Orthologs versus paralogs
The concepts of orthology and paralogy, accompanied by precise
definitions (see below), were introduced by Walter Fich in a seminal
paper published as early as 1970 [1]. Over the next three decades or so,
the use and abuse (not infrequently misuse) of these terms gave rise to
interesting discussions about the fruitfulness of such concepts [2-4], a
debate in which Fitch himself took part [5].
Currently, in the post-genome era, with hundreds of genomes
sequenced, there is little doubt of the usefulness, I would even say
necessity, of these concepts. Thus, as stated by Eugene V Koonin, a
researcher at the Natioonal Center for Biotecnhology Information (NCBI),
“a clear distinction between orthologs and paralogs is critical for the
construction of a robust evolutionary classification of genes and
reliable functional annotation of newly sequenced genomes” [6].
Before adding another word, let’s remember the original definitions
of homology, orthology and parlogy in the context of gene/protein
evolution.
- Homology
-
Refers to genes (or proteins), that are descended from a gene (or
protein) present in a common ancestor.
- Orthology
-
Genes or proteins separated (that have diverged) by speciaciation events
are called orthologs.
- Paralogy
-
Genes of proteins separated (that have diverged) by duplication events
are called paralogs.
Much of the confusion arises from the misconception that two
paralogous genes must be present in the same genome (the same species),
when nothing in the definition imposes such a restriction. Such
confusion may have been favored by the fact that at the time during
evolution when paralogs originate by duplication, both copies will be
present, of course, in the same genome (Figure 1), but it does not imply
that these paralogs will always be restricted to the same genome
(organism) as evolution progresses.
Figure 1. Duplications events are marked by
orange squares while speciation events are indicated by orange
circles.
Along time, gene families expand (gene duplication) and contract
(gene loss), so we can end up finding paralogs in different species.
Sometimes these paralogs are easy to recognize, for instance, multiple
homologs in the same genome will always be paralogs. However, many other
times paralogs are easy to confuse with orthologs (Figure 2).
Figure 2. Certain sequences of events
(duplications: orange squares, speciations: orange circles, and losses:
orange crosses) can lead to confusing paralogs as ortholgs.
Before we move forward and jump into the specifics of the GS
evolutionary history in gymnosperms, one last general cautionary note.
Another misconception that we often found in the literature, is related
to the use of the term ‘ortholog’ for proteins present in different
species but fulfiling the same function, while in these msleading
context, the term ‘paralogs’ is reserved for proteins found in the same
species but performing different functions. However, as pointed by
Jensen [4]: “Although plenty of examples exist for which this
evolutionary scenario has indeed played out, it is quite possible for
orthologs to acquire different catalytic (or regulatory) properties and
for paralogs to retain the same function”. In fact, recent data suggests
that this last possibility occurs very frequently [7].
Glutamine synthetase from conifers
The enzyme glutamine synthetase (GS, EC 6.3.1.2) catalyzes the
incorporation of ammonium to glutamate to form glutamine in an
ATP-dependent fashion. This enzyme plays a crucial role in plant
nitrogen assimilation, being responsible for up to 95 % of the ammonium
assimilated in plants.
Conifers present two well differentiated families of cytosolic GS
isoforms: GS1a, mainly expressed in photosynthetic tissues, and GS1b
being relevant in nonphotosynthetic tissues. These GS1 lineages have
been detected in all gymnosperms as well as in basal groups of
angiosperms [8]. In general, GS1a is encoded by a single gene (locus).
In contrast, it is not uncommon to find different loci contributing to
the variability of GS1b, both in gymnosperms and angiosperms, although
in a much more pronounced way in the latter.
To illustrate the points we aim to develop and exposed in the current
vignette, we have selected 12 conifer species whose phylogenetic
relationships are well established (Figure 3).
Figure 3. Species tree of 12 representative
conifer species, covering the orders Cupresales (sky blue color),
Aracariales (lemon color) and Pinales (green color).
Together, these species contribute 29 GS isoforms: 12 GS1a proteins
(one per species) and 17 GS1b proteins (Figure 4).
Next, we point a few instructions to plot both the species and
protein family trees using the data included into the orthG package.
# Plot species tree:
str <- ape::read.tree(text = "((((Pa,Psm),(Pp,Pin)),(Abi,Ap)),((Ara,(Pod,Nag)),(Sci,(Tba,Tax))));")
plot(str)
# Load GS sequence data:
agf <- agf
# Aligning sequences and building an unrooted tree
# (remember you need the MUSCLE software in your path):
conif <- agf[which(agf$short %in% str$tip.label), ]
ptr <- mltree(msa(sequences = conif$prot,
ids = conif$phylo_id,
inhouse = TRUE)$ali)$tree
# Rooting and plotting an ultrametric tree :
ptr <- madRoot(ptr)
plot(ptr, use.edge.length = FALSE, cex = 0.6)
Figure 4. Protein tree of the GS family from the
conifer species shown in the previous figure. The phylogenetic
relationship in conflict with the species tree has been emphasized by
pink rectangles.
As it is evident from previous figures, there are some conflicts
between the well established species tree (Figure 3) and the GS protein
tree (Figure 4) we have build up. For instance, the genus Picea is
sister to the genus Pinus, however GS1a from Picea species are more
closely related to the GS1a from Abies than Pinus, which obviously
demand reconciliation!
When we look at the genomes of these 12 conifer species, we find that
GS1a is a single-copy gene. That is, that each species has one and only
one GS1a gene. We could naively thing that GS1a is a good gene to infer
the phylogenetic relationships of these species. However, we have seen
that this is not the case. The reason is that we have assumed that the
12 GS1a proteins we have been analyzing are orthologs, when in reality,
as we shall see below, we have a mixture of orthologs and paralogs.
Phylogenetic reconciliation
In phylogenetics, reconciliation is the term used to encompass a wide
and varied range of inference techniques that aim to find the historical
events (gene duplication, transfer, loss, etc) that best explain the
inconsistencies between gene tree and species tree. Although many
approaches have been developed in the last few years (see [9] for a
review), the one implemented in the orthGS package is
the method described in [10], which is based on the parsimony principle:
the best gene/protein family history is the history with the fewest
events.
We will focus in the six species belonging to the order Pinales
(green rectangle from Figure 3), which phylogenetic relationships seem
to be in conflict with that suggested in the protein tree (Figure 4, top
pink rectangle). So, let’s start by collecting the relevant data
regarding the GS isoforms present in this set of species:
data <- subsetGS(sp = c("Ap", "Abi", "Pin", "Pp", "Psm", "Pa"))[, 2:9]
data$phylo_id
#> [1] "Pp_GS1a" "Pa_GS1a" "Ap_GS1a" "Abi_GS1a" "Psm_GS1a"
#> [6] "Pin_GS1a" "Pp_GS1b_1" "Pp_GS1b_2" "Pa_GS1b_1" "Pa_GS1b_2"
#> [11] "Ap_GS1b" "Abi_GS1b_1" "Psm_GS1b_1" "Psm_GS1b_2" "Pin_GS1b_1"
#> [16] "Pin_GS1b_2"
We can check that there are 16 forms of GS: 6 GS1a proteins (one per
species) and 10 GS1b isoforms. Next, we are going to plot the orthology
network for this set of proteins. For this purpose we will use the
function orthG that will take as argument the set of
species that interest us:
o <- orthG(set = c("Ap", "Abi", "Pin", "Pp", "Psm", "Pa"))
A <- o[[1]] # Adjacency matrix
g <- o[[2]] # igraph object (orthology network)
It should be noted that the function orthG return a list
with two objects. The first one is an adjacency matrix. In our case a
square matrix of order 16 whose entries are 1 for orthology or 0 for
paralogy. The second element of the returned list is an igraph
network.
If desired, we can plot a subnetwork indicating the nodes of
interest. For instance, let’s focus our interest on GS1a:
gs1a <- data$phylo[grepl("GS1a", data$phylo_id)] # selected nodes
plot(igraph::subgraph(g, vids = gs1a))
We observe that GS1a isoforms from the genus Pinus are paralogs of
the GS1 from the genus Picea and Abies. On the other hand, all the GS1a
from Picea and Abies are orthologs among them. In this way, the
phylogenetic relationships inferred for the GS1a proteins from Pinales
(Figure 4), are be reconciled with the phylogenetic tree of the species
(Figure 3). Furthermore, the sequence of evolutionary events leading to
such a reconcilation can be inferred and plotted (Figure 5)
Figure 5. Species and protein trees
reconciliation. The focus has been placed on the order Pinales. The
duplicantions (squares) and losses (crosses) has been emphisized with
colored circles (turquoise for GS1a and brawn for GS1b).
GS1a orthology network beyond conifers
As already mentioned, GS1a is an isoform described to be present not
only in all gymnosperms, including Aracaurales, Cupressales, Pinales,
Gnetales, Ginkgoales and Cycadales, but also in basal angiosperms such
as species from the orders Amborellales and Magnoliales.
The data included in the orthGS package cover 34
GS1a proteins from as many species.
data <- sdf
gs1a <- data[which(data$gs == "GS1a" & data$tax_group != "Ferns"), ]
This parity between the number of proteins and the number of species
is rather misleading, since it might induce us to think that all these
proteins are orthologs and that they have diverged exclusively due to
speciation processes. However, if we plot the orthology network inferred
by tree reconciliation, the picture is quite different: gene duplication
within the GS1a lineage is frequent, but unlike what happens with the
GS1b lineage, in the case of GS1a duplications seem to be balanced by
losses.
o <- orthG()
g <- o[[2]]
plot(igraph::subgraph(g, vids = gs1a$Sec.Name_))
Figure 6. GS1a orthology network. Note that the
GS1a forms from basal angiosperms are placed in the center and conected
with gymnosperms GS1a, which are distributed around the periphery of the
graph.
It would be interesting to investigate in the future whether this
balance between duplications and losses responds to selective forces or,
on the contrary, can be explained based on the stochastic nature of
these processes.