The smallest minimized genomes from the top-down approach will likely be produced by minimizing the already smallest genome, that of M. genitalium. JCVI has been pursuing this plan in 6 ambitious steps:

JCVI has completed steps (i)–(iv). Step (iv) was particularly challenging because of slow growth rates and because bacterial genomes engineered in yeast have DNA restriction/modification systems that are incompatible with the Mycoplasma host cell [26]. To learn how to transplant and express chemically synthesized genomes (iv), JCVI “booted-up” a synthetic, essentially wild-type, computer-specified, Mycoplasma mycoides genome (1,080 kbp) in a closely related cell to yield “Synthia” [13]. This technological milestone marks the dawn of “synthetic genomics” and will undoubtedly accelerate the engineering of microbial factories, once costs are significantly lowered, producing fuels, pharmaceuticals, chemicals, and novel biomaterials (see Prospects for Biotechnology). Notwithstanding the importance of this achievement, it should not be overinterpreted as synthesis of a cell or life, as standard usage of “synthetic” would imply either cell-free synthesis of the whole cell (rather than its genome) or generation of something very unnatural (rather than a genetically-modified organism). The published plan for steps (v) and (vi) is to synthesize a M. genitalium-based genome lacking all dispensable genes to boot up a “Mycoplasma laboratorium” cell (last paragraph of ref. [20]). However, though virtually all genes that are individually dispensable in M. genitalium have been determined, it is recognised that a major hurdle is synthetic lethals (i.e. non-viable cells when two individually viable mutations are combined [20]).

How may cellular complexity and synthetic lethality be circumvented to allow top-down production of a minimal genome? One route is step-wise deletion of the 100 individually dispensable genes, perhaps aided by directed evolution [24]. However, the number of combinations is astronomical, rational choice of combinations is limited by poor understanding (e.g., the functions of one fifth of the genes of M. genitalium remain to be determined), and considerably less than 100 of the 525 genes are likely dispensable in combination. There will also be multiple different minimal genome “solutions,” depending on the temporal order of deletion. Nevertheless, this will teach us much about redundancy in biology.

A second route is evident from tables of M. genitalium genes involved in the core replicative functions of DNA, RNA and protein syntheses [4]: these genes are in the minority, with the majority of M. genitalium genes being involved in functions such as metabolism of small molecules. Thus, if additional nutrients were supplied in the extracellular medium (and perhaps their uptake aided by encoding extra transmembrane transporters) it may be feasible to delete many more genes. This could take us down to a truly minimal, protein-coding cell: one sufficient for replication but not for metabolism of most small molecules.

Interestingly, development of such extreme metabolic dependence without loss of genetic independence may have already occurred in the reductive evolution of the intracellullar bacterial endosymbionts of insects [27]. These recently-sequenced symbiont genomes include the smallest non-organellar, non-viral genomes, Carsonella ruddii (160 kbp; 213 genes [28]) and Hodgkinia cicadicola (144 kbp; 188 genes [29]). In contrast to mitochondrial and chloroplast evolution, there is no evidence so far of gene transfer from bacterial symbiont to host [27]. Almost all of the core replicative functions have been predicted computationally to reside in the symbiont genome, although notable exceptions are several essential tRNAs and aminoacyl-tRNA synthetases [27,30]. Ultimate proof of genetic independence can only come from development of a defined in vitro system for replication of either a bacterial symbiont or a derivative engineered to encode any missing essential genes. Such experimental verification would constitute our third envisioned top-down route to a minimal genome.

Of all the macromolecular components from E. coli and its bacteriophages, only 151 were hypothesized to be sufficient for the MCP, constituting a minigenome of 113 kbp [4]. Of these 151, it is striking that 96% are for protein synthesis and that there is considerable similarity in gene number and content and genome size to the recently-sequenced, extremely-metabolic-dependent, bacterial endosymbionts of insects (see above). An RNA/protein-based transcription/translation system has been reconstituted from purified components [40], but the omission of DNA does not simplify the number of genes that ultimately will be necessary to encode the whole system for self-replication. Rather, it creates a new set of challenges unsolved in the modern world: production of a functional large RNA genome that avoids inhibitory double-stranded RNA structures and replicative mutations [35].

Progress in step (i) has been rapid for E. coli (but slow for M. genitalium [41]). Of the missing 1–4 key ribosomal RNA (rRNA) modification genes, 3 have just been discovered [42–44]. The gene for modifying transfer RNA (tRNA) A37 to t⁶A has also been found and shown to be essential for E. coli viability [45]. This only leaves as little as one other gene to find, involved in modifying tRNA U34 to cmo⁵U, with 2 genes in that pathway already known [46]. Thus, reconstitution from purified components of every subsystem of the MCP is tantalizingly near. In an attempt to close perhaps the biggest remaining gap, we are over-expressing the 5 known key rRNA modification enzymes [4] to test for activation of unmodified 23S rRNA transcripts necessary for synthesis of ribosomes in vitro.

Less progress has been reported on steps (ii)–(iv). With regard to step (ii), though the E. coli translation apparatus and ribosome were reconstituted separately from purified cellular components 3 decades ago, their translational accuracy is poorly characterized and in vitro efficiencies of protein synthesis and ribosome turnover remain low in both purified and crude systems (Table 1). The break-even milestone for ribosomes making all of the proteins in the proposed minigenome [4] is synthesis of ~35,000 peptide bonds by each ribosome (including 7491 peptide bonds for the ribosomal proteins). Towards the integration required for steps (iii) and (iv), bacterial transcription initiation has been reconstituted in a purified translation system [47], purified DNA-dependent transcription and translation has been performed within liposomes [48], and membrane proteins involved in phospholipid synthesis have been synthesized in active form in liposomes [49]. But some of the other subsystems require unphysiological conditions that preclude integration. Simple systems for DNA replication require thermocycling and oligo primers (PCR or circle-to-circle amplification [50]), while self-assembly of the E. coli ribosome from natural components requires low and high Mg²⁺ concentrations, high temperatures and long incubation times [51]. Nevertheless, physiological conditions for E. coli ribosome assembly have now been found and rRNA synthesis, ribosome assembly and translation (Fig. 1) have been integrated under batch conditions (Jewett and Church, submitted). The next steps will be substitution of the E. coli cells and extracts used for the macromolecule syntheses by purified subsystems.

How might the efficiencies and utilities of purified systems be improved? There are some recent indications that adding genes not on the minimal list [4] should help. Inclusion of translation elongation factors not present in PURE kits might improve efficiency and/or accuracy: EF-P facilitates formation of the first peptide bond by positioning fMet-tRNA_i^fMet [52], and LepA promotes back translocation of the mRNA-tRNA complex [53,54]. Comprehensive analysis of the individual effects of every E. coli protein on purified translation showed that 344 (8%) were stimulatory [55]. Most beneficial were ATP-dependent RNA helicase, HrpA, and trigger factor, increasing yields by ~80% and ~30%, respectively. More than 20 different auxiliary factors are thought to facilitate ribosome assembly, including chaperones, GTPases and helicases [56]. For example, ATP-dependent RNA helicase, DbpA, has specificity for 23S rRNA [57], and RimJ functions in ribosomal protein acetylation and in 30S subunit assembly [58]. Choices for gene addition will be informed by studies such as the measurement of kinetic effects on 30S assembly of Era, RimM and RimP [59]. Also, cytoplasmic mimicry has been shown to be a powerful guiding principle. Mimicking combined energy metabolism, oxidative phosphorylation and protein synthesis in crude extracts increased protein synthesis yields (Table 1; [60,61]). Activating natural energy metabolism in crude extracts reduces costs and suggests that incorporating metabolic modules [62] into the MCP could further increase utility.

The MCP mostly involves synthesis and optimization of purified translation systems. Such systems have a number of advantages over alternative methods of protein synthesis such as lack of RNases/proteases/inclusion bodies, high compatibility with cytotoxic proteins, flexibility of incorporation of unnatural amino acids, ease of product purification, and direct control of reaction conditions. The main hurdle preventing application of the PURE system in biotechnology is the high cost (Table 1) due to its production from >30 different fermentations. To address this limitation, we are developing a cost-effective method for over-expressing the entire system in a single E. coli cell followed by single batch purification [18,19,63]. Selection of variant 23S rRNAs for improved unnatural amino acid incorporation [64] could be uncoupled from cell viability by synthesizing ribosomes in vitro; such variants would facilitate the directed evolution of peptidomimetic drug candidates [65].