Are There Bugs in Our Genome?

For evolutionary biologists working on the exchange of genes between species (lateral gene transfer), the most exciting news from the human genome sequencing project has been the claim by the "public effort" (1) that between 113 and 223 genes have been transferred from bacteria to humans (or to one of our vertebrate ancestors) over the course of evolution. We, and probably many others wanting to test whether this result is really solid (2), have been beaten to the punch by Salzberg and colleagues (3). Their analysis, appearing on page 1903 of this week's issue, suggests that the actual number of bacterial genes in our genome may be lower than the predicted 223. These authors argue that there are other biologically plausible explanations besides lateral gene transfer that could account for the presence of bacterial genes in our genome.

The claim for lateral gene transfer from bacteria into vertebrates (as exemplified by our own species' genome) was based on similarity searches. Such searches involve screening vertebrate genomes for sequences that are very similar to bacterial genes but are absent from other eukaryotic genomes. Genes shared by vertebrates and bacteria that are not found in other eukaryotes are considered to be probable bacteria-to-vertebrate transfers (BVTs). The 113 to 223 BVTs in question have significant similarity to bacterial sequences but no "comparable similarity" to genes in other completed eukaryotic genomes (1).

Salzberg et al. (3) now provide a careful reanalysis of these data (1) with a similar but more conservative approach that includes the addition of data from Celera's version of the human genome (4). As in the original study (1), the investigators' goal was to detect possible transfer of genes by analyzing gene distribution across taxa. They found 135 genes in the public effort's data set of 31,780 protein-encoding sequences (Ensembl proteome) and 89 genes in the Celera proteome of 26,544 proteins that were possible BVTs (3). This is similar to the public effort's conservative estimate of 113 possible bacterial genes in the human genome (1). Lateral gene transfer is not the only factor that could explain these results. For instance, differential gene loss (that is, random independent loss of genes in different eukaryotic lineages) may yield similar gene distribution patterns. The Salzberg et al. reanalysis demonstrates that the calculation of the number of bacterial genes in the human genome is highly dependent on how many nonvertebrate genomes were screened against the human genome. These authors found a downward trend in the number of BVTs observed when the human genome was screened against an increasing number of nonvertebrate genomes. Such a pattern is indeed

consistent with differential gene loss in eukaryotes, and it is reasonable to assume that the downward trend might continue as more nonvertebrate genomes become available for screening. Indeed, after removing possible BVTs that are more similar to genes in eukaryotic genomes for which only partial sequences are available, the number of possible BVTs drops to 114 and 68 for the Ensembl and Celera proteomes, respectively.

Another factor yielding false BVTs is differences in evolutionary rates among different vertebrate and nonvertebrate lineages. Salzberg et al. investigated the effect of differing evolutionary rates by relaxing the similarity criteria for inclusion of nonvertebrate homologs. Again, they found a reduction in the number of BVTs to 74 (Ensembl) and 56 (Celera). After additional trimming of the data--removing two mitochondrial genes, searching a further curated version of the Ensembl data, checking for annotation errors, and comparing the two data sets--the authors calculated the final number of possible BVTs to be 41 (Ensembl) or 46 (Celera).

So, the original description of 223 BVTs is probably overenthusiastic. But even 41 (or 46) BVTs is sufficient cause for excitement. The statistical arguments of the sort Salzberg et al. present can never eliminate the possibility that some of these BVT candidates (or others that they eliminated with their parsimonious broad-brush approach) really are true BVTs. In fact, although the Salzberg study criticizes conclusions about evolution based on similarity searches, this study, too, is based on a similarity search approach! The best way to determine whether some of the genes in the final list are real BVTs is to construct molecular phylogenetic trees for each of the possible instances. If a vertebrate gene sequence is nested within a robust cluster of bacterial sequences, the most probable explanation is that the vertebrate gene was laterally transferred from bacteria. Salzberg and colleagues mention that they have constructed phylogenetic trees for some genes (one is depicted in figure 2 of their paper), but they state that "most did not show patterns consistent with bacterial-to-vertebrate gene transfer." Yet they don't tell us how many "most" is, nor do they state which genes do show patterns consistent with BVTs.

We have prepared phylogenetic trees for seven of the genes listed in the supplementary information provided by Salzberg et al. Among these, we found one probable case of lateral gene transfer between bacteria and vertebrates: the gene encoding a putative N-acetylneuraminate lyase (see the figure). This gene was previously shown (very convincingly) to have been transferred from bacteria into the protozoan parasite Trichomonas vaginalis (5). Indeed, the gene found in Trichomonas is nested within a robust bacterial cluster, which would be expected for a gene that has undergone lateral transfer. The vertebrate version of this gene clusters unequivocally with Vibrio cholerae and Yersinia pestis genes, indicative of lateral gene transfer involving bacteria and eukaryotes that is independent of the bacteriato-Trichomonas transfer. In this case, however, it is not possible to infer the direction of the transfer because neither the bacterial genes nor the vertebrate genes are nested within the other.

Is lateral gene transfer between prokaryotes and eukaryotes an extremely rare event? Vertebrates are multicellular organisms, and so any evolutionarily stable incorporation of foreign DNA must take place in the germ cells that give rise to eggs and sperm. Unicellular eukaryotes, on the other hand, often live close to prokaryotes and frequently use them as food, which means that they have a much greater exposure to prokaryotic DNA than do vertebrate germ cells. Inevitably, this might lead to a gradual replacement of ancient eukaryotic genes with bacterial homologs (6). Indeed, pathogenic protozoa that live in environments rich in prokaryotes demonstrate that this does happen. T. vaginalis, a parasite of vertebrate epithelial cells, no doubt acquired its N-acetylneuraminate lyase gene from bacteria in its environment (see the figure) (5). The protozoan Entamoeba histolytica possesses fermentation enzymes that must have come from different anaerobic prokaryotes, and another protozoan, Giardia lamblia, encodes an enzyme in the mevalonate pathway that is unquestionably of bacterial origin (7, 8).

The fact that the ancestors of mitochondria and chloroplasts (DNA-containing cellular organelles) have contributed genes to the eukaryotic nucleus poses a serious problem for the detection of prokaryote-to-eukaryote lateral gene transfer. The endosymbiont bacteria that gave rise to mitochondria and chloroplasts have been a major source of bacterial genes in eukaryotic nuclear genomes, and their ancestral lineages are the alpha-proteobacteria and cyanobacteria, respectively. It seems sensible to infer that nuclear genes of similar ancestry are of endosymbiont origin, whereas those that cluster within other bacterial groups result from independent lateral transfers. However, independent transfers from alpha-proteobacteria and cyanobacteria subsequent to the original endosymbioses surely have also occurred, and these lineages themselves have been recipients of transferred genes (9). Reconstructing the history of prokaryote genes in the eukaryotic nucleus will be a formidable but exciting challenge (10).

To determine the extent to which lateral gene transfer is an important evolutionary force within eukaryotic evolution, we need to move beyond BLAST-based analysis to large-scale phylogenetic analysis. This is a realistic task: The methods are available, and several eukaryotic genome sequencing projects from a relatively broad range of eukaryotes are well under way (11, 12). We would not be surprised to see eukaryotes distributed in multiple places in a prokaryotic background. Such a finding could only be accounted for by multiple lateral gene transfer events (independent of genes donated by organelles). We expect, however, that the vast majority of these transfer events happened before the evolution of multicellularity. Our multicellularity probably saved us from participating in the dirty business of lateral gene transfer so beloved by microbes.

CHART: Gene swapping among friends and neighbors. Phylogeny of the gene encoding N-acetylneuraminate lyase (Ensembl ID IGI_M1_ctg1425_20). Phylogenetic relationships for this gene among prokaryotes, protozoans, and vertebrates were estimated using TREE-PUZZLE (13) from an amino acid alignment generated by CLUSTALX (14). Bullets on the nodes indicate that bootstrap support values--obtained using neighbor joining of PAM-based distances estimated in PHYLIP (15)--and puzzle support values were above 95%. Bacteria donated this gene to the protozoan parasite T. vaginalis. Vertebrates (human, mouse, and pig) together with two bacterial lineages (Vibrio and Yersinia) also show a branching pattern indicative of gene transfer, although it is not possible to infer the direction of the transfer.

1. International Human Genome Sequencing Consortium, Nature 409, 860 (2001).

• 2. C. P. Ponting, Trends Genet. 17, 235 (2001).
• 3. S. L. Salzberg, O. White, J. Peterson, J. A. Eisen, Science 292, 1903 (2001); published online 17 May 2001 (10.1126/science.1061036).
• 4. J. C. Venter et al., Science 291, 1304 (2001).
• 5. A. P. de Koning, F. S. Brinkman, S. J. Jones, P. J. Keeling, Mol. Biol. Evol. 17, 1769 (2000).
• 6. W. F. Doolittle, Trends Genet. 14, 307 (1998).
• 7. J. Field, B. Rosenthal, J. Samuelson, Mol. Microbiol. 38, 446 (2000).
• 8. Y. Boucher, W. F. Doolittle, Mol. Microbiol. 37, 703 (2000).
• 9. T. Rujan, W. Martin, Trends Genet. 17, 113 (2001).
• 10. C. G. Kurland, S. G. E. Andersson, Microbiol. Mol. Biol. Rev. 64, 786 (2000).
• 11. T. Sicheritz-Ponten, S. G. E. Andersson, Nucleic Acids Res. 29, 545 (2001).
• 12. TIGR Microbial Database (www.tigr.org/tdb/mdb/mdbinprogress.html).
• 13. K. Strimmer, A. von Haeseler, Mol. Biol. Evol. 13, 964 (1996).
• 14. J. D. Thompson, T. J. Gibson, F. Plewniak, F. Jeanmougin, D. G. Higgins, Nucleic Acids Res. 25, 4876 (1997).
• 15. J. Feisenstein, Cladistics 5, 166 (1989).

Published online 17 May 2001; 10.1126/science.1062241 Include this information when citing this paper.