Surprisingly, the complexity of gene products is not a barrier to transfers of the genes encoding them

Uri Gophna

Uri Gophna is a Senior Lecturer at Tel- Aviv University, Tel Aviv, Israel.

Summary
● Despite lateral gene transfers, bacterial genera remain distinct, suggesting limits to gene flow between species.
● According to the complexity hypothesis, only those genes encoding relatively simple proteins transfer readily between species.
● Genes that encode monomeric proteins have values for phylogenetic discordance that are similar to those for genes that encode subunits of multimers, providing evidence that gene product complexity is not a barrier to gene transfer.
● Transfers of complex protein-encoding genes, many of which are located on operons and gene clusters, could be very common.
● Loss of interacting residues by proteins encoded by laterally transferred genes is also likely to be a common occurrence.


The capacity of microorganisms for exchanging genetic material horizontally with distantly related species endows them with a source of nearly endless genetic variety. From the clinical perspective, lateral gene transfers (LGT) can generate new pathogens, including the etiologic agents of anthrax and the black plague, and also can spread antibiotic resistance. From an evolutionary standpoint, some researchers consider microbial LGT to be rampant, and there is increasing evidence that genes can be transmitted across biological domains, including from Bacteria to Archaea and from Eukarya to Bacteria.

Despite this apparent genetic promiscuousness, however, different genera of cyanobacteria or α-proteobacteria consistently form clusters in phylogenetic trees, no matter what phylogenetic reconstruction method is used. This distinctiveness strongly suggests that there are factors that limit gene flow and help to maintain a strong vertical species signal, at least in some lineages. The factors that obstruct foreign genes from integrating into the genome of a new host are called barriers to LGT.

Gene Barriers Appear To Vary By Family and Function

Barriers to transfer likely differ across gene families and functional categories. For example, informational genes-those whose products are involved in transcribing and translating genetic information from DNA into messenger RNA (mRNA) and proteins, respectively- are rarely transferred between species. This rarity contrasts to transfers of metabolic genes, such as those encoding biosynthetic enzymes, according to Ravi Jain, Maria Rivera, and James Lake at the University of California, Los Angeles (UCLA), whose report on this subject was published more than a decade ago. Many hundreds of genomes later, this observation still holds.

The relative rarity of transfers involving informational genes, especially those encoding ribosomal proteins or other parts of the translation apparatus, provides a major basis for assigning individual microorganisms to particular phylogenies, including by means of the popular interactive tree of life (iTOL). But why are these informational genes transferred so seldom? What is the barrier to their transfer?

Lake and his collaborators at UCLA suggested that the proteins that are involved in transcription or translation belong to large complexes and thus depend on multiple interactions with other partners to function. However, simpler proteins encoded by operational genes generally require fewer encoded partners, if any, to function. A protein encoded by a recently transferred gene is not likely to interact properly with a host partner. More generally, the more connections a laterally acquired protein must make with other proteins, the smaller its chances of making all those new contacts.

Of course, if the transferred protein is not functional, acquisition of the gene that encodes it is likely to be neutral for the new host. In the absence of positive selection, mutation will likely lower the chances for neutral transferred genes to be retained.

The Complexity Hypothesis Is Powerful

The complexity hypothesis-which claims that genes encoding relatively simple proteins transfer readily between species, whereas genes encoding more complex proteins, including those that depend on interactions with one or several other proteins, do not-is a powerful one. Moreover, it could help to explain a well-known exception to the rarity of transfer of informational genes.

Specifically, aminoacyl-tRNA synthetases, which are part of the "informational genes" set, transferred frequently between organisms during the course of evolution, and some of those transfers occurred between species in different domains. When loading amino acids onto tRNA molecules, however, these enzymes do not interact with the translation apparatus in a host species. Thus, they serve as modular components of protein synthesis, perhaps accounting for why they are relatively easily transferred between species, as several researchers, including Carl Woese of the University of Illinois, Urbana, and W. Ford Doolittle at Dalhousie University in Halifax, Nova Scotia, Canada, point out.

Until recently, however, no one put the complexity hypothesis to experimental test. For example, is it true that the number of interactions a gene product makes will affect the transferability of its cognate gene, regardless of function? Do the type and stability of the interactions matter? For example, do the same evolutionary rules apply for interactions between subunits belonging to a stable protein complex and for those between other, more transiently connected, proteins?

There is an alternative explanation for the rarity of transfers of genes encoding complex protein subunits. It posits that acquiring certain subunits is deleterious to cells, rather than being neutral from not benefiting the recipient. Such a damaging effect of a foreign defective allele on its native wild-type homolog will be the lateralgene- transfer equivalent of a dominant-negative phenotype. This alternative hypothesis thus might explain why inactive or disrupted forms of informational genes are so seldom seen in microbial genomes.

In some cases, increased gene dosage via duplication could lead to higher levels in a cell of a complex subunit and thus, in itself, may be deleterious, as suggested by Bala´ zs Papp and Csaba Pal while working with Laurence Hurst at the University of Bath, Bath, United Kingdom. They called this phenomenon the "balance hypothesis." Indeed, consistent with their argument, some foreign bacterial genes, including those encoding several ribosomal proteins, cannot be over-expressed in Escherichia coli or duplicated in other bacterial genomes, according to Rotem Sorek and coworkers at the Department of Energy Joint Genome Institute in Walnut Creek, Calif.

Finding Ways To Put These Hypotheses to Experimental Test

We began to explore systematically the link between protein-protein interactions (PPIs) and LGT, with the goal of uncovering those forces that govern laterally acquired genes. For this purpose we studied cells of E. coli, drawing on protein-protein interaction data in the Swiss- Prot database. Its contents are manually curated, and its annotations distinguish between merely transient interactions and those between and among subunits within stable complexes.

Because we are relying solely on E. coli PPI data, we cannot draw general conclusions regarding propensities for LGT among other species of microorganisms. For instance, if E. coli inherited a particular gene, that fact says nothing about whether that gene was transferred among other taxa. To overcome this hurdle, we estimated phylogenetic discordance, using the phylogenetic discordant sequences (PDS) estimator (see box).

PDS measures to what extent the phylogenetic signal of a coding gene matches the bulk signal of other proteins in a particular genome. It does so by looking at the ranks of pairwise similarity values across the entire dataset, spanning hundreds of genomes. A gene family that was frequently transferred will show greater phylogenetic discordance, with a PDS value close to 0. Even if that gene family is ancestral in a particular genome, its pattern across many genomes will be discordant.

If either the complexity or the balance hypothesis is correct, genes that encode complex subunits should be transferred rarely and, thus, would have higher PDS scores. However, genes that encode monomeric proteins, which we expect to have undergone more frequent LGT, have a slightly higher PDS than do those that encode subunits of multimers, according to our analysis. This surprising result suggests that the complexity of gene products is not a barrier to transfers of those genes.

Effects of Particular Gene Transfers into E. coli from other Microorganisms

Puzzled with those results from our bioinformatic analysis, we next examined transfers of genes encoding complex protein subunits in vivo, focusing on the E. coli accA gene. This essential gene encodes the alpha subunit of carboxyl- transferase, a part of the larger acetyl- CoA carboxylase (ACC) complex, which is located in the cytoplasm and catalyzes the first and rate-limiting step of fatty-acid synthesis. It is conserved from bacteria to animals.

We began by generating four E. coli strains, each carrying a different plasmid-encoded accA homolog under the control of an inducible promoter. Because each recipient strain has its own intrinsic accA gene, our gene transfers belong to a type of LGT in which an additional exogenous ortholog, or xenolog, is introduced into the recipient organism. One of the accA variants contained the native E. coli gene, as a control for the effect of increased gene dosage, while the other three strains received foreign homologs from different bacterial species, each of which encode proteins that are about 50% identical to the E. coli AccA.

When we induced the genes for the exogenous subunits, three of the four strains experienced equivalent growth rates for 30 generations. The fourth strain, which contained the gene encoding the AccA subunit from Bacillus subtilis experienced only a minor impairment in its growth. Notably, we could delete the chromosomal accA gene only from the strain carrying the plasmid-encoded copy of the E. coli genes, indicating that none of the foreign subunits was biologically active. Furthermore, unlike the native subunit, foreign subunits do not form stable interactions with the host carboxyl-transferase β subunit, according to our coimmunoprecipitation experiments.

Thus, these experimental findings support the idea that, without correct interactions, LGT is effectively neutral, which is consistent with the complexity hypothesis. Recall that the balance hypothesis postulates that incorrect interactions stemming from LGT would be deleterious to recipient organisms.

Effects on E. coli of Transferred Genes Encoding Informational Proteins

The question remains whether LGT effects will be neutral when the transferred genes encode proteins with a high degree of similarity with native protein subunits, which is typical for several kinds of informational proteins. To evaluate this issue, we targeted genes encoding proteins within the RNA polymerase (RNAP) complex, again using a plasmid vector but in this case cloning the genes encoding RNA polymerase β subunit, namely rpoB, from B. subtilis and from E. coli.

The B. subtilis RpoB is nearly 60% identical and 80% similar to its E. coli homolog. Unlike what occurs in the accA coimmunoprecipitation experiment, the B. subtilis RpoB subunit binds both host interaction partners, RpoC and RpoA, as well as several proteins that do not associate with RNAP. Despite these interactions, however, the presence of rpoB from B. subtilis and the protein that it encodes leads to no growth disadvantage for that strain.

This example further supports the notion that LGT neutrality can be maintained even in the presence of multiple interactions, including novel ones, from the proteins encoded by those transferred genes. This observation also reinforces the view that barriers to LGT are not very large. Moreover, if novel interactions are neutral at first, they may be selected for when the environment changes, promoting genetic novelty. Furthermore, carrying a large excess of one native subunit protein of a complex does not appear to have any deleterious effect on the cell making that protein, confirming our other observations that, for E. coli at least, the balance hypothesis is not supported.

Evaluating Transient Interactions

We next tested what we consider to be external interactions, those that are more transient such as those involving proteins that are not part of a stable complex. Strikingly, genes with more than one recorded interaction in Swiss-Prot have a significantly higher mean PDS score than do those that mediate only a single interaction. Therefore, the number of interactions between one protein with others is likely a better predictor of its gene transfer capability than is the number of contacts it makes within a multimer.

Thus, we hypothesize that transfers of complex protein-encoding genes, many of which are located on operons and gene clusters, are very common. Consider, for example, the genes encoding the Type III secretion system (T3SS), which is a large and highly complex molecular structure, involving at least 20 genes for its assembly. The genes encoding T3SS apparently were transferred multiple times in bacterial history. The structural T3SS genes are typically clustered in one location, either on plasmids or in pathogenicity islands. More broadly, 95% of hetero-oligomeric protein complexes in E. coli are encoded by operons or gene clusters, facilitating their cotransfer. Conversely, genes encoding transiently interacting protein partners tend to be encoded in trans, making their cotransfer far less likely.

Decay of Interaction Regions in Transferred Genes

Are genes that must interact with multiple partners seldom retained, as the complexity hypothesis proposes? Perhaps transferred genes are retained, even if they have fewer connections with proteins in their new, compared to their original, host cell. We find that recently transferred genes encode proteins with fewer interactions than do proteins encoded by native genes. However, did they also have few interactions in their native context?

Because the organism from which laterally transferred genes come rarely is known and might be extinct, one can address the question of gene origins only indirectly. One idea is to look at the interaction potential of a protein, rather than directly at its interactions.

In collaboration with Yanay Ofran of Bar Ilan University in Ramat Gan, Israel, we are now examining protein interaction potentials by using ISIS, an algorithm developed by him with Burkhard Rost at Columbia University in New York, N.Y. ISIS that measures the interaction potential of any particular protein by identifying highly interacting residues from its sequence. A recently acquired gene presumably has an intact interaction potential, assuming it takes time for formerly interacting residues to disappear via mutations. If recipient cells seldom retain genes encoding highly interacting proteins, recently transferred genes should encode proteins with relatively low interaction potentials. However, we observed the opposite. Thus, on average, recently transferred genes encoded proteins with higher interaction potentials than those encoded by native genes, whereas ancient transfers had intermediate potential. Apparently, proteins that had more interactions, and consequently have more interacting residues, are quicker to integrate into new protein-interaction networks. However, some interactions decay because so many original interactions no longer form, and are likely to outnumber the new contacts.

To test this hypothesis, we looked at 10 genes acquired by E. coli subsequent to its split from Salmonella, focusing on genes that rarely experience LGT (PDS < 0.95). When looking at their homologs in other genomes, where they presumably were vertically inherited, 9 of the 10 cases had higher average interaction potentials among those other bacteria than did their laterally acquired counterparts in E. coli. This finding suggests that losses of interacting residues from proteins encoded by laterally transferred genes are common.

Concluding Thoughts

The evidence for cotransfers of genes encoding proteins that are part of functional modules is consistent with the great genomic plasticity that is found among bacteria. Further, there is an inherent flexibility among proteins themselves that enables them to form new contacts with other proteins while older contacts are lost. The combination of genomic and experimental studies has made us realize that the barriers to gene transfers are low and leaky.


SUGGESTED READING

Clarke, G. D., R. G. Beiko, M. A. Ragan, and R. L. Charlebois. 2002. Inferring genome trees by using a filter to eliminate phylogenetically discordant sequences and a distance matrix based on mean normalized BLASTP scores. J. Bacteriol. 184:2072-2080.

Doolittle, W. F., Y. Boucher, C. L. Nesbo, C. J. Douady, J. O. Andersson, and A. J. Roger. 2003. How big is the iceberg of which organellar genes in nuclear genomes are but the tip? Phil. Trans. R Soc. Lond. B Biol. Sci. 358:39-57 (discussion 57-38).

Gophna, U., Ron, E.Z., and D. Graur. 2003 Bacterial type III secretion systems are ancient and evolved by multiple horizontal transfer events. Gene 312:151-163.

Jain, R., M. C. Rivera, and J. A. Lake. 1999. Horizontal gene transfer among genomes: the complexity hypothesis. Proc. Natl. Acad. Sci. USA 96:3801-3806.

Ofran, Y., and B. Rost. 2007. ISIS: interaction sites identified from sequence. Bioinformatics 23:e13-16.

Papp, B., C. Pal, and L. D. Hurst. 2003. Dosage sensitivity and the evolution of gene families in yeast. Nature 424:194-197.

Sorek, R., Y. Zhu, C. J. Creevey, M. P. Francino, P. Bork, and E. M. Rubin. 2007. Genome-wide experimental determination of barriers to horizontal gene transfer. Science 318:1449-1452.

Wellner, A., M. N. Lurie, and U. Gophna. 2007. Complexity, connectivity, and duplicability as barriers to lateral gene transfer. Genome Biol. 8:R156.

Wellner, A., and U. Gophna. 2008. Neutrality of foreign subunits in an experimental model of lateral gene transfer. Mol. Biol. Evol. 25:1835-1840.

Woese, C. R., G. J. Olsen, M. Ibba, and D. Söll. 2000. Aminoacyl-tRNA synthetases, the genetic code, and the evolutionary process. Microbiol. Mol. Biol. Rev. 64:202-236.