The Public Goods Hypothesis for the evolution of life on Earth.

HYPOTHESIS Open Access
The public goods hypothesis for the evolution of
life on Earth
James O McInerney1*, Davide Pisani1, Eric Bapteste2 and Mary J O’Connell3
Abstract
It is becoming increasingly difficult to reconcile the observed extent of horizontal gene transfers with the central
metaphor of a great tree uniting all evolving entities on the planet. In this manuscript we describe the Public
Goods Hypothesis and show that it is appropriate in order to describe biological evolution on the planet.
According to this hypothesis, nucleotide sequences (genes, promoters, exons, etc.) are simply seen as goods,
passed from organism to organism through both vertical and horizontal transfer. Public goods sequences are
defined by having the properties of being largely non-excludable (no organism can be effectively prevented from
accessing these sequences) and non-rival (while such a sequence is being used by one organism it is also available
for use by another organism). The universal nature of genetic systems ensures that such non-excludable sequences
exist and non-excludability explains why we see a myriad of genes in different combinations in sequenced
genomes. There are three features of the public goods hypothesis. Firstly, segments of DNA are seen as public
goods, available for all organisms to integrate into their genomes. Secondly, we expect the evolution of
mechanisms for DNA sharing and of defense mechanisms against DNA intrusion in genomes. Thirdly, we expect
that we do not see a global tree-like pattern. Instead, we expect local tree-like patterns to emerge from the
combination of a commonage of genes and vertical inheritance of genomes by cell division. Indeed, while genes
are theoretically public goods, in reality, some genes are excludable, particularly, though not only, when they have
variant genetic codes or behave as coalition or club goods, available for all organisms of a coalition to integrate
into their genomes, and non-rival within the club. We view the Tree of Life hypothesis as a regionalized instance
of the Public Goods hypothesis, just like classical mechanics and euclidean geometry are seen as regionalized
instances of quantum mechanics and Riemannian geometry respectively. We argue for this change using an
axiomatic approach that shows that the Public Goods hypothesis is a better accommodation of the observed data
than the Tree of Life hypothesis.
Background
The “Tree of Life” hypothesis has been in existence for
most of the last two centuries and is one of several
hypotheses that have been put forward to explain the
diversity of life on the planet [1]. In its recent practice,
this theory focusses on the vertical inheritance of genes
from parent to offspring and the continuous division of
lineages in the process of speciation. The theory was lar-
gely formulated in the days when the science of sys-
tematics was mostly concerned with the analysis of
plant, fungi and animals and the study of evolution was
essentially focussed on the study of these three
eukaryotic lineages. Indeed, arguably the greatest contri-
butor to the definition of the species concept, Ernst
Mayr, was so adamant in pinpointing that his views spe-
cifically applied to sexual organisms that he clearly titled
his major work on speciation mechanisms “Systematics
and the origin of species from the viewpoint of a Zoolo-
gist“ [2].
Microbiologists engaged in a long and largely unsatis-
fying search for the tree of prokaryotic life for most of
the 20th Century [3,4] (also see [5] for review). Woese
detailed this search in his treatise on bacterial evolution
[6], where he also wrote about the ideas and false-starts
that arose from time to time in the earlier part of the
century. The 1970s had resulted in significant develop-
ments in the sequencing of genes and this led Woese
and others to the belief that there was a new tool
* Correspondence: james.o.mcinerney@nuim.ie
1Molecular Evolution and Bioinformatics Unit, Department of Biology,
National University of Ireland Maynooth, Co. Kildare, Ireland
Full list of author information is available at the end of the article
McInerney et al. Biology Direct 2011, 6:41
http://www.biology-direct.com/content/6/1/41
© 2011 McInerney et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

available to systematists that would ultimately lead to a
satisfactory and detailed resolution of the entire Tree of
Life [7-9]. The fine-grained picture of the tree of life
was rapidly being brought into focus by the use of ribo-
somal RNA sequencing and analysis. Indeed, so power-
ful was this line of argument that today, the world’s
most targeted gene for sequencing is the small subunit
ribosomal RNA gene. However, and remarkably so, the
most sequenced and most ubiquitous kind of gene in
the world is not the small subunit ribosomal RNA gene.
Transposases, typically known to function by frequent
movement from one genetic element to another are the
most abundantly discovered genes in metagenomic stu-
dies where sampling of genes is undirected [10]. This
discrepancy between the most abundant marker
sequenced in the traditional phylogenetic framework
and the most abundant genes actually obtained in nat-
ure by chance, suggests that the scope of phylogenetic
analysis is focussed on the analysis of certain genes,
while if a different perspective was taken - to focus on
the most abundant genes - then a different interpreta-
tion of evolutionary history might be more readily
obtained.
Commonly, by the 1990s, the optimism surrounding
the reconstruction of a tree of life was giving way to a
more realistic picture of life on the planet. Hilario and
Gogarten [11] and Martin et al. [12] pointed out
remarkable inconsistencies in molecular phylogenies
derived from ATPase and Glyceraldehyde-3-phosphate
dehydrogenase genes when these phylogenies were com-
pared with ribosomal RNA phylogenies. A genome ana-
lysis showed that a substantial portion of the known E.
coli genome was acquired by horizontal gene transfer
since its separation from Salmonella [13]. At the end of
the century, Doolittle [14,15] and Martin [16] summar-
ized these growing problems with the tree of life model.
Soon afterwards, serious efforts were being made to
identify interspecies gene transfer events and to quantify
the extent of HGT in prokaryotic genomes [17-19] and
phylogenetic tree diagrams have been increasingly giving
way to network models of genome evolution in prokar-
yotes [20,21]. It must be pointed out, however, that the
focus on HGT events has been strongly criticized
[22,23] and some congruence between gene trees is
easily found [24].
It is not the case, however, that network diagrams are
being universally employed and indeed they still appear
in only a minority of studies that deal with the molecu-
lar systematics of prokaryotes. Tree diagrams, inferred
from a subset of genes with a wide or universal distribu-
tion are still the most commonly used models for pro-
karyotic evolution [25], though it has been pointed out
that usually these diagrams are only constructed from
less than one percent of the genome of these organisms
[26]. When a larger portion of the genome is used, then
the resulting tree diagram is highly dependent on the
method used in its construction [27] and in any case, no
strongly-supported nodes are found when moving
towards the base of this tree [24,27]. In the middle
ground are studies that try to reconstruct a tree or for-
est of trees in the presence of HGT events [28-33].
What is becoming increasingly obvious is the need to
either improve the Tree of Life model, if that is indeed
possible, or replace it with one (or several) hypotheses
that better fit the data. At the moment the interpreta-
tions of this model seem to be straddling the middle
ground - there is a great Tree of Life (sensu Darwin,
Lamarck, etc.) but it has annotations and complications
superimposed on its great frame, caused by interspecies
gene transfer. The problem with this model is that it is
becoming increasingly implausible. Recent estimates
show, for instance, that Escherichia coli as a species uses
approximately 18,000 genes [34,35], while the percent of
gene families that are now known to be found in every
E. coli is just 6% of the total (see also Beauregard-Racine
et al. [36]), though a typical E. coli strain only possesses
4,000-5,500 genes. This kind of scenario is seen again
and again during genome resequencing projects where
multiple strains of the same species are sequenced. A
minority of the genes that are found in a prokaryotic
species are found in just one genome of that species. As
a consequence, the Tree of Life hypothesis has been
modified extensively from its original description, in
order to avoid its rejection. We now know - unlike the
originators of the hypothesis almost two centuries ago -
that the main process of genome innovation for many
of the evolving entities on this planet is not vertical des-
cent, rather it is recombination and gene acquisition
[37]. Genes and genomes did not form part of the origi-
nal formulation of the Tree of Life hypothesis and pro-
cesses such as horizontal gene transfer and mobile
genetic elements such as viruses, plasmids and transpo-
sons obviously did not feature. In order to accommodate
these newly discovered, important features, the hypoth-
esis has been stretched to fit the data, however, given
our knowledge of the data, it seems that the elastic limit
of the original hypothesis has been passed.
In Darwin’s formulation of the Tree of Life hypothesis,
he said that he attempted “[...] to show that there is a
constant tendency in the forms that are increasing in
number and diverging in character, to supplant and
exterminate the less divergent, the less improved, and
preceding forms.” This particular quotation gets to the
heart of tree-thinking: that evolving entities would
always diverge away from one another and that evolu-
tion is a process of divergence. To put it another way, all
formulations of the Tree of Life hypothesis have at their
core the basic tenet that the pattern of diversity that we
McInerney et al. Biology Direct 2011, 6:41
http://www.biology-direct.com/content/6/1/41
Page 2 of 17

see on the planet is caused by a tree-like evolutionary
process and the differences in these formulations is to
be found in how much they allow deviation from this
central idea. Almost without exception, efforts to
describe the diversity of life on the planet have focussed
on the construction of this tree. The tree has been por-
trayed as a strictly bifurcating tree, as a fuzzy tree, as a
tree with cobwebs hanging from it and so forth. Like-
wise, the number and kinds of evolving entities has
changed over time, prokaryotes being largely ignored
initially (see [5]), mainly due to the difficulties of gener-
ating interpretable trees from the available data.
The current attempts at constructing the tree of life
can be roughly divided into four approaches (though,
other classifications of the approaches are easily con-
structed). Firstly, there is the tree as exemplified by the
small subunit ribosomal RNA gene [6]. Next, there is
the multi-gene approach using widely distributed genes,
usually of informational function [25]; the third
approach is to search for the biggest observable trend
embedded in the data [31], and the fourth is to con-
struct phylogenetic supertrees, the so-called tree-from-
trees method. These approaches have different meanings
and care must be taken to interpret what they say.
When the ribosomal RNA approach was first advocated
by Woese and co-workers, the ubiquity of the gene and
its attractive properties in terms of rapidly and slowly-
evolving sites, its conserved structure and its supposed
recalcitrance to horizontal gene transfer meant that it
was simply being used as a surrogate for the evolution
of the entire organism, a position that is no longer ten-
able. Using multiple genes in a concatenated superalign-
ment is designed to overcome the limitations of using a
single gene, however the interpretation of this tree is
somewhat similar to the interpretation of the rRNA tree
and this approach has been criticised as a “Tree of 1%”,
not a tree of life [26]. The Statistical Tree of Life
(STOL) and phylogenetic supertrees are constructed
from a much larger sample of genomes and attempt to
either construct a single tree (supertree approaches) or a
statistical trend that is tree-like from a large sample of a
genome. The interpretation of these structures is some-
what difficult, though they may approximate a “Tree of
Cells”. Unfortunately, it is now necessary to carefully
read each manuscript to find out what the authors are
calling the “Tree of Life”.
Presentation of the Hypothesis
Clearly, if it is our ambition to question or test the Tree
of Life hypothesis, then many approaches are inappropri-
ate because they conflate the explanandum and the
explanans (as already pointed out by Bapteste and Doolit-
tle [38]) - constructing a tree, however slender, weak or
fuzzy and using this tree to provide proof that there is a
tree. However, with few exceptions [21,39], all
approaches to date suffer from starting at the same place
- assuming that there is a fundamental tree-like structure
at the heart of biological evolution and then invoking ad
hoc criteria to explain data that does not conform (the
presence of plasmids, viruses, horizontal gene transfer,
genome fusion, endosymbioses and so forth). Tree of Life
efforts have generally had no regard for mobile genetic
elements such as plasmids and viruses and their classifi-
cation and evolution is typically discussed in a completely
separate body of literature (e.g. [40]).
Tree-like patterns undeniably exist but a tree is the
incorrect starting proposition and no matter how much
we customize the interpretation of this metaphor, we
will not be able to make it fit the observed data. In this
manuscript, inspired by patterns observed in genomic
data, we present a model of evolution explaining both
the reason why genomes have an almost endless combi-
nation of genes and why modest tree-like patterns are
seen when some small subsets of the data are compared
with each other. First of all we present two axioms
(uncontroversial starting points that are self-evidently
true) on which our model is based.
Axiom 1
All evolving entities should be included in a model that
aims at providing the highest-level evolutionary picture.
Axiom 2
Genes move both horizontally and vertically.
For axiom 1, we include all nucleotides that are part
of a replicating system - chromosomal, plasmid, viral
and so forth. Given that these two conditions are
uncontroversial, any hypothesis or model that purports
to describe biological evolution on the planet should be
compatible with these two desiderata. With these
axioms in mind, we move on to describing the public
goods hypothesis, which is our replacement for the Tree
of Life hypothesis.
The Public Goods model of evolution
While John Locke wrote about ownership of property
and its governance in 1690 [41], it is usually Paul
Anthony Samuelson, the first American to win the
Nobel Prize in Economics who is credited with being
the person who introduced the notion of public goods
and private goods [42]. In drawing the distinction
between different kinds of goods, Samuelson said:
“[...] I explicitly assume two categories of goods:
ordinary private consumption goods [...] which can be
parcelled out among different individuals [...] and
collective consumption goods [...] which all enjoy in
common in the sense that each individual’s consump-
tion of such a good leads to no subtraction from any
other individual’s consumption of that good [...].”
McInerney et al. Biology Direct 2011, 6:41
http://www.biology-direct.com/content/6/1/41
Page 3 of 17

Today four categories of goods have been character-
ized (see table 1). These goods differ in two of the fea-
tures that can be said to describe goods: whether the
good is rival and/or whether the good is excludable.
Rival refers to the availability of the good. A good is
non-rival if the consumption of the good by one indivi-
dual does not reduce the availability of that good for
another individual. A good is non-excludable if it is
impossible or at least very difficult to exclude the good
from being available to everybody. The air we breathe is
a good example of a public good. The consumption of
clean air by one individual does not greatly reduce the
availability of air for other individuals, so this makes air
a non-rival good and it is impossible to effectively
exclude all other individuals from accessing air, so this
makes air a non-excludable good. Consequently, we
think of air as a public good.
In order to elaborate on our hypothesis, we must first
of all explore genes to show that some may be consid-
ered as public goods. For the rest of this manuscript we
will use the term “gene”, but in reality we are talking
about any sequence of nucleotides that can reasonably
be considered to be a functioning unit or good (a gene,
portion of gene, promoter sequence, etc.). Genes can be
classified into homologous families and sub classified
into orthologous and paralogous genes (also other clas-
sifications such as ohnologs and xenologs exist), but for
the purposes of simplification we will speak generally
about homologs, defined by Richard Owen as “the same
organ in different animals under every variety of form
and function” (Owen, 1843), though of course, we are
speaking about genes and not animal organs. As far as
we know, there is no limit to the number of gene copies
that can exist for any given family of genes on Earth,
though obviously we see that individual evolving entities
have finite genome sizes. We see some gene families
widely distributed (say, transposases, informational
genes [25] or enzymes such as Rubisco [43]) and con-
versely, many genes have so far been only seen in a sin-
gle sequenced genome (personal observation). In
addition, a particular gene copy from a homologous
gene family can be used without “using up” the gene
family, in the same way that molecules of air can be
used by some organisms without appreciably using-up
the air. When gene copies are moved laterally, they are
usually moved using a copy-and-paste mechanism (e.g.
in plasmids, phage, class I transposons, outer membrane
vesicles and gene transfer agents), not a ‘stealing’
mechanism such as cut-and-paste (e.g. class II Transpo-
sons, nanotubes [44]). These features seem to suggest
that some genes are non-rival - the use of such a gene
by one organism does not preclude its use by another
organism.
If we furthermore consider the observations that have
been made on the widespread and easy gene availability
to single-celled prokaryotes: conjugation by plasmids
[4,45], the transformation of prokaryotes by phages [46],
the natural competence of some organisms like Neis-
seria [47], the secretion of outer membrane vesicles
packaging DNA in bacteria [48,49], the ability of gene
transfer agents to release small sections of DNA into
the environment [50], particularly, it seems in the
oceans [51], the sharing of DNA material through nano-
tubes [44] and the experimental demonstration that bar-
riers to forced gene transfer by cloning are almost non-
existent [52,53], these empirical observations all point to
many genes having the property of being non-exclud-
able. That is to say, it is very difficult for prokaryotes to
completely prevent other prokaryotes from obtaining a
particular gene. Indeed, the evolution of CRISPR ele-
ments [54] and restriction-modification systems [55],
themselves laterally transferred [56] testifies to the evo-
lutionary success of (and obvious need for) mechanisms
that protect against the (passive or aggressive) acquisi-
tion of genes.
Genuine public goods (non-excludable and non-rival)
are relatively rare, often somewhat intangible and
sometimes context-dependent. Knowledge is consid-
ered to be a public good - it is difficult to prevent the
spread of knowledge and facts don’t get “used up” if
many people know them. In an earlier example, we
cited clean air as a public good. However, in the con-
text of scuba divers under water, bottled air is not a
public good, it is a private good. In another example -
this time somewhat analogous to gene sharing - it is
asserted that file-sharing on the internet has the hall-
mark of a public good (the files are spread by a copy-
ing mechanism and once a file is freely available on
the internet, it is very difficult to completely prevent
people from accessing that file and making copies
available for further sharing) [57]. Legislation and
other means could conceivably restrict file-sharing on
the internet and therefore, while it might be consid-
ered to be a public good, its classification might be
context dependent.
We may conclude, therefore, that if some genes are
non-excludable and non-rival, they are public goods.
Likewise, it can be argued (and empirically tested) that
some genes might be de facto non-rival but excludable
to some extent and in some contexts, therefore, they
Table 1 The four categories of goods classified according
to the criteria of whether they are rival and/or
excludable
Excludable Non-excludable
Rival Private goods Common goods
Non-Rival Club goods Public goods
McInerney et al. Biology Direct 2011, 6:41
http://www.biology-direct.com/content/6/1/41
Page 4 of 17

might be better described as club goods rather than
public goods.
A protein-coding gene could be excludable, for
instance, if it uses a genetic code that is unique to a par-
ticular group of organisms. This DNA sequence could
produce a defective protein in an organism with a more
orthodox (universal) genetic code and consequently the
gene would not be fixed in that population, but would
be lost. Excludability of the gene in this case would
come from an intrinsic characteristic of the molecular
sequence and would be somewhat independent of func-
tion. In another scenario, a protein might only function,
say, in the absence of oxygen and therefore all aerobic
organisms would be excluded from using the gene that
encodes this protein. In this case, the function of the
encoded gene is the feature that would make a gene
into a club good. In a recent study it has been shown
that the connectivity of a protein in a protein-protein
interaction network plays a large part in whether it can
be successfully transferred to another species [58]. It is
possible that genes encoding very highly connected pro-
teins are effectively excludable and therefore could be
considered club goods.
We might also consider the case where the composi-
tion of the club plays a central role in whether a gene
remains a public good or becomes a club good. Symbio-
tic association of multiple lineages, have frequently been
observed and indeed, these organisms might be
“addicted” to one another. For instance, phylogenetically
heterogeneous communities (or coalitions [59]) found in
gut microbiomes, comprised of archaebacteria and
eubacteria, have converged in their repertoires of carbo-
hydrate-active enzymes to adapt to shared challenges.
This is in large part thanks to lateral gene transfer
mediated by mobile elements [60]. Such genes whose
exchanges are restricted between members of the coali-
tion, but that are not so widely spread outside the coali-
tion, constitute a particular type of club goods, because
of the original nature of the club itself.
We can call these genes “coalition goods”, because
they might represent a class of genes that, although they
are potentially public goods, are generally excluded from
being used by cells outside of the coalition, which is
commonly illustrated by metagenomic studies reporting
stronger functional signatures in microbial communities
over phylogenetic signatures, i.e. in which the presence
of particular genes coding for a certain function is more
essential than the presence of a particular lineage for
the community to evolve in a sustainable way [61]. The
dependency of the different lineages in the coalition on
one another thriving under particular selective pressures
may be sufficient to set up a barrier to gene availability.
Therefore such coalition goods are ultimately bounded
by the spatial distribution and the ecological interactions
of the members of the coalition. The genes might be
theoretically public goods, but practically, they are club
or coalition goods.
Most importantly, the notion of a gene as a public
good or a coalition good is entirely at odds with the
Tree of Life hypothesis, which implies that genes are
either private goods (excludable and rival) or club goods
(excludable and non-rival) as long as we define the club
to be a single clade (monophyletic group).
How are the goods kept in public ownership?
To be sustainable, public goods need to last and remain
available. In the case of genes, the maintenance of these
public goods and coalition goods is taken on board by
the organisms and mobile genetic elements. There is a
single genetic code (allowing for the numerous variants,
which are mostly, though not always, found in limited
kinds of organisms - Mycoplasma and Spiroplasma for
instance have a variant of the genetic code). This genetic
code is maintained by almost all the organisms that use
it. Mutations to genes that introduce stop codons are
generally weeded out by natural selection. Genes that
have proved useful to organisms, mobile elements, or
their coalitions have been replicated, while acquisition
of deleterious genes has led to the extinction of their
carriers. Therefore, organisms and mobile genetic ele-
ments patrol the public goods and these goods are
maintained over some evolutionary time by natural
selection. Nonetheless, a turnover in the composition of
the gene families is expected, because the nature of
selective advantages can evolve over time, i.e. after oxy-
gen level raised in the atmosphere, numerous gene
families involved in O2 metabolism likely became public
goods [62]. The same logic applies for antibiotic resis-
tance genes, etc. This is another feature of genes that
legitimise the classification of some genes as public
goods, though it is not a necessary feature. If genes were
private goods, then we might expect alternative genetic
codes to evolve over time thereby helping to keep the
private goods private. This is the case to some extent
with epigenetic modifications of DNA, yet overall this is
not what we see.
Implications of the Hypothesis
One of the consequences of tree-thinking [63] is that
genes are thought of as private goods or lineage goods
(i.e. club goods in which the club strictly matches a
monophyletic group). That is to say that if inheritance
of genes proceeds in a tree-like fashion, with genes only
passed down through time from ancestor to direct des-
cendent, then genes are seen as being available only to
one particular lineage, with other lineages being effec-
tively excluded from using these genes (this would
define genes as being excluded). The absence of
McInerney et al. Biology Direct 2011, 6:41
http://www.biology-direct.com/content/6/1/41
Page 5 of 17