Content uploaded by Michael Behe
Author content
All content in this area was uploaded by Michael Behe on Aug 22, 2014
Content may be subject to copyright.
EXPERIMENTAL EVOLUTION, LOSS-OF-FUNCTION MUTATIONS,
AND “THE FIRST RULE OF ADAPTIVE EVOLUTION”
Michael J. Behe
Department of Biological Sciences, Lehigh University, Bethlehem, Pennsylvania 18015 USA
e-mail: mjb1@lehigh.edu
keywords
experimental evolution, adaptation, mutation, loss of function, malaria,
Yersinia pestis
abstract
Adaptive evolution can cause a species to gain, lose, or modify a function; therefore, it is of
basic interest to determine whether any of these modes dominates the evolutionary process under
particular circumstances. Because mutation occurs at the molecular level, it is necessary to
examine the molecular changes produced by the underlying mutation in order to assess whether
a given adaptation is best considered as a gain, loss, or modification of function. Although that
was once impossible, the advance of molecular biology in the past half century has made it
feasible. In this paper, I review molecular changes underlying some adaptations, with a partic-
ular emphasis on evolutionary experiments with microbes conducted over the past four decades.
I show that by far the most common adaptive changes seen in those examples are due to the loss
or modification of a pre-existing molecular function, and I discuss the possible reasons for the
prominence of such mutations.
Adaptation by Gain, Loss, or
Modification of Function
INTHE ORIGIN OF SPECIES, Darwin
(1859) emphasized the relentlessness of
natural selection:
[N]atural selection is daily and hourly scru-
tinising, throughout the world, every varia-
tion, even the slightest; rejecting that
which is bad, preserving and adding up all
that is good; silently and insensibly work-
ing, whenever and wherever opportunity
offers, at the improvement of each organic
being in relation to its organic and inor-
ganic conditions of life. (p. 84)
Yet he realized that the changes that were
selected to adapt an organism to its envi-
ronment did not have to be ones that con-
ferred upon it a new ability, such as sight or
flight. For instance, while observing some
barnacles, Darwin discovered unexpected
cases of the gross simplification of an or-
ganism (Stott 2003):
The male is as transparent as glass...In
the lower part we have an eye, & great testis
& vesicula seminalis: in the capitulum we
have nothing but a tremendously long pe-
nis coiled up & which can be exserted.
There is no mouth no stomach no cirri, no
The Quarterly Review of Biology, December 2010, Vol. 85, No. 4
Copyright © 2010 by The University of Chicago Press. All rights reserved.
0033-5770/2010/8504-0002$15.00
Volume 85, No. 4December 2010THE QUARTERLY REVIEW OF BIOLOGY
419
proper thorax! The whole animal is re-
duced to an envelope...containing the
testes, vesicula, and penis. (p. 213)
Adaptive evolution can just as easily lead to
the loss of a functional feature in a species
lineage as to the gain of one. Over the
course of evolutionary history, snakes have
lost legs, cavefish have lost vision, and the
parasitic bacterium Mycoplasma genitalium
has lost its ability to live independently in
the wild, all in an effort to become better
adapted to their environments. Whatever
variation that sufficiently aids a particular
species at a particular moment in a partic-
ular environment—be it the gain or loss of
a feature, or the simple modification of
one—may be selected. Since species can
evolve to gain, lose, or modify functional
features, it is of basic interest to determine
whether any of these tends to dominate
adaptations whose underlying molecular
bases are ascertainable. Here, I survey the
results of evolutionary laboratory experi-
ments on microbes that have been con-
ducted over the past four decades. Such
experiments exercise the greatest control
over environmental variables, and they
yield our most extensively characterized re-
sults at the molecular level.
distinctions among adaptive
mutations
Adaptation is often viewed from two dis-
tinct aspects: the phenotypic aspect and
the molecular aspect. In order to avoid
confusion concerning how adaptive evolu-
tion proceeds, these two aspects must be
kept separate. A phenotypic adaptation
might have any number of underlying mo-
lecular bases, resulting from either the
gain, loss, or modification of a molecular
feature. As an example, consider a bacte-
rium that evolves virulence in a mamma-
lian host. If the adaptive phenotype is
deemed to be the gain of pathogenicity,
then that could arise in several ways. For
instance: 1) a gene for a bacterial protein
that is the target of the mammalian im-
mune response might be deleted, which
would be the loss of a functional molecular
feature of the bacterium; 2) a gene for a
protein that is the target of the mammalian
immune response might sustain a missense
mutation, causing an amino acid substitu-
tion at a recognized epitope of the protein,
thereby decreasing or eliminating the im-
mune response while preserving protein
function, which could be considered a sim-
ple modification of a pre-existing mo-
lecular element; 3) the bacterium might
acquire a gene or gene cluster by horizon-
tal transfer that leads to expression of a
new surface feature that hides the epitope
from the mammalian immune system,
which would be a gain of a functional mo-
lecular feature. The focus of this review is
on the gain, loss, or modification of func-
tional molecular features underlying adap-
tation, no matter whether the phenotypic
manifestation strikes us as a loss or gain of
a property.
There are, of course, myriad functional
molecular features of a cell, including those
of all the major biomolecular categories: nu-
cleic acids, proteins, polysaccharides, lipids,
small metabolites, and so forth. However, the
functional molecular features of only two cat-
egories—nucleic acids and proteins—are di-
rectly coded by the genome, and thus only
these features can be directly affected by mu-
tation; all other biological features are only
affected indirectly by mutations, through
their effects on nucleic acids and proteins. In
this review, I focus on adaptive evolution by
gain, loss, or modification of what I term
Functional Coded elemenTs (FCTs). An FCT is a
discrete but not necessarily contiguous re-
gion of a gene that, by means of its nucleo-
tide sequence, influences the production,
processing, or biological activity of a particu-
lar nucleic acid or protein, or its specific
binding to another molecule. Examples of
FCTs are: promoters; enhancers; insulators;
Shine-Dalgarno sequences; tRNA genes;
miRNA genes; protein coding sequences;
organellar targeting- or localization-
signals; intron/extron splice sites; codons
specifying the binding site of a protein for
another molecule (such as its substrate,
another protein, or a small allosteric regu-
lator); codons specifying a processing site
of a protein (such as a cleavage, myristoyl-
ation, or phosphorylation site); polyade-
420 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
nylation signals; and transcription and
translation termination signals. Examples
of general nucleic acid features that are
not FCTs include base composition, strand
asymmetry, distance between coded ele-
ments, and the like. They may enter into
consideration indirectly only if they cause
an FCT to be lost or gained—for example,
if altering the base composition also elim-
inates the activity of an enhancer.
With these considerations in mind, I
classify adaptive mutations as belonging to
one of three modes:
1) A “loss-of-FCT” adaptive mutation is a
mutation that leads to the effective loss of
the function of a specific, pre-existing,
coded element, while adapting an organ-
ism to its environment. The loss of the
ability of a frame-shifted gene to produce a
functional product, of an altered promoter
to bind a transcription factor, or of a mu-
tated protein to bind its former ligand, are
examples of loss-of-FCT mutations.
2) A “gain-of-FCT” adaptive mutation is
a mutation that produces a specific, new,
functional coded element while adapting
an organism to its environment. The con-
struction by mutation of a new promoter,
intron/exon splice site, or protein process-
ing site are gain-of-FCT mutations. Also in-
cluded in this category is the divergence by
mutation of the activity of a previously du-
plicated coded element.
3) A “modification-of-function” adap-
tive mutation is a mutation whose defining
property is negative—while adapting an or-
ganism to an environment, it does not lead
to the loss or gain of a specific FCT. This
definition is intended to be broad enough
to act as a “catch-all” for anything that falls
outside the first two modes. It includes
point mutations as well as other mutations
that have a quantitative effect on a pre-
existing FCT, increasing or decreasing its
strength, for instance, or shifting its activity
somewhat (such as allowing a protein to
bind a structurally-related ligand at the
same site as its normal substrate), but
without effectively eliminating it. The
category “modification-of-function” also
includes the simple duplication of fea-
tures, without divergence of activity, such as
a gene or regulatory region, since this pro-
cess does not by itself produce a new func-
tional element, although it may, like other
events classified here as “modification-of-
function,” be a step to future gain-of-FCT
events. It also includes mutations that may
act by more amorphous means, such as rear-
ranging gene order, or changing the dis-
tance or orientation between two interacting
elements (e.g., the deletion of a stretch of
DNA that brings two interacting transcrip-
tion factors closer, or that affects their orien-
tation with respect to each other). This
category is dubbed “modification-of-function”
rather than “modification-of-FCT ” in order
to emphasize that the mutation may not nec-
essarily exert its influence through alteration
of a discrete coded element, as in the above
amorphous examples.
In order to fall under any of the above
three categories, a mutation must be adap-
tive. Therefore, if, for instance, a mutation
causes a transcription factor binding site to
be formed or lost, but the mutation is ei-
ther neutral or deleterious, it is excluded
from consideration. Although assignment
of an adaptive mutation to one of the
modes above may occasionally be ambigu-
ous, the assignment is straightforward in
the great majority of cases in which molec-
ular changes are elucidated, as we shall see
below.
illustrations of modes of adaptation
To better our understanding of the
kinds of mutations that fall into these three
modes, it is useful to look at illustrations
from outside the realm of the experimen-
tal evolution of microbes. Consider, for in-
stance, human mutations that have been
selected in the past 10,000 years under the
pressure of the malarial parasite Plasmo-
dium falciparum. The reason for using such
examples as these is that, because of its
threat to human health, the interaction of
malaria with humans has been very well-
studied at the molecular level.
If a person of typical northern European
descent visited a malarious region of the
world and contracted the disease, she
would likely view the resistance to malaria
among many of the native peoples as an
December 2010 421EXPERIMENTAL EVOLUTION
unquestionably positive trait: the ability to
withstand a debilitating disease. However,
because of the detailed knowledge of the
molecular bases of malaria resistance that
has been accumulated over past decades,
that is not the way I would categorize it
here. Many persons indigenous to malari-
ous regions are resistant, but they have ac-
quired their resistance through different
molecular mechanisms (Table 1). Al-
though the mutations are all adaptive, dis-
tinctions may be made among them. The
most well-known population of resistant
persons is heterozygous for sickle hemo-
globin. In one of the two genes that they
inherited from their parents for the

chain of hemoglobin, the sixth codon des-
ignates valine rather than the glutamic
acid that the other

chain gene codes.
When the red blood cell is invaded by the
malarial parasite, the mutant, hydrophobic
valine residue allows hemoglobin mole-
cules to aggregate with each other into
long microtubular structures. This poly-
merization is associated with a measure of
resistance to malaria (Friedman 1978), al-
though the exact mechanism for resistance
in vivo is still a matter of debate (Verra et
al. 2007). This is an example of an adaptive
gain-of-FCT mutation because a codon
helping to specify a new, albeit weak—with
Kd⬃1 mM (Behe and Englander 1979)—
protein binding site has appeared.
Another population indigenous to some
malarial regions has a different point mu-
tation in their hemoglobin. In this in-
stance, the sixth codon of the

chain has
mutated from glutamic acid to lysine. Al-
though the altered hemoglobin (HbC)
does not aggregate as sickle hemoglobin
does, it confers resistance to malaria for
reasons that are unclear. Because appar-
ently no new, discrete, coded molecular
feature has been developed, this is catego-
rized as a “modification-of-function” adap-
tive mutation; no FCT has been lost or
gained.
A third population of natives in some
malarial regions is comparatively healthy
due to a different kind of mutation. They
have a condition called thalassemia, which,
like sickle hemoglobin and HbC, also con-
fers a measure of resistance to malaria. In a
thalassemic person, however, one of the
␣
or

chain hemoglobin genes that is inher-
ited from a parent is nonfunctional, for
any one of a variety of reasons. In some
thalassemic individuals, a whole gene has
been deleted, whereas, in others, a frame-
shift mutation has caused the gene to
produce a nonfunctional protein. In still
others, the genetic control region near
TABLE 1
Human mutations selected for resistance to malaria
Clinical phenotype Underlying mutation
Mode of
adaptation Reference
Hb S

6 glu val G Cavalli-Sforza et al. (1994)
Hb C

6 glu lys M Modiano et al. (2001)
Hb E

26 glu lys M Weatherall et al. (2002)
Hereditary persistence
of fetal hemoglobin
Deletion/point mutations in control regions of
hemoglobin gamma chain gene
L Forget (1998)
Thalassemia Deletion/point mutations in either
␣
or

hemoglobin genes
L Flint et al. (1986)
G6PD deficiency Point mutations, deletions of G6PD gene L Ruwende et al. (1995)
Loss of Duffy antigen in
red blood cells
Point mutation switching off production of Duffy
antigen specifically in red blood cells
L Pogo and Chaudhuri (2000)
Band 3 protein
deficiency
Deletion L Kennedy (2002)
Abbreviations:
G - adaptive gain of functional coded element
L - adaptive loss of functional coded element
M - adaptive modification of function
422 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
one of the globin genes has suffered a
change that makes it nonfunctional, and so
it produces no protein. Any of these would
be an example of a loss-of-FCT mutation.
Other populations are also malaria-
resistant due to loss-of-FCT mutations,
but in genes other than the globin genes.
Take, for instance, the gene that codes
for glucose-6-phosphate dehydrogenase
(G6PD). Any one of hundreds of separate
known mutations in the coding and con-
trol regions of the gene has eliminated the
functional enzyme, or has significantly de-
creased its catalytic activity. Other loss-of-
FCT mutations include the deletion of a
gene for Band 3 protein, and the loss of a
promoter site for expression of Duffy anti-
gen in red blood cells.
In summary, human genetic lines in ma-
larious regions of the world have adapted
to an environment that includes the malar-
ial parasite, thereby giving rise to a number
of molecular changes (Table 1). They have
done so by different modes, which can be
classified as adaptive mutations involving
loss- or gain-of-FCT, or modification-of-
function. As we shall see below, microbes
in laboratory evolution experiments also
adapt in these ways.
Evolution Experiments with Bacteria
Rather than performing an exhaus-
tive review of laboratory evolution exper-
iments, I will use several recent, influential
reviews as jumping-off points for my survey
of the relevant literature, starting with
“Evolution experiments with microorgan-
isms: the dynamics and genetic bases of
adaptation” (Elena and Lenski 2003). In
their review, Elena and Lenski (2003) were
particularly interested in laboratory studies
that were “open-ended and long-term,”
such as they themselves had conducted.
They eschewed reviewing experiments that
“targeted specific, often new, metabolic
functions, even though this work is of great
interest” (Elena and Lenski 2003:458). In-
stead, for that topic, they referred readers
to two previous reviews. Interestingly, even
though their own review was quite up-to-
date, the two reviews on specific metabolic
functions that they cited—“Evolution of
new metabolic functions in laboratory or-
ganisms” (Hall 1983) and Microorganisms as
Model Systems for Studying Evolution (Mort-
lock 1984a)—were both about twenty years
old at the time, thus indicating a lack of
recent work in that subfield.
metabolism of unusual compounds
I will begin by examining the early
work reported in Microorganisms as Model
Systems for Studying Evolution, edited by
Mortlock (1984a) (Table 2). Four of the
first five chapters concern several species
of bacteria in which cultures are evolved
to metabolize an unusual chemical food
source that the unevolved species cannot
grow upon. In general, in each of the
investigations, the first mutation selected
is either one that inactivates a repressor
gene, thus turning the gene it regulates
into a constitutively-expressed gene, or,
less often, one that increases the copy
number of a particular metabolic gene.
(The identities of some mutations in this
early work were inferred by the authors,
not demonstrated.) The amplified or de-
regulated gene is one that typically codes
for an enzyme that metabolizes a related
compound and that has a limited ability
to metabolize the novel substrate. The
increase in the concentration of enzyme
brought about by deregulation or gene
amplification allows it to metabolize
enough of the novel material to permit
the mutated bacterium to grow, albeit
sometimes slowly. Similar results were re-
ported by Clarke (1984) in Chapter 7,
which focuses on mutations in the meta-
bolic pathway of Pseudomonas that metab-
olizes amides. During the course of many
experiments, she obtained a variety of
point mutations in the regulatory protein
for the amidase gene and in the pro-
moter region for the amidase gene, as
well as in the gene itself.
Constitutive mutations are loss-of-FCT mu-
tations because the organism has lost the
function of a coded feature that controls syn-
thesis of the enzyme. Once the bacterium
has mutated, allowing it to utilize a novel
substrate sufficiently enough to grow, fur-
ther selection can often improve the growth
December 2010 423EXPERIMENTAL EVOLUTION
TABLE 2
Work reported in Microorganisms as Model Systems for Studying Evolution (Mortlock 1984)
Chapter Laboratory phenotype Underlying mutation
Mode of
adaptation Reference
1 Xyltitol catabolism by Klebsiella Inactivation of a ribitol
dehydrogenase repressor gene
L Mortlock
(1984c)
l-arabitol catabolism by Klebsiella Inactivation of a ribitol
dehydrogenase repressor gene
L
Increased catabolism of xylitol Apparent point mutations of ribitol
dehydrogenase
M
Increased catabolism of xylitol Apparent alteration of ribitol
dehydrogenase gene promoter
M
Increased catabolism of xylitol Apparent increase in copy number
of ribitol dehydrogenase gene
M
Xylitol transport across
membrane
Deregulation of d-arabitol operon L
2 Increased xylitol catabolism by
Klebsiella
Apparent increase in copy number
of ribitol dehydrogenase gene
M Hartley
(1984)
Increased xylitol catabolism by
Klebsiella
Point mutation in ribitol
dehydrogenase protein
M
Increased xylitol catabolism by
E. coli
Transfer of Klebsiella rdh operon to
E. coli plus gene amplification
M
Increased xylitol catabolism by
E. coli
Transfer of Klebsiella rdh operon to
E. coli plus point mutation
M
4d-lyxose isomerase activity by
Klebsiella
Uncertain; perhaps due to
constitutive d-mannose
isomerase
L Mortlock
(1984b)
d-lyxose isomerase activity in E.
coli
Uncertain; perhaps due to
constitutive d-mannose
isomerase
L
d-arabinose isomerase activity by
Klebsiella
Constitutive l-fucose isomerase L
d-arabinose isomerase activity in
E. coli
Point mutation-induced l-fucose
pathway
M
5 Growth on l-1,2-propanediol of
E. coli
Possible alteration of
oxidoreductase promoter
M Lin and Wu
(1984)
Growth on d-arabinose of E. coli Either constitutive expression of
fucose genes, or dual control of
regulator protein by l-fucose
and d-arabinose
L
G
Growth on d-arabitol of E. coli Constitutive production of
oxidoreductase in already-
mutated strain 3
L
Growth on xylitol of E. coli Constitutive production of d-xylose
pathway
L
Growth on ethylene glycol of E.
coli
Constitutive use of l-1,2-
propanediol pathway
L
6 Growth of E. coli on lactose
(after deletion of lacZ)
Point mutation in cryptic

-
galactosidase ebgo; also rendered
constitutive by frameshifts and
IS-disruption of ebg repressor
M
L
Hall (1984)
Growth on lactulose Second point mutation in ebgoM
Increased growth on lactulose Point mutation in ebg repressor
allowing lactulose as inducer
M
Growth on lactobionate Third point mutation in ebgoM
continued
424 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
rate through point mutations in the gene for
the overproduced enzyme. These additional
mutations are modification-of-function mu-
tations, because no FCT is gained or lost.
Increased-copy-number mutants are also
modification-of-function mutants, because
there is no evidence that any of the extra
copies have diversified, which would then
place it in the class of gain-of-FCT muta-
tions.
An interesting variation on this pattern
is reported in Chapter 5 by Lin and Wu
(1984). Mutants of E. coli and S. typhi-
murium (S. enterica var Typhimurium) gain
the ability to metabolize the unusual sub-
strate d-arabinose by altering the specificity
of a regulatory protein. Normally, the en-
zyme fucose isomerase is induced in these
bacteria when some fucose enters the cell
and binds to a positive regulatory protein,
which then turns on the gene for the isom-
erase. The regulatory protein of some mu-
tants, however, responds to both fucose
and d-arabinose. It turns out that the un-
usual substrate d-arabinose can be metab-
olized by enzymes of the fucose pathway,
and, because the protein has apparently
gained an additional binding site for the
novel substrate, the mutation is classed as
gain-of-FCT.
recruitment of cryptic proteins
The evolution of the ability of the bacte-
ria described above to metabolize novel
compounds depends on mutations in
known proteins of known metabolic path-
ways. Several chapters in Microorganisms as
Model Systems for Studying Evolution (Mort-
lock 1984a), however, describe examples
where previously unsuspected, cryptic pro-
teins were recruited to take over the func-
tion of a known protein whose gene had
been deliberately deleted.
In Chapter 9, Kemper (1984) describes
the ability of the product of a gene of Sal-
monella typhimurium, which he called newD,
to replace a protein coded by the leuD gene
of the leucine biosynthesis pathway. NewD
ordinarily binds tightly to a larger protein
subunit that Kemper dubbed SupQ, just as
LeuD binds to a larger subunit, LeuC. In
order to allow NewD to restore leucine
biosynthetic function when leuD is deliber-
ately deleted, the gene for SupQ also has to
be deliberately knocked out. Kemper ar-
gues that supQ/newD was not simply a gene
duplication of leuC/leuD (the genes of the
leucine pathway) (Stover et al. 1988). How-
ever, at that time, sequence data was more
difficult to come by than it is today. It is
now known that the S. typhimurium genome
contains a homolog of LeuD with 38%
TABLE 2
Continued
Chapter Laboratory phenotype Underlying mutation
Mode of
adaptation Reference
7 Growth in presence of
butyramide of P. aeruginosa
Possible point mutation in
regulator protein gene amiR
M Clarke
(1984)
Growth in presence of
succinate/lactamide
Possible point mutation in
regulator protein gene amiR plus
catabolite repression mutant
M
Increased production of
amidase
Mutation in promoter M
Growth on butyramide,
valeramide, acetanilide
Point mutations in amidase M
8 Resistance of yeast to allyl
alcohol
Point mutations in alcohol
dehydrogenase
M Wills (1984)
9 Rescue of leuD deletion Availability of cryptic leuD-like gene M Kemper
(1984)
Abbreviations:
L - adaptive loss of functional coded element
M - adaptive modification of function
December 2010 425EXPERIMENTAL EVOLUTION
amino acid sequence identity. It seems
likely that the protein that Kemper discov-
ered that could substitute for LeuD was
this homolog. The S. typhimurium genome
also contains a homolog of LeuC with 52%
amino acid sequence identity; it seems
likely that SupQ was this homolog. No such
LeuC or LeuD homologs exist in E. coli.
Because newD is a homolog of LeuD that
already binds a homolog of LeuC, this is a
modification-of-function event.
In Chapter 6, Hall (1984) discusses his
work over the previous decade with an en-
zyme he dubbed Ebg, which stands for
evolved

-galactosidase. He had shown that
E. coli that had had its lacZ gene (which
codes for a

-galactosidase) intentionally
deleted could be mutated so that the al-
tered Ebg protein metabolized the sugar
lactose in its stead. In subsequent years,
however, Hall (1995) showed unexpect-
edly that Ebg was an unrecognized homo-
log of LacZ, with 38% sequence identity.
The Ebg active site was very similar to that
of LacZ, and acquired its ability to metab-
olize lactose only when the Ebg active site
mutated to one identical to that of an ac-
tive LacZ. Because the unevolved enzyme
could already metabolize lactose at a low
rate, the adaptive point mutations that
increased its activity are categorized as
modification-of-function ones. Mutations
in the ebg repressor that allowed the ebg

-galactosidase protein to be constitutively
expressed are loss-of-FCT.
In further experiments, both Kemper
(1984) and Hall (2003) deleted the homol-
ogous recruited protein genes ebg and
newD, but were unable to recruit any of the
thousands of other cellular proteins to re-
place them. Recently, Patrick et al. (2007)
employed functional genomics tools to sur-
vey 107 single-gene knockout strains of E.
coli that could not grow on M9-glucose me-
dium, transfecting each one with a plasmid
library that contained every E. coli open-
reading frame. They discovered that about
20% of the knockout strains could be res-
cued by overexpression of at least one
other E. coli gene. Patrick et al. (2007)
found that the missing gene product and
its suppressor were generally not homo-
logs, unlike the previous cases of LacZ/
Ebg and LeuD/NewD.
The work of Patrick et al. (2007) is not in
itself an experimental evolution study, as
all of the manipulations were performed
by the investigators and no spontaneous
mutations arose. Nonetheless, it does dem-
onstrate much potential redundancy in the
E. coli genome that may contribute to its
evolvability, as the authors suggest. It may
be instructive to consider how newer work
would be classified if the events that
Patrick et al. (2007) investigated were to
occur in the wild. If, to adapt an organism
in nature, a gene were initially lost by mu-
tation, that of course would be a loss-of-
FCT event. If subsequent rescue in nature
required the overexpression of a poten-
tially compensating gene, as in Patrick et
al.’s (2007) experimental cases, then that
could occur in at least several ways: 1) a
repressor element of the compensating
gene could be deleted, rendering the gene
constitutive, which would be a loss-of-FCT
mutation; 2) an existing promoter of the
compensating gene could have its se-
quence altered, binding a transcription
factor more strongly and thus leading to
overexpression, which would be classed as
a modification-of-function mutation, as an
existing FCT was altered, not gained or
lost; 3) a new promoter could be con-
structed by mutation to increase the tran-
scription rate of the gene, which would be
a gain-of-FCT mutation; 4) the compensat-
ing gene could increase in copy number,
which by itself would be classed as a mod-
ification-of-function mutation; or 5) a copy
of the compensating gene could diverge in
sequence to more efficiently assume the
activity of the deleted gene, which would
be classified as a gain-of-FCT mutation. Fi-
nally, the effect of the deleted gene may be
compensated in the wild not by increasing
the expression of its potential comple-
ment, but instead by deleting or reducing
the expression of one or more other genes,
which would be an adaptive loss-of-FCT
mutation (this possibility was not investi-
gated in the excellent work of Patrick et al.
[2007]). As these examples make clear,
any of multiple molecular evolutionary
426 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
pathways could underlie the rescued
phenotype. To understand whether an
adaptation represents a gain, loss, or
modification of a function, the molecular
events underlying the adaptation must
first be understood.
the work of lenski and colleagues
In the 1990s, investigators began to per-
form the “open-ended and long-term” evo-
lution experiments that are the main focus
of a review by Elena and Lenski (2003).
Certainly, the most extensive and longest
running investigation has been under-
taken by Lenski himself, who has been con-
tinuously growing cultures of E. coli in his
laboratory since the late 1980s and moni-
toring the facets of their evolution (Lenski
2004). Diluting a portion of the previous
day’s culture a hundred-fold, each day can
potentially see 6 to 7 new generations of
bacteria (26.6 ⬇100). At intervals, Lenski
freezes a portion of a bacterial generation
and stores it, so that, later, the descendant
generations can be straightforwardly ana-
lyzed and compared head-to-head with
their ancestors. Over the years, the num-
ber of generations has approached 50,000.
With a cumulative population size of about
1014 cells, Lenski’s investigation is large
enough and long enough to give solid, re-
liable answers to many questions about
evolution.
The results of a number of Lenski’s pa-
pers over the term of the experiment are
very briefly summarized in Table 3, includ-
ing his early work (not discussed here) in
which the underlying molecular bases of
adaptive phenotypes had not yet been
identified. Schneider et al. (2000) exam-
ined multiple insertion sequence (IS) mu-
tations that became fixed in the bacterial
populations within ten thousand genera-
tions. The mutations were of the kind
caused by IS elements, including insertions
and recombinations, that led to genetic
inversions and deletions. By examining the
DNA sequence of the E. coli in the neigh-
borhood surrounding the IS elements, the
investigators saw that several genes involved in
central metabolism were knocked out, as well
as some cell wall synthesis genes and sev-
eral others. In subsequent work, Cooper et
al. (2001) discovered that twelve of twelve
cell lines showed adaptive IS-mediated de-
letions of their rbs operon, which is in-
volved in making the sugar ribose. Thus,
the adaptive mutations that were initially
tracked down all involved loss-of-FCT.
Several years later, when the cultures
had surpassed their 20,000th generation,
Lenski’s group at Michigan State brought
more advanced techniques to bear on the
problem of identifying the molecular
changes underlying the adaptation of the
E. coli cultures. Using DNA expression pro-
files, they were able to reliably track down
changes in the expression of 1300 genes of
the bacterium, and determined that 59
genes had changed their expression levels
from the ancestor, 47 of which were ex-
pressed at lower levels (Cooper et al 2003).
The authors stated that “The expression
levels of many of these 59 genes are known
to be regulated by specific effectors includ-
ing guanosine tetraphosphate (ppGpp)
and cAMP-cAMP receptor protein (CRP)”
(Cooper et al 2003:1074). They also noted
that the cellular concentration of ppGpp is
controlled by several genes including spoT.
After sequencing, they discovered a non-
synonymous point mutation in the spoT
gene. When the researchers examined ten
other populations that had evolved under
the same conditions for 20,000 genera-
tions, they found that seven others also had
fixed nonsynonymous point mutations in
spoT, but with different substitutions than
the first one that had been identified, thus
suggesting that the mutations were de-
creasing the protein’s activity.
The group then decided to concentrate
on candidate genes suggested by the phys-
iological adaptations that the cells had
made over 20,000 generations. One such
adaptation was a change in supercoiling
density; therefore, genes affecting DNA to-
pology were investigated (Crozat et al.
2005). Two of these genes, topA and fis, had
sustained point mutations. In the case of
topA, the mutation coded an amino acid
substitution, whereas, with fis, a transver-
sion had occurred at the fourth nucleotide
before the starting ATG codon. The topA
December 2010 427EXPERIMENTAL EVOLUTION
mutation decreased the activity of the en-
zyme, while the fis mutation decreased the
amount of fis gene product produced.
Moving the mutations into the ancestor
improved its fitness in minimal glucose me-
dia.
The genes that had earlier been ob-
served to be disrupted by IS elements,
thereby leading to greater fitness in the
growth medium, were then sequenced in
twelve culture lines that had evolved for
20,000 generations (Woods et al. 2006). All
lines were observed to have point muta-
tions in or near one or more of the genes;
TABLE 3
Selected results of Richard Lenski’s long-term E. coli evolution project
Laboratory phenotype Underlying mutation
Mode of
adaptation Reference
Mean fitness of evolved strains improves by
37% on minimal glucose medium over
2000 generations; most improvement
comes early.
unknown — Lenski et al. (1994)
Over 10,000 generations, fitness and cell
volume increase rapidly, then slow down;
fitness increases in discrete steps; authors
liken this to “macroevolution.”
unknown — Lenski and Travisano
(1994)
Over 10,000 generations, mutator strains
evolve in 3 of 12 populations; probably
hitchhike with beneficial mutations.
Probable defect in methyl-directed
mismatch repair pathway
—* Sniegowski et al.
(1997)
After 6000 generations, two phenotypes, L
and S, coexist in balanced polymorphism.
Fitness is frequency dependent.
unknown — Rozen and Lenski
(2000)
Study of nine IS mutants that are fixed after
10,000 generations. IS had inserted into
several genes (pykF, nadR, pbpA-rodA, hokB/
sokB) or caused rearrangements.
IS mediated insertions, deletions,
rearrangements
L,L
L,L
Schneider et al.
(2000)
12 of 12 lines show deletion of rbs operon
mediated by IS150. Selective value of
deletion is 1%-2%.
Repeated deletion of rbs operon L Cooper et al. (2001)
DNA expression arrays showed 59 gene
changes in parallel in two strains grown
for 20,000 generations. Many genes
controlled by CRP and ppGpp.
Eight separate point mutations in
spoT, which controls cellular
concentration of ppGpp
M Cooper et al. (2003)
Sequence random genes from 10,000 and
20,000 generations
No mutations in nonmutator
strains. Some point mutations
in mutator strains, but no
adaptive effect
—* Lenski et al. (2003b)
Mutations affecting DNA topology: topA
mutant decreases protein activity; fis
mutant decreases amount of protein made
topA H33Y; fis A-C transversion
four nucleotides before start
ATG
M
M
Crozat et al. (2005)
Four genes (pykF, nadR, pbpA-rodA,
hokB/sokB) first identified by IS insertions
are examined in other populations.
Same four genes repeatedly suffer
selectable mutations at different
sites
M,M
M,M
Woods et al. (2006)
Comparison of protein profiles of evolved E.
coli vs. ancestor
malT locus suffers multiple
deletions, substitutions
L Pelosi et al. (2006)
Citrate utilization under aerobic conditions unknown — Blount et al. (2008)
*Identified or inferred mutations are not categorized because they are neutral, not adaptive.
Abbreviations:
G - adaptive gain of functional coded element
L - adaptive loss of functional coded element
M - adaptive modification of function
428 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
however, the genes were mutated at a vari-
ety of points. In other words, there was
strong parallelism at the level of which
gene it is adaptive to mutate, but little or
no parallelism for a particular mutation in
a particular gene. The fact that multiple
point mutations in each gene could serve
an adaptive role—and that disruption by IS
insertion was beneficial—suggests that the
point mutations were decreasing or elimi-
nating the protein’s function.
In an investigation of global protein pro-
files of the evolved E. coli, Lenski’s group
discovered that the MalT protein of the
maltose operon had suffered mutations in
8 out of 12 strains (Pelosi et al. 2006).
Several mutations were small deletions
while others were point mutations, thus
suggesting that decreasing the activity of
the MalT protein was adaptive in minimal
glucose media. This was tested by moving
the mutation into the ancestral strain,
which subsequently gained in fitness.
Recently, Lenski’s group reported the
isolation of a mutant E. coli that had
evolved a Cit⫹phenotype. That is, the
strain could grow under aerobic condi-
tions in a culture of citrate (Blount et al.
2008). Wild E. coli cannot grow under such
conditions, as it lacks a citrate permease to
import the metabolite under oxic condi-
tions. (It should be noted that, once inside
the cell, however, E. coli has the enzymatic
capacity to metabolize citrate.) The pheno-
type, whose underlying molecular changes
have not yet been reported, conferred an
enormous growth advantage because the
culture media contained excess citrate but
only limited glucose, which the ancestral
bacteria metabolized. Blount et al. (2008)
marshaled evidence to show that multiple
mutations were needed in the population
before the final mutation conferred the
ability to import citrate; the activating mu-
tation did not appear until after the
30,000th generation.
As Blount et al. (2008) discussed, several
other laboratories had, in the past, also
identified mutant E. coli strains with such a
phenotype. In one such case, the underly-
ing mutation was not identified (Hall
1982); however, in another case, high-level
constitutive expression on a multicopy
plasmid of a citrate transporter gene, citT,
which normally transports citrate in the
absence of oxygen, was responsible for elic-
iting the phenotype (Pos et al. 1998). If the
phenotype of the Lenski Cit⫹strain is
caused by the loss of the activity of a nor-
mal genetic regulatory element, such as a
repressor binding site or other FCT, it will,
of course, be a loss-of-FCT mutation, de-
spite its highly adaptive effects in the pres-
ence of citrate. If the phenotype is due to
one or more mutations that result in, for
example, the addition of a novel genetic
regulatory element, gene-duplication with
sequence divergence, or the gain of a new
binding site, then it will be a noteworthy
gain-of-FCT mutation.
The results of future work aside, so far,
during the course of the longest, most
open-ended, and most extensive laboratory
investigation of bacterial evolution, a num-
ber of adaptive mutations have been iden-
tified that endow the bacterial strain with
greater fitness compared to that of the an-
cestral strain in the particular growth me-
dium. The goal of Lenski’s research was not
to analyze adaptive mutations in terms of
gain or loss of function, as is the focus here,
but rather to address other longstanding evo-
lutionary questions. Nonetheless, all of the
mutations identified to date can readily be
classified as either modification-of-function
or loss-of-FCT.
an ambiguous case
A fascinating case of concurrent gain-
and loss-of FCT can be seen in the work of
Zinser et al. (2003). The authors examined
a mutant of E. coli that thrived when the
culture was starved, but that was outcom-
peted under conditions of growth. They
demonstrated that the result depended on
two mutational events: 1) an initial inser-
tion of an IS element between the tran-
scriptional promoter of the cstA gene and a
nearby CRP box that regulated it; and
2) inversion of the sequence flanked by the
new IS element and another IS element
⬃60 kb distant and upstream of the ybeJ-
gltJKL-ybeK operon. The inversion brought
the ybeJ gene of the mutant, which encodes a
December 2010 429EXPERIMENTAL EVOLUTION
portion of an amino acid transporter, under
the influence of the CRP box that previously
regulated the cstA gene and that encodes a
starvation-inducible oligopeptide permease.
In the new arrangement, the ybeJ gene was
actively transcribed, while the cstA gene
no longer was. Interestingly, the ybeJ gene
product helped the bacterium grow under
conditions of starvation as expected, but
the authors demonstrated that expression
of the cstA gene itself would also be bene-
ficial under the same conditions. Nonethe-
less, the inverted arrangement was selected
because it possessed the greatest net fitness.
This example points to unavoidable am-
biguity in the classification of some adap-
tive evolutionary events. One can view the
adaptation investigated by Zinser et al.
(2003) in several ways: 1) as a modification-
of-function event (i.e., the insertion of a
duplicate IS element) followed, as a result
of the inversion, by the loss of a functional
coded element controlling the cstA gene
(the CRP box) and the gain of a functional
coded element controlling the ybeJ gene
(the same CRP box); or 2) since no FCT
was gained or lost in the inversion but
rather was simply rearranged, the inversion
sums to a second modification-of-function
mutation, albeit one in which there is now
a potential for a recombination switch un-
der conditions of starvation vs. growth.
Evolution Experiments with Viruses
Because of their ability to rapidly re-
produce to enormous population sizes,
viruses, like bacteria, are well-suited to
experimental evolutionary studies, and
such studies have been well-reviewed in
recent years (Elena and Lenski 2003; Elena
and Sanjuan 2007; Bull and Molineux
2008; Elena et al. 2008). Several general
features distinguish viruses from bacteria
and alter evolutionary expectations. 1) Viral
genomes are usually much smaller than
bacterial genomes, often by a thousand-
fold, and one consequence of this is that
mutations are much easier to identify by
sequencing for viruses than for bacteria.
Another consequence, however, is that vi-
ruses do not have deep reserves of homol-
ogous genes or other genetic material to
recruit as bacteria do, except that which
comes from their cellular hosts. 2) Because
of their small genome size, most of their
genome is required for their life cycle; rel-
atively few viral genes are dispensable.
3) For RNA viruses, the mutation rate is
greatly increased, as much as a million-fold
over that of cells. This allows various mu-
tation “solutions” to a selective challenge
to present themselves much more rapidly
than for cells.
Table 4 summarizes evolution experi-
ments with viruses discussed in the reviews
above or in related papers. Most of the
experiments discussed below disturbed vi-
rus growth in various ways and followed
their evolutionary reaction to the distur-
bance. The papers are categorized by the
general selective pressure that was brought
to bear on the viruses.
recovery from multiple bottleneck-
induced deleterious mutations
Two laboratories (Burch and Chao 1999;
Bull et al. 2003) subjected two different
viruses—an RNA virus (
6) and a DNA
virus (T7)—to repeated population bottle-
necks (i.e., severe reductions in population
size that cause the loss of genetic variation)
of just one or a few viruses. Because of the
high mutation rate of RNA viruses, each
virus has a high probability of having a
mutation, most likely deleterious, and,
through many bottlenecks, the researchers
were able to accumulate many mutations
in the virus. Because of the lower rate of
mutation in DNA viruses, the workers us-
ing phage T7 added a mutagen to the
growth medium in order to increase its
mutation rate.
Burch and Chao (1999) showed that
6
suffered a ten-fold loss of fitness after 20
bottleneck generations of repeatedly pick-
ing individual plaques grown on a plate of
bacterial host, and then using the plaque
to seed a new plate. When subsequently
allowed to grow at larger population
sizes, the phage was able to recover fit-
ness by compensatory mutations; how-
ever, because the molecular natures of
the mutations were not determined, they
cannot be classified here.
430 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
Bull et al. (2003) were able to accumu-
late hundreds of mutations in bacterio-
phage T7, and to reduce its fitness to about
0.0001% of the wild type. During the recov-
ery phase of the experiment, two separate
populations of phage gained a factor of
about 100-fold and 10,000-fold in fitness dur-
ing 115 and 82 passages, respectively. Both
populations were continuing to recover
fitness when the experiment was halted.
Among the mutations accumulated during
the bottleneck phase of the experiment were
several small and large deletions in nones-
sential genes, one small insertion, and about
400 point mutations, almost equally divided
between missense mutations and silent mu-
tations. There were also several nonsense
mutations in nonessential genes.
In the recovery phase, there were also sev-
eral large and small deletions, as well as a few
dozen point mutations, most of which were
missense mutations that were likely subject
to positive selection. Recovery-phase muta-
tions involving large or frame-shifting dele-
tions in genes can, with reasonable confi-
dence, be classified as loss-of-FCT. Selectable
point mutations are more difficult to classify
definitively. None were reported to have
formed identifiable, functional coded ele-
ments, such as new promoter sites or protein
processing sites. However, the possibility re-
mains that some new functional coded ele-
ment that is not readily identifiable from its
nucleic acid sequence may have arisen in
one or more cases. To rigorously test such
a possibility, careful biochemical analysis
would be required. In the absence of ev-
idence for gain-of-FCT, selectable point
mutations are tentatively classified as
“modification-of-function” in Table 4.
adapting to a new environment
A second category of selective pressure is
that of viruses placed in new environments.
Cuevas et al. (2002) described the compe-
tition of two neutrally-marked strains of
the single-stranded RNA vesicular stomati-
tis virus (VSV) when grown on the same
medium and the same cells, but at varying
population sizes. There was remarkable
parallelism in the exact nucleotide changes
for the two strains over multiple experi-
ments, thus suggesting that the mutations
compensate for growth in the particular
environment. All mutations were nucleo-
tide substitutions; no deletions or inser-
tions were seen. Twelve substitutions were
synonymous, fifteen were nonsynonymous,
and three were in intergenic regions.
Bull et al. (1997) have studied the adapta-
tion of replicate lineages of the single-stranded
DNA virus
X174 to high temperature
(43.5°C) on several hosts. In one study of 119
total substitutions at 68 sites in the virus, over
half were identical with substitutions in other
lineages, and no deletions or insertions were
reported. Adaptation improved fitness in pri-
mary lines by about 4000-fold with S. typhi-
murium as host, and by about 100,000-fold with
E. coli as host. In another study, Wichman et al.
(1999) grew
X174 on S. typhimurium at ele-
vated temperature (43.5°C). Half of the muta-
tions in one line also appeared in a second line
grown under identical conditions. Most
changes were amino acid substitutions, but
one common mutation was an intergenic de-
letion of 27 nucleotides. Since no discrete mo-
lecular features were attributed to the deleted
intergenic region, that deletion can be consid-
ered a modification-of-function—no func-
tional coded element was lost. No discrete new
molecular features were identified within mu-
tated coding regions, but further biochemical
analysis would be required to determine if a
functional coded feature—such as a new pro-
tein binding site—were gained. In the absence
of such evidence, selectable point mutations
are tentatively classified as “modification-of-
function” in Table 4.
adapting to a new host
Another category of selective pressure is
that of viruses forced to adapt to a novel
host organism in the laboratory. The adap-
tation of a virus to a new host in the wild
can, of course, be a major—even cata-
strophic—epidemiological event, as the ex-
amples of HIV and, potentially, H1N1 show.
Nonetheless, as emphasized throughout this
review, to understand how adaptation pro-
ceeds, the phenotypic and molecular as-
pects must be kept separate. The ability
of a virus to grow on an alternate host
is a phenotype that may have a variety of
December 2010 431EXPERIMENTAL EVOLUTION
TABLE 4
Summary of selected viral evolution experiments
Investigator action Underlying mutation
Mode of
adaptation Reference
Viruses subject to drift
Bacteriophage
6 subject to drift in
bottleneck populations
Recovery by growth in larger populations
showed many small steps; no sequence
data
— Burch and Chao (1999)
Bacteriophage T7 repeatedly passed
through single-virus population
bottlenecks
Several dozen point mutations; several
large and small deletions in recovery
phase
L,M Bull et al. (2003)
Viruses adapting to new environments
Two neutrally-marked strains of
VSV competed at different
population ratios
Extensive parallel convergence at the
nucleotide and amino acid level; some
co-variations noted
M Cuevas et al. (2002)
Bacteriophage
X174 adapted to
high temperature growth
Multiple convergent substitution
mutations across several lineages
M Bull et al. (1997)
Two populations of bacteriophage
X174 adapted to high
temperature and novel host
Multiple convergent mutations, mostly
substitutions, in two populations
accumulated in different orders
M Wichman et al. (1999)
Viruses adapting to new hosts
X174 adapted to growth on
Eschericia and Salmonella hosts
Amino acid substitutions at 5 sites in
major capsid attachment protein affect
host preference
M Crill et al. (2000)
X174 adapted to growth on E. coli
strains differing in
lipopolysaccharide attachment
site
Amino acid substitutions at 10 sites in
major capsid attachment protein affect
host preference
M Pepin et al. (2008)
6 adapted to growth on 15
Pseudomonas syringae strains
Amino acid substitutions at 9 sites in
spike attachment protein affect host
range
M Duffy et al. (2006)
Broad-specificity
6 adapted to
growth on single Pseudomonas
syringae strain
Single substitution in spike attachment
protein gene narrows host range
M Duffy et al. (2007)
Viruses manipulated to be defective
Deletion of 19 intercistronic
nucleotides from RNA virus MS2
containing Shine-Dalgarno
sequence and two hairpins
One revertant deleted 6 nucleotides;
another duplicated an adjoining 14-
nucleotide sequence; missing
functional coded elements
substantially restored
G,G Olsthoorn and van
Duin (1996)
4 nucleotide deletion in lysis gene
of MS2
Reading frame restored by deletions,
insertions
G,G Licis and van Duin
(2006)
Randomized operator sequence of
MS2
Modification-of-function mutations
appeared, but operator hairpin not
restored
M Licis et al. (2000)
Deletion of T7 viral ligase gene Loss-of-FCT and modification-of-function
mutations in several genes; new ligase
activity not obtained from host
L,M Rokyta et al. (2002)
Bacteriophage T7 forced to
replicate using T3 RNA
polymerase
9 of 16 promoters altered sequence;
several missense modification-of-
function changes
M Bull et al. (2007)
T7 RNA polymerase gene moved
toward end of viral genome
Loss of one E. coli polymerase
termination site; facilitated
rearrangement of RNAP gene closer
to the beginning of the viral genome
L,M Springman et al. (2005)
continued
432 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
possible underlying molecular bases. As
reviewed below and as observed in exper-
imental evolutionary studies, viruses have
presumably adapted to new hosts by mod-
ification-of-function mutations rather than
by gain- or loss-of-FCT ones.
Many experimental studies have been
performed investigating the adaptation of
viruses to new hosts, and several recent
papers illustrate the topic of gain- or loss-
of-FCT that this review is concerned with.
Commonly, although not exclusively (see
Yin and Lomax 1983; Subbarao et al. 1993;
Agudelo-Romero et al. 2008), mutations in
coat proteins underlie the phenotypic vari-
ation. Crill et al. (2000) studied the adap-
tation of
X174 to Eschericia and Salmonella
hosts. Over 11 days of selection on a host,
up to 28 substitutions, as well as a 2-base
insertion and 27-base deletion, were ob-
served in all genes except one. The authors
determined that nonsynonymous nucleo-
tide substitutions at 5 sites in the major
capsid protein gene F affected host prefer-
ence. The altered amino acid residues lie
on the virion surface and correlate strongly
with attachment efficiency but, intrigu-
ingly, are at sites on the protein other than
the putative carbohydrate binding site
identified by X-ray crystallographic studies
(McKenna et al. 1994). Pepin et al. (2008)
revisited host adaptation of
X174, but in a
more specific context. They examined the
adaptation of the virus to three strains of E.
coli that differed from each other only by a
single sugar group in the lipopolysaccha-
ride site of attachment. Of 21 mutations, all
were nucleotide substitutions; seven of those
were synonymous, and, of the remaining 14,
10 occurred in the major capsid protein
gene F. Since the substitutions apparently
did not form a new FCT or cause the loss of
one, but simply modified the strength of at-
tachment, they are modification-of-function
mutations.
Duffy et al. (2006) examined the evo-
lution of the RNA bacteriophage
6on
15 Pseudomonas syringae strains. Nine mu-
tations were isolated that affected host
range, and all were amino acid substitu-
tions in the host-attachment spike-protein
P3. One virus strain carrying a substitution
of P3 that broadened the
6 host range was
used for further studies (Duffy et al. 2007).
When the mutant strain was grown exclu-
sively on an alternate permissive host, it
acquired a further single nucleotide substitu-
tion in the gene for the spike protein. The
substitution conferred increased growth on
the alternate host and re-narrowed the host
range.
The work reviewed above presents an op-
portunity to further clarify the classification
categories introduced in this paper. Host ad-
aptations could potentially be classified in
several ways. Some investigators might cate-
gorize the adaptations reported by Duffy et
al. (2007), for instance, as gains and losses of
the function of binding very specific host
ligands. As I discussed earlier in this paper,
however, I classify shifts in ligand preference
as “modification-of-function” events if the
binding occurs at the same site on the pro-
tein, as in the examples discussed here. From
TABLE 4
Continued
Investigator action Underlying mutation
Mode of
adaptation Reference
Viruses forced to be interdependent
Separate viruses, f1 and IKe,
engineered to carry distinct
antibiotic resistance markers
In media containing both antibiotics,
phages co-packaged into f1 protein
coats; two-thirds of IKe genome
deleted, second antibiotic gene
captured by f1
L,M
G
Sachs and Bull (2005)
Abbreviations:
G - adaptive gain of functional coded element
L - adaptive loss of functional coded element
M - adaptive modification of function
December 2010 433EXPERIMENTAL EVOLUTION
this view, the functional coded element is the
protein binding site, which can be modified
either by amino acid substitution directly at
the site or elsewhere in the protein. Consid-
ered in this way, a functional site is not
gained or lost as host preference is altered,
but is instead modified.
recovery from introduced defects:
the rna bacteriophage ms2
Another category of selective pressure is
the intentional introduction by the investiga-
tor of defined defects in a virus. Olsthoorn
and van Duin (1996) deleted a 19-nucleotide
intercistronic segment of RNA phage MS2
between the stop codon for the maturation
protein gene and the start codon for the coat
protein gene; this segment contains the
Shine-Dalgarno sequence and usually forms
two hairpins. The synthesis of the coat pro-
tein is especially sensitive to the presence of
a hairpin containing its initiator codon, and
the investigators reported that the titer of the
altered phage dropped ten orders of magni-
tude (Olsthoorn and van Duin 1996). Two
separate kinds of revertants were isolated—
one that had deleted 6 nucleotides that
coded the final two residues of the matura-
tion protein, and another that contained a
duplication of an adjoining 14-nucleotide
sequence. Remarkably, the deletion rever-
tant substantially restored the coat pro-
tein initiator hairpin and the Shine-
Dalgarno sequence and, subsequently,
approximated them more closely by ac-
cumulating several point mutations. The
duplication revertant partially restored both
native hairpins and the Shine-Dalgarno
sequence, and subsequently accumulated
point mutations to more closely approxi-
mate the starting phage. The titers of the
reconstructed phages were very similar to the
starting phage.
Both the 6-nucleotide deletion and the
14-nucleotide duplication are gain-of-
FCT mutations since they both produced
new coded molecular features in the vi-
rus that did not exist in the immediate
precursor; that is, the virus that sustained
the deliberately-deleted 19 nucleotides.
The 6-nucleotide deletion did not inacti-
vate the maturation protein, so it is not a
loss-of-FCT mutation. The 14-nucleotide
duplication led to the formation of new
functional coded elements (it did not
simply repeat pre-existing elements), so
it is not just a modification-of-function
mutation. The subsequent point muta-
tions that improve activity are modifica-
tion-of-function.
Another paper from van Duin’s labora-
tory investigated the evolution of phage
MS2 in which a 4-nucleotide sequence was
deleted from a 38-nucleotide intercistronic
region between the coat protein gene and
the replicase gene (Licis and van Duin
2006). The gene for the lysis protein spans
the intercistronic region and overlaps into
the coat protein and replicase genes, so
that the deletion introduces a frameshift
into it. Furthermore, the deletion destabi-
lizes the operator loop, which is needed to
bind a dimer of coat protein, initiating cap-
sule formation, as well as regulating ex-
pression of the replicase gene. The mutant
phage decreased in fitness by a factor of
106.
Most initial revertants at relatively low ti-
ters first repaired the frameshift by inserting
a nucleotide in this region. One deleted two
more nucleotides, which also restored the
correct reading frame. Several further rever-
tants were obtained by increasing the initial
population size of mutant viruses and passag-
ing the virus. In particular, insertion rever-
tants were obtained, one of which (revIN4)
had inserted four nucleotides exactly at the
site where they initially had been deleted.
These are all gain-of-FCT mutations. After
further passages, most revertants acquired
point mutations that repaired damaged fea-
tures, including the operator loop. Only one
revertant achieved a fitness equal to the wild
type phage after multiple passages, and this
was a derivative of revIN4 that went on to
revert completely to the wild type.
A third paper from van Duin’s lab con-
cerning MS2 examined the evolutionary
consequences of partially randomizing the
sequence of the operator hairpin (Licis et
al. 2000). The fitness of the initial mutants
decreased anywhere from 103to 107. Most
revertants failed to reconstruct the opera-
tor. The two best revertants, with 2% and
434 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
20% of the fitness of the wild type, respec-
tively, did not restore the operator hairpin.
Thus, the mutations in this study are
modification-of-function ones.
recovery from introduced defects:
the dna bacteriophage t7
Bull, Wichman, Molineux, and colleagues
(Bull et al. 1997, 2003, 2007; Bull and Mo-
lineux 2008) introduced large changes
into the double stranded DNA bacterio-
phage T7 to observe how it might recover.
The DNA polymerase of T7 is much more
accurate than the RNA polymerase of MS2,
the system used by van Duin. Nonetheless,
they were able to show considerable evolu-
tion of the DNA phage.
In one experiment, the gene for DNA
ligase from phage T7 was intentionally de-
leted and the phage’s evolutionary recov-
ery was observed (Rokyta et al. 2002). The
authors noted that the deletion exerts little
selective pressure if the phage is grown on
a host that contains an active cellular li-
gase. They grew the phage in a cell line
whose ligase had been previously mutated
to decrease its activity. Initially, the mutant
T7 had very low fitness, but it eventually
recovered to approximately 20% of wild
type. Of the ten mutations recovered, five
were determined to compensate for the
loss of ligase activity. Three mutations were
in genes, like ligase, involved in DNA me-
tabolism: single amino acid substitutions
in phage DNA polymerase and primase-
helicase, and formation of a stop codon at
position 26 of phage endonuclease. One
mutation was a deletion of 18 nucleotides
in gene 1.5—a gene of unknown function.
The final mutation was the insertion of a
nucleotide into gene 2.8, causing a frame-
shift. The authors determined, however,
that the likely beneficial effect was to in-
crease the transcription of neighboring
gene 3. These can be classified as loss-of-
FCT and modification-of-function muta-
tions. Rokyta et al. (2002) remarked that
they initially expected the phage to acquire
a new ligase activity, either by recombina-
tion or by gene duplication and divergence
by point mutation. If that had happened, it
would have been a notable gain-of-FCT
mutation; however, ligase activity was not
recovered.
In another study, the gene for RNA poly-
merase (RNAP) was deleted from bacterio-
phage T7, and the modified phage was
used to infect cells that harbored a plasmid
carrying a gene for bacteriophage T3
RNAP (Bull et al. 2007). The bacterio-
phages T3 and T7 are closely related, and
the two polymerases differ in sequence by
only 18%. T3 RNAP transcribes from T7
promoters only at low levels, but a single
nucleotide change in the T7 consensus
promoter sequence allows significant activ-
ity by the T3 polymerase. Bull et al. (2007)
used this system to follow the expected
evolution in T7 promoters to accommo-
date T3 polymerase. The initial fitness of
the phage was a factor of about 1010 less
than wild type, but, after adaptation, the
fitness approached that of wild type.
Sequencing revealed that 9 of 16 T7 pro-
moters had acquired mutations—mostly
nucleotide substitutions and one deletion.
Seven missense mutations accumulated in
proteins that do not bind RNAP, and these
were determined to be generally beneficial
and not strictly compensatory for T3
RNAP. Several missense mutations also oc-
curred in some genes to compensate for
T3 RNAP. All of the mutations can be cat-
egorized as modification-of-function.
In a third study, the order of genes in
bacteriophage T7 was rearranged, with the
T7 RNAP gene, which normally enters the
cell early, placed at an ectopic position
near the opposite end of its genome, thus
enabling it to enter the cell late. Ordi-
narily, the phage polymerase would assist
in pulling the phage DNA into the cell by
its transcriptional activity. By placing the
gene toward the further end of the phage,
that activity was precluded, and the fitness
of the phage dropped nine orders of mag-
nitude. Springman et al. (2005) studied
three revertants. Two revertants recovered
only a comparatively small portion of the
fitness they lost, increasing by about 1000-
fold. Both of these revertants abolished an
E. coli polymerase termination site, thereby
allowing the cellular polymerase, which
usually only transcribes early genes of the
December 2010 435EXPERIMENTAL EVOLUTION
phage, to continue down to the ectopic site
of the phage polymerase gene and tran-
scribe it. These are loss-of-FCT mutants, as
they have lost a coded regulatory site. A
third revertant recovered a much larger
amount of fitness, increasing by about a
million-fold. The originally constructed
mutant that gave rise to this revertant had
had some flanking DNA sequence of the
phage RNA polymerase gene purposely left
behind in its original position, in order to
allow for subsequent recombination of the
ectopic gene at its original site. After about
five passages, the expected recombination
occurred with a concomitant large increase
in fitness. This recombination can be catego-
rized as a modification-of-function mutation,
because coded features were rearranged, not
gained or lost.
similar viruses constrained to
occupy the same cell
Sachs and Bull (2005) examined the evo-
lution of two filamentous single-stranded
DNA bacteriophages of E. coli, f1 and IKe.
The two phages share 55% nucleotide
identity and contain ten similar genes. The
investigators inserted different antibiotic
resistance genes into each phage, so that
cells growing in a culture containing both
antibiotics would need to be infected by
both phages in order to survive. Similarly,
one kind of phage could only survive in a
cell if the other kind of phage were there
also. Over the course of 50 passages, the
fitness of the phages grew as they increas-
ingly came to be packaged together into
single f1 coats. During the evolutionary
process, f1 acquired eight point mutations
and IKe acquired nine point mutations, as
well as two large deletions. A deletion of
206 nucleotides of noncoding DNA
around passage 15 gave the largest in-
crease in fitness, and allowed virtually all
IKe to be packaged into an f1 coat. Subse-
quently, around passage 40, IKe suffered a
massive deletion, losing two-thirds of its
genome and retaining only the antibiotic
resistance gene, plus two genes controlling
copy number and co-interference. This re-
sulted in a small, additional increase in
phage fitness.
One can view these adaptations from sev-
eral points of view. When considering the
mutations individually, the large deletion
in IKe is a loss-of-FCT mutation, and the
smaller deletion in IKe and the point mu-
tations in both phages are likely simple
modification-of-function. When consider-
ing the phages as competing organisms,
however, IKe was essentially destroyed in
the evolutionary process, and f1acquired
the use of its critical antibiotic resistance
gene. This could be considered a pseudo-
horizontal-gene-transfer event from IKe to f1
and, therefore, a gain-of FCT event for f1.
Mutational Saturation
The longer an evolution experiment is
run, and the larger the population of mi-
crobes it harbors, the greater the chance for
mutations to appear that are rare and partic-
ularly beneficial. Lenski’s (2004) long-term
experiment with E. coli, which is approaching
50,000 generations and a cumulative popu-
lation size of about 1014 organisms, is the
clear leader in such projects.
There are several other such experi-
ments to note, which involve particularly
large numbers of microbes or generations.
Wichman et al. (2005) described the adap-
tive evolution of the bacteriophage
X174
over the course of 13,000 generations,
which, for the rapidly reproducing mi-
crobe, required only 180 days. Although it
was not reported, the total population size
of phage over the course of the experi-
ment likely reached 1013-1014. Couñago et
al. (2006) replaced the essential gene for ade-
nylate kinase in Geobacillus stearothermophilus—a
moderate thermophile—with that of Bacillus
subtilis—a mesophile—which they then grew
in a turbidostat at increasing temperatures.
Over the course of 1500 generations, they iso-
lated six thermostable mutants of the en-
zyme—one single point mutant and five dou-
ble point mutants derived from the single
mutant. The cumulative number of bacteria
over the term of the experiment was 1013-1014.
Table 5 shows the calculated saturation
of genomes with mutations in these large-
scale experiments, which can be estimated
for E. coli,G. stearothermophilus, and
X174
by using the value for the mutation rate of
436 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
Drake et al. (1998): approximately 0.003
per genome per generation for DNA-based
microbes. On average in the population
during the course of long-term experi-
ments, each nucleotide in each genome is
expected to be substituted from 7⫻103to
6⫻107times; of course, the exact rate of
substitution could vary considerably from
nucleotide to nucleotide. Thus these exper-
iments plumb the depths of what adaptive
mutation can accomplish in these systems
in a single step, or in a series of single,
related steps—such as Couñago et al.’s
(2006) double mutants and Blount et al.’s
(2008) Cit⫹phenotype.
Implications
circumstances of gain- and
loss-of-fct mutations
Tables 2 through 4 summarize results
from the past four decades of evolutionary
experiments with microbes, categorizing
the adaptive mutations as loss- or gain-of-
FCT, or modification-of-function. As can
be seen, only one of the adaptive muta-
tions from bacteria (Tables 2 and 3) is
gain-of-FCT, yet several adaptive mutations
from experiments with viruses (Table 4)
belong to that class as well. Why the differ-
ence? One reason may be that, except for
the capture of an antibiotic resistance gene
by phage f1, the viral gain-of-FCT muta-
tions all reconstruct functional coded ele-
ments that had been deliberately removed
from—or rendered inactive in—the ances-
tral virus, thus restoring pieces of a once-
integrated system. That is, they began at a
point that was known to be able to benefit
from a gain-of-FCT mutation. The lysis
gene of MS2 was rescued from a frame-
shift deletion mutation by adding or delet-
ing additional nucleotides to restore the
correct reading frame. The mutant that
achieved the greatest fitness was the one
that reverted completely to the wild type
sequence. Similarly, the very deleterious
effects of a 19-nucleotide deletion of MS2
containing important functional coded
elements (several hairpins and the Shine-
Dalgarno sequence) was overcome by gain-of-
FCT mutations that restored those same ele-
ments to a greater or lesser degree.
Considering the time- and population-
scale constraints of the experiments, it is
not surprising that, when large experimen-
tal deletions were constructed that re-
moved the coding sequences of whole
genes (rather than just frame shift muta-
tions or short control elements), the de-
leted genes were not restored. However, it
was surprising that more modest adaptive
gain-of-FCT mutations were not seen ei-
ther. The removal of T7 ligase resulted in
point mutations and deletions in other
genes involved in DNA metabolism,
which are loss-of-FCT and modification-
of-function mutations. Intentional dele-
tion of the gene for T7 RNA polymerase
and infection of a cell harboring a T3 poly-
merase gene yielded mutations that appar-
ently strengthened weak T3 promoters,
which are modification-of-function
changes. Rearrangement of the order of
bacteriophage T7 genes, thereby decreas-
ing its fitness, did not provoke the evolu-
tionary construction of new coded control
elements. Rather, one existing element was
lost (an E. coli polymerase termination site)
and the gene order reverted, guided by
flanking DNA that Springman et al. (2005)
intentionally left in the viral DNA se-
quence.
A second reason why several adaptive
gain-of-FCT viral mutations but few bacte-
TABLE 5
Saturation coverage of mutations in three microbes
Microbe Generations
Cumulative
population size N
Estimated mutation
rate per nucleotide
Average fold-saturation
with mutations
E. coli 50,000 1013–1014 7⫻10⫺10 7⫻103–7 ⫻104
G. stearothermophilus 1,500 1013–1014 5⫻10⫺10 5⫻103–5 ⫻104
X174 13,000 1013–1014 6⫻10⫺76⫻106–6 ⫻107
December 2010 437EXPERIMENTAL EVOLUTION
rial ones were identified might be that
RNA viruses have increased mutation rates
compared to those of bacteria. This factor
may also partially explain the occurrence
of gain-of-FCT mutations in experiments
with the RNA virus MS2, but not in exper-
iments with the DNA virus T7, which has a
much lower mutation rate.
Tables 2–4 also show that adaptive loss-
of-FCT mutations are less common for
small viruses than for bacteria. No such
mutations were reported for the small bac-
teriophages
6,
X174, VSV, or MS2 in the
reviewed laboratory experiments. The small
DNA virus IKe suffered a massive loss when
two-thirds of its genome was deleted. How-
ever, it had been co-infected into cells with
the similar small DNA virus f1, which likely
could have supplied all the functions that
IKe lost. No adaptive loss-of-FCT mutations
were sustained by f1 itself. The larger DNA
virus T7 did acquire several loss-of-FCT
mutations in various experiments. It seems
likely that the general explanation for this
pattern is that the smaller the virus, the
smaller the percentage of its genome that
is dispensable for its basic life cycle. The
larger a viral genome, the greater the per-
centage that is dispensable. For larger viral
genomes, and for cells, more coded ele-
ments are available for adaptation by loss-
of-FCT mutations.
the first rule of adaptive evolution
As seen in Tables 2 through 4, the large
majority of experimental adaptive mutations
are loss-of-FCT or modification-of-function
mutations. In fact, leaving out those experi-
ments with viruses in which specific genetic
elements were intentionally deleted and then
restored by subsequent evolution, only two
gain-of-FCT events have been reported: the de-
velopment of the ability of a fucose regulatory
protein to respond to d-arabinose (Lin and
Wu 1984), and the antibiotic gene capture by
f1 (Sachs and Bull 2005). Why is this the case?
One important factor is undoubtedly that the
rate of appearance of loss-of-FCT mutations is
much greater than the rate of construction of
new functional coded elements. Suppose an
adaptive effect could be secured by diminish-
ing or removing the activity of a certain pro-
tein. If the gene for the protein were, for
instance, 1000 nucleotides in length, then
there would be numerous targets of opportu-
nity for a loss-of-FCT mutation. The deletion of
any single nucleotide in the coding sequence
would alter the reading frame and likely de-
stroy or greatly diminish protein activity. The
insertion of a nucleotide anywhere in the cod-
ing sequence would do the same. Longer in-
sertions or deletions would commonly have
the same effect, as would alteration of a codon
from sense to nonsense. All these would fall
into the category of loss-of-FCT mutations.
Nucleotide substitutions resulting in mis-
sense mutations, although not likely to com-
pletely eliminate protein activity, are very likely
to diminish activity to a greater or lesser extent,
as, in multiple experiments, the majority of
amino acid substitutions have been found to
decrease a protein’s activity (Reidhaar-Olson
and Sauer 1988; Bowie and Sauer 1989; Lim
and Sauer 1989; Bowie et al. 1990; Reidhaar-
Olson and Sauer 1990; Axe et al. 1996; Huang
et al. 1996; Sauer et al. 1996; Suckow et al.
1996). Although, if residual protein activity
remained, these would be categorized as mod-
ification-of-function mutations under the ac-
counting system used here—that is, the partial
diminishment of the function provides the
adaptive effect. (A caveat: in the particular case
of microbes that were first allowed to accumu-
late deleterious mutations and then recover,
such as in Bull et al. (2003), it is likely that
many adaptive point mutations are compensa-
tory for initial deleterious mutations and there-
fore increase (mutated) protein function.
Biochemical analysis would be needed in order
to prove that protein function increased or
decreased and was the basis of the adaptive
effect.) If the basic point mutation rate per
nucleotide per generation were 10⫺9, then the
rate of appearance of an adaptive loss-of-FCT
mutation would likely be on the order of
10⫺6, because of the many ways possible to de-
crease the activity of a protein. Indeed, an
adaptive mutation rate of ⬃10⫺5was recently
measured in E. coli (Perfeito et al. 2007).
Contrast this with a situation in which a
particular nucleotide in a gene for a cer-
tain protein has to be mutated in order to
gain an adaptive effect. (The particular
mutation can be thought to help code for
438 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
a new binding site in the protein or to
construct a new genetic control element
from a sequence that was already a near-
match, in addition to other possibilities.
These would be gain-of-FCT mutations.) If
the basic point mutation rate per nucleotide
were 10⫺9, then that would also be the rate of
appearance of the beneficial mutation. Even
if there were several possible pathways by
which to construct a gain-of-FCT mutation,
or several possible kinds of adaptive gain-of-
FCT features, the rate of appearance of an
adaptive mutation that would arise from the
diminishment or elimination of the activity
of a protein is expected to be 100–1000
times the rate of appearance of an adaptive
mutation that requires specific changes to a
gene.
This reasoning can be concisely stated as
what I call “The First Rule of Adaptive Evo-
lution”:
Break or blunt any functional coded element
whose loss would yield a net fitness gain.
It is called a “rule” in the sense of being a
rule of thumb. It is a heuristic, useful gen-
eralization, rather than a strict law; other
circumstances being equal, this is what is
usually to be expected in adaptive evolu-
tion. Since the rule depends on very gen-
eral features of genetic systems (that is, the
mutation rate and the probability of a loss-
of-FCT versus a gain-of-FCT mutation), it is
expected to hold for organisms as diverse
as viruses, prokaryotes, and multicellular
eukaryotes. It is called the “first” rule be-
cause the rate of mutations that diminish
the function of a feature is expected to be
much higher than the rate of appearance
of a new feature, so adaptive loss-of-FCT or
modification-of-function mutations that
decrease activity are expected to appear
first, by far, in a population under selec-
tive pressure.
illustrations of the first rule
The first rule, gleaned from laboratory
evolution experiments, can be used to
interpret data from evolution in nature,
including human genetic mutations in re-
sponse to selective pressure by malaria (Ta-
ble 1). Hundreds of distinct mutations are
known that diminish the activities of G6PD
or the
␣
-or

- chains of hemoglobin, leading
to thalassemia. Yet it is estimated that the
gain-of-FCT mutation leading to sickle he-
moglobin has arisen independently only a
few times, or perhaps just once, within the
past 10,000 years (Cavalli-Sforza et al.
1994). Thus, loss-of-FCT adaptive muta-
tions in this situation appeared several or-
ders of magnitude more frequently than
did a gain-of-FCT mutation. Nonetheless,
the sickle hemoglobin mutation did arise
and spread in a regional population.
Therefore, if a gain-of-FCT mutation such
as the sickle gene has a sufficiently large
selection coefficient, then, even though
adaptive loss-of-FCT mutations arrive more
rapidly and in greater numbers, it is possi-
ble for the gain-of-FCT mutation to out-
compete them.
Another illustration from nature of the
first rule can be seen in the adaptive evo-
lution of the plague bacterium Yersinia pes-
tis over the past 1,500–20,000 years (Wren
2003). A likely evolutionary scenario for its
great virulence is that the plague bacte-
rium serially acquired several plasmids that
conferred upon it the ability to be trans-
ferred by flea bite (Carniel 2003). The 101
kb pFra plasmid carries the yplD gene,
which codes for a phospholipase D that is
necessary for the survival of the bacterium
in the flea proventriculus. The 9.6 kb pPla
plasmid codes for a plasminogen activator,
which allows the bacterium to move in its
host unhindered by blood clotting. The
acquisitions of these genes are gain-of-FCT
events. Y. pestis has also subsequently lost a
large number of chromosomal genes—a
general estimate is that 150 genes have
been lost (Carniel 2003; Chain et al. 2006).
A number of discarded genes have activi-
ties in other Yersinia species that allow
pathogen-host adhesion. The plague bac-
terium has also acquired hundreds of
missense mutations (Carniel 2003). The
discarded genes are of course loss-of-FCT
mutations, and many missense mutations are
likely to diminish activity. Thus, the organism
adapted relatively quickly to its new lifestyle—
first made possible by several gain-of-FCT
events—through much more numerous loss-
December 2010 439EXPERIMENTAL EVOLUTION
of-FCT and modification-of-function muta-
tions.
Two recent laboratory studies also illus-
trate the first rule. Ferenci (2008) discusses
the adaptive value of mutations to the gene
for a specialized E. coli RNA polymerase
factor that contributes to the general stress
response. The author observed that “rpoS
mutations occurred, and indeed spread at
rapid rates within a few generations of
establishing glucose-limited chemostats”,
and also that “The majority of rpoS muta-
tions accumulating in glucose-limited cul-
tures are loss-of-function mutations with
little or no residual RpoS protein. ...The
mutations include stop codons, deletions,
insertions as well as point mutations” (Fe-
renci 2008:447). A second study showed
that the loss of mating genes during asex-
ual growth in Saccharomyces cerevisiae pro-
vided a 2% per-generation growth-rate
advantage (Lang et al. 2009). The authors
noted that “in bacteria, gratuitous gene ex-
pression reduces growth rate....We sus-
pect that the cost of gene expression is not
specific to bacterial enzymes or genes in
the yeast mating pathway, but rather re-
flects a universal cost of gene expression
and that this cost must be borne in all
environments where the gene is ex-
pressed” (Lang et al. 2009:5758). Thus in
any environment in which a gene becomes
superfluous or, more generally, in any en-
vironment where its loss would yield a net
fitness gain, the frequent mutations occur-
ring in the population that tend to elimi-
nate the functional coded element will
turn adaptive.
how frequently are loss-of-fct and
gain-of-fct mutations adaptive?
Although the rates of appearance of
loss-of-FCT and modification-of-function
mutations that degrade protein activity
are always expected to be much greater
than the rate of appearance of gain-of-FCT
mutations, a separate question concerns
what fractions of the mutations in those
categories are adaptive. That is, although
loss-of-FCT mutations might appear rap-
idly, if they do not yield a selective bene-
fit—i.e., if they are not adaptive—then they
will not usually spread in a population. In
the same vein, a gain-of-FCT mutation may
eventually appear that builds some new ge-
netic feature such as a transcription factor
binding site, for instance; yet if the feature
is not adaptive within the organism’s ge-
netic context, it will not be selected (Stone
and Wray 2001).
These are empirical questions that are
difficult to answer conclusively. However,
the data and experiments discussed in
this review offer some insights. In the most
open-ended laboratory evolution experi-
ment (Lenski 2004), in which no specific
selection pressure was intentionally brought
to bear, all of the adaptive mutations that
have been so far identified have either been
loss-of-FCT or modification-of-function mu-
tations, and there is strong reason to believe
that most of the modification-of-function
mutations diminished protein activity. Ex-
cept in cases where specific genetic features
were first removed, as well as in the case of
antibiotic gene capture by f1, all adaptive
mutations in laboratory evolution experi-
ments with viruses seem to be loss-of-FCT or
modification-of-function mutations. Thus, in
general laboratory evolutionary situations
(that is, where a microorganism was under a
general selective pressure rather than a
specific one), adaptive loss-of-FCT or
modification-of-function mutations were
always available. This cannot be said for
gain-of-FCT mutations.
One objection might be that the above
examples are artificial. They concern labo-
ratory evolution, and it may be that dimin-
ished expression of some pre-existing,
commonly-held genes gives an organism
an advantage over its conspecifics in such a
constant environment, but not in the var-
ied and changing environments of nature.
It is true that the laboratory is an artificial
environment, and the opportunities for
some events that occur at irregular inter-
vals in the wild—such as lateral gene trans-
fer—are essentially nonexistent. This is
clearly an area that needs to be addressed
in more detail. Further laboratory evolu-
tion studies with more complex environ-
ments or cultures of mixed species would
serve to shed light on the extent to which
440 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
such factors affect opportunities for adap-
tation by gain- or loss-of-FCT mutations.
Nonetheless, results arguably similar to
those that have been seen in laboratory
evolution studies to date have also been
seen in nature, such as the loss of many
genes by Yersinia pestis (after, of course, the
acquisition of new genetic material in the
form of several plasmids), and the loss-of-
FCT mutations that have spread in human
populations in response to selective pres-
sure from malaria. A tentative conclusion
suggested by these results is that the com-
plex genetic systems that are cells will often
be able to adapt to selective pressure by
effectively removing or diminishing one or
more of their many functional coded ele-
ments.
A second possible objection is that many
of the reviewed experiments were conducted
on comparatively small populations of mi-
crobes for relatively short periods of time, so
that although loss-of-FCT and modification-
of-function mutations might be expected to
occur, there simply was not much opportu-
nity to observe gain-of-FCT mutations. After
all, one certainly would not expect new
genes with complex new properties to arise
on such short time-scales. Although it is true
that new complex gain-of-FCT mutations are
not expected to occur on short time-
scales, the importance of experimental
studies to our understanding of adapta-
tion lies elsewhere. Leaving aside gain-of-
FCT for the moment, the work reviewed
here shows that organisms do indeed
adapt quickly in the laboratory—by loss-of-
FCT and modification-of-function muta-
tions. If such adaptive mutations also arrive
first in the wild, as they of course would be
expected to, then those will also be the kinds
of mutations that are first available to selec-
tion in nature. This is a significant addition
to our understanding of adaptation. That
knowledge is also a necessary prerequisite for
elucidating the nature of long-term adapta-
tion, as consideration of how long-term ad-
aptation proceeds must take into account
how organisms adapt in the short-term.
Furthermore, although complex gain-of-
FCT mutations likely would occur only on
long time-scales unavailable to laboratory
studies, simple gain-of-FCT mutations need
not take nearly as long. As seen in Table 1,
a gain-of-FCT mutation in sickle hemoglo-
bin is triggered by a simple point mutation,
which helps code for a new protein bind-
ing site. It has been estimated that new
transcription-factor binding sites in higher
eukaryotes can be formed relatively quickly
by single point mutations in DNA se-
quences that are already near matches
(Stone and Wray 2001). In general, if a
sequence of genomic DNA is initially only
one nucleotide removed from coding for
an adaptive functional element, then a sin-
gle simple point mutation could yield a
gain-of-FCT. As seen in Table 5, several
laboratory studies have achieved thousand-
to million-fold saturations of their test or-
ganisms with point mutations, and most of
the studies reviewed here have at least
single-fold saturation. Thus, one would ex-
pect to have observed simple gain-of-FCT
adaptive mutations that had sufficient selective
value to outcompete more numerous loss-of-
FCT or modification-of-function mutations in
most experimental evolutionary studies, if they
had indeed been available.
A third objection could be that the time
and population scales of even the most
ambitious laboratory evolution experi-
ments are dwarfed when compared to
those of nature. It is certainly true that,
over the long course of history, many crit-
ical gain-of-FCT events occurred. However,
that does not lessen our understanding,
based upon work by many laboratories
over the course of decades, of how evolu-
tion works in the short term, or of how the
incessant background of loss-of-FCT muta-
tions may influence adaptation.
Conclusion
Adaptive evolution can cause a species to
gain, lose, or modify a function. Therefore, it
is of basic interest to determine whether any
of these modes dominates the evolutionary
process under particular circumstances. The
results of decades of experi-mental labora-
tory evolution studies strongly suggest that,
at the molecular level, loss-of-FCT and di-
minishing modification-of-function adaptive
mutations predominate. In retrospect, this
December 2010 441EXPERIMENTAL EVOLUTION
conclusion is readily understandable from
our knowledge of the structure of genetic
systems, and is concisely summarized by the
first rule of adaptive evolution. Evolution has
myriad facets, and this one is worthy of some
notice.
REFERENCES
Agudelo-Romero P., Carbonell P., Perez-Amador
M. A., Elena S. F. 2008. Virus adaptation by ma-
nipulation of host’s gene expression. PLoS ONE
3:e2397.
Axe D. D., Foster N. W., Fersht A. R. 1996. Active
barnase variants with completely random hydro-
phobic cores. Proceedings of the National Academy of
Sciences USA 93:5590–5594.
Behe M. J., Englander S. W. 1979. Mixed gelation
theory. Kinetics, equilibrium and gel incorpora-
tion in sickle hemoglobin mixtures. Journal of Mo-
lecular Biology 133:137–160.
Blount Z. D., Borland C. Z., Lenski R. E. 2008. His-
torical contingency and the evolution of a key
innovation in an experimental population of Esch-
erichia coli.Proceedings of the National Academy of
Sciences USA 105:7899–7906.
Bowie J. U., Sauer R. T. 1989. Identifying determi-
nants of folding and activity for a protein of un-
known structure. Proceedings of the National Academy
of Sciences USA 86:2152–2156.
Bowie J. U., Reidhaar-Olson J. F., Lim W. A., Sauer
R. T. 1990. Deciphering the message in protein
sequences: tolerance to amino acid substitutions.
Science 247:1306–1310.
Bull J. J., Badgett M. R., Rokyta D., Molineux I. J.
2003. Experimental evolution yields hundreds of
mutations in a functional viral genome. Journal of
Molecular Evolution 57:241–248.
Bull J. J., Badgett M. R., Wichman H. A., Huelsenbeck
J. P., Hillis D. M., Gulati A., Ho C., Molineux I. J.
1997. Exceptional convergent evolution in a virus.
Genetics 147:1497–1507.
Bull J. J., Molineux I. J. 2008. Predicting evolution
from genomics: experimental evolution of bacte-
riophage T7. Heredity 100:453–463.
Bull J. J., Springman R., Molineux I. J. 2007. Com-
pensatory evolution in response to a novel RNA
polymerase: orthologous replacement of a cen-
tral network gene. Molecular Biology and Evolution
24:900–908.
Burch C. L., Chao L. 1999. Evolution by small steps
and rugged landscapes in the RNA virus
6. Ge-
netics 151:921–927.
Carniel E. 2003. Evolution of pathogenic Yersinia:
some lights in the dark. Pages 3–12 in The Genus
Yersinia: Entering the Functional Genomic Era, edited
by M. Skurnik et al. New York: Kluwer Academic/
Plenum.
Cavalli-Sforza L. L., Menozzi P., Piazza A. 1994. The
History and Geography of Human Genes. Princeton
(NJ): Princeton University Press.
Chain P. S., Hu P., Malfatti S. A., Radnedge L., La-
rimer F., Vergez L. M., Worsham P., Chu M. C.,
Andersen G. L. 2006. Complete genome sequence
of Yersinia pestis strains Antiqua and Nepal516:
evidence of gene reduction in an emerging patho-
gen. Journal of Bacteriology 188:4453–4463.
Clarke P. H. 1984. Amidases of Pseudomonas aerugi-
nosa. Pages 187–232 in Microorganisms as Model
Systems for Studying Evolution, edited by R. P. Mort-
lock. New York: Plenum Press.
Cooper T. F., Rozen D. E., Lenski R. E. 2003. Parallel
changes in gene expression after 20,000 generations
of evolution in Escherichia coli.Proceedings of the Na-
tional Academy of Sciences USA 100:1072–1077.
Cooper V. S., Schneider D., Blot M., Lenski R. E.
2001. Mechanisms causing rapid and parallel
losses of ribose catabolism in evolving populations
of Escherichia coli B. Journal of Bacteriology
183:2834–2841.
Couñago R., Chen S., Shamoo Y. 2006. In vivo molec-
ular evolution reveals biophysical origins of organ-
ismal fitness. Molecular Cell 22:441–449.
Crill W. D., Wichman H. A., Bull J. J. 2000. Evolution-
ary reversals during viral adaptation to alternating
hosts. Genetics 154:27–37.
Crozat E., Philippe N., Lenski R. E., Geiselmann J.,
Schneider D. 2005. Long-term experimental evo-
lution in Escherichia coli. XII. DNA topology as a
key target of selection. Genetics 169:523–532.
Cuevas J. M., Elena S. F., Moya A. 2002. Molecular
basis of adaptive convergence in experimental
populations of RNA viruses. Genetics 162:533–542.
Darwin C. 1859. On the Origin of Species by Means of
Natural Selection, or the Preservation of Favoured Races
in the Struggle for Life. London (UK): John Murray.
Drake J. W., Charlesworth B., Charlesworth D., Crow
J. F. 1998. Rates of spontaneous mutation. Genetics
148:1667–1686.
Duffy S., Burch C. L., Turner P. E. 2007. Evolution of
host specificity drives reproductive isolation among
RNA viruses. Evolution 61:2614–2622.
Duffy S., Turner P. E., Burch C. L. 2006. Pleiotropic
costs of niche expansion in the RNA bacterio-
phage
6. Genetics 172:751–757.
Elena S. F., Agudelo-Romero P., Carrasco P., Codoner
F. M., Martin S., Torres-Barcelo C., Sanjuan R.
2008. Experimental evolution of plant RNA vi-
ruses. Heredity 100:478–483.
442 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
Elena S. F., Lenski R. E. 2003. Evolution experiments
with microorganisms: the dynamics and genetic
bases of adaptation. Nature Reviews Genetics 4:457–
469.
Elena S. F., Sanjuan R. 2007. Virus evolution: Insights
from an experimental approach. Annual Review of
Ecology, Evolution, and Systematics 38:27–52.
Ferenci T. 2008. The spread of a beneficial mutation
in experimental bacterial populations: the influ-
ence of the environment and genotype on the
fixation of rpoS mutations. Heredity 100:446–452.
Flint J., Hill A. V., Bowden D. K., Oppenheimer S. J.,
Sill P. R., Serjeantson S. W., Bana-Koiri J., Bhatia
K., Alpers M. P., Boyce A. J. 1986. High frequen-
cies of
␣
-thalassaemia are the result of natural
selection by malaria. Nature 321:744–750.
Forget B. G. 1998. Molecular basis of hereditary per-
sistence of fetal hemoglobin. Annals of the New York
Academy of Sciences 850:38–44.
Friedman M. J. 1978. Erythrocytic mechanism of
sickle cell resistance to malaria. Proceedings of the
National Academy of Sciences USA 75:1994–1997.
Hall B. G. 1982. Chromosomal mutation for citrate
utilization by Escherichia coli K-12. Journal of Bacte-
riology 151:269–273.
Hall B. G. 1983. Evolution of new metabolic functions
in laboratory organisms. Pages 234–257 in Evolu-
tion of Genes and Proteins, edited by M. Nei and R. K.
Koehn. Sunderland (MA): Sinauer Associates.
Hall B. G. 1984. The evolved

-galactosidase system of
Escherichia coli. Pages 165–186 in Microorganisms as
Model Systems for Studying Evolution, edited by R. P.
Mortlock. New York: Plenum Press.
Hall B. G. 1995. Evolutionary potential of the ebgA
gene. Molecular Biology and Evolution 12:514–517.
Hall B. G. 2003. The EBG system of E. coli: origin and
evolution of a novel

-galactosidase for the metab-
olism of lactose. Genetica 118:143–156.
Hartley B. S. 1984. Experimental evolution of ribitol
dehydrogenase. Pages 23–54 in Microorganisms as
Model Systems for Studying Evolution, edited by R. P.
Mortlock. New York: Plenum Press.
Huang W., Petrosino J., Hirsch M., Shenkin P. S.,
Palzkill T. 1996. Amino acid sequence determi-
nants of

-lactamase structure and activity. Journal
of Molecular Biology 258:688–703.
Kemper J. 1984. Gene recruitment for a subunit of
isopropylmalate isomerase. Pages 255–284 in Mi-
croorganisms as Model Systems for Studying Evolution,
edited by R. P. Mortlock. New York: Plenum Press.
Kennedy J. R. 2002. Modulation of sickle cell crisis by
naturally occurring band 3 specific antibodies—a
malaria link. Medical Science Monitor 8:HY10–HY13.
Lang G. I., Murray A. W., Botstein D. 2009. The cost
of gene expression underlies a fitness trade-off in
yeast. Proceedings of the National Academy of Sciences
USA 106:5755–5760.
Lenski R. E. 2004. Phenotypic and genomic evolution dur-
ing a 20,000-generation experiment with the bacte-
rium Escherichia coli.Plant Breeding Reviews 24:225–265.
Lenski R. E., Simpson S. C., Nguyen T. T. 1994. Genetic
analysis of a plasmid-encoded, host genotype-specific
enhancement of bacterial fitness. Journal of Bacteriology
176:3140–3147.
Lenski R. E., Travisano M. 1994. Dynamics of adapta-
tion and diversification: a 10,000-generation ex-
periment with bacterial populations. Proceedings of
the National Academy of Sciences USA 91:6808–6814.
Lenski R. E., Winkworth C. L., Riley M. A. 2003. Rates
of DNA sequence evolution in experimental pop-
ulations of Escherichia coli during 20,000 genera-
tions. Journal of Molecular Evolution 56:498–508.
Licis N., Balklava Z., van Duin J. 2000. Forced retro-
evolution of an RNA bacteriophage. Virology 271:
298–306.
Licis N., van Duin J. 2006. Structural constraints and
mutational bias in the evolutionary restoration of
a severe deletion in RNA phage MS2. Journal of
Molecular Evolution 63:314–329.
Lim W. A., Sauer R. T. 1989. Alternative packing
arrangements in the hydrophobic core of lambda
repressor. Nature 339:31–36.
Lin E. C. C., Wu T. T. 1984. Functional divergence of
the l-fucose system in mutants of Escherichia coli.
Pages 135–164 in Microorganisms as Model Systems
for Studying Evolution, edited by R. P. Mortlock.
New York: Plenum Press.
McKenna R., Ilag L. L., Rossmann M. G. 1994. Anal-
ysis of the single-stranded DNA bacteriophage
X174, refined at a resolution of 3.0 Å. Journal of
Molecular Biology 237:517–543.
Modiano D., Luoni G., Sirima B. S., Simpore J., Verra
F., Konate A., Rastrelli E., Olivieri A., Calissano C.,
Paganotti G. M., D’Urbano L., Sanou I., Sawadogo
A., Modiano G., Coluzzi M. 2001. Haemoglobin C
protects against clinical Plasmodium falciparum ma-
laria. Nature 414:305–308.
Mortlock R. P. 1984a. Microorganisms as Model Systems
for Studying Evolution. New York: Plenum Press.
Mortlock R. P. 1984b. The development of catabolic
pathways for the uncommon aldopentoses. Pages
109–134 in Microorganisms as Model Systems for
Studying Evolution, edited by R. P. Mortlock. New
York: Plenum Press.
Mortlock R. P. 1984c. The utilization of pentitols in
studies of the evolution of enzyme pathways. Pages
1–21 in Microorganisms as Model Systems for Studying
Evolution, edited by R. P. Mortlock. New York:
Plenum Press.
Olsthoorn R. C., van Duin J. 1996. Evolutionary re-
construction of a hairpin deleted from the ge-
nome of an RNA virus. Proceedings of the National
Academy of Sciences USA 93:12256–12261.
Patrick W. M., Quandt E. M., Swartzlander D. B.,
December 2010 443EXPERIMENTAL EVOLUTION
Matsumura I. 2007. Multicopy suppression under-
pins metabolic evolvability. Molecular Biology and
Evolution 24:2716–2722.
Pelosi L., Kuhn L., Guetta D., Garin J., Geiselmann J.,
Lenski R. E., Schneider D. 2006. Parallel changes
in global protein profiles during long-term exper-
imental evolution in Escherichia coli.Genetics 173:
1851–1869.
Pepin K. M., Domsic J., McKenna R. 2008. Genomic
evolution in a virus under specific selection for
host recognition. Infection, Genetics and Evolution
8:825–834.
Perfeito L., Fernandes L., Mota C., Gordo I. 2007.
Adaptive mutations in bacteria: high rate and
small effects. Science 317:813–815.
Pogo A. O., Chaudhuri A. 2000. The Duffy protein: a
malarial and chemokine receptor. Seminars in He-
matology 37:122–129.
Pos K. M., Dimroth P., Bott M. 1998. The Escherichia coli
citrate carrier CitT: A member of a novel eubacterial
transporter family related to the 2-oxoglutarate-
malate translocator from spinach chloroplasts. Jour-
nal of Bacteriology 180:4160–4165.
Reidhaar-Olson J. F., Sauer R. T. 1988. Combinatorial
cassette mutagenesis as a probe of the informa-
tional content of protein sequences. Science 241:
53–57.
Reidhaar-Olson J. F., Sauer R. T. 1990. Functionally
acceptable substitutions in two
␣
-helical regions of
lambda repressor. Proteins 7:306–316.
Rokyta D., Badgett M. R., Molineux I. J., Bull J. J.
2002. Experimental genomic evolution: extensive
compensation for loss of DNA ligase activity in a
virus. Molecular Biology and Evolution 19:230–238.
Rozen D. E., Lenski R. E. 2000. Long-term experimental
evolution in Escherichia coli. VIII. Dynamics of a bal-
anced polymorphism. American Naturalist 155:24–35.
Ruwende C., Khoo S. C., Snow R. W., Yates S. N., Kwi-
atkowski D., Gupta S., Warn P., Allsopp C. E., Gilbert
S. C., Peschu N. 1995. Natural selection of hemi- and
heterozygotes for G6PD deficiency in Africa by resis-
tance to severe malaria. Nature 376:246–249.
Sachs J. L., Bull J. J. 2005. Experimental evolution
of conflict mediation between genomes. Proceed-
ings of the National Academy of Sciences USA 102:
390–395.
Sauer R. T., Milla M. E., Waldburger C. D., Brown
B. M., Schildbach J. F. 1996. Sequence determi-
nants of folding and stability for the P22 Arc re-
pressor dimer. FASEB Journal 10:42–48.
Schneider D., Duperchy E., Coursange E., Lenski
R. E., Blot M. 2000. Long-term experimental evo-
lution in Escherichia coli. IX. Characterization of
insertion sequence-mediated mutations and rear-
rangements. Genetics 156:477–488.
Sniegowski P. D., Gerrish P. J., Lenski R. E. 1997.
Evolution of high mutation rates in experimental
populations of E. coli.Nature 387:703–705.
Springman R., Badgett M. R., Molineux I. J., Bull J. J.
2005. Gene order constrains adaptation in bacte-
riophage T7. Virology 341:141–152.
Stone J. R., Wray G. A. 2001. Rapid evolution of
cis-regulatory sequences via local point mutations.
Molecular Biology and Evolution 18:1764–1770.
Stott R. 2003. Darwin and the Barnacle. New York:
W. W. Norton.
Stover C. K., Kemper J., Marsh R. C. 1988. Molec-
ular cloning and characterization of supQ/newD,
a gene substitution system for the leuD gene of
Salmonella typhimurium.Journal of Bacteriology
170:3115–3124.
Subbarao E. K., London W., Murphy B. R. 1993. A
single amino acid in the PB2 gene of influenza A
virus is a determinant of host range. Journal of
Virology 67:1761–1764.
Suckow J., Markiewicz P., Kleina L. G., Miller J.,
Kisters-Woike B., Muller-Hill B. 1996. Genetic
studies of the Lac repressor. XV: 4000 single
amino acid substitutions and analysis of the result-
ing phenotypes on the basis of the protein struc-
ture. Journal of Molecular Biology 261:509–523.
Verra F., Simpore J., Warimwe G. M., Tetteh K. K.,
Howard T., Osier F. H., Bancone G., Avellino P.,
Blot I., Fegan G., Bull P. C., Williams T. N., Con-
way D. J., Marsh K., and Modiano D. 2007. Hae-
moglobin C and S role in acquired immunity
against Plasmodium falciparum malaria. PLoS ONE
2:e978.
Weatherall D. J., Miller L. H., Baruch D. I., Marsh
K., Doumbo O. K., Casals-Pascual C., Roberts
D. J. 2002. Malaria and the red cell. Hematology:
American Society of Hematology Education Program
2002:35–57.
Wichman H. A., Badgett M. R., Scott L. A., Boulianne
C. M., Bull J. J. 1999. Different trajectories of
parallel evolution during viral adaptation. Science
285:422–424.
Wichman H. A., Millstein J., Bull J. J. 2005. Adaptive
molecular evolution for 13,000 phage genera-
tions: a possible arms race. Genetics 170:19–31.
Wills C. 1984. Structural evolution of yeast alcohol
dehydrogenase in the laboratory. Pages 233–254
in Microorganisms as Model Systems for Studying Evo-
lution, edited by R. P. Mortlock. New York: Plenum
Press.
Woods R., Schneider D., Winkworth C. L., Riley M. A.,
Lenski R. E. 2006. Tests of parallel molecular evo-
lution in a long-term experiment with Escherichia
coli.Proceedings of the National Academy of Sciences
USA 103:9107–9112.
Wren B. W. 2003. The Yersiniae –a model genus to
444 Volume 85THE QUARTERLY REVIEW OF BIOLOGY
study the rapid evolution of bacterial pathogens.
Nature Reviews Microbiology 1:55–64.
Yin F. H., Lomax N. B. 1983. Host range mu-
tants of human rhinovirus in which nonstructural
proteins are altered. Journal of Virology 48:410–
418.
Zinser E. R., Schneider D., Blot M., Kolter R. 2003.
Bacterial evolution through the selective loss of
beneficial genes. Trade-offs in expression involv-
ing two loci. Genetics 164:1271–1277.
Handling Editor: Daniel E. Dykhuizen
December 2010 445EXPERIMENTAL EVOLUTION