self-organization, why not the entire cell? Yes, indeed—but as we ascend the hierarchy of biological organization, the meaning assigned to self-organization and its underlying mechanisms undergo significant changes. Cells do not construct themselves from pre-fabricated standard parts; instead, they grow. And that mode of self-organization is not purely chemical, for it must produce parts that have biological functions, performed in the service of a larger entity that can compete and thrive in the wide world.

We are quite well informed about just how cells grow, and it is clear now that they do so by modeling themselves upon the existing structural framework, which is thereby transmitted to the next generation. To be sure, all the macromolecules involved in this sort of self-organization are gene-specified, but spatial order is not. There appear to be very few individual genes that prescribe dimensions, position, or orientation on the cellular scale. Instead, thanks to the ways in which cells grow, spatial cues are sometimes inherited in a manner quite independent of genes, (a well-established phenomenon known as “structural inheritance.” The form and organization of cells thus stem from two distinct informational roots: the genomic instructions that specify the parts, and the continuity of cellular architecture that guides their placement. Specified proteins and cellular guidelines operate synergistically, reinforcing each other to generate form and organization. As Rudolf Virchow might have said, it takes a cell to make a cell.

      Evidence to support such a holistic view of what happens during growth is scattered, but continues to accumulate. Let’s glance at some examples. First, while many sub-cellular structures can be envisaged as products of self-construction from preformed parts, others cannot. A familiar instance is the peptidoglycan wall of bacteria, which consists of a network as large as the cell, made up of covalently-linked subunits. Enlargement during growth calls for extensive cutting, splicing, and cross-linking, even while keeping the wall physically continuous from one generation to the next. Second, even self-organizing structures must do so in a manner that ensures their correct placement in cellular space. A particularly neat example comes from recent work on the role of microtubules in cell morphogenesis of the fission yeast, Schizosaccharomyces pombe (reviewed by Martin). Microtubules define the poles of elongating cells by depositing there various members of the Tea complex, which in turn recruit additional factors. Cells of a certain mutant, orb6, grow as spheres even though they possess all this machinery.