Why Chloroplasts and Mitochondria Contain Genomes

Chloroplasts and mitochondria originated as bacterial symbionts. The larger, host cells acquired genetic information from their prokaryotic guests by lateral gene transfer. The prokaryotically-derived genes of the eukaryotic cell nucleus now function to encode the great majority of chloroplast and mitochondrial proteins, as well as many proteins of the nucleus and cytosol. Genes are copied and moved between cellular compartments with relative ease, and there is no established obstacle to successful import of any protein precursor from the cytosol. Yet chloroplasts and mitochondria have not abdicated all genes and gene expression to the nucleus and to cytosolic translation. What, then, do chloroplast- and mitochondrially-encoded proteins have in common that confers a selective advantage on the cytoplasmic location of their genes? The proposal advanced here is that co-location of chloroplast and mitochondrial genes with their gene products is required for rapid and direct regulatory coupling. Redox control of gene expression is suggested as the common feature of those chloroplast and mitochondrial proteins that are encoded in situ. Recent evidence is consistent with this hypothesis, and its underlying assumptions and predictions are described.

Why do mitochondria and chloroplasts require their own separate genetic systems when other organelles that share the same cytoplasm, such as peroxisomes and lysosomes, do not? . . .. The reason for such a costly arrangement is not clear, and the hope that the nucleotide sequences of mitochondrial and chloroplast genomes would provide the answer has proved unfounded. We cannot think of compelling reasons why the proteins made in mitochondria and chloroplasts should be made there rather than in the cytosol.

Hypotheses concerning the function of cytoplasmic genetic systems
The lock-in hypothesis According to the lock-in hypothesis [12], core components of multisubunit complexes must be synthesized, de novo, in the correct compartment. However, mechanisms of protein import and targeting [16,23,34,44] now effectively address the problem of specifying the correct locations for subunits of membrane protein complexes. Locking assembly into specific cellular compartments does not now seem likely as the primary function of organellar genomes.
The frozen accident 'The frozen accident hypothesis' states that the evolutionary process of gene transfer from organelle to nucleus was under way when something happened that stopped it. One example [50] suggests that successful import into mitochondria of the precursor proteins arising from newly translocated genes proceeded for long enough, after the original endosymbiosis, for most of the symbiontderived genes to be lost to the cell nucleus. The 'accident' that halted the process was the evolutionary origin of exocytosis and protein secretion. The 'frozen accident' seems incompatible with the precision and specificity of protein targeting [16,23,34,44].

Hydrophobicity
The hydrophobicity hypothesis is that intrinsic membrane proteins must be synthesized de novo within organelles [13,37,50]. This hypothesis has some very clear counter-examples in chloroplasts. One is the large subunit of the enzyme ribulose-1,5bisphosphate carboxylase-oxygenase (RubisCO), an abundant but water-soluble protein. RubisCO has a chloroplast-encoded large subunit and a nuclear-encoded small subunit [18], and serves as a model for coordination of nuclear and chloroplast gene expression. Other counter-examples from chloroplasts are the nuclear-encoded but hydrophobic subunits of the chloroplast light-harvesting complexes, LHC II and LHC I. Hydrophobicity is, in many cases, a feature of proteins that are encoded in situ, but does not seem to be, in itself, the issue.

Some proteins cannot be imported
Protein import mechanisms appear mostly to rely on specific molecular chaperones for guided unfolding, prior to membrane insertion, and refolding, following translocation of the polypeptide into its destined compartment [18,22,24]. Since polypeptides are mostly 'threaded' through membranes, no special constraint on transport should be expected to arise from three-dimensional structure or surface properties of a protein. Holoproteins containing co-factors can be transported along with the apoprotein by the twin-arginine translocation (Tat) system [11,40,43]. The presence of a vesicular transport system in chloroplasts [51] also counts against the unimportability hypothesis. If vesicular transport into chloroplasts is possible it is difficult to see why there is any protein that is universally forbidden from changing compartments within the cell.

Some genes cannot be moved
Mitochondria show departures from the otherwise universal genetic code [2,10]. There is, therefore, an obstacle to correct cytosolic translation of mRNA that is transcribed from unchanged nuclear copies of mitochondrial genes. Distinct mitochondrial 'dialects' of the genetic code must present a barrier to expression of nuclear copies of mitochondrial protein-coding genes. However, differences in the genetic code are unlikely to be the primary reason why genes are still present in mitochondria, since it is not clear why the code has to be different for some genes and not for others -the problem remains of why there are mitochondrial genes at all. The significance of this heterogeneity in coding is most unlikely to rest in the early origin of mitochondria, since the 'universal' code is employed by bacteria. Furthermore, the diversity of departures from the 'universal' code amongst mitochondria of different eukaryotic lineages suggests that these departures had independent origins.

Co-location of genes and gene products for redox regulation of gene expression
The hypothesis of co-location for redox regulation of gene expression, here termed CORR, was systematically proposed in two articles [4,5], developed [9] and has been independently reviewed [38]. A more detailed, recent analysis is presented elsewhere [7]. Figure 1 illustrates the general idea of redox regulatory control giving rise to retention of genes in organelles. The CORR hypothesis is based on ten assumptions, or principles, as listed below.
1. Bioenergetic organelles evolved from freeliving bacteria. The endosymbiont hypothesis The CORR hypothesis assumes that it is beyond the ability of the nuclear-cytoplasmic system to respond rapidly and directly to changes in light, carbon dioxide and oxygen concentration. Such responses require continued redox regulation of gene expression. This regulation has therefore been retained from the eubacterial endosymbionts that were ancestral to chloroplasts and mitochondria, and its continued operation requires co-location of the genes concerned, with their gene products, within bioenergetic organelles was once controversial [27] but is now widely accepted [28,30]. 2. Gene transfer between the symbiont, or organelle, and nucleus may occur in either direction and is not selective for particular genes. Nothing certain seems to be known about the mechanism of intra-cellular gene relocation [21,29,31,32], so there is no basis for asserting a mechanistic argument for selectivity. Gene transfer between organelle and nucleus seems to be a special case of lateral, or horizontal, gene transfer [14]. 3. There is no barrier to the successful import of any precursor protein, nor to its processing and assembly into a functional, mature form. This principle forbids the existence of a protein, even a synthetic one, that cannot be imported by some means, as a precursor.

Direct redox control of expression of certain genes was present in the bacterial progenitors of chloroplasts and mitochondria, and was vital for selectively advantageous cell function before, during and after the transition from bacterium to organelle. The mechanisms of this control have been conserved.
Experiments carried out in vitro on the products of protein synthesis [3,17] or RNA synthesis [25,33,53] are consistent with a general phenomenon of redox dependency of gene expression in bioenergetic organelles. Redox control of chloroplast transcription has been demonstrated [8,35,36,48], and was prompted by this principle. 5. For each gene under redox control, it is selectively advantageous for that gene to be retained and expressed only within the organelle. The twin selective advantages of redox control are likely to be energetic efficiency and the suppression of the harmful side effects of electron transport operating on the wrong substrates. 6. For each bacterial gene that survives and is not under redox control, it is selectively advantageous for that gene to be located in the nucleus and expressed only in the nucleus and cytosol. If the mature gene product functions in chloroplasts or mitochondria, the gene is first expressed in the form of a precursor for import. There are several possible reasons for the selective advantage of nuclear location of genes for organellar proteins. One is the decreased probability of mutation arising from free-radical by-products of 'incorrect' electron transport [9,39]. Another is that organellar genes do not undergo recombination and are present in relatively small copy numbers, so that disadvantageous mutations will spread relatively quickly through a clonal population of organelles [41]. 7. For any species, the distribution of genes between organelle and nucleus is the result of selective forces that continue to operate. Mitochondria that lose their function in aerobic respiration also lose their genomes, as seen in the relict mitochondria of microsporidia [15,49] and the 'mitosome' of Entamoeba histolytica [47]. In chloroplasts, loss of photosynthesis results in loss of photosynthetic genes. The examples of Epifagus (a parasitic higher plant) and the residual apicomplexan plastids of Plasmodium [52] and Toxoplasma, where some plastidic genetic system is retained, mean that some role for their gene products must be found which requires redox regulation of gene expression, if this principle is correct. 8. Those genes for which redox control is always vital to cell function have gene products involved in, or closely connected with, primary electron transfer. These genes are always contained within the organelle. This principle would be violated if any photosynthetic reaction centre core subunit or respiratory chain subunit currently organelle-encoded is found which functions in primary, vectorial electron transport and is nevertheless encoded in the nucleus and synthesized cytosolically as a precursor for import, all without any selective cost to the individual. 9. Genes whose products contribute to the organelle genetic system itself, or whose products are associated with secondary events in energy transduction, may be contained in the organelle in one group of organisms but not in another, depending on the physiology and biochemistry of photosynthesis and respiration in the species concerned. Even phylogenetically moderatelyrelated species (within kingdoms) show differences in the location of some genes, e.g. the ATP synthase subunits of Neurospora and Saccharomyces [10]. 10. Components of the redox-signalling pathways upon which co-location for redox regulation depends are themselves not involved in primary electron transfer and so their genes have been relocated to the nucleus. This principle states that the redox signalling pathways that are required for CORR should not themselves be expected to utilize components whose biosynthesis must be under direct redox control. Such components should therefore fall into the major category of organellar proteins and be imported as precursors from the cytosol.

Conclusion and prospects
The CORR hypothesis is that the function of chloroplast and mitochondrial genomes is to provide co-location (of gene and gene product) for (evolutionary) continuity of redox regulation of gene expression. This hypothesis seems to be consistent with available evidence and applies equally to chloroplasts and mitochondria. CORR predicts regulatory properties of known components of chloroplasts and mitochondria that suggest flexibility in energy metabolism. These components include the RubisCO large subunit and CF o and F o subunits of coupling ATPase in both chloroplasts and mitochondria. Redox regulation of the relative stoichiometry of components, as seen in regulation of chloroplast photosystem stoichiometry [35], may extend to (C)F o -(C)F 1 -ATPase, suggesting that the function of retention of genes for Why some organelles have genomes 35 one or more of these components is flexibility in the stoichiometries of H + to e − and H + to ATP in chemiosmotic coupling [6,45].
The central proposal of CORR is that chloroplasts and mitochondria contain genes for proteins whose function in electron transfer demands rapid, direct and unconditional redox regulatory control of their biosynthesis. This testable hypothesis makes many predictions, and may explain the otherwise puzzling distribution of genes between the nucleus and the cytoplasm of eukaryotic cells.