Minor Fitness Costs in an Experimental Model of Horizontal
Gene Transfer in Bacteria
Anna Knöppel,y,1 Peter A. Lind,y,z,1 Ulrika Lustig,1 Joakim Näsvall,1 and Dan I. Andersson*,1
1
Department of Medical Biochemistry and Microbiology, Uppsala University, Uppsala, Sweden
Present address: New Zealand Institute for Advanced Study, Massey University, Auckland, New Zealand
y
These authors contributed equally to this work.
*Corresponding author: E-mail: [email protected].
Associate editor: Michael Nachman
z
Abstract
Genes introduced by horizontal gene transfer (HGT) from other species constitute a significant portion of many bacterial
genomes, and the evolutionary dynamics of HGTs are important for understanding the spread of antibiotic resistance and
the emergence of new pathogenic strains of bacteria. The fitness effects of the transferred genes largely determine the
fixation rates and the amount of neutral diversity of newly acquired genes in bacterial populations. Comparative analysis
of bacterial genomes provides insight into what genes are commonly transferred, but direct experimental tests of the
fitness constraints on HGT are scarce. Here, we address this paucity of experimental studies by introducing 98 random
DNA fragments varying in size from 0.45 to 5 kb from Bacteroides, Proteus, and human intestinal phage into a defined
position in the Salmonella chromosome and measuring the effects on fitness. Using highly sensitive competition assays,
we found that eight inserts were deleterious with selection coefficients (s) ranging from & 0.007 to 0.02 and 90 did
not have significant fitness effects. When inducing transcription from a PBAD promoter located at one end of the insert, 16
transfers were deleterious and 82 were not significantly different from the control. In conclusion, a major fraction of the
inserts had minor effects on fitness implying that extra DNA transferred by HGT, even though it does not confer an
immediate selective advantage, could be maintained at selection-transfer balance and serve as raw material for the
evolution of novel beneficial functions.
Key words: bacterial evolution, horizontal gene transfer, lateral gene transfer, fitness effects.
Introduction
Article
Horizontal gene transfer (HGT) is a major force in bacterial
evolution (Ochman et al. 2000; Creevey et al. 2011; Baltrus
2013). The percentage of HGT genes vary widely dependent
on lifestyle and genome size, ranging from close to 0% in
symbiotic intracellular organisms with reduced genomes,
such as Buchnera aphidicola, up to 25% in free-living bacteria
with large genomes (Snel et al. 1999; Ochman et al. 2000;
Koonin et al. 2001; Nakamura et al. 2004; Ge et al. 2005).
High levels of HGT have also been linked to a parasitic or
pathogenic lifestyle and are of major importance for the
spread of antibiotic resistance genes (Ambur et al. 2009;
Barlow 2009; Juhas et al. 2009).
Apart from ecological constraints (i.e., likelihood of sharing
growth niches), mechanistic constraints on HGT likely influence the transfer rate in natural populations. For example,
host range restrictions will limit the transfer potential of
conjugation and transduction (Thomas and Nielsen 2005),
and once inside the recipient cell, restriction systems can
destroy transferred fragments. Finally, the HGT genes must
be integrated into a replicon by homologous or illegitimate
recombination (Thomas and Nielsen 2005) to be stably maintained. Of the transfers that integrate into the genome, only a
subset will fix, and the rate of fixation will depend on the
fitness effects of the transferred gene(s), the potential for
compensatory mutations and the effective population size
(Lind et al. 2010). Most insertional HGT events are expected
to be deleterious given that most bacteria have a coding
density of 95% and that disruption of many genes or regulatory regions reduces fitness (Elena et al. 1998). The costs of
replication and transcription of the extra DNA are relatively
small, whereas costs of translation of the foreign genes are of
much larger importance (Ingraham et al. 1983; Lawrence and
Hendrickson 2005) In addition, the expressed protein could
be involved in a metabolic reaction that is costly (Koskiniemi
et al. 2012). Finally, increased levels of an RNA or protein
could lead to imbalances in gene dosage that results in, for
example, improper gene regulation or unwanted molecular
interactions, thereby disrupting native systems. Such effects
imply that successful transfers are probably on the edges of
functional networks (Kurland 1992; Woese 1998; Jain et al.
1999; Comas et al. 2007; Wellner et al. 2007; Stoebel et al.
2008). Thus, HGT is expected to be more common between
related species where the mechanisms of regulation of transcription and translation are conserved and the successful
integration of the alien genes into the functional networks
of the cell is more likely.
Experimental tests of the fitness effects of HGT are necessary to to identify the mechanistic causes of the dominant fitness constraints and to determine to what degree deleterious
effects will reduce the HGT diversity in bacterial populations.
Sorek et al. (2007) reported the first major investigation of
ß The Author 2014. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For permissions, please
e-mail: [email protected]
1220
Mol. Biol. Evol. 31(5):1220–1227 doi:10.1093/molbev/msu076 Advance Access publication February 17, 2014
Minor Fitness Costs of Horizontal Gene Transfer . doi:10.1093/molbev/msu076
MBE
Fig. 1. Genetic system for experimental studies of HGT on the Salmonella chromosome. (A) The 0.45–5-kb random inserts from Bacteroides fragilis,
Proteus mirabilis, and human intestinal phage were cloned into a template plasmid (pBADcam). (B) The inserts were amplified together with an
arabinose-inducible promoter (PBAD), a Cam-resistance marker, and a terminator, using primers with homology extensions for the cbiA locus in
Salmonella. Through -red recombineering, the fragments were then transformed into the Salmonella cbiA-locus. (C and D) The inserts were transduced into strains expressing cyan fluorescent protein (CFP) or yellow fluorescent protein (YFP), and the fitness of the inserts was measured through
competition experiments followed by flow cytometry analysis.
fitness constraints on random HGTs, where they searched for
genes that could not be cloned in Escherichia coli and found
that most genes resistant to HGT are informational core
genes and that the fitness costs (which were not quantified)
were largely dependent on toxicity through gene dosage
effects (Sorek et al. 2007). Here, we examined genes that are
clonable in E. coli, excluding genes with lethal or very large
effects on fitness (s > 0.875), and introduced the foreign
DNA into a neutral position in the Salmonella typhimurium
chromosome. This allowed us to measure the fitness effects of
defined insertion mutants and examine what fraction of HGT
events are strongly deleterious, have minor effects, or are
highly advantageous in the absence of the cost of insertional
mutagenesis or abnormal copy number increases associated
with chromosomal genes carried on multicopy plasmids
(Sorek et al. 2007).
Results
Experimental Rationale
Using lambda red recombination, random fragments of DNA
from Bacteroides fragilis, Proteus mirabilis, and human intestinal phage (Reyes et al. 2010) were inserted into the cbiA gene
in the S. typhimurium chromosome, together with a chloramphenicol resistance marker, an arabinose-inducible promoter and a transcriptional terminator (fig. 1). Inactivation
of the cbiA gene has no fitness effects on S. typhimurium
under the growth conditions used (data not shown). By selecting one single position for insertion, we exclude the unpredictable fitness effects of random insertional mutagenesis and
focus on the fitness effects conferred only by the inserted
fragment. B. fragilis is a gram-negative obligate anaerobe
(GC content 43%) that is a commensal resident of the
lower human gastrointestinal tract (Cerdeno-Tarraga et al.
2005). Proteus mirabilis is a gram-negative enteric bacterium
with a genomic GC content of 39% (Pearson et al. 2008). Both
S. typimurium and P. mirabilis are Gammaproteobacteria,
whereas B. fragilis is classified in the distantly related
phylum Bacteroidetes. Based on previous experiments (Lind
et al. 2010), we expect the P. mirabilis gene products to be
functional in S. typhimurium and disrupt cellular function to a
lesser degree than the phylogenetically distant B. fragilis insert,
provided these are successfully expressed.
Bacteriophages are present in large numbers in most environments (Suttle 2005; Srinivasiah et al. 2008; Reyes et al.
2012) and are a major source of HGT, both by acting as agents
of gene transfer (transduction) and by lysogen formation by
temperate phages (Canchaya et al. 2003). Phage-related elements are very common in recently acquired HGTs (Daubin
et al. 2003), but it is not clear whether this is solely due to their
assumingly higher transfer rate or that they more commonly
encode genes that are less deleterious or advantageous. Very
little is known about the probability of random phage open
reading frames (ORFs) to be successfully expressed in a novel
host, and our experiments investigate whether these inserts
commonly have large effects on fitness. Human bacteriophage DNA was obtained from the gut (feces) of healthy
unrelated female adults (Reyes et al. 2010). In total, 52 inserts
were constructed from B. fragilis, 29 from P. mirabilis, and 17
from the intestinal phage. Because the human intestinal
phages used in this study were not previously sequenced
and annotated, stepwise sequencing followed by ORFprediction and BlastX (NCBI) searches were performed
to map putative genes and their expected functions (supplementary fig. S1 and table S1, Supplementary Material
online).
Fitness Measurements
To examine the fitness effects of the inserts, we performed
competition experiments in M9 glucose using strains marked
with fluorescent proteins (cfp and yfp, or bfp and syfp2) as
previously described (Lind et al. 2010). Overnight cultures
grown in M9 glucose of the insertion mutants and an isogenic
1221
Knöppel et al. . doi:10.1093/molbev/msu076
MBE
Fig. 2. Mean fitness of insert mutants. All data were generated through competition experiments and are reported relative to wild-type controls (wildtype strains DA22555 for cfp/yfp and DA27153 for bfp/syfp2; s = 0). * indicates deleterious inserts. (A) Fitness (mean ± SD) in M9 glucose. (B) Fitness
(mean ± SD) in M9 glucose supplemented with 0.2% arabinose. The mutants are sorted in the same order as in A. (C and D) Fitness distributions
showing the data presented in A and B, respectively. (E) Fitness in M9 glucose plotted against fitness in M9 glucose 0.2% arabinose.
wild type (marked with cfp or yfp, or bfp or syfp2) were mixed
at equal ratios and were cycled for 50 generations by diluting
the culture 1,000-fold each day (ten generations/day). Each
day, the ratio of insertion mutant to wild type was determined by counting 105 cells using a fluorescence-activated
cell sorter (BD FACS Aria), and the selection coefficients were
determined using the regression model s = [ln R(t)/R(0)]/t,
where R is the ratio of mutant to wild type, as previously
described (Dykhuizen 1990), and all mutants were assayed in
both the cfp and yfp background or bfp and syfp2 background. This assay allows detection of differences in fitness
1222
as small as j s j > 0.007 and measures composite fitness over
the whole growth cycle (lag phase, exponential growth, and
stationary phase survival) in contrast to typical growth rate
measurements that only assay the exponential growth phase.
Fitness data and information on the inserted genes are presented in supplementary tables S1–S3, Supplementary
Material online. Using this assay, 8/98 inserts were significantly deleterious and 90/98 were not significantly different
from the control ( j s j < 0.007; fig. 2A). When arabinose was
added to induce transcription from the PBAD promoter, 16/98
inserts were deleterious and 82/98 were not significantly
Minor Fitness Costs of Horizontal Gene Transfer . doi:10.1093/molbev/msu076
MBE
Fig. 3. Analysis of the mean fitness of insert mutants. Fitness measurements for M9 glucose are marked in turquoise, and M9 glucose supplemented
with 0.2% arabinose are marked in red. All data were generated through competition experiments and are reported relative to wild-type controls (s = 0).
(A) A box plot presenting fitness of insert mutants, separated by insert origin (B. fragilis, P. mirabilis, and phage). Boxes represent the interquartile range
(25th–75th percentile) and whiskers the range without outliers. Outliers are marked with circles. Wilcoxon Signed-Rank test, failed to show any
difference between any of the groups (P > 0.05). (B) Fitness in relation to GC content of the insert. (C) Fitness in relation to length of the insert. (D)
Fitness in relation to number of complete ORFs. Linear regressions show no correlation (P > 0.05).
different from the control (fig. 2B). When induced, overall
fitness was lower than when no arabinose was added to
the growth medium (Wilcoxon signed-rank test,
P = 4 107). Typically, the inserts with deleterious effects
without arabinose were increasingly deleterious under arabinose induction, whereas inserts with no significant fitness
effects remained at a similar fitness level as without arabinose
(fig. 2 and supplementary tables S2 and S3, Supplementary
Material online). The ORFs encoded by deleterious inserts are
listed in supplementary table S3, Supplementary Material
online.
Further analysis of the inserts revealed no correlation between fitness effect and insert origin (Wilcoxon signed-rank
test, P > 0.05; fig. 3A), neither was there any correlation
between fitness effect and GC content, size of insert, or
number of complete ORFs (linear regression P > 0.05;
fig. 3B–D). The GC content of the inserts ranged from
28.6% to 51.6%, where the mean GC content (±standard
deviation) for B. fragilis inserts was 43.9 (±4.6)%, P. mirabilis
40.5 (±2.7)%, and phage 42.5 (±4.2)%. The mean GC content
of the S. typhimurium genome is 52.2% (McClelland et al.
2001) and E. coli is 50.8% (Riley et al. 2006)). The lengths of
the inserts ranged from 0.45 to 5 kb (supplementary table S2,
Supplementary Material online) with the mean length
(±standard deviation) of 2.2 (±0.9) kb for B. fragilis, 2.9
(±1.1) kb for P. mirabilis, and 2.6 (±0.9) kb for the phages.
On average, each insert contained 3.1 ORFs (complete and
truncated) and 1.5 complete ORFs. The genes included in the
1223
Knöppel et al. . doi:10.1093/molbev/msu076
inserts are listed in supplementary table S2, Supplementary
Material online (B. fragilis and P. mirabilis) and supplementary
table S1, Supplementary Material online (phage). The genes
found in B. fragilis and P. mirabilis inserts were analyzed for
connectivity, using STRING 9.1 (Franceschini et al. 2013).
Twelve of 81 fragments in the Bacteroides and Proteus
groups contained full sequences of high connectivity genes
(10 predicted interaction partners with high confidence
[Franceschini et al. 2013]) and 21/81 fragments contained
genes that are likely to be expressed from PBAD (full genes
in forward direction from PBAD with no genes in reverse in
between). Further analyses revealed no differences in mean
fitness between inserts containing at least one full gene with
high connectivity (10 predicted interaction partners with
high confidence using STRING 9.1) and those not including
such (t-tests ara: P = 0.978, + ara: P = 0.806). Neither was
there any difference in mean fitness in the presence of arabinose of inserts, including at least one full gene in the forward
direction from PBAD and those that did not (t-test: P = 0.806).
Discussion
Among the 98 horizontally transferred DNA fragments, 90
had fitness effects not significantly different from the control
( j s j < 0.007). The high number of inserts with j s j < 0.007
suggests the possibility that a relatively large fraction of HGT
events could be nearly neutral, that is, their fate is determined
by genetic drift in populations with Ne j s j 1, where Ne is
the effective population size (Maruyama and Kimura 1980).
However, it should be noted that experimental methods to
study fitness effects are limited by the fact that transfers with
small negative and positive effects cannot reliably be distinguished from truly neutral events (fig. 4). Natural selection
can act on j s j < 1/Ne, and as Ne is >107 for many bacteria,
j s j values as small as ~107 could be seen by selection
(Kimura 1983; Charlesworth and Eyre-Walker 2006). In our
experimental setup, deleterious fitness effects that range from
s = 0.007 to 0.875 were detectable. The lower limit was set
by the detection limit in the competition assay, where j s jvalues <0.007 cannot be distinguished from wild-type fitness.
The higher limit was estimated from the first cloning step of
the inserts into plasmids in E. coli. To be seen by the naked
eye, a colony needs to contain ~106 cells, corresponding
to about 20 generations of growth from a single cell.
Transformants were incubated for 48 h, meaning that a cell
with a generation time of up to 2.4 h (48 h/20 generations) is
detectable as a colony. The wild-type cell, without an alien
fragment, has a generation time of 0.3 h. Thus, cells with a
relative growth rate of 0.3/2.4 = 0.125 of the wild-type form
visible colonies, meaning that very slow-growing mutants
(with an s value as low 0.875 where 1 is lethal) can still
be detected. Although this limitation clearly excludes inserts
that are very strongly deleterious and lethal from the distribution of fitness effects, there is reason to believe these would
be a minor fraction of randomly transferred fragments given
that >99% of genes from 79 genomes have previously been
shown to be clonable in E. coli using standard methods (Sorek
et al. 2007). Even if our experimental setup was designed to
reduce the influence of fitness during strain construction,
1224
MBE
Fig. 4. Comparison of analyzable s values between different experimental approaches to study fitness constraints of HGTs. In our study,
s values j s j > 0.007 and s > 0.875 were detectable. The lower
limit, j s j > 0.007, was set by the detection limit in the competition
assay and the higher limit, j s j < 0.875, was estimated from the first
cloning step of the inserts into plasmids into Escherichia coli. In the
Sorek et al. (2007) study, HGTs that are lethal when carried on a multicopy plasmid in E. coli were analyzed.
described in the methods section below, we acknowledge
that there is always a bias toward higher fitness variants in
any experiment where different mutants compete with each
other and that this will introduce a bias that will somewhat
reduce the frequency of low fitness inserts. Similarly, we
cannot exclude that some inserts that had too high fitness
cost to be found in the E. coli cloning strain would have had a
smaller effect on S. typhimurium physiology.
The fitness assays conducted in this study measures composite fitness over the entire growth cycle, including lag
phase, exponential phase, and stationary phase in a minimal
glucose medium. This is clearly an artificial environment that
is relatively benign, and the magnitude of the fitness effects
reported here might be very different from a natural environment or under more stressful conditions where fitness costs
could be either amplified or reduced (Elena and de Visser
2003). In our experimental setup, we also exclude the fitness
effects of insertional mutagenesis by inserting the fragments
into a defined position with no detectable effect on fitness.
This is done to separate the fitness effects of disrupting native
genes with the effects conferred by the content of the transferred DNA fragments. On the basis of experimental data in
E. coli (Elena et al. 1998) and the high coding density of bacterial genomes, we would expect that the large majority
(>90%) of random HGTs inserted into the genome would
be strongly deleterious regardless of gene content and be
rapidly lost from the population. It seems likely that insertional effects will dominate the distribution of fitness effects
of integrative HGT events, but we also require information of
the fitness effects that are dependent on gene content to
understand the potential for HGTs to reach high frequencies
in bacterial populations.
If a significant fraction of HGTs, when inserted into a neutral position, have selection coefficients <0.001 and each
import will produce a total of 1/ j s j descendants before
being lost from the population (Lind et al. 2010), these
nearly neutral HGTs could be relatively common in natural
populations if the mechanistic rate of HGT is high enough.
A neutral HGT will rarely fix in the population, but when it
does, this will take in the order of Ne generations (Kimura
1983) in an ideal population with constant population size.
For an effective population size for enteric bacteria of 107
(Charlesworth and Eyre-Walker 2006) and assuming a generation time of 1 day, this corresponds to >10,000 years, but
Minor Fitness Costs of Horizontal Gene Transfer . doi:10.1093/molbev/msu076
this time could be substantially shortened for more realistic
conditions where populations might experience bottlenecks
and hitchhiking with linked beneficial loci. Perhaps more importantly, the presence of a substantial pool of horizontally
acquired genetic diversity might provide raw material for selection to act upon when challenged by changing environmental conditions. So even if the early evolutionary dynamics
of the HGT genes is largely determined by genetic drift, once
under selection they can be important for the evolution of
novel beneficial functions and acquisition of novel traits.
The fragments were determined deleterious in this study if
the confidence interval for independent fitness measurements did not overlap with the standard deviation for the
wild type. In total, 8/98 inserts were deleterious, and when the
PBAD promoter induced, the number doubled (16/98; fig. 2A
and B). It should be noted that we have not demonstrated
that any of the inserts actually express a gene product, although we strongly believe this to be the cause of the deleterious fitness effects. This is supported by a finding that
seven of eight inserts that are significantly deleterious only
under arabinose induction contain at least one ORF in the
forward direction (supplementary tables S2 and S3,
Supplementary Material online). Likewise, the arabinose induction of the PBAD promoter does not necessarily mean
increased transcription of the alien ORFs, even though this
was demonstrated for an YFP control strain. As these inserts
contain ORFs in both directions relative to the PBAD promoter, we do not expect any increase in gene expression
for ORFs in the opposite direction, although this is possible
due to reduced binding of H-NS protein that have been
shown to silence foreign genes (Navarre et al. 2006). It is
also possible that differences in fitness effects are solely due
to the change in environment by adding arabinose that is not
related to increased transcription from the PBAD promoter.
Despite these reservations, we interpret the increases in fitness cost of many of the deleterious inserts under arabinose
induction as supporting the case that the fitness costs are
mainly dependent on the expression of gene products
encoded on the inserts. The inserts with deleterious effects
contained ORFs that encoded proteins of diverse functions
(supplementary table S3, Supplementary Material online) including BF0844 (putative transmembrane protein), upper
and lower collar connector proteins, kdpC (potassium-transporting ATPase subunit c), BF0969 (putative RNA polymerase
ECF-type sigma factor), dapB (dihydrodipicolinate reductase),
ispH (4-hydroxy-3-methylbut-2-enyl diphosphate reductase),
and menE (putative O-succinylbenzoate-CoA ligase). The inserts also encoded truncated genes such as: prfA (putative
peptide chain release factor 1), clpB (negative regulator of
genetic competence), and BF0970 (putative antisigma factor).
Deleterious effects caused by insertion of HGTs into a neutral chromosomal position can be conferred at different levels.
First, the extra DNA may cause a burden on replication.
Second, expression of the extra genes may be costly both at
the level of transcription and translation. However, as translation is by far more energy consuming than replication and
transcription, it is expected to dominate the cost of gene expression (Ingraham et al. 1983; Lawrence and Hendrickson
MBE
2005; Bragg and Wagner 2009). In addition, the expressed
protein could be involved in an enzymatic reaction that is
costly (Diaz Ricci and Hernandez 2000; Koskiniemi et al. 2012).
Finally, increased levels of a gene product could lead to imbalances in protein/RNA dosage that results in, for example,
improper gene regulation or unwanted molecular interactions, thereby disrupting regulatory networks or complexes
(Pal et al. 2005; Sorek et al. 2007; Park and Zhang 2012). From
the present experiments, we cannot identify the reason for
the cost of the eight deleterious inserts (s values ranging from
& 0.007 to 0.02) because no correlation was seen between fitness effect and size of the inserts, or number of genes
(fig. 3C and D), which is expected if costs of gene expression
predominated, and this suggests that the deleterious effect
predominantly is due to, largely unpredictable, toxic gene
dosage effects caused by specific gene products.
Our demonstration that most HGTs events have minor
effect on fitness when introduced into a neutral position in
the genome suggests that most horizontally acquired DNAs
are only temporary residents in the genome. However, this
does not exclude the presence of strongly selected advantageous transfers that are important under strong selective
pressures, including transfer of antibiotic resistance genes
and genes allowing invasion of new ecological niches.
Although we did not find any inserts with strong, statistically
significant positive effects on fitness, it is possible that some
inserts confer beneficial effects that are smaller than 0.7%.
The only other experimental study of random HGTs to
date, to our knowledge (Sorek et al. 2007), focused on genes
that are not clonable in E. coli (in pUC plasmids with ~200
copies per cell) using a purely qualitative classification:
unclonable (lethal; 0.57%) or clonable (viable; 99.43%), covering 246,045 genes from 79 bacterial and archaeal genomes.
Figure 4 describes the major differences in which types of
HGT events would be discovered in our and their study. In
the study of Sorek et al., unclonable genes were most frequently found among bacteria belonging to the phylum
Firmicutes and the class Gammaproteobacteria, groups that
also coincided with lower GC content than the average
among the 79 genomes. In our study, no correlation between
GC content and fitness was found (fig. 3B). In agreement with
our findings, GC content seemed to play no role in the fitness
effect of exchanged ribosomal genes (Lind et al. 2010).
Potential explanations for the difference between the results
of those of Sorek et al. and ours is that our material was too
small to detect a GC effect or that the GC effect is more
pronounced when the genes are expressed at very high
levels from a high-copy number plasmid (as in the study of
Sorek et al. [2007]). It is also possible that the GC effect is only
relevant for classes of fragments that are lethal or have large
effects on fitness, and these are not included in this study.
The relatively small fitness effects of HGTs in our experimental model suggest there could be a major hidden genetic
diversity in bacterial populations. Whether or not this is the
case will depend on if the rates of conjugation, transduction,
and transformation in natural populations are high enough to
reintroduce HGT variation lost by genetic drift.
1225
MBE
Knöppel et al. . doi:10.1093/molbev/msu076
Materials and Methods
Strains and Media
All experiments used S. enterica subsp. enterica serovar
Typhimurium str. LT2 (designated S. typhimurium in the
text) and derivatives thereof except for the cloning of
random DNA fragments that was performed using E. coli
cloning strains (TOP10 [Invitrogen] or NEB 5-alpha [New
England Biolabs]). Luria-Bertani agar (LA) was used as solid
media and supplemented with ampicillin (200 mg/l) or chloramphenicol (10/15 mg/l) where needed for selection and
plasmid maintenance. Luria-Bertani/Lysogeny broth (LB)
and M9 minimal medium with 0.2% glucose or SOC (New
England Biolabs) were used as liquid media.
Construction of Strains with Random Insertions
A chloramphenicol resistance cassette was PCR-amplified
with high-fidelity polymerase (Phusion high-fidelity DNA polymerase, Finnzymes) with plasmid pKD3 (Datsenko and
Wanner 2000) (GenBank AY048742) as template and using
primers with added 50 -end restriction sites (oligonucleotide
primers used in strain construction are listed in supplementary table S4, Supplementary Material online). The PCR product was ligated into plasmid pBAD30 (Guzman et al. 1995)
between the SphI and HindIII restriction sites to form plasmid
pBADcam. Genomic DNA was isolated from B. fragilis and
P. mirabilis using Genomic Tip 100G (Qiagen). The phage
DNA was obtained from human feces from two healthy
and unrelated female adults (Reyes et al. 2010). Briefly, after
separation of the phage from bacteria and large particles present in feces, the samples were run on a CsCl gradient, and
viral particles in the 1.2–1.5 g ml1 range were collected and
used for DNA extraction. The phage DNA was amplified with
Genome Phi V2 (GE Healthcare) according to the manufacturer’s instructions before cloning. The amplified DNA was
restriction digested with EcoRI and KpnI (Fermentas), subjected to agarose gel electrophoresis, and fragments (2–5 kb
in size) were purified using GeneJET Gel Extraction Kit
(Fermentas). Plasmid pBADcam was isolated using E.Z.N.A.
plasmid purification kit (Omega Bio-Tek) and restriction digested with EcoRI and KpnI and treated with Shrimp Alkaline
Phosphatase (Fermentas). The gel-purified DNA fragments
were then ligated into the pBADcam plasmid and transformed into E. coli by electroporation, and suitable dilutions
were plated after 20-min shaking incubation at 37 C on
nonselective LA plates (fig. 1). The LA plates were incubated
overnight at 37 C before replica printing to selective chloramphenicol plates. This protocol was used to reduce bias
toward fast-growing transformants during recovery and
reduce the probability of obtaining multiple clones with identical inserts. Transformants were scored after 24 and 48 h and
used as template in a PCR to amplify the insert, the PBAD
promoter, and the chloramphenicol resistance cassette using
primers with 50 -homology regions for insertion into cbiA in
the S. typhimurium chromosome using the using the Red
system as previously described (Datsenko and Wanner 2000;
Lind et al. 2010). Correct inserts were then confirmed by PCR
1226
using primers outside the insert region and sequenced. The
inserts, including four independently constructed controls
containing only the PBAD promoter, the chloramphenicol
marker and the terminator, without a foreign DNA insert,
were moved by phage P22 transduction into the wild-type
S. typhimurium LT2 background and the two cfp and yfp
containing preadapted strains used for the competition experiments. For five strains, bfp and syfp2 were instead used as
markers.
Competition Experiments
Selection coefficients (s) were estimated by competing the
mutants with insert against an isogenic wild-type control
without a foreign insert but containing the arabinose promoter, chloramphenicol resistance marker, and terminator.
The inserts and controls were moved by P22 transduction
into strains that had previously been serially passaged under
the experimental conditions for 1,000 generations and encoding CFP and YFP fluorescent proteins, as previously described
(Lind et al. 2010). HGT mutants and wild-type controls carrying either a cfp or yfp marker or a bfp or syfp2 marker were
mixed at equal volumes from an overnight culture and maintained by 1,000-fold serial dilution (ten generations per serial
passage) every 24 h. 105 cells were counted each cycle by flow
cytometry (BD FACS Aria) to determine the ratio of mutant
to wild type for five cycles (50 generations). Estimation of
selection coefficients (s) were performed according to the
regression model s = [ln R(t)/R(0)]/t, as previously described
(Dykhuizen 1990), where R is the ratio of mutant to wild type,
and all mutants were assayed in both the cfp and yfp background, or the bfp and syfp2 background. Pearson correlation
between the cfp/bfp and yfp/syfp2 strains was r = 0.83 without
arabinose and r = 0.91 with arabinose induction, showing that
fitness was dependent on the insert. Independently constructed wild-type controls were used to correct for the
small difference in fitness between the cfp/bfp and yfp/syfp2
alleles. Supplementary figure S3, Supplementary Material
online is a control experiment that demonstrates that arabinose induces the PBAD promoter under the conditions used.
Statistical Analysis
The statistical analysis was carried out using the software R
(version 3.0.2) and RStudio (version 0.97.551). P values < 0.05
are referred to as statistically different and >0.05 as not different. An insert was determined to be deleterious if the
confidence interval for independent fitness measurements
of the specific strain did not overlap with the SD for control
strain measured in + ara. If the confidence interval overlapped with the SD for control strain, the strain was instead
referred to as “not significantly different from the control”
(supplementary fig. S2, Supplementary Material online). For
connectivity analysis, STRING 9.1 (string-db.org, [Franceschini
et al. 2013]) was used to list the number of interaction partners for each gene in the B. fragilis and P. mirabilis inserts. The
cutoff for included interaction partners was set to 0.7 (combined score) that refers to “predicted functional partners”
with “high confidence.”
Minor Fitness Costs of Horizontal Gene Transfer . doi:10.1093/molbev/msu076
Supplementary Material
Supplementary figures S1–S3 and tables S1–S4 are available at
Molecular Biology and Evolution online (http://www.mbe.
oxfordjournals.org/).
Acknowledgments
This work was supported by grants from the Swedish
Research Council to D.I.A. The authors thank Hervé
Nicoloff for technical support and insightful discussions.
They also thank Forest Rowher for kindly providing us with
the phage DNA used in this study.
References
Ambur OH, Davidsen T, Frye SA, Balasingham SV, Lagesen K, Rognes T,
Tonjum T. 2009. Genome dynamics in major bacterial pathogens.
FEMS Microbiol Rev. 33:453–470.
Baltrus DA. 2013. Exploring the costs of horizontal gene transfer. Trends
Ecol Evol. 28:489–495.
Barlow M. 2009. What antimicrobial resistance has taught us about
horizontal gene transfer. Methods Mol Biol. 532:397–411.
Bragg JG, Wagner A. 2009. Protein material costs: single atoms can make
an evolutionary difference. Trends Genet. 25:5–8.
Canchaya C, Fournous G, Chibani-Chennoufi S, Dillmann ML, Brussow
H. 2003. Phage as agents of lateral gene transfer. Curr Opin Microbiol.
6:417–424.
Cerdeno-Tarraga AM, Patrick S, Crossman LC, Blakely G, Abratt V,
Lennard N, Poxton I, Duerden B, Harris B, Quail MA, et al. 2005.
Extensive DNA inversions in the B. fragilis genome control variable
gene expression. Science 307:1463–1465.
Charlesworth J, Eyre-Walker A. 2006. The rate of adaptive evolution in
enteric bacteria. Mol Biol Evol. 23:1348–1356.
Comas I, Moya A, Gonzalez-Candelas F. 2007. Phylogenetic signal and
functional categories in Proteobacteria genomes. BMC Evol Biol.
7(Suppl 1):S7.
Creevey CJ, Doerks T, Fitzpatrick DA, Raes J, Bork P. 2011. Universally
distributed single-copy genes indicate a constant rate of horizontal
transfer. PLoS One 6:e22099.
Datsenko KA, Wanner BL. 2000. One-step inactivation of chromosomal
genes in Escherichia coli K-12 using PCR products. Proc Natl Acad Sci
U S A. 97:6640–6645.
Daubin V, Lerat E, Perriere G. 2003. The source of laterally transferred
genes in bacterial genomes. Genome Biol. 4:R57.
Diaz Ricci JC, Hernandez ME. 2000. Plasmid effects on Escherichia coli
metabolism. Crit Rev Biotechnol. 20:79–108.
Dykhuizen DE. 1990. Experimental studies of natural selection in bacteria. Annu Rev Ecol Syst. 21:373–398.
Elena SF, de Visser JA. 2003. Environmental stress and the effects of
mutation. J Biol. 2:12.
Elena SF, Ekunwe L, Hajela N, Oden SA, Lenski RE. 1998. Distribution of
fitness effects caused by random insertion mutations in Escherichia
coli. Genetica 102–103:349–358.
Franceschini A, Szklarczyk D, Frankild S, Kuhn M, Simonovic M, Roth A,
Lin J, Minguez P, Bork P, von Mering C, et al. 2013. STRING v9.1:
protein-protein interaction networks, with increased coverage and
integration. Nucleic Acids Res. 41:D808–D815.
Ge F, Wang LS, Kim J. 2005. The cobweb of life revealed by genome-scale
estimates of horizontal gene transfer. PLoS Biol. 3:e316.
Guzman LM, Belin D, Carson MJ, Beckwith J. 1995. Tight regulation,
modulation, and high-level expression by vectors containing the
arabinose PBAD promoter. J Bacteriol. 177:4121–4130.
Ingraham JL, Maaløe O, Neidhardt FC. 1983. Growth of the bacterial cell.
Sunderland (MA): Sinauer Associates.
Jain R, Rivera MC, Lake JA. 1999. Horizontal gene transfer among
genomes: the complexity hypothesis. Proc Natl Acad Sci U S A. 96:
3801–3806.
MBE
Juhas M, van der Meer JR, Gaillard M, Harding RM, Hood DW, Crook
DW. 2009. Genomic islands: tools of bacterial horizontal gene transfer and evolution. FEMS Microbiol Rev. 33:376–393.
Kimura M. 1983. The neutral theory of molecular evolution. Cambridge:
Cambridge University Press.
Koonin EV, Makarova KS, Aravind L. 2001. Horizontal gene transfer in
prokaryotes: quantification and classification. Annu Rev Microbiol.
55:709–742.
Koskiniemi S, Sun S, Berg OG, Andersson DI. 2012. Selection-driven gene
loss in bacteria. PLoS Genet. 8:e1002787.
Kurland CG. 1992. Translational accuracy and the fitness of bacteria.
Annu Rev Genet. 26:29–50.
Lawrence JG, Hendrickson H. 2005. Genome evolution in bacteria: order
beneath chaos. Curr Opin Microbiol. 8:572–578.
Lind PA, Tobin C, Berg OG, Kurland CG, Andersson DI. 2010.
Compensatory gene amplification restores fitness after inter-species
gene replacements. Mol Microbiol. 75:1078–1089.
Maruyama T, Kimura M. 1980. Genetic variability and effective population size when local extinction and recolonization of subpopulations
are frequent. Proc Natl Acad Sci U S A. 77:6710–6714.
McClelland M, Sanderson KE, Spieth J, Clifton SW, Latreille P, Courtney
L, Porwollik S, Ali J, Dante M, Du F, et al. 2001. Complete genome
sequence of Salmonella enterica serovar Typhimurium LT2. Nature
413:852–856.
Nakamura Y, Itoh T, Matsuda H, Gojobori T. 2004. Biased biological
functions of horizontally transferred genes in prokaryotic genomes.
Nat Genet. 36:760–766.
Navarre WW, Porwollik S, Wang Y, McClelland M, Rosen H, Libby SJ,
Fang FC. 2006. Selective silencing of foreign DNA with low GC
content by the H-NS protein in Salmonella. Science 313:236–238.
Ochman H, Lawrence JG, Groisman EA. 2000. Lateral gene transfer and
the nature of bacterial innovation. Nature 405:299–304.
Pal C, Papp B, Lercher MJ. 2005. Adaptive evolution of bacterial metabolic networks by horizontal gene transfer. Nat Genet. 37:1372–1375.
Park C, Zhang J. 2012. High expression hampers horizontal gene transfer.
Genome Biol Evol. 4:523–532.
Pearson MM, Sebaihia M, Churcher C, Quail MA, Seshasayee AS,
Luscombe NM, Abdellah Z, Arrosmith C, Atkin B, Chillingworth
T, et al. 2008. Complete genome sequence of uropathogenic
Proteus mirabilis, a master of both adherence and motility.
J Bacteriol. 190:4027–4037.
Reyes A, Haynes M, Hanson N, Angly FE, Heath AC, Rohwer F, Gordon JI.
2010. Viruses in the faecal microbiota of monozygotic twins and
their mothers. Nature 466:334–338.
Reyes A, Semenkovich NP, Whiteson K, Rohwer F, Gordon JI. 2012.
Going viral: next-generation sequencing applied to phage populations in the human gut. Nat Rev Microbiol. 10:607–617.
Riley M, Abe T, Arnaud MB, Berlyn MK, Blattner FR, Chaudhuri RR,
Glasner JD, Horiuchi T, Keseler IM, Kosuge T, et al. 2006. Escherichia
coli K-12: a cooperatively developed annotation snapshot–2005.
Nucleic Acids Res. 34:1–9.
Snel B, Bork P, Huynen MA. 1999. Genome phylogeny based on gene
content. Nat Genet. 21:108–110.
Sorek R, Zhu Y, Creevey CJ, Francino MP, Bork P, Rubin EM. 2007.
Genome-wide experimental determination of barriers to horizontal
gene transfer. Science 318:1449–1452.
Srinivasiah S, Bhavsar J, Thapar K, Liles M, Schoenfeld T, Wommack KE.
2008. Phages across the biosphere: contrasts of viruses in soil and
aquatic environments. Res Microbiol. 159:349–357.
Stoebel DM, Dean AM, Dykhuizen DE. 2008. The cost of expression of
Escherichia coli lac operon proteins is in the process, not in the
products. Genetics 178:1653–1660.
Suttle CA. 2005. Viruses in the sea. Nature 437:356–361.
Thomas CM, Nielsen KM. 2005. Mechanisms of, and barriers to, horizontal gene transfer between bacteria. Nat Rev Microbiol. 3:711–721.
Wellner A, Lurie MN, Gophna U. 2007. Complexity, connectivity, and
duplicability as barriers to lateral gene transfer. Genome Biol. 8:R156.
Woese C. 1998. The universal ancestor. Proc Natl Acad Sci U S A. 95:
6854–6859.
1227