Molecular genetic studies of plastid genomes show that they encode only 60-200 proteins, while perhaps as many as 5,000 nuclear-coded gene products are targeted to plastids [6]. From complete sequences it is known that each cyanobacterial genome codes for at least 1,500 proteins, and they are therefore at least an order of magnitude larger than plastid genomes. It is perhaps surprising that the size of the proteome of a free-living cyanobacterium is not greatly different from that of a subcellular organelle. Genomic studies have been very important in showing the evolutionary fate of the cyanobacterial genes that originated from the endosymbiotic pre-plastids. The genes in pre-plastids were either retained, lost, or transferred to the nucleus. The process of transfer of genes to the nucleus would have involved duplication of each plastid gene, and a nuclear copy of the gene becoming able to produce a functional product in the cytosol or, with appropriate targeting sequences, in other compartments.

The cyanobacterial ancestor of the plastids was, relative to the three cyanobacteria with completed genome sequences that were examined by Martin et al. [6], closer to N. punctiforme than to P. marinus or Synechocystis sp. Although three genomes is not a large sample size, it is of interest that N. punctiforme is a diazotroph, so the plastid ancestor could also have been a nitrogen-fixer. Were early plastids perhaps also able to fix atmospheric nitrogen?

Figure ​Figure11 illustrates the various endosymbiotic events described here. Amongst eukaryotes, the apicomplexan parasitic pathogens Toxoplasma and Plasmodium have curious cytoplasmic organelles bounded by three membranes, namely 'apicoplasts', which genome sequencing has established as bona fide plastids complete with a characteristic inverted repeat within the plastid genome [9]. The presence of three membranes, as is found around chloroplasts of dinoflagellates and euglenoids, betrays an ancestry from a secondary symbiosis, as does the presence of four membranes surrounding the plastids of, for example, photosynthetic heterokonts (a diverse group, some of which are algae) such as diatoms and brown algae. The function of the apicoplast is not clearly understood, but one suggestion is that it is indispensable for the synthesis of iron-sulfur proteins. The function of the residual plastid genome is even less clear, and it provides a test case for any theory for the function of organellar genes. Although Plasmodium has a plastid genome that some think is on the way out, trypanosomes, which are also non-photosynthetic, have no plastid or plastid genome at all, but are now clearly seen to be former euglenoids because of the remaining genes for a variety of plant-like enzymes, including sedoheptulose-1,7-bisphosphatase (otherwise found only in the Benson-Calvin cycle) [10,11].

An important contribution to bridging the evolutionary gap between Mesostigma and Marchantia is the work of Turmel et al. [15] on a member of the charophytes sensu stricto (that is, excluding Mesostigma) Chaetosphaeridium globosum. Before the work of Turmel et al. [15] only fragmentary data addressed the issue of gene content and organization of the eight charophytes sensu stricto. The complete plastid genome sequence of Chaetosphaeridium globosum [15] shows that most of the embryophyte characteristics were present in the charophyte alga, so that the major changes had occurred between the branch to Mesostigma and that to Chaetosphaeridium. The common features shared by the plastid DNA of Chaetosphaeridium and of embryophytes include the gene content, the intron composition, and the gene order. Thus, the Chaetosphaeridium chloroplast genome has 124 genes (compared to 136 in Mesostigma and 110-120 in embryophytes), one Group I intron (there are none in Mesostigma and one in embryophytes), 16 cis-spliced Group II introns (none in Mesostigma and 18-19 in embryophytes) and one trans-spliced Group II intron (none in Mesostigma, one in embryophytes). Genome size (118-155 kilobases) is relatively constant among Mesostigma, Chaetosphaeridium and higher plant plastids. By contrast, the mitochondrial genome of Chaetosphaeridium is closely similar to that of Mesostigma in terms of size (57 kb and 42 kb, respectively), gene content and, perhaps, intron content. Chaetosphaeridium has a much smaller genome size than the obese mitochondrial genomes of Marchantia (187 kb) or Arabidopsis (367 kb), and many more cis-spliced Group II introns (18-25 rather than two). The apparently different tempo of evolution in mitochondria and plastids of the charophytes deserves further investigation. An important point about the functional genomics of the plastid is the determinant of which genes essential for plastid function are retained in the plastid genome. Higher plant plastid genomes have slightly fewer genes than in the plastids of the charophytes sensu lato (that is, the charophytes sensu stricto plus Mesostigma).

The endosymbiont hypothesis took a long time to graduate from wild and untestable speculation to an accepted view of plastid origins and evolution. In contrast, comparative genomics has quickly elevated the kinship of chloroplasts and cyanobacteria to a keystone of our understanding of the most abundant of cells, the primary producers on which life now depends, not to mention some vicious and enterprising pathogens whose exploits are a global burden to human health. The title of this article asks what the cyanobacteria have done for plants. "What have they not done?" is a question perhaps more easily addressed.