Clade and Cladogram

Originally a tool of biological systematics, the cladogram is today also used in breeding genetics. It shows relationships of descent as a branching tree: every branching point (node) stands for the last common ancestor of the lineages emerging from it; every branch, with everything attached to it, is a clade.

The logic applies regardless of the level considered: clades can be formed, for example, for species, for breeds, for breeding lines and for individual DNA variants (haplotypes). The basis of a cladogram is DNA information: SNP arrays, mitochondrial DNA sequences (maternal lines), Y chromosome markers (paternal lines).

Please note: A cladogram shows the structure of the branching; it gives no information about the timing of the branching events, the similarity of the representatives shown or the occurrence of particular traits.

The clade

The term (Greek klados = branch) goes back to phylogenetic systematics (cladistics), founded by Willi Hennig from 1950 onwards. A clade in this sense is a monophyletic group: it comprises the "founding ancestor" and every one of its descendants.

What counts, therefore, is common descent. A clade becomes recognisable through traits that arose anew in the common ancestor and were passed on by it to its descendants, for example a particular mutation. The trait is the evidence, not the definition: on average, a mutation is passed on to only half of the offspring and can be lost again over the generations. Descendants that do not carry it still belong to the clade.

Within a population whose animals mate freely with each other, recombination reshuffles the genome in every generation. Individual animals can therefore only be arranged cleanly into clades for genetic material that does not recombine (mitochondrial DNA, Y chromosome) or for short genome segments (haplotypes). For species, and for breeds that have developed separately for a long time, the clade model is, by contrast, a good approximation.

The following groupings are not considered clades:

Clades are nested within each other: every clade contains smaller clades and is itself part of a larger one. Whether a group is a clade therefore depends on which section of the tree is being considered.

Traits that indicate a clade

Not all shared traits are of equal value as evidence of a clade:

• Synapomorphy: a shared derived trait that arose anew in the last common ancestor. Only a synapomorphy indicates a clade. In molecular terms it corresponds to a shared mutation at a defined position.
• Symplesiomorphy: a shared ancestral trait that was already present before. It says nothing about the branching within the group.
• Homoplasy: the same trait without a shared origin, arising through convergence, back mutation or repeated mutation at the same position. It is the most important source of error in building trees.

The cladogram

Components of a cladogram:

• Terminal nodes (leaves): the units examined, i.e. species, breeds, individuals or haplotypes.
• Internal nodes: hypothetical last common ancestors. They are a reconstruction, not a sampled animal.
• Root: gives the tree a direction and thus the statement of what is ancestral and what is derived. Rooting is done via an outgroup that certainly lies outside the group examined, e.g. the wolf for dog breeds, Przewalski's horse for horse breeds.

Three points are often misunderstood when reading a cladogram:

1. Only the topology counts. In a true cladogram, branch lengths carry no information.
2. Nodes can be rotated freely. Two breeds drawn next to each other need not be more closely related than breeds drawn far apart; what matters is only the node at which they meet.
3. The tree is a model. It represents a hypothesis about the course of descent, not a measurement.

Distinction from related representations

The last row is particularly important in breeding practice: many breed overviews are dendrograms from a cluster analysis of genetic distances. They look like family trees, but at first they only say "similar", not "related".

How a cladogram is built

1. Build the character matrix: today molecular, i.e. SNP genotypes from arrays or sequencing, mitochondrial sequences, Y chromosome haplotypes, microsatellites.
2. Calculate the tree: distance based (neighbour joining, UPGMA) or character based (maximum parsimony, maximum likelihood, Bayesian inference).
3. Assess the nodes: using bootstrap values (Felsenstein 1985) or posterior probabilities. They indicate how stable a branching is when the data are repeatedly resampled at random. Values from about 70 % are usually regarded as well supported, values from 95 % as robust. A node with a low value is an open question, not a statement.
4. Root the tree: via the outgroup.

Limit of the tree model: A tree assumes that lineages split and then remain separate. For domestic animal breeds this applies only to a limited extent, because crossbreeding continuously moves genes between branches. Such networks are therefore additionally represented as median-joining networks or admixture graphs.

Breed analysis and breed assignment

The most comprehensive analysis of this kind to date (Parker et al. 2017) examined 1,346 dogs from 161 breeds and assigned them to 23 clades. The clades group breeds that go back to shared regional origins and purposes of use, such as terriers, spitz-type dogs or herding dogs.

For practical breed analysis this has a direct consequence: breeds within one clade share large parts of their genome and are statistically harder to separate than breeds from different clades. This explains why a breed test distributes proportions across several breeds of the same clade for closely related breeds, while distantly related breeds are distinguished reliably.

Our DNA tests for breed analysis

• Dog:Breed Composition Analysis Dog
• Dog, bundle:ProLine Bundle Dog: DNA identity, inbreeding and diversity analysis and breed analysis in one order
• Horse:Breed Analysis incl. Identity Horse
• Horse, bundle:Bundle Breed Analysis, Identity, Inbreeding Assessment: breed analysis, inbreeding assessment and genetic identity in one order

Maternal and paternal lines as haplogroups

Mitochondrial DNA (inherited purely maternally) and the Y chromosome (purely paternally) do not recombine. Their variants therefore form clean clades, which are called haplogroups.

In the horse, 18 haplogroups (A–R) were derived from 83 complete mitogenomes (Achilli et al. 2012). The finding shows that a large number of mare lines entered domestication. In breeding, the assignment serves to

• assign an animal to a mare family or sire line,
• verify studbook records on the maternal or paternal line,
• preserve rare lines in a targeted way.

Origin and spread of individual hereditary variants

A particular mutation usually arises only once, on a particular chromosome segment. All carriers of this mutation then form a clade with respect to this segment, recognisable by the shared surrounding haplotype. Exceptions are mutations that arise repeatedly at the same position (homoplasy, see above); they show up in carriers having different surrounding haplotypes. This allows

• proof that a disease variant has a single origin and has been passed on across breed boundaries,
• an estimate of its age and distribution,
• the recognition that the same clinical picture in another breed rests on a different mutation: the carriers then do not belong to the same clade. Practical consequence: a genetic test validated for breed A is in this case not informative for breed B.

Conservation breeding and genetic diversity

The branching pattern of a population shows which lines separated early and therefore carry the greatest differences. Matings across clade boundaries raise the expected heterozygosity of the offspring and lower their genomic inbreeding coefficient. In small populations this leads to the recommendation to preserve the lines that form branches of their own, not those that are numerically most frequent.

Detecting crossbreeding

Animals whose position in the tree does not match their documented descent can indicate undocumented crossbreeding or incorrect pedigree records. Low bootstrap values alone are no proof of this, as they have many possible causes, for example too few informative markers. Crossbreeding is detected more specifically with methods such as admixture analysis.

Limits of interpretation

• A cladogram says nothing about the performance, breeding value or health of an animal.
• Proximity in the tree means common descent, not similarity in appearance.
• The result depends on the choice of markers and the reference sample: a breed missing from the reference cannot be placed correctly.

Lexikon:

• Genomic Inbreeding Coefficient (GIC)

Literature:

• Parker, H.G., Dreger, D.L., Rimbault, M. et al. Genomic Analyses Reveal the Influence of Geographic Origin, Migration, and Hybridization on Modern Dog Breed Development. Cell Reports19 (4), 697–708 (2017). https://doi.org/10.1016/j.celrep.2017.03.079
• Achilli, A., Olivieri, A., Soares, P. et al. Mitochondrial genomes from modern horses reveal the major haplogroups that underwent domestication. PNAS109 (7), 2449–2454 (2012). https://doi.org/10.1073/pnas.1111637109
• Felsenstein, J. Confidence limits on phylogenies: an approach using the bootstrap. Evolution39 (4), 783–791 (1985). https://doi.org/10.1111/j.1558-5646.1985.tb00420.x
• Hennig, W. Phylogenetic Systematics. University of Illinois Press, Urbana (1966), foundation of cladistics.