Horizontal gene transfer, evolutionary networks and uncertainty

State which copies and events a gene tree represents

A species tree describes divergence among the represented species lineages. A gene tree follows the ancestry of sampled copies at a locus or region. Within that history, gene-copy lineages can join backwards in time at a shared ancestral copy, a coalescence. Depending on the dataset and history, gene trees can also include duplications and transfers. Not every gene-tree node denotes a species split.

Homologous copies share ancestry, but the event separating them matters. Orthologous relationships refer to separation by speciation; paralogous relationships refer to gene duplication. Paralogs can occur within one species or across different species. These terms describe history, not a guarantee of identical function.

Our two-species model begins with one gene copy that duplicates into lineages x and y before species A and B diverge. Without further duplication or loss, A and B each retain x and y. A-x and B-x are orthologs in the model, as are A-y and B-y. Comparing A-x with B-y instead crosses the older duplication event.

A later loss of A-y would leave three observed copies, not undo the duplication. A study that sampled only A-x and B-y could mistakenly treat its older gene divergence as the A/B species divergence. Copy identification, gene arrangements and wider sampling help test that interpretation.

Gene ancestry can join before species divergence

Tracing two sampled copies backwards through a population does not require them to meet immediately at a species divergence. Their ancestral copies can remain distinct in an ancestral population. Incomplete lineage sorting describes such failures to coalesce within the relevant ancestral interval; it can lead to a gene topology that differs from the species topology.

In the constructed species history, A and B share an ancestral population from model time 2 to 5 before joining C’s ancestry at time 5. Times increase backwards from the present. The sampled A and B gene lineages remain distinct through that interval. In the older common population, A’s sampled lineage joins C’s at time 6, and B’s joins them at time 7.

The resulting gene tree groups A’s and C’s copies first, while the species tree groups A and B. No gene moved across species boundaries in this example. The different history comes from how ancestral copies coalesced within populations, not convergence in the final sequences or a species relabelling.

The paths are assigned to make the possibility visible; they do not estimate its probability, ancestral population size or duration in nature. Not every incomplete-sorting history creates a discordant topology, and discordance alone does not diagnose sorting. Mutation and limited data also affect how reliably a gene history can be inferred.

A lateral connection changes a region’s ancestry

Horizontal gene transfer moves genetic material between lineages outside an ordinary parent–offspring relationship. A transferred region can subsequently be inherited by the recipient’s descendants. The recipient genome then includes regions with different histories. Transfers need not replace the recipient’s entire genome or affect every gene in the same way.

Hybridisation and introgression provide other routes to ancestry contributions between species lineages. Recombination can make different parts of one region have different histories. A network can represent such connections when a single branching tree would omit a contribution. The type of edge must be named: descent, gene transfer and endosymbiotic integration are not interchangeable.

Our transfer sketch begins with sister species A/B and a separate C lineage. A designated gene region moves from a contemporaneous C lineage to B at time 1 before the present and replaces B’s model copy. B inherits that transferred region thereafter. The rest of the drawn species history remains A/B sisterhood.

A gene tree for that region can consequently associate B’s sampled copy with C’s rather than A’s. But a similar disagreement could arise from other causes, including duplication/loss, sorting, alignment error or poorly inferred trees. Compare alternative explanations, inspect sequence context and independent loci, and assess support before assigning a transfer.

Keep a supported backbone and test conflicting histories

A network does not mean that all evolutionary relationships are unknowable, and a tree does not mean every gene followed one clean branching path. For a specified set of lineages, a branching backbone may summarise much of the history while particular regions require additional connections.

Combining genes without examining their histories can conceal conflict behind a single strongly supported result. Methods designed for species-tree or network inference explicitly model some sources of discordance, but every such model has assumptions. Wider sampling, copy checks and sensitivity analyses help assess which conclusions survive.

Uncertainty can be stated constructively: this split is supported under these analyses; this region favours a conflicting gene history; transfer and sorting remain alternatives. That is more useful than presenting either a perfectly resolved universal diagram or an unexplained collection of contradictory trees.

Worked example

Check your understanding

1. In the duplication model, are A-x and B-y orthologs merely because they are in different species?

   Show reasoning

   No. Their separating event is the earlier x/y duplication, so they are paralogs under the supplied history. A-x/B-x and A-y/B-y trace separation by the A/B speciation event. Different species alone does not define orthology.

2. Why does the A/C gene pairing at time 6 not require horizontal transfer in the sorting example?

   Show reasoning

   At time 6, all three sampled lineages are inside the older shared ancestral population, entered at time 5. A and B failed to coalesce during their exclusive ancestral interval from 2 to 5. A/C can therefore join first without crossing a species boundary.

3. One region groups B with C, while other evidence groups A with B. What should be checked before declaring transfer?

   Show reasoning

   Check copy identity and loss/duplication possibilities, alignment and sequence quality, support and sampling, sorting or introgression models and the region’s genomic context. A discordant gene tree is evidence to investigate, not a unique signature of one mechanism.

What to carry forward

Name the units and events of each history: gene-copy ancestry, species divergence and lateral contributions can differ without contradicting descent.