Although de Duve's Perspective, “The origin of eukaryotes: a reappraisal”1 is masterful and relevant, certain lesser known but important work was overlooked. Support for his idea abounds: phagocytotic intracellular motility preceded the 'adoption' of mitochondria and plastids (Fig. 1). What de Duve calls “coherent collections of enzymes” are 'semes', the units of evolutionary analysis2.

Figure 1: Did this cell originate from an archaebacterium?
figure 1

The image shows a porcine epithelial cell (line LLCPK), with microtubules stained in red and the microtubule tip-binding protein EB1 stained in green. The nucleus is stained in blue. Image courtesy of Lynne Cassimeris, Department of Biology, Lehigh University, Bethlehem, Pennsylvania, USA.

Full size image

The archaebacterial–eubacterial merger (Gupta's chimaera3) resulted in membrane fusion; archaebacterial lipids and proteins formed the endoplasmic reticulum whereas the Golgi components evolved from eubacterial membrane biosynthesis4. The archaebacterial–eubacterial symbiotic merger of a thermoacidophilic sulphidogenic heterotroph5,6 with a motile sulphide-to-sulphur oxidizing heterotroph occurred under the threat of oxygen toxicity. From this syntrophy, hundreds of protists evolved. Their descendants thrive in anoxic habitats (for example, pelomyxids, mastigamoebae, devescovinids, oxymonads, trichomonads and other parabasalids). The presence of phagocytosis, mitosis and endocytosis in these amitochondriates testifies to the evolution of cytoskeletal motility (Fig. 2) before mitochondria2. The contributor of motility to the chimaera was the ancestral 'Perfil'ieva', a free-living, aerotolerant, sulphurous mud-scum-mat Spirochaeta-like eubacterium of geochemical significance7,8 (now banked in the German culture collection in Braunschwieg, with strain accession numbers Str. P=DSMZ 19205 and Str. SR=DSMZ19230). By use of Hall's new algorithm, more than 50 genes for the synthesis of eukaryotic enzymes and lipids were acquired by the chimeric eukaryotes in the transition from a Spirochaeta-like eubacterium to the [9(2)+2] motility organelle (J. L. Hall & L.M., unpublished observations). At lower stringency, even more eukaryotic sequences in Perfil'ieva should be detected.

Figure 2
figure 2

Extant microbes exemplify each step in this karyomastigont model of the origin of the nucleocytoskeleton.

Full size image

Permanent bacterial conjugation-nucleoid membrane formation9 generated the nucleus tethered to [9(2)+2] motility organelles (such as undulipodia including cilia) and its attachment apparatus, which became the centriole–centrosome system (Mixotricha is analogous10). The intron-less eubacterial DNA (now in the nucleolinus of the nucleolus) became the centrosome–centriole DNA11. Specific centrosomal RNAs and proteins for assembly and maintenance of the centriole–centrosome system12 are discussed in Ref. 13. Redundancy reduction followed fusion. Genetic and metabolic systems acquired from intracellular motile symbionts were integrated and redeployed (comparable with what occurred in the evolution of Staurojoenina14, Peridinium balticum, Mesodinium rubrum and Hatena)15,16. Natural selection in microoxic habitats maintained heterotrophic chimaeras.

Information-molecule loss from centriole-kinetosomes (such as gene loss in plastids and mitochondria) occurred in the sulphurous Proterozoic eon (2500–541 million years ago17) during which time peroxisomes were acquired. Besides peroxisomes, organelles that are probably of bacterial origin that continued the trend of loss of genes to the nucleus until completion include some hydrogenosomes, γ-particles of Blastocladiella, and mitosomes.

Genome analysis of hydrogenase distribution is explicable only if hydrogen gas production entered anaerobic protists via at least two distinct events18: acquisition of a cytosolic enzyme complex or of a symbiotic bacterium (or both). The hydrogenase seme derives from the chimaera's eubacterial partner (aerotolerant sulphide-oxidizing Perfil'ieva) or some other 'adopted' eubacterium. The hydrogenosomal Fe–Fe hydrogenases — for example, those of Trichomonas vaginalis and most parabasalids (many of which are multinucleate, but none of which are mitochondriate) — evolved from Clostridium-like bacteria. But the hydrogenases of Spironucleus, Giardia and Entamoeba histolytica are cytosolic; presumably they retained enzymes from the eubacterial ancestor (for example, the sulphur syntrophic Perfil'ieva). Anaerobic chytrids, even in the same genus, differ markedly (Neocallimastix ovalis versus Neocallimastix frontalis), confirming hydrogenase–hydrogenosome polyphyly19. The origin of Fe–Fe hydrogenases that are incapable of generating hydrogen gas in 'crown taxa eukaryotes' (animals, plants and fungi) noted by Meyer17 is implied by the data that de Duve discussed. The eubacterial cytosolic or periplasmic hydrogenase complex that was acquired from eubacterial ancestors changed during eukaryosis as intracellular motility evolved in amitochondriates. The hydrogenase system hypertrophied, mutated or was lost in response to the rising oxygen threat. Dispensable hydrogen gas production was not selected for, but rather hydrogenases and/or hydrogenosomes and their components were retained for myriad other semes.

Semes must be identified. Amino-acid or nucleotide homologies without seme identification lead to systematic inaccuracy in evolutionary reconstruction. Molecular sequencing techniques may resolve origins, but not in absence of the knowledge of whole organisms in their paleoenvironments.