Current Biology 17, R373–R386, May 15, 2007 ª2007 Elsevier Ltd All rights reservedDOI 10.1016/j.cub.2007.03.032
ReviewA Systematics for Discovering
the Fundamental Units of Bacterial Diversity
Frederick M. Cohan and Elizabeth B. Perry
Bacterial systematists face unique challenges when
trying to identify ecologically meaningful units of
biological diversity. Whereas plant and animal sys-
tematists are guided by a theory-based concept of
species, microbiologists have yet to agree upon
a set of ecological and evolutionary properties that
will serve to define a bacterial species. Advances in
molecular techniques have given us a glimpse of
the tremendous diversity present within the micro-
bial world, but significant work remains to be done
in order to understand the ecological and evolution-
ary dynamics that can account for theorigin, mainte-
nance, and distribution of that diversity. We have
developed a conceptual framework that uses eco-
logical and evolutionary theory to identify the DNA
sequence clusters most likely corresponding to the
fundamental units of bacterial diversity. Taking into
account diverse models of bacterial evolution, we
argue that bacterial systematics should seek to
identify ecologically distinct groups with evidence
of a history of coexistence, as based on interpreta-
tion of sequence clusters. This would establish a
theory-based species unit that holds the dynamic
properties broadly attributed to species outside of
‘‘The sure and definite determination (of species
of bacteria) requires so much time, so much acu-
men of eye and judgment, so much of persever-
ance and patience that there is hardly anything
else so difficult.’’
— Otto F. Mu ¨ller
For decades, the International Journal of Systematic
Bacteriology featured on its cover this testimonial to
the challenge of studying bacterial diversity. Indeed,
compared to zoologists and botanists, bacterial sys-
tematists face unique difficulties when beholding a
small phylogenetic group to identify its ecologically
distinct populations and the different roles that the
populations play within a community. Bacterial sys-
tematists are handicapped to some extent by the pau-
city of morphological differences that could help us
demarcate closely related species. More profoundly,
bacteriologists cannot predict with confidence what
will be the traits determining the ecological differences
between closely related species; this is because
prokaryotes often adapt to new niches by acquiring
genesfrom distantrelatives through horizontal genetic
transfer . Consequently, the ecological differences
among closely related bacteria are often invisible to
fared if Charles Darwin had arrived on the Galapagos
Islands with the handicaps of a bacterial systematist.
Would he have noticed 13 distinct finch species, each
with a bill morphology adapted for consuming a differ-
have appeared as a flock of related organisms — all
much of a muchness of finchdom? The challenge of
not seeing bacterial species — with the help of mor-
phology — has been surmounted by systematists in
several ways, but we shall see that bacterial systemat-
ics still suffers deeply for not readily sensing the
ecological differences among close relatives.
Closely related bacterial species were first distin-
guished by careful analysis of phenotype (typically
metabolism). In recent decades, systematists have
adopted molecular approaches that have allowed
standardized species demarcation and have ensured
that each taxon is a monophyletic group — a true evo-
lutionary group,including alland only thedescendants
of a given ancestor . Ironically, just as these molecu-
fident systematics, they have also revealed a daunting
task ahead. Surveys of gene sequence diversity from
environmental DNA have indicated that fewer than
1% of bacterial species are cultivable at present ;
methods have indirectly estimated that bacterial spe-
cies may number in the millions or even billions [4,5].
This massive expansion of the scope of systematics,
us as a Sisyphusian curse. However, we will show how
molecular and genomic approaches, when combined
with advances in ecological and evolutionary theory,
characterize our planet’s biological diversity.
The key to molecular discovery of biodiversity is to
find organisms that fall into highly distinct sequence
clusters for a given gene or set of genes. Because
such clusters have each had a long history of separate
evolution, they are likely to have evolved unique adap-
tations shared by the entire cluster. Systematists have
applied this sequence-based approach to discover
prokaryote diversity at all levels — from the urking-
doms, such as the Archaea, to species, and to even
lower levels of diversity [6,7]. By identifying and then
characterizing ever smaller phylogenetic groups,
each with its unique history and adaptations, system-
atists have reached a fuller understanding of the
ways that bacteria can make a living.
Department of Biology, Wesleyan University, Middletown,
Connecticut 06459-0170, USA.
But how far must we delve up the tree of life, identi-
fying smaller and smaller clades, before we have
fully characterized ecological diversity within the bac-
terial world? A comprehensive study of any type of
biological diversity would be complicated beyond fea-
sibility if nearly every individual organism were ecolog-
ically unique. Fortunately, organisms from all walks of
life — including bacteria, fungi, plants, and animals —
and from every known community, appear to fall into
discrete clusters of ecologically interchangeable indi-
viduals [8,9]. We will argue that systematics should
seek to recognize and characterize all these irreduc-
ible, ecologically distinct groups within a clade or
playing unique roles in community assembly, ecosys-
tem function and biotic interactions [10,11]. We will
explain that this is not the present aim of bacterial sys-
systematics frequently contain a diversity of popula-
tionsthatare distinctintheir biochemistry, physiology,
genome content and ecology; classifying an unknown
the organism’s way of life. Here we propose a system-
atics that demarcates and names the fundamental,
ecologically distinct groups within the bacteria. This
approach aims to satisfy Hutchinson’s central mission
an organism can inform us precisely about the organ-
ism’s ecological and physiological properties .
Theory-Free Approaches to Bacterial Systematics
Species demarcation in bacteria has been historically
handicapped by lacking a theory-based conceptual
framework. Systematists have yet to agree upon
a set of ecological and evolutionary properties that
can be expected for the set of organisms within a bac-
marcated empirically as clusters of similar organisms.
Bacterial species were first demarcated as phenotypic
the 1970s, bacterial systematics began incorporating
molecular methods to help distinguish closely related
species. Beginning with whole-genome comparisons
via DNA–DNA hybridization, systematists established
molecular criteria that would correspond to the spe-
cies groups that had already been defined by their
metabolic characteristics. Through whole-genome hy-
bridization, members of different named species were
usually found to share less than 70% of their genome
content, while members of the same species share
greater than 70% of their genome content . In
1987, this molecular cutoff became part of the canon
of species demarcation .
Systematists have recently sought to replace the
DNA hybridization standard of divergence with criteria
based on sequence divergence of homologous genes
. One principal advantage to a sequence-based
criterion is that any newly discovered organism can
be compared, in silico, to every existing sequence in
a growing data base for 16S ribosomal (r)RNA se-
quences . Stackebrandt and Goebbel  found
that, if two strains are at least 2.5% divergent in 16S
rRNA, they are sure to fall into different species on
the basis of DNA–DNA hybridization (although the
converse is not necessarily true). More recently,
a 16S divergence level of w1% has been deemed suf-
ficient to consider strains as sufficiently divergent to
be in different species . Similarly, Konstantinidis
and Tiedje  have found that a genome-wide aver-
age nucleotide identity of at least 94% in homologous
protein-coding genes is typical for members of the
species recognized by bacterial systematics. Efforts
are now underway to make possible a universal classi-
fication based on protein-coding sequences [20–22].
Another empirical approach is to find molecular cri-
teria that yield clear clusters of closely related organ-
isms, without the constraint that the clusters need to
coincide with existing species. For example, Hanage
et al.  have shown that phylogeny based on a con-
catenation of several genes can yield sequence clus-
ters that are robust with respect to recombination,
even within bacterial groups with relatively high re-
combination rates. Also, studies of genomic content
may have the potential to help identify clusters of
We must note that these molecular and genomic
approaches, however promising, are not designed to
infuse a theory of species into systematics. They are
merely adding new empirical criteria for dividing or-
ganisms into clusters with little attempt to correlate
clusters with the fundamental, ecologically distinct
populations within a natural community.
What is wrong with demarcating bacterial species
without a theory-based concept of species? To illus-
trate the importance of theory, we turn to the system-
species was developed long ago . While species
concepts for macrobes are by no means without dis-
pute, all modern species concepts embrace the fol-
lowing dynamic attributes for species : a species
is a cohesive group (there are forces that limit genetic
diversity within a species); a species is monophyletic
(invented only once); different species are irreversibly
separate; and different species are ecologically dis-
tinct (allowing them to coexist into the future).
In the case of animals and plants, there is one quin-
within a given species: successful interbreeding in
nature. Interbreeding acts as a force of genetic and
or plant species, and loss of the ability to interbreed al-
lows two species to diverge phenotypically and to
follow irreversibly separate evolutionary paths .
Different macrobial species, so defined, are typically
distinct in morphology, behavior, physiology and
ecology; organisms of the same species are expected
to be functionally interchangeable. Thus, zoologists
and botanists benefit from a systematics that satisfies
Hutchinson’s  fundamental mission of systemat-
ics — classifying an organism to an animal or plant
species yields detailed and specific information about
The Enormous Diversity within Named
Regrettably, lacking a concept of species, bacterial
systematics has failed to identify taxa that might sat-
isfy this mission. That is, a typical named bacterial
speciescontains hugediversityatalllevelsof analysis;
so species identification does not provide specific
ecological information about any of its members.
Even though bacterial species were originally demar-
cated as phenotypic (usually metabolic) clusters,
named bacterial species hold an enormous amount
of phenotypic diversity [27–29]. Species also show
a significant amount of genomic diversity. DNA–DNA
hybridization studies have demonstrated that mem-
bers of a named species frequently share only 80–
90% of their genes . These results have been
more recently corroborated by physical mapping of
genomes [31,32], and by genome sequence compari-
sons [33–36]. Even for genes shared across an entire
named species, there exists a great deal of sequence
variation within the species. At the 16S rRNA locus, se-
quence diversity within a recognized bacterial species
is frequently at 1%. This is equal to the level of diver-
gence typically found between orders of mammals at
the homologous nuclear gene 18S rRNA . When
we correct for the faster rate of molecular evolution
in bacteria, we can estimate the time of divergence
among conspecific bacteria to be about five times
greater than that for eukaryotic species .
Recentecological studies show that anamedbacte-
rial species is typically an assemblage of closely re-
lated but ecologically distinct populations [39–41].
For example, a species of free-living heterotrophs
may contain numerous sequence clusters that differ
in the carbon sources they can utilize, as seen in the
aquatic Shewanella putrefaciens . Clusters within
a species of free-living soil heterotrophs may differ in
the solar radiation (and covarying parameters) to
which they are adapted, as seen in Bacillus simplex
 and in B. licheniformis . Very closely related
phototrophs may partition resources by light quantity
and spectral quality, as seen in Synechococcus
[10,44], and in the mineral nutrient resources they
can utilize or store, as seen in Prochlorococcus mari-
nus . Within a pathogen species, populations can
differ in their host ranges , their target tissues
, or in the mode of transmission .
Even though fine-scale ecological differences are
frequently overlooked by modern systematics, we
should note that these differences within named spe-
cies are extremely interesting from an evolutionary
point of view. These differences tell us about the kinds
of ecological adaptations that can accrue over very
short time periods, and that can foster coexistence
among the most closely related of populations. We ex-
pect that finding the basis of coexistence among clos-
est relatives will be a rewarding challenge for the next
generation of bacterial community ecologists and
There is one realm of bacteriology where systemat-
ics has been scrupulous about naming every ecologi-
cally distinct sequence cluster — medical microbiol-
ogy. When one bacterial sequence cluster can kill us,
but a close relative cannot, it becomes worth our while
to put this distinction into our taxonomy . This is
seen in the naming of Bacillus anthracis (causing an-
thrax) as distinct from B. cereus , and the naming
of Yersinia pestis (causing plague) as distinct from
We would prefer that all bacteria be as aptly demar-
cated into ecologically distinct groups as are the most
virulent agents of human death. We believe an under-
standing of bacterial population dynamics can help
us identify these groups possessing the fundamental
attributes of species.
The Peculiarities of Bacterial Population Dynamics
In contrast to most animals and plants, prokaryotes
reproduce clonally, and the genetic exchange among
prokaryotes proceeds by various parasexual pro-
cesses that are not tied to reproduction. When genetic
exchange occurs in bacteria, a short segment from
a ‘donor’ individual replaces the homologous segment
in a ‘recipient’ individual. Recombination in bacteria
and plants, with the recombination rate ranging from
less than the mutation rate (per gene segment) to
about ten times the mutation rate in the most fre-
quently recombining organisms [50–52]. Only one bac-
terial species, Helicobacter pylori, is known to un-
dergo recombination many orders of magnitude
more frequently than mutation .
The rarity of recombination is expected to simplify
the formation of new species of bacteria. In order for
adaptive divergence to occur between highly sexual
animal or plant populations, gene flow must be re-
stricted by some geographical or ecological barrier
. In contrast, recombination in bacterial popula-
tions is not frequent enough to hinder adaptive diver-
gence . Because of the extremely low rates of re-
combination in prokaryotes, the influx of genes from
other populations is unlikely to disrupt the integrity of
each population’s specific adaptations. Evolution of
sexual isolation is thus not a necessary step toward
permanent divergence between ecologically distinct
bacterial populations , and sympatric speciation
may be a common occurrence.
The typical rarity of recombination in prokaryotes is
also expected to lead to a recurrent purging of diver-
sity within an ecological population. When natural se-
lection increases the frequency of an adaptive muta-
tion at a particular gene locus within a population,
there is little opportunity for separation of the adaptive
mutation from the genome in which it originated. Thus,
the entire genome of the adaptive mutant increases in
frequency and diversity is purged at nearly all loci. This
kind of diversity-purging event, called ‘periodic selec-
tion’ [55,56], can be an important agent limiting the
tion (excluding the frequently recombining H. pylori
[53,57]). In contrast, in the highly sexual animals and
plants, each bout of natural selection has only very
limited ability to purge diversity — generally affecting
only the chromosomal region near the locus under
What then are the forces limiting genetic diversity
within an animal or plant species, and to what extent
do they apply to bacteria? Genetic drift is the primary
force limiting sequence diversity within a highly sexual
population of animals or plants . Random elimina-
tion of lineages limits sequence diversity. While ge-
of genetic drift are quite weak in populations much
larger than 109. In large bacterial populations, if
drift were the only factor limiting the sequence diver-
sity, we would expect to see saturation in variation at
every neutral nucleotide site . When we do not
see evidence of such saturation in large populations,
we can conclude that a factor other than drift (such
as periodic selection) is operating to limit sequence
In some bacteria, however, population sizes may be
The power of genetic drift may be especially important
when considering obligate pathogens and commen-
sals, which must be transmitted from host to host. In
cases where very few bacterial cells colonize an indi-
vidualhost, and very few cells successfully leave anin-
dividual host, the effective population size of the bac-
teria is reduced to nearly that of the host species.
The Ecotype Model of Bacterial Species
Takinginto account thesefeatures ofbacterial popula-
tionbiology, we considerfirstamodel of bacterial spe-
cies based on the ‘ecotype’. An ecotype is defined
here as a group of bacteria that are ecologically similar
to one another, so similar that genetic diversity within
the ecotype is limited by a cohesive force, either peri-
odic selection or genetic drift, or both. The present
concept of ecotype is more general than in our previ-
ous work, where periodic selection was viewed as
the only cohesive force [38,54].
In this model, diversity within an ecotype is ephem-
eral, persisting only until the next periodic selection
event, when diversity is next crushed to near zero at
all loci, or until purged by genetic drift (Figure 1A).
What, then, is the source of permanent divergence
among closely related bacteria? Divergence can be-
come permanent when a mutation (or recombination
event) places the organism into a new ecological niche
and founds a new ecotype. Because the new ecotype
is ecologically distinct from the parental ecotype, peri-
odic selection events in the parental ecotype cannot
extinguish the founding organism and its descendants
(Figure 1B). The new ecotype thus escapes the peri-
odic selection of the parental ecotype, and the two
new ecotypes are free to diverge indefinitely.
erties attributed to species . Each ecotype is a co-
hesive group whose diversity is limited by periodic
reversibly separate because they are out of range of
one another’s periodic selection events, and because
recombination is too rare to prevent their adaptive di-
vergence [59,60]; they are by definition ecologically
distinct, which allows them to coexist into the future.
Finally, such ecotypes are monophyletic groups be-
cause they are founded by a single individual (except
in the Recurrent Niche Invasion model, below).
A variant of the ecotype model is particularly prom-
ising for its utility in systematics. This is the ‘Stable
Figure 1. Three classes of mutation and recombination events that determine ecotype diversity in bacteria.
Circles represent different genotypes, and asterisks represent adaptive mutations. (A) Periodic selection mutations. These improve
the fitness of an individual such that the mutant and its descendants out-compete all other cells within the ecological niche (ecotype);
these mutations do not affect the diversity within other ecotypes because ecological differences prevent direct competition. Periodic
selection leads to the distinctness of ecotypes by purging the divergence within but not between ecotypes. (B) Ecotype-formation
mutations. Here a mutation or recombination event allows the cell to occupy a new ecological niche, founding a new ecotype. The
ecotype-formation mutant, as well as its descendants, can now escape periodic selection events from its former ecotype. (C) Speci-
ation-quashing mutation. Even if two ecotypes have sustained a history of separate periodic selection events, an extraordinarily
adaptive mutant genotype may out-compete to extinction another ecotype. Competitive extinction of another ecotype (Ecotype 2)
is possible only if all of Ecotype 2’s resources are also utilized by Ecotype 1. Speciation-quashing mutations are expected in the
Nano-niche model . (Used with permission from Landes Publishers.)
Ecotype’ model, where ecotypes are created and ex-
tinguished at a very low rate, and during its long life-
time an ecotype is recurrently purged of its diversity
by periodic selection events [2,9,61] (Figure 2A).
Most such ecotypes (of sufficient age) should be dis-
tinguishable from other ecotypes as separate se-
quence clusters under most rates of recombination
encountered in bacteria [7,60,62]. Ideally, we may
identify ecotypes as DNA sequence clusters, provided
that the Stable Ecotype model applies. However, the
Stable Ecotype model does not take into account sev-
example, geographic isolation, adaptation through
plasmid transfer and loss, genetic drift and very rapid
speciation. We will examine each of these complicat-
ing factors in turn, but we will first consider whether
the basic Stable Ecotype model appears correct.
Guttman and Dykhuizen provided the first evidence
for the diversity-purging effects of selection in natural
populations of bacteria . They found a chromo-
somal region (near gapA) in Escherichia coli that is
nearly homogeneous, even while the rest of the chro-
mosome shows substantial heterogeneity and forms
discrete sequence clusters. They originally interpreted
this pattern as evidence that an adaptive mutation in
the gapA region swept through the ‘‘population’’ of
E. coli, but that recombination rescued existing varia-
tion in all areas outside of this chromosomal region.
Such an interpretation could lead to the generalization
that in bacteria, as in the highly sexual animals, each
adaptive mutation purges diversity only in the chromo-
somal region near the adaptive mutation.
We disagree with this interpretation, because we
have previously shown recombination to be rare
enough to support periodic selection in nearly all bac-
teria, provided that the rate of recombination is within
an order of magnitude of mutation (per gene) ; this
certainly includes E. coli, as well as all bacteria studied
for recombination rates, with the exception of Helico-
bacter pylori . Also, this interpretation begs the
question of the origin of the sequence clusters within
E. coli. If all of E. coli is one population through which
one adaptive mutation can advance, why should there
be multiple, apparently coexisting sequence clusters?
We have previously proposed an alternative model
for the small chromosomal regions of homogeneity
that one encounters in genomic comparisons, as well
as in the original gapA survey. In our ‘Adapt Globally
Act Locally’ model [65,66], we interpret the various se-
quence clusters of E. coli as distinct ecotypes that un-
dergo their own periodic selection events (Figure 3).
Certain mutations arise, however, that are adaptive in
many different ecotypes, and they can be passed to
them through recombination. The adaptive mutation
can then trigger a periodic selection event within
each ecotype into which it is transferred. A series
of periodic selection events in different ecotypes
results in homogenization of the small chromosome
region that is transferred between ecotypes, but the
sequence clusters corresponding to the different
ecotypes are otherwise left intact.
As many closely related genomes are compared,
we believe that the Adapt Globally model will be the
most compelling explanation for small chromosomal
regions of homogeneity. For example, we have found
diverse, with different ecotypes (revealed as sequence
clusters) living at different temperatures and in differ-
ent photic zones [10,44]. However, this clade is homo-
geneous within and between ecotypes in the nitrogen-
fixing region of their chromosomes . Clearly, the
homogeneity in this region is not due to one adaptive
mutant outcompeting cells in the entire, ecologically
heterogeneous Synechococcus clade. Rather, an
adaptive mutant was most likely able to extinguish
of the adaptive mutation to other ecotypes through
recombination, with subsequent periodic selection
within each recipient ecotype.
We next address other arguments that periodic
selection may not be a significant force in bacterial
evolution. Roumagnac et al.  noted that existing
sequence variation within Salmonella enterica serovar
Typhi appears neutral with respect to fitness — non-
synonymous and synonymous substitution rates are
similar — and claimed that this result was inconsistent
with a history of periodic selection. However, periodic
selection is expected simply to reduce the sequence
of sequence variation after periodic selection does not
constitute a test of periodic selection.
Also, Roumagnac et al.  have noted that a partic-
ular adaptive mutant clone of S. enterica serovar Typhi
(H58), conferring resistance to the antibiotic nalidixic
acid, has rapidly increased in frequency but does not
appear to be headed toward 100% frequency; they
have argued that this failure to reach 100% is evidence
againstperiodic selection. However, whenanadaptive
likely because the clade contains multiple ecotypes
[69,70]. In the case of H58, the inability to reach
100% within the S. enterica serovar Typhi clade, either
in local geographic regions or globally, provides evi-
dence of at least two ecotypes within serovar Typhi,
perhaps one adapted to the niche of humans treated
with nalidixic acid, and one or more ecotypes where
nalidixic acid is not part of the environment.
We are aware of only one taxon where periodic se-
lection is known not to effect a genome-wide purging
of diversity. This is the extremely frequently recombin-
ing H. pylori , an obligate pathogen of humans that
shows nearly the same pattern of geographical diver-
sity as its human host ; all other bacterial groups
studied appear clonal enough to support periodic se-
lection . What is not clear is the extent to which
genetic drift might also limit the diversity within a
given ecotype. This is an issue that will probably not
be resolved by analysis of sequence diversity, be-
cause periodic selection and drift can both, separately
ortogether, limit sequence diversity within anecotype.
In order to assess the relative contribution of these
forces, we must instead consider the population
size and recombination rates of the organism in its
natural environment. Genetic drift can be ruled out as
a major factor in populations larger than 109, and peri-
odic selection is unlikely to be a major cohesive force
in populations with extremely high recombination
Figure 2. A diversity of evolutionary models for the relationship between ecologically distinct populations and DNA sequence clusters.
Ecotypes are represented by different colors; periodic selection events are indicated by asterisks; extinct lineages are represented by
dashed lines; clades that may be perceived as sequence clusters are marked by a horizontal black line at the top of the phylogeny.
(A) The Stable Ecotype model. This model is marked by a much higher rate of periodic selection than ecotype formation, such that
each ecotype endures many periodic selection events during its lifetime. The Stable Ecotype model generally yields a one:one corre-
spondence between ecotypes and sequence clusters. (B) The Geotype-plus-Boeing model. A history of geographic isolation, followed
byrecent human-aided transport(during the ‘ageofBoeing’),canleadtomultiplesympatricsequence clusterswithin asingleecotype.
Lineages from different regions are represented by different line thicknesses. (C) Genetic Drift model. Genetic drift can also produce
multiple clusters within each ecotype, provided that effective population sizes are relatively small. (D) The Speedy Speciation model.
Herenew ecotypes areformedat ahigh rate,such thatmanyyoung ecotypes havenot diverged sufficiently enough tobedistinguished
as separate sequence clusters. For example, the olive and tan ecotypes within the rectangle form a single sequence cluster when an-
alyzed using a gene with a slow evolutionary clock. However, when analyzed by a more rapidly evolving molecular marker (the ‘fast
gene’ indicated), a one:one correspondence between ecotypes and sequence clusters may be seen. (E) The Species-less model.
Here, invention and extinction of ecotypes occur very frequently. Diversity within an ecotype may be constrained primarily by the short
amount of time between the ecotype’s founding by a single mutant (or recombinant) to the time the population goes extinct. (F) The
Nano-niche model. Here an ecotype contains a set of subtly ecologically distinct subpopulations that may use the same set of re-
sources but in different proportions. While each of the subpopulations may undergo its own periodic selection events for some
time (small asterisks), eventually a particularly adaptive mutation (large asterisk) extinguishes all but one of the subpopulations and
purges diversity throughout the whole ecotype. While a slowly evolving gene might not give enough resolution to distinguish the sub-
populations, a faster gene might (inset box). (G) Recurrent Niche Invasion model. Here a lineage may move, frequently and recurrently,
from one population to another, usually by acquisition and loss of niche-determining plasmids. (H) Cohesive Recombination model.
With recurrent recombination between ecotypes, distinct clusters may never be seen in genes that are not niche-determining.
Systematics and the Diversity of Models
of Bacterial Evolution
We next consider alternatives to the Stable Ecotype
model, where the correspondence between ecotypes
and sequence clusters is not expected to be 1:1. In
some alternative models of bacterial evolution, one
ecotype may contain multiple sequence clusters. Con-
sider first a model in which ecotypes are long-lived, as
in the Stable Ecotype model, but there is only rare mi-
gration among the geographic regions of the ecotype.
Ecologically identical populations in different regions
can thus diverge into different sequence clusters.
to closely related populations from different geo-
graphic regions that have diverged as a result of their
geographic isolation and not ecological differences.
How likely is the formation of geotypes in the bacte-
rial world? For nearly a century, the theory of bacterial
biogeography was dominated by the idea that ‘‘every-
That is, what determines the presence of a species at
a location is not its ability to get there, but only its abil-
ity to thrive once it arrives [73,74]. However, recent
studies suggest that at least some bacteria are se-
verely restricted in their migration. Some extremo-
philes show only rare dispersal across uninhabitable,
mesic habitats to other favorable, extreme locales
[71,75,76], and rare migration is not limited to extrem-
ophiles . Some environments are especially pro-
hibitive of dispersal , for example, the highly
viscous substrate of deep-rock bacteria and the pe-
rennially frozen lakes of Antarctica. Many pathogens
and commensals can be only as mobile as their hosts.
A history of geographic separation among ecologi-
cally interchangeable populations can lead to difficul-
ties for sequence-based taxonomy. In a model we
call the ‘Geotype-plus-Boeing’ model [2,9,61], geo-
graphically isolated populations of the same ecotype
could diverge into separate sequence clusters in the
time before rapid human transport; then, in recent de-
cades (or centuries), human shipping and travel could
carry all the endemic geotypes within a single ecotype
into each region of the world (Figure 2B). In this transi-
tional era when air travel (and even transoceanic sea
travel) is still relatively new, we may see multiple se-
quence clusters (the pre-‘Boeing’ geotypes) within
one ecotype at one place, as seen in the plague bacte-
rium Yersinia pestis  and in the tuberculosis agent
Mycobacterium tuberculosis . Therefore, we can-
not conclude from sequence clustering alone that
two sympatric clusters are separate ecotypes.
Genetic drift can also yield subclusters of closely
related organisms of the same ecotype, especially in
pathogens and endosymbionts, where effective popu-
lation size can be severely limited. Thus, under either
the Geotype-plus-Boeing or the Drift models (Fig-
ure 2C), the relationship between ecotypes and
sequence clusters can be one:many.
We next consider models yielding a many:1 corre-
spondence between ecotypes and sequence clusters,
beginning with the ‘Speedy Speciation’ model, where
new ecotypes form at a rapid rate, and new ecotypes
are expected to coexist and diverge from one another
into the indefinite future (Figure 2D). While the numer-
ous new ecotypes are expected to form their own se-
quence clusters eventually, at any one moment a large
number of newly formed ecotypes may be invisible to
detection as sequence clusters for a given molecular
marker. In these cases of rapid speciation, it may be
important to use more rapidly evolving molecular
markers to detect the most recently formed ecotypes,
such as variable number tandem repeated (VNTR) se-
quences . Slowly evolving genes will not highlight
tution is much less than the rate of ecotype formation
and periodic selection . A Speedy Speciation
(I) The Animal-like model. Recombination is frequent enough that periodic selection does not limit sequence diversity within ecotypes.
The effective population size is low enough that genetic drift can effectively limit diversity within ecotypes. Also, migration is infrequent
enough that geographic regions have their own geotypes within each ecotype.
Figure 3. The Adapt Globally Act Locally model.
Adaptive mutants compete only with cells within their ecotype,
but the mutation may be adaptive in many ecotypes and can be
transferred through recombination. (A) An adaptive mutation
(asterisk) occurs in Ecotype 1. Periodic selection purges diver-
sity (black circles) within the ecotype. (B) The adaptive allele is
transferred toacellinEcotype 2viarecombination. (C)Because
the mutation is adaptive in this ecological niche as well, a peri-
odic selection event occurs within Ecotype 2 . (Used with
permission from the Society of Systematic Biologists.)
regime that generates a many:one relationship be-
tween ecotypes and sequence clusters, as seen
through protein-coding gene diversity, might actually
generate a one:one relationship when VNTR or inser-
tion sequences are used to reveal more microdiverse
Under what circumstances might the Speedy Speci-
many empty ecological niches are available, following
an evolutionary innovation , although it is not clear
what evolutionary innovations might precipitate rapid
speciation in bacteria. Rapid speciation in bacteria
. HGT-facilitated speciation may be promoted by
a bacterium’s intrinsic ability to absorb a genetic
shock — such as a metabolic imbalance caused by
stantial ill effect  and to the bacterium’s ability to
evolve changes that ameliorate adaptive changes ini-
tially carrying some deleterious side-effect [84,85].
Also rapid speciation may be promoted by a bacte-
rium’s ability to independently tweak the expression
of newly acquired genes, so as to maintain metabolic
balance , as may be the case in Pseudomonas .
Geographic isolation may be a necessary step facil-
itating bacterial speciation, depending on the extent to
which evolution of a new adaptation diminishes the
population’s previously existing adaptations. When
there are intrinsic trade-offs between new and preex-
isting adaptations, a set of adaptive mutations can im-
mediately produce a new population that can coexist
with the parental population. However, when evolution
of new adaptations does not diminish old adaptations,
evolution of ecologically distinct, coexisting popula-
tions can occur only in geographic isolation, where
each population loses the subset of adaptations not
needed in its respective environment . Thus, rapid
evolution of new species in bacteria could depend on
low dispersal and geographic isolation. Finally, a rapid
increase in the number of ecotypes may also be due to
a lower rate of extinction, but little is known about ex-
tinction in the bacterial world. Thus, for a number of
reasons, some bacteria may engage in rapid specia-
tion, perhaps too quickly to ever discern through clus-
tering of DNA sequences.
We next consider a variant of the Speedy Speciation
model, the ‘Species-Less’ model [2,9,60,61,89], where
there is a high rate of both formation and extinction of
ecotypes, most likely when environments change rap-
idly. In the Species-Less model, the diversity within an
ecotype need not be constrained into the indefinite fu-
constrained principally by the short amount of time
from the ecotype’s founding from a single mutant (or
recombinant) to the time the ecotype goes extinct
(Figure 2E). Here a single sequence cluster might con-
tain multiple very young ecotypes.
We next consider two models in which there is a
rapid and frequent invasion of ecological niches, but
the ecologically distinct populations are not expected
to diverge indefinitely. Both models yield a many:one
correspondence betweenecologically distinct
populations and sequence clusters. The ‘Nano-niche’
model postulates a great diversity of ephemeral habi-
tats, for example, small particles in the marine water
column in the case of Vibrio splendidus (M. Polz, per-
sonal communication) (Figure 2F). In this model, sub-
groups within one ecotype each become adapted in
nuanced ways to the subtleties of their own habitats;
they may even have their own separate periodic selec-
tion events. Nevertheless, it may be possible for one
especially adaptive mutant to outcompete to extinc-
tion all the ecological diversity among the ecotype’s
subgroups (a ‘speciation-quashing’ event; Figure 1C)
. In this case, the various ecologically distinct sub-
groups within an ecotype are not irreversibly separate,
and do not have a chance to diverge into separate se-
We note that recombination may have the potential
to foster an extended coexistence of these ecologi-
cally distinct subgroups. If the adaptive mutation aris-
ing in one subgroup can be transferred to other sub-
groups and confer adaptation there, as in the Adapt
Globally Act Locally model , the first subgroup
might fail to outcompete the other subgroups to
In the ‘Recurrent Niche Invasion’ model  (Fig-
ure 2G), members of each ecotype frequently and re-
currently lose the adaptations of their present ecotype
and acquire the adaptations of another. Recurrent
niche invasion is most likely when the populations
owe their ecological distinctness entirely to the facile
gain or loss of a plasmid. For example, in Bacillus thur-
ingiensis, some clades can host a number of alterna-
tive ‘crystal toxin’ plasmids, each adapted to killing
a different order of insect. A lineage may recurrently
move from one ecological niche to another by losing
one crystal toxin plasmid and then acquiring another.
If these reciprocal ecological conversions recur re-
peatedly, then the populations are notirreversibly sep-
arate lineages, and they may never appear as separate
A complex relationship between ecotypes and se-
quence clusters may occur if recurrent recombination
into separate sequence clusters for most genes — the
‘Cohesive Recombination’ model (Figure 2H) [54,90].
However, recombination between ecotypes is ex-
pected to decrease over time, owing to a positive
feedback between sexual isolation and sequence
divergence among ecotypes [60,62].
Finally, we propose the ‘Animal-like’ model (Fig-
ure 2I), inspired by the population dynamics of Helico-
bacter pylori [53,57]. In this model, recombination is
extremely frequent, such that periodic selection is
not a cohesive force. Effective population sizes are
low enough to allow genetic drift to maintain sequence
diversityat alow level inthe absence ofperiodic selec-
tion. Also, dispersal is rare enough such that there is
a trace of migration history in the sequence record.
In other words, the recombination, population size,
and migration parameters are as expected for many
One consequence of the high recombination rates in
the Animal-like model is that speciation would require
recombination between nascent ecotypes to be
reduced by some form of sexual isolation, as is the
case for animal speciation. Possible mechanisms of
sexual isolation are: adoption of different restriction
tors of recombination (phage or plasmids) across eco-
types ; DNA sequence divergence, which hinders
recombination [93–95]; and most importantly, ecolog-
ical differences preventing access to other popula-
tions’ DNA .
Incorporating Ecology and Evolution
into a Systematics of Ecotypes
Animal and plant systematists have the advantage
that, for a given clade, the traits determining niche dif-
ferences can be anticipated. Finch species are spe-
cialized by differences in the shapes and sizes of their
bills; herbaceous species on the American prairie can
specialize to different growth seasons through differ-
ences in their root architecture, and so on. Thus, ani-
mal and plant systematists can demarcate species
by identifying clades that are homogeneous for the
traits most likely to be niche specifying .
In the case of bacteria, horizontal genetic transfer
very difficult, if not impossible, to predict. Indeed,
close relatives can infect entirely different hosts
time the traits distinguishing the most closely related
ecotypes, we suggest that identification of ecotypes
should begin with a sequence-based approach to
formulate hypotheses about putative ecotypes, fol-
lowed by an ecological approach to confirm these
We begin by assuming the Stable Ecotype model,
where ecotypes are in principle discoverable as se-
quence clusters, and later take into account that other
models may apply instead. Even under the Stable Eco-
type model, however, ecological interpretation of se-
quence-based phylogenies is not straightforward.
Any sequence-based phylogeny is likely to contain a
hierarchy of subclusters within clusters, and it is not
clear which level of cluster corresponds to ecotypes
. We have therefore developed and tested a con-
ceptual framework for identifying ecotypes from se-
quence clusters under the assumptions of the Stable
Ecotype model [10,38,44,100].
This ‘ecotype simulation’ approach begins by char-
acterizing the sequence diversity within a clade as
the number of sequence clusters (or bins) present for
different sequence-identity criteria, following Martin
 and Acinas et al.  (Figure 4). The number of
sequence clusters at a particular sequence identity
level represents the number of lineages at some point
in the past that have survived to the present . The
simulation estimates the rates of periodic selection
and drift, the net rate of ecotype formation (taking
into account ecotype extinction), and the number of
ecotypes so as to yield the clade sequence diversity
pattern of Figure 4 with maximum likelihood (Figure 5).
to examine three clades whose ecological diversity
ing isolates of Bacillus primarily from ‘Evolution
Canyon’ III of the Negev Desert , sequences
from uncultured members of the Synechococcus A-
A0clade from Yellowstone hot springs [9,10,104], and
world-wide clinical and environmental isolates of
Legionella pneumophila .
Ecotype simulation estimated more ecotypes than
the number of existing species and subspecies, for
each clade analyzed, corroborating previous evidence
that the demarcations of bacterial systematics fre-
quently lump many ecologically distinct populations
into a single species [9,40,41,43,105]. We extended
the ecotype simulation analysis to identify all the indi-
vidual ecotypes of a clade resolvable with DNA se-
quence data [38,44]. The rationale was first to quantify
the sequence diversity within a given subclade, and
then to determine the number of ecotypes that yields
the subclade’s sequence diversity pattern with maxi-
mum likelihood. We then demarcated putative eco-
types as the largest clades that were each consistent
with a single ecotype.
In each clade we have analyzed, ecotype simulation
demarcated some putative ecotypes that appear eco-
ecological distinctions were inferred from differences
were most frequently isolated. In Synechococcus, mi-
crohabitat distribution suggested that very closely re-
lated putative ecotypes are specialized to different
temperatures and photic zones [10,44]. In Bacillus,
very closely related putative ecotypes were found to
be adapted to different conditions of solar insolation
(or co-varying factors) on different faces of the canyon
[38,44] (Figure 6). In the case of Legionella, putative
ecotypes differed in the species of amoebae they
can infect .
Sequence identity criterion
Number of bins
Figure 4. Clade sequencediversity, for the B. subtilis–B. lichen-
iformis clade, for gene rpoB.
A set of 88 sequences from the clade was binned into clusters
with different levels of minimum pairwise identity. The log-linear
portion of the curve to the left of w0.99 identity indicates a con-
stant net rate of ecotype formation; the flair of diversity to the
right of w0.99 indicates the facile sequence diversity within
ecotypes . The ecotype simulation analysis yielded esti-
mates of the parameters for periodic selection rate, net ecotype
and the estimated quartet of parameter values generated the
‘model’ values shown .
Ecotype simulation promises to be an effective way
to identify the fundamental units of bacterial ecology
and evolution. This approach has the advantage that
it analyzes a particular clade’s population dynamics
to obtain the appropriate demarcation threshold for
that clade. It is unlikely that the universal molecular
thresholds routinely used in systematics [7,15,18], or
indeed any universal molecular threshold, could
demarcate these fundamental units .
Nevertheless, sequence-based hypotheses about
putative ecotypes must take into account the great di-
sequence clusters, even if identified by ecotype simu-
lation or a similar algorithm, do not always correspond
to ecotypes as in the Stable Ecotype model. To this
end, we suggest that ecotypes should be demarcated
as the smallest groups that: (1) show a history of coex-
istence as separate, ecologically distinct lineages, as
inferred from ecotype simulation (or an equivalent
sequence-based approach); and (2) show a prognosis
for future coexistence, as inferred from the ecological
distinctness of the groups in nature [38,44].
Why should we not demarcate ecotypes solely by
sequence clustering? To the extent that the Stable
Ecotype model is correct, different sequence clusters
are indeed likely to represent different ecotypes. How-
ever, to the extent that the Geotype-plus-Boeing
model applies, different clusters could represent
formerly isolated populations of the same ecotype
that have recently been flown or shipped to the same
locations. Alternatively, in cases where drift is likely
to be an important force, an ecotype could contain
multiple sequence clusters caused by genetic drift.
Therefore, sequence clusters must be verified to be
ecologically distinct before they can be declared
Why is ecological distinctness alone insufficient to
demarcate ecotypes? First, given the potential for hor-
izontal genetic transfer, any two closely related iso-
lates or populations are likely to differ somewhat in
their physiology [106,107]. Clearly what we want to
know goes beyond laboratory assessment of physio-
logical differences that have no bearing on ecological
niche in nature. Rather, we need to ascertain that pop-
ulations are ecologically distinct in a way that allows
them to partition resources in nature, and thereby co-
exist into the future. Sequence data provide a means
for inferring that ecological differences observed in
the laboratory are important in nature. When two eco-
clusters, we may infer that the populations are long-
standing in their coexistence, possibly owing to their
ecological differences (alternatively to previous geo-
graphic separation) .
There is a second problem in identifying populations
as different ecotypes when they are not yet separate
Most recent common ancestor of all sampled organisms
Figure 5. The ecotype simulation algorithm.
A computer algorithm simulates the evolutionary history of the n organisms sampled from nature, under different quartets of values for
the net rate of ecotype formation (EF), the rates of periodic selection (PS) and drift (D), and the number of ecotypes in the sample. In the
coalescence approach taken, the algorithm considers only the lineages that are directly ancestral to the n sampled organisms (circles).
These focal lineages are represented by solid lines; the many contemporary lineages not sampled from each ecotype are indicated by
light dashed lines (E1, E2, and E3); the lineages extinguished by past periodic selection and drift are represented by bold, short-dashed
lines and long-dashed lines, respectively, with each extinction represented by a square. The program begins with a ‘backward’ sim-
ulation that stochastically produces a phylogenetic representation of the history of the community, establishing nodes of coalescence
of lineages (indicated by grey circles) and time between nodes (t1, t2, etc.); this phylogeny is then taken as a scaffold for the forward
simulation. The purpose of the forward simulation is to produce nucleotide substitutions throughout the history of the clade, according
to the phylogenetic scaffold. To begin a simulation, a set of n contemporary organisms (representing the n organisms sampled from
nature) are distributed randomly among the n ecotypes. (In the case of this figure, n = 14 and n = 3.) Working backwards from the n
organisms in the present, the processes of ecotype formation, periodic selection, and drift occur stochastically in time according to
their respective rates (U, s, and d). The backwards phase of the simulation ends when all of the branches have coalesced into a single
node; this represents the most recent common ancestor of all the sampled organisms. Then the forward simulation begins when a se-
quence is assigned to this most recent common ancestor. Nucleotide substitutions then occur stochastically, going forward in time,
between each pair of nodes in the phylogeny derived from the backward simulation, according to the time between the events deter-
mining the nodes .
sequence clusters. The populations might not be irre-
versibly separate, a case most likely when populations
owe their ecological distinctness entirely to the gain or
or when ecological differences between populations
are not sufficient to indefinitely evade one another’s
periodic selection events (Nano-niche model). How-
ever, if we demarcate ecotypes only when ecologically
distinct groups form separate sequence clusters, we
can be assured that the populations have had a history
of divergence as separate lineages [38,61].
Finally, we must take into account models where
ecologically distinct populations are frequently too
new to be distinguishable as sequence clusters. We
believe that we will not normally want to grant ecotype
status to a new population that has not yet demon-
strated its ability to coexist with others (by forming
a separate sequence cluster), but there are clearly
some cases where we might waive the sequence clus-
ter requirement. Some newly arisen pathogens, for ex-
ample, are difficult to distinguish from closely related
populations by sequences of protein-coding genes
(for example, Bacillus anthracis versus B. cereus)
, but the ecological distinctness we observe (re-
garding virulence) is clearly relevant to the ways that
the bacteria make a living in nature. When the ecolog-
ical distinctness of such groups is not readily revers-
ible (for example, with the gain or loss of a plasmid),
it is reasonable to give a prognosis for the continued
coexistence of these populations as separate lineages
and to declare them ecotypes .
In summary, to accommodate models yielding
a one:many correspondence between ecotypes and
sequence clusters, systematists would need to con-
firm that putative ecotypes identified as sequence
clusters are ecologically distinct from one another.
To accommodate models with a many:1 correspon-
dence, we will need to confirm that each putative
ecotype is ecologically homogeneous within itself.
How will systematists determine whether putative
ecotypes are ecologically distinct? We anticipate that
in future applications of ecotype simulation, microbiol-
ogists will confirm the ecological distinctness of puta-
tive ecotypes through microhabitat distribution stud-
ies, as well as comparisons of genome content and
analyses of genome-wide gene expression and com-
prehensive metabolic phenotype . Much of the ev-
idence will draw on the existing skills of taxonomists,
who are trained intesting the capabilities of the growth
of organisms with different resources and under differ-
ent growth conditions. We recommend a new charge
for taxonomists, to move from finding diagnostic phe-
notypic characters  to using the broadest diver-
sity of techniques to assess ecological differences
. The challenge of an ecotype-based systematics
will thus be no less demanding than the ‘‘acumen of
eye and judgment’’ described by O. F. Mu ¨ller two
centuries ago, but as we discuss, the rewards should
be worth the investment.
An Ecotype-Based Systematics
We have proposed a systematics for identifying eco-
types, the fundamental units of bacterial ecology and
evolution. We recommend that these ecotypes be rec-
ognized also as the fundamental units of bacterial sys-
tematics , by issuing a name for every ecotype with
a history of coexistence and a prognosis for future co-
existence with other ecotypes. In the case when a leg-
acy, named species is found to have multiple eco-
types, we recommend that each confirmed ecotype
be given a trinomial name with an ecovar epithet, for
example, Legionella pneumophila ecovar pneumo-
phila for the ecotype containing the originally de-
scribed Philadelphia strain [2,100]. We recommend
that newly discovered clades be demarcated such
that each confirmed ecotype is named as a separate
species (with a binomial name). We believe that the
fullness of ecological diversity within the bacterial
world will be taken most seriously when each ecotype
is given its own name .
ics will allow microbiologists to focus on groups most
likely to differ in adaptations of physiology, genome
content, and gene expression. For example, an eco-
their choices of organisms to be fully sequenced.
Because comparative genomics can yield details of
ecological divergence between organisms [45,67], it
is important to choose organisms from different eco-
types for genome sequencing — then differences in
gene contentwill havea greater chance of determining
the niche. Ecotype-based systematics will help avoid
Putative Ecotype 1
n = 1(1-2)
Putative Ecotype 2
n = 1(1)
Putative Ecotype 3=
n = 1(1-2)
Putative Ecotype 4
n = 1(1)
Figure 6. Phylogeny and ecotype demarcation of the B. lichen-
iformis subclade of Bacillus.
The four putative ecotypes of this subclade differsignificantly in
their associations with the three major microhabitats of ‘Evolu-
tion Canyon’ in the Negev Desert. Microhabitat sources were
the south-facing slope (B), the north-facing slope (C), and
the canyon bottom (V); asterisks indicate strains isolated else-
where. Ecotypes were also demarcated based on the more rap-
idly evolving gyrA gene, and what appeared to be one ecotype
(Ecotype 2) for rpoB was further split into three more ecotypes
(as illustrated schematically in Figure 2D) .
choosing organisms that have a high chance of being
physiology, we should choose model organisms from
different ecotypes. For example, in the case of Legion-
ella pneumophila, three strains have been chosen as
models (Lens, Philadelphia and Paris), but Paris and
Philadelphia appear to be from the same ecotype and
indeed are very similar in their physiology [100,109].
In preparation for future epidemics, whether natural
or the result of biowarfare, we should try to discover all
pathogenic species. We could then anticipate and
prepare for future epidemics by characterizing the
disease-causing properties of each ecotype.
Biotechnologists may also take advantage of an
ecotype-based systematics. After discovering a strain
with a valuable enzyme, one could then search for ho-
mologs in each ecotype closely related to the strain.
This may allow discovery of similar enzymes with
different substrates or with optima at different
An ecotype-based systematics would simplify the
burden of industrial testing of bacterial strains for their
safety and efficacy in agricultural applications. For ex-
ample, for any named species that is heterogeneous
for characteristics of safety concern — for example,
secreted metabolites and persistence in the environ-
ment — the European Union requires that any new
strain developed for release be tested for these char-
acteristics of concern; however, individual strains
from a species known to be homogeneous for these
features need not be tested . Thus, demarcating
or at least lessen the burden of these tests.
An ecotype-based systematics will allow quantifica-
tion of the ecological diversity within a community. Re-
cent studies have demarcated and counted bacterial
taxa in a community by binning DNA sequences for
a given gene (often 16S rRNA) into sequence clusters
and then attributing each sequence cluster to an oper-
ational taxonomic unit . The operational taxo-
nomic units typically used in these studies have no
theoretical justification, but ecotype simulation offers
a quantification of ecological diversity as the number
of ecotypes present in a sample of sequences from
Finally, a systematics of ecotypes will allow us to
identify and characterize the ecologically distinct
groups of bacteria, a critical step forward in our ven-
ture to understand the myriad ecological interactions
within natural microbial communities.
We are grateful to Mark Achtman, Brendan Bohannan, William
Hanage, Alex Koeppel, Danny Krizanc, Jeffrey Lawrence,
Eviatar Nevo, Martin Polz, Johannes Sikorski, Alan Templeton,
David Ward, and Stephen Wessels for many valuable
discussions that have profoundly shaped our view of bacterial
and by the Exobiology program of NASA.
1. Gogarten,J.P., Doolittle, W.F.,and Lawrence,J.G. (2002). Prokary-
otic evolution in light of gene transfer. Mol. Biol. Evol. 19, 2226–
2.Gevers, D., Cohan, F.M., Lawrence, J.G., Spratt, B.G., Coenye,
T., Feil, E.J., Stackebrandt, E., Van de Peer, Y., Vandamme,
P., Thompson, F.L., et al. (2005). Opinion: Re-evaluating pro-
karyotic species. Nat. Rev. Microbiol. 3, 733–739.
Giovannoni, S.J., and Stingl, U. (2005). Molecular diversity and
ecology of microbial plankton. Nature 437, 343–348.
Dykhuizen, D.E. (1998). Santa Rosalia revisited: why are there so
many species of bacteria? Antonie Van Leeuwenhoek 73, 25–33.
Gans, J., Wolinsky, M., and Dunbar, J. (2005). Computational im-
provements reveal great bacterial diversity and high metal toxicity
in soil. Science 309, 1387–1390.
Pace, N.R. (1997). A molecular view of microbial diversity and the
biosphere. Science 276, 734–740.
Palys, T., Nakamura, L.K., and Cohan, F.M. (1997). Discovery and
classification of ecological diversity in the bacterial world: the
role of DNA sequence data. Int. J. Syst. Bacteriol. 47, 1145–1156.
Claridge, M.F., Dawah, H.A., and Wilson, M.R. (1997). Species: The
Units of Biodiversity (London: Chapman & Hall).
Ward, D.M., and Cohan, F.M. (2005). Microbial diversity in hot
spring cyanobacterial mats: pattern and prediction. In Geother-
mal Biology and Geochemistry in Yellowstone National Park,
W.P. Inskeep and T. McDermott, eds. (Bozeman: Thermal Biology
Institute), pp. 185–202.
Ward, D.M., Bateson, M.M., Ferris, M.J., Ku ¨hl, M., Wieland, A.,
Koeppel, A., and Cohan, F.M. (2006). Cyanobacterial ecotypes in
the microbial mat community of Mushroom Spring (Yellowstone
National Park, Wyoming) as species-like units linking microbial
community composition, structure and function. Phil. Trans. Roy.
Soc. Ser. B 361, 1997–2008.
Staley, J.T. (2006). The bacterial species dilemma and the geno-
mic-phylogenetic species concept. Philos. Trans. R. Soc. Lond.
B. Biol. Sci. 361, 1899–1909.
Hutchinson, G.E. (1968). When are species necessary? In Popula-
University Press), pp. 177–186.
Jones, D., Sackin, M.J., and Sneath, P.H. (1972). A numerical
taxonomic study of streptococci of serological group D. J. Gen.
Microbiol. 72, 439–450.
Johnson, J. (1973). Use of nucleic-acid homologies in the
Wayne, L.G.,Brenner,D.J., Colwell,R.R., Grimont,P.A.D.,Kandler,
O., Krichevsky, M.I., Moore, W.E.C., Murray, R.G.E., Stackebrandt,
E., Starr, M.P., et al. (1987). Report of the ad hoc committee on rec-
onciliation of approaches to bacterial systematics. Int. J. Syst.
Bacteriol. 37, 463–464.
uncultured bacterial phylogenetic division OP11. Appl. Environ.
Microbiol. 70, 845–849.
Stackebrandt, E., and Ebers, J. (2006). Taxonomic parameters
revisited: tarnished gold standards. Microbiol. Today 33, 152–155.
Konstantinidis, K.T., and Tiedje, J.M. (2005). Genomic insights that
advance the species definition for prokaryotes. Proc. Natl. Acad.
Sci. USA 102, 2567–2572.
Konstantinidis, K.T., Ramette, A., and Tiedje, J.M. (2006).
Toward a more robust assessment of intraspecies diversity,
using fewer genetic markers. Appl. Environ. Microbiol. 72, 7286–
Santos, S.R., and Ochman, H. (2004). Identification and phyloge-
neticsorting ofbacteriallineageswith universallyconservedgenes
and proteins. Environ. Microbiol. 6, 754–759.
Zeigler, D.R. (2003). Gene sequences useful for predicting related-
ness of whole genomes in bacteria. Int. J. Syst. Evol. Microbiol. 53,
Hanage, W.P., Fraser, C., and Spratt, B.G. (2006). Sequences,
sequence clusters and bacterial species. Phil. Trans. Roy. Soc.
Ser. B 361, 1917–1927.
Mayr, E. (1944). Systematics and the Origin of Species from the
Viewpoint of a Zoologist (New York: Columbia Univ. Press).
de Queiroz, K. (2005). Ernst Mayr and the modern concept
of species. Proc. Natl. Acad. Sci. USA 102 (Suppl. 1), 6600–
Coyne, J.A., and Orr, H.A. (2004). Speciation (Sunderland: Sinauer
De Clerck, E., Rodriguez-Diaz, M., Forsyth, G., Lebbe, L., Logan,
N.A., and DeVos, P. (2004). Polyphasic characterization of Bacillus
coagulans strains, illustrating heterogeneity within this species,
and emended description of the species. Syst. Appl. Microbiol.
28. Logan, N.A., and Berkeley, R.C. (1984). Identification of Bacillus
strains using the API system. J. Gen. Microbiol. 130(Pt 7), 1871–
microarrays for high-throughput phenotypic testing and assay of
gene function. Genome Res. 11, 1246–1255.
30.Feldgarden, M., Byrd, N., and Cohan, F.M. (2003). Gradual evolu-
tion in bacteria: evidence from Bacillus systematics. Microbiology
31.Bergthorsson, U., and Ochman, H. (1998). Distribution of chromo-
some length variation in natural isolates of Escherichia coli. Mol.
Biol. Evol. 15, 6–16.
32.Thompson, J.R., Pacocha, S., Pharino, C., Klepac-Ceraj, V., Hunt,
D.E., Benoit, J., Sarma-Rupavtarm, R., Distel, D.L., and Polz, M.F.
(2005). Genotypic diversity within a natural coastal bacterioplank-
ton population. Science 307, 1311–1313.
33. Boucher, Y., Douady, C.J., Sharma, A.K., Kamekura, M., and
Doolittle, W.F. (2004). Intragenomic heterogeneity and intergeno-
mic recombination among haloarchaeal rRNA genes. J. Bacteriol
34. Lindsay, J.A., and Holden, M.T. (2004). Staphylococcus au-
reus: superbug, super genome? Trends Microbiol. 12, 378–
35. Nelson, K.E., Fouts, D.E., Mongodin, E.F., Ravel, J., DeBoy,
R.T., Kolonay, J.F., Rasko, D.A., Angiuoli, S.V., Gill, S.R.,
Paulsen, I.T., et al. (2004). Whole genome comparisons of
serotype 4b and 1/2a strains of the food-borne pathogen
Listeria monocytogenes reveal new insights into the core
genome components of this species. Nucleic Acids Res. 32,
36.Goris, J., Konstantinidis, K.T., Klappenbach, J.A., Coenye, T.,
Vandamme, P., and Tiedje, J.M. (2007). DNA-DNA hybridization
values and their relationship to whole-genome sequence similari-
ties. Int. J. Syst. Evol. Microbiol. 57, 81–91.
bial Diversity and Bioprospecting, A.T. Bull, ed. (Washington, D.C.:
American Society for Microbiology Press), pp. 40–48.
38.Cohan, F.M. (2006). Toward a conceptual and operational union of
bacterial systematics, ecology, and evolution. Proc. Roy. Soc.
Lond. Series B 361, 1985–1996.
39. Lopez-Lopez, A., Bartual, S.G., Stal, L., Onyshchenko, O., and
Rodriguez-Valera, F. (2005). Genetic analysis of housekeeping
genes reveals a deep-sea ecotype of Alteromonas macleodii in
the Mediterranean Sea. Environ. Microbiol. 7, 649–659.
40. Schloter, M., Lebuhn, M., Heulin, T., and Hartmann, A. (2000). Ecol-
41.Smith, N.H., Kremer, K., Inwald, J., Dale, J., Driscoll, J.R., Gordon,
S.V., van Soolingen, D., Hewinson, R.G., and Smith, J.M. (2006).
Ecotypes of the Mycobacterium tuberculosis complex. J. Theor.
Biol. 239, 220–225.
42.Ho ¨fle, M.G., Ziemke, F., and Brettar, I. (2000). Niche differentiation
of Shewanellaputrefacienspopulations from the Balticas revealed
by molecular and metabolic fingerprinting. In Microbial Bio-
systems: New Frontiers, Proc. Eighth Int. Symp. Micr. Ecol., C.R.
Bell, M. Brylinsky, and P. Johnson-Green, eds. (Halifax: Atlantic
Canada Society for Microbial Ecology), pp. 135–142.
43. Sikorski, J., and Nevo, E. (2005). Adaptation and incipient sympat-
ric speciation of Bacillus simplex under microclimatic contrast at
‘‘Evolution Canyons’’ I and II Israel. Proc. Natl. Acad. Sci. USA
44. Cohan, F.M., Perry, E., Koeppel, A., Krizanc, D., Ward, D.M., Bate-
son, M., Rooney, A., Sikorski, J., Nevo, E., and Ratcliff, R.M. (2007).
Identifying the fundamental units of bacterial diversity. (in
45.Coleman, M.L., Sullivan, M.B., Martiny, A.C., Steglich, C., Barry, K.,
Delong, E.F., and Chisholm, S.W. (2006). Genomic islands and the
ecology and evolution of Prochlorococcus. Science 311, 1768–
46. Enright, M.C., Spratt, B.G., Kalia, A., Cross, J.H., and Bessen, D.E.
(2001). Multilocus sequence typing of Streptococcus pyogenes
and the relationships between emm type and clone. Infect. Immun.
J.H., and Bourtzis, K. (2006). Toward a Wolbachia multilocus
sequence typing system: discrimination of Wolbachia strains pres-
ent in Drosophila species. Curr. Microbiol. 53, 388–395.
48.Stackebrandt, E., Frederiksen, W., Garrity, G.M., Grimont, P.A.,
Kampfer, P., Maiden, M.C., Nesme, X., Rossello-Mora, R., Swings,
re-evaluation of the species definition in bacteriology. Int. J. Syst.
Evol. Microbiol. 52, 1043–1047.
49.Helgason, E., Okstad, O.A., Caugant, D.A., Johansen, H.A.,
Fouet, A., Mock, M., Hegna, I., and Kolsto, A.B. (2000). Bacillus
anthracis, Bacillus cereus, and Bacillus thuringiensis–one spe-
cies on the basis of genetic evidence. Appl. Environ. Microbiol.
50.Cohan, F.M. (2002). Population structure and clonality of bacteria.
In Encyclopedia of Evolution, Volume 1, M. Pagel, ed. (New York:
Oxford University Press), pp. 161–163.
51.Feil, E.J., Maiden, M.C., Achtman, M., and Spratt, B.G. (1999). The
relative contributions of recombination and mutation to the diver-
gence of clones of Neisseria meningitidis. Mol. Biol. Evol. 16,
How clonal are bacteria?Proc. Natl.Acad. Sci. USA 90, 4384–4388.
53.Falush, D., Kraft, C., Taylor, N.S., Correa, P., Fox, J.G., Achtman,
M., and Suerbaum, S. (2001). Recombination and mutation during
long-term gastric colonization by Helicobacter pylori: estimates
of clock rates, recombination size, and minimal age. Proc. Natl.
Acad. Sci. USA 98, 15056–15061.
54.Cohan, F.M. (1994). The effects of rare but promiscuous genetic
exchange on evolutionary divergence in prokaryotes. Am. Nat.
55.Koch, A.L. (1974). The pertinence of the periodic selection
phenomenon to prokaryote evolution. Genetics 77, 127–142.
56.Levin, B.R. (1981). Periodic selection, infectious gene exchange
and the genetic structure of E. coli populations. Genetics 99, 1–23.
57.Linz, B., Balloux, F., Moodley, Y., Manica, A., Liu, H., Roumagnac,
P., Falush, D., Stamer, C., Prugnolle, F., van der Merwe, S.W., et al.
(2007). An East African origin for the intimate association between
humans and Helicobacter pylori. Nature 445, 915–918.
58.Kimura, M. (1983). The Neutral Theory of Molecular Evolution
(Cambridge: Cambridge Univ. Press).
59.Cohan, F.M. (2002). What are bacterial species? Annu. Rev. Micro-
biol. 56, 457–487.
species? Theor. Popul. Biol. 61, 449–460.
61.Godreuil, S., Cohan, F., Shah, H., and Tibayrenc, M. (2005). Which
species concept for pathogenic bacteria? An E-Debate. Infect.
Genet. Evol. 5, 375–387.
62.Cohan, F.M. (1995). Does recombination constrain neutral diver-
gence among bacterial taxa? Evolution 49, 164–175.
63.Guttman, D.S., and Dykhuizen, D.E. (1994). Detecting selective
sweeps in naturally occurring Escherichia coli. Genetics 138,
64. Cohan, F.M. (2005). Periodic selection and ecological diversity in
bacteria. In Selective Sweep, D. Nurminsky, ed. (Georgetown,
Texas: Landes Bioscience), pp. 78–93.
66. Majewski, J., and Cohan, F.M. (1999). Adapt globally, act locally:
the effect of selective sweeps on bacterial sequence diversity.
Genetics 152, 1459–1474.
67.Bhaya, D., Grossman, A.R., Steunou, A.S., Khuri, N., Cohan, F.M.,
Hamamura, N., Melendrez, M., Bateson, M.M., Ward, D.M., and
Heidelberg, J.F. (2007). Genomic, metagenomic and functional
analyses of cyanobacteria from hot-spring microbial mats reveal
an unexpected diversity in nutrient utilization strategies. ISME J.
68.Roumagnac, P., Weill, F.X., Dolecek, C., Baker, S., Brisse, S.,
Chinh, N.T., Le, T.A., Acosta, C.J., Farrar, J., Dougan, G., et al.
(2006). Evolutionary history of Salmonella typhi. Science 314,
69. Rainey, P.B., and Travisano, M. (1998). Adaptive radiation in a
heterogeneous environment. Nature 394, 69–72.
70.Rozen, D.E., and Lenski, R.E. (2000). Long-term experimental
phism. Am. Nat. 155, 24–35.
Geographical isolation in hot spring cyanobacteria. Environ.
Microbiol. 5, 650–659.
72.Beijerinck, M.W. (1913). Jaarboek van de Koninklijke Akademie v.
Wetenschappen (Amsterdam: Mu ¨ller).
73.Finlay, B.J. (2002). Global dispersal of free-living microbial eukary-
ote species. Science 296, 1061–1063.
74.Roberts, M.S., and Cohan, F.M. (1995). Recombination and migra-
tion rates in natural populations of Bacillus subtilis and Bacillus
mojavensis. Evolution 49, 1081–1094.
75.Staley, J.T., and Gosink, J.J. (1999). Poles apart: biodiversity and
biogeography of sea ice bacteria. Annu. Rev. Microbiol. 53, 189–
76. Whitaker, R.J., Grogan, D.W., and Taylor, J.W. (2003). Geographic Download full-text
archaea. Science 301, 976–978.
Papke, P.T., and Ward, D.M. (2004). The importance of physi-
cal isolation in microbial evolution. FEMS Microbiol. Ecol. 48,
Achtman, M., Morelli, G., Zhu, P., Wirth, T., Diehl, I., Kusecek, B.,
Vogler, A.J., Wagner, D.M., Allender, C.J., Easterday, W.R., et al.
(2004). Microevolution and history of the plague bacillus, Yersinia
pestis. Proc. Natl. Acad. Sci. USA 101, 17837–17842.
Gutacker, M.M., Mathema, B., Soini, H., Shashkina, E., Kreiswirth,
B.N., Graviss, E.A., and Musser, J.M. (2006). Single-nucleotide
polymorphism-based population genetic analysis of Mycobacte-
rium tuberculosis strains from 4 geographic sites. J. Infect. Dis.
Keim, P., and Smith, K.L. (2002). Bacillus anthracis evolution and
epidemiology. Curr. Top. Microbiol. Immunol. 271, 21–32.
Palys, T., Berger, E., Mitrica, I., Nakamura, L.K., and Cohan, F.M.
(2000). Protein-coding genes asmolecularmarkers for ecologically
distinct populations: the case of two Bacillus species. Int. J. Syst.
Evol. Microbiol. 50(Pt 3), 1021–1028.
MaynardSmith,J.,andSzathma ´ry,E.(1995).The MajorTransitions
in Evolution (Oxford: Oxford Univ. Press).
Wagner, A. (2005). Robustness and Evolvability in Living Systems
(Princeton: Princeton Univ. Press).
Cohan, F.M., King, E.C., and Zawadzki, P. (1994). Amelioration of
the deleterious pleiotropic effects of an adaptive mutation in Bacil-
lus subtilis. Evolution 48, 81–95.
Levin, B.R., Perrot, V., and Walker, N. (2000). Compensatory muta-
tions, antibiotic resistance and the population genetics of adaptive
evolution in bacteria. Genetics 154, 985–997.
Vermeij, G.J. (1973). Adaptation, versatility, and evolution. Syst.
Zool. 22, 466–477.
Stover, C.K., Pham, X.Q., Erwin, A.L., Mizoguchi, S.D., Warrener,
P., Hickey, M.J., Brinkman, F.S., Hufnagle, W.O., Kowalik, D.J.,
Lagrou, M., et al. (2000). Complete genome sequence of Pseudo-
monas aeruginosa PA01, an opportunistic pathogen. Nature 406,
in Pseudomonas. Am. Nat. 160, 569–581.
genomics. In Microbial Genomes, C.M. Fraser, T.D. Read, and K.E.
Nelson, eds. (Totowa, New Jersey: Humana), pp. 175–194.
Cohan, F.M. (1996). The role of genetic exchange in bacterial
evolution. ASM News 62, 631–636.
Cohan, F.M., Roberts, M.S., and King, E.C. (1991). The potential for
genetic exchange by transformation within a natural population of
Bacillus subtilis. Evolution 45, 1393–1421.
Turner, K.M., and Feil, E.J. (2007). The secret life of the multilocus
sequence type. Int. J. Antimicrob. Agents 29, 129–135.
Majewski, J., and Cohan, F.M. (1999). DNA sequence similarity re-
quirements for interspecific recombination in Bacillus. Genetics
Majewski, J., Zawadzki, P., Pickerill, P., Cohan, F.M., and Dowson,
C.G. (2000). Barriers to genetic exchange between bacterial spe-
cies: Streptococcus pneumoniae transformation. J. Bacteriol.
keys to speciation: DNA polymorphism and the control of genetic
exchange in enterobacteria. Proc. Natl. Acad. Sci. USA 94, 9763–
Cohan, F.M. (2002). Sexual isolation and speciation in bacteria.
Genetica 116, 359–370.
Templeton, A.R., Maskas, S.D., and Cruzan, M.B. (2000). Gene
trees: a powerful tool for exploring the evolutionary biology of
species and speciation. Plant. Species Biol. 15, 211–222.
Dobrindt, U. (2005). (Patho-)Genomics of Escherichia coli. Int. J.
Med. Microbiol. 295, 357–371.
man and avian pathogens. Curr. Opin. Microbiol. 9, 28–32.
Cohan, F.M., Koeppel, A., and Krizanc, D. (2006). Sequence-based
discovery of ecological diversity within Legionella. In Legionella:
State of the Art 30 Years after Its Recognition, N.P. Cianciotto, Y.
Abu Kwaik, P.H. Edelstein, B.S. Fields, D.F. Geary, T.G. Harrison,
ington, D.C.: ASM Press), pp. 367–376.
Martin, A.P. (2002). Phylogenetic approaches for describing and
comparing the diversity of microbial communities. Appl. Environ.
Microbiol. 68, 3673–3682.
Acinas,S.G.,Klepac-Ceraj,V.,Hunt, D.E.,Pharino,C.,Ceraj, I.,Dis-
tel, D.L., and Polz, M.F. (2004). Fine-scale phylogenetic architec-
ture of a complex bacterial community. Nature 430, 551–554.
spatiotemporal distribution of soil microfungi in ‘‘Evolution
Canyon’’ III, Nahal Shaharut, extreme southern Negev Desert,
Israel. Biol. J. Linn. Soc. 90, 263–277.
Ferris, M.J., Ku ¨hl, M., Wieland, A., and Ward, D.M. (2003). Cyano-
bacterial ecotypes in different optical microenvironments of a 68
ternal transcribed spacer region variation. Appl. Environ Microbiol.
Gordon, D.M., and Cowling, A. (2003). The distribution and genetic
structure of Escherichia coli in Australian vertebrates: host and
geographic effects. Microbiology 149, 3575–3586.
ing bacterial populations: a post-genomic reformation. Nat. Rev.
Genet. 3, 462–473.
Lawrence, J.G., and Hendrickson, H. (2003). Lateral gene transfer:
when will adolescence end? Mol. Microbiol. 50, 739–749.
Vandamme, P., Pot, B., Gillis, M., de Vos, P., Kersters, K., and
Swings, J. (1996). Polyphasic taxonomy, a consensus approach
to bacterial systematics. Microbiol. Rev. 60, 407–438.
Bru ¨ggemann, H., Hagman, A., Coppe ´e, J., Dillies, M., Heuner, K.,
Steinert, M., Gouyette, C., and Buchrieser, C. (2006). Bringing the
genome of L. pneumophila to life: Gene expression patterns of
the replicative and transmissive phase in vitro. In Legionella, N.P.
Cianciotto, Y. Abu Kwaik, P.H. Edelstein, B.S. Fields, D.F. Geary,
T.G. Harrison, C. Joseph, R.M. Ratcliff, J.E. Stout, and M.S. Swan-
son, eds. (Washington, D.C.: ASM Press).
European Commission (2005). Guideline developed within the
Standing Committee on the Food Chain and Animal Health on the
taxonomic level of micro-organisms to be included in Annex I to
Bohannan, B.J., and Hughes, J. (2003). New approaches to analyz-
ing microbial biodiversity data. Curr. Opin. Microbiol. 6, 282–287.