M." ...Ih..||0‘!.
.l duUuunllc; I .H‘dr‘ ..
.. H “I‘HHJLF: U

‘

p331!
L. 'i‘

 

. 0 IL. ‘1‘
30“.“...tv
. \IO.

. A. .47 » Iw
l.|\\;l109’..l .
.ut...‘a't§ :1. O

1..L..0.0
53v
0

 

 

 

lap-Fl... . .1
.u , p» 5. v.3
:57]
III; ‘
{5:}...
.I11 ‘ 5,33,»... .iil.
, , is f
. [.53.
. .1 A
at...
. .51» . I .1
, p .21 s

m“ I

“2‘36 llllllllllllllllllllllllllllllHllllllllll'l‘llllllllllll

3 1293 01565 92

l LIBRARY

Michigan State
Unlvorslw

 

 

 

 

This is to certify that the

dissertation entitled

COMPARATIVE AND EXPERIMENTAL TESTS FOR A TRADEOFF
IN BACTERIAL FITNESS AT LOW AND HIGH
SUBSTRATE CONCENTRATION

presented by

Gregory Jon Velicer

has been accepted towards fulfillment
of the requirements for

 

 

 

Ph.D. degree in Genetics
EAJWJZ S. W
Major professor

Date

 

MSU i: an Affirmative Action/Equal Opportunity Institution 0- 12771

PLACE ll RETURN BOX to remove thle checkout from your record.
TO AVOID FINES return on or betore due due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

——l| l l

MSU Ie An Affirmative Action/Equal Opportunity Intuition
Wane-m

 

 

COMPARATIVE AND EXPERIMENTAL TESTS FOR A TRADEOFF IN
BACTERIAL FITNESS AT LOW AND HIGH SUBSTRATE CONCENTRATION

By

Gregory Jon Velicer

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Genetics Program

1997

ABSTRACT

COMPARATIVE AND EXPERIMENTAL TESTS FOR A TRADEOFF IN
BACTERIAL FITNESS AT LOW AND HIGH SUBSTRATE CONCENTRATION

By

Gregory Jon Velicer

It is commonly assumed that microorganisms that are
superior competitors under abundant resource conditions are
relatively inferior under scarce resource conditions, and
vice versa. The studies described in this dissertation test
this tradeoff hypothesis by both comparative and experimental
approaches. In the comparative study, growth rates of seven
natural strains of bacteria capable of degrading the
herbicide 2,4-dichlorophenoxyacetic acid (2,4-D) were
measured at both high and low concentrations of 2,4-D and
succinate. The results are suggestive of a positive
correlation between growth rates at high and low substrate
concentrations, contrary to the negative correlation
predicted by the tradeoff hypothesis. For the experimental
study, numerous replicate populations of 2,4—D degrading
bacteria were allowed to evolve independently under both
abundant (batch culture) and scarce (chemostat culture)
resource regimes. The competitive performance of each
evolved strain relative to its ancestor was measured in both
types of selective regime. The tradeoff hypothesis predicts
that adaptation to one selective regime will cause fitness
loss in the alternative regime. Competition results

demonstrate that this is not always the case. Numerous

evolved lines simultaneously improved their fitness in both
abundant and scarce resource selective regimes, refuting the
tradeoff hypothesis in its most general form. Changes in
performance in the alternative regime after adaptation to a
selective regime appear to be caused primarily by pleiotropic
side-effects (both beneficial and harmful) of selected
mutations. Also, experimental lines that adapted
significantly to one substrate (either 2,4-D or succinate) in
the batch regime show heterogeneous correlated performance
responses on the alternative substrate, with some lines
gaining and other lines losing fitness on the alternative

substrate.

This dissertation is dedicated to
my parents - Dr. Leland and Janet Velicer,
my brothers - Mark and Daniel,
my wife - Yuen—tsu,
and my daughter — Irene Marie.

iv

ACKNOWLEDGMENTS

I thank Professor Richard E. Lenski, my graduate
advisor, for his many hours of guidance and encouragement
over the past five years. The members of my guidance
committee, Larry Forney, Tom Schmidt, Larry Snyder and Jim
Tiedje, were very helpful and supportive throughout my
graduate school years in numerous ways.

Several people gave scientific assistance that was
important to the completion of this project, including
Brendan Bohannan, John Dunbar, Ryszard Korona, Cathy McGowan,
Judy Mongold, Cindy Nakatsu, Rob Sanford, and Tim Sassanella.
I thank Shenandoah Oden and Elizabeth Smalley for many long
hours of technical assistance and Lynette Ekunwe for helping
keep the Lenski lab functional and enjoyable to work in. The
friendliness and helpfulness of Paul Sniegowski, Doug Taylor,
Mike Travisano, Paul Turner and Farida vasi during my early
years in the Lenski lab is much appreciated.

Thanks to my grandparents, George and Elsie Velicer, for
giving generously to my education from hard—earned farm
money. Thanks to my parents, Dr. Leland and Janet Velicer,
for their many years of love and encouragement of my studies.
And thanks to my wife and daughter, Yuen—tsu and Irene Marie,
for making it a true joy to head home from the lab at the end

of each day.

TABLE OF CONTENTS

LIST OF TABLES ........................................... viii
LIST OF FIGURES ............................................ ix
INTRODUCTION ................................................ l
The microbial system ................................... 2
The comparative study .................................. 4
The experimental study ................................. 5
Effects of adaptation to a single substrate on
performance on a different substrate .............. 7
CHAPTER 1

APPLICATION OF TRADITIONAL AND PHYLOGENETIC COMPARATIVE
METHODS TO TEST FOR A TRADEOFF IN BACTERIAL GROWTH RATES

AT LOW VERSUS HIGH SUBSTRATE CONCENTRATION .................. 8
Introduction ........................................... 8
Materials and Methods ................................. ll

Bacterial strains ................................ 11
Culture medium ................................... 12
Estimation of growth rates ....................... 12
16S ribosomal DNA sequencing ..................... 15
Construction and scaling of phylogeny ............ 16
Results ............................................... 16
Traditional comparative method: application ...... 16

Phylogenetically independent contrasts: theory...26
Phylogenetically independent contrasts:

application ................................. 31
An indirect test of the preadaptation hypothesis.34

Discussion ............................................ 37
CHAPTER 2

EXPERIMENTAL TESTS FOR A TRADEOFF IN BACTERIAL FITNESS AT LOW

VERSUS HIGH SUBSTRATE CONCENTRATION ........................ 46

Introduction .......................................... 46

Experimental evolution overview .................. 49

Materials and Methods ................................. 53

Bacterial strains and culture media .............. 53

Stage I batch evolution .......................... 54

Stage I chemostat evolution ...................... 55

Stage II batch evolution ......................... 57

Stage II chemostat evolution ..................... 58

Batch competition experiments .................... 59
Measurement of selection rate constant for batch

competitions ................................ 63

vi

Experimental design of batch competitions ........ 64

Protocol and SRC for chemostat competitions ...... 65
Results ............................................... 66
Adaptation to selective regime ................... 66
Performance in alternative regime ................ 72
Homogeneous and heterogeneous responses to
selection .................................. 73
Interpretation of chemostat data ................. 76
Discussion ............................................ 82
Implications for the tradeoff hypothesis ......... 82
Strain history and evolutionary change ........... 87
Potential mechanisms of adaptation ............... 88
The strengths and limitations of this study ...... 90
CHAPTER 3
BACTERIAL COMPETITIVE PERFORMANCE ON ALTERNATIVE SUBSTRATES
AFTER ADAPTATION TO A SINGLE CARBON SOURCE ................ 96
Introduction .......................................... 96
Materials and Methods ................................ 101
Results .............................................. 103
Adaptation to selective substrate ............... 103
Performance on alternative substrate ............ 110
Discussion ........................................... 112
Population dynamics of adaptation ............... 112
Multiple adaptive pathways ...................... 114
Potential physiological mechanisms of
adaptation ................................. 116

Possible relevance to the adaptation of
genetically modified organisms in natural
habitats ...................................... 121

REFERENCE LIST ............................................ 124

vfi

Table
Table

Table

Table

Table

Table

Table

Table

Table

Table

10.

LIST OF TABLES

Phylogenetic tree branch lengths ................. 20
Growth rates of strains and ancestral nodes ...... 33

Chemostat dilution rates relative to ancestral
maximum growth rates ............................. 56

Dilution rates of competition experiments ........ 62

Results of competition experiments for evolved
lines from Stage I selection in succinate ........ 68

Results of competition experiments for evolved
lines from Stage II selection in 2,4-D ........... 70

Frequency of tradeoff and non-tradeoff patterns..86

Dilution factors and rates for batch competition
experiments ..................................... 102

Results of competition experiments for evolved
lines from Stage I selection in succinate and
Stage II selection in 2,4-D ..................... 105

Frequency of four types of correlated response
to selection ................................... 111

viii

Figure
Figure

Figure

Figure

Figure

Figure
Figure

Figure

Figure

Figure
Figure
Figure
Figure

Figure

Figure

10.
11.
12.
l3.
14.

15.

LIST OF FIGURES

Exponential growth of strains TFD2 and TFD20....14
Phylogenetic relationships ...................... 18
Correlations using the older comparative
method .......................................... 23
Phylogenetic effects on correlation analyses....25
Equidistant and hierarchical strain
phylogenies ..................................... 28
Correlations using new comparative methods ...... 36
Derivation of bacterial strains ................. 51
Tradeoff and non-tradeoff patterns of
adaptation ...................................... 74
Heterogeneous and homogeneous selection
responses ....................................... 75
Four types of chemostat competition data ....... 78
Hypothetical adaptive threshold for tradeoff...94
Derivation of bacterial strains ................ 98
Performance changes of Stage I lines .......... 107
Performance changes of Stage II F—derived
lines ......................................... 108

Performance changes of Stage II S-derived

lines ......................................... 109

ix

INTRODUCTION

The physiological structure and function of organisms
may involve tradeoffs in the allocation of limiting resources
to alternative selective demands. These tradeoffs can be
studied at a variety of levels, including genetic,
organismal, and population levels. The studies described in
this dissertation have taken population and evolutionary
approaches to test for a tradeoff between competitive fitness
at low versus high resource conditions.

To what extent can organisms optimize their ability to
compete both when their primary carbon source is abundant and
when it is scarce? Conventional wisdom holds that it is
impossible for an organism to be a superior competitor under
both of such radically different resource conditions.

Holding this view, microbial ecologists have often assumed
that microorganisms with high maximum growth rates under
nutrient rich conditions are generally poor competitors for
sparse resources, and Vice versa.

The evidence supporting this position is largely
anecdotal and comparative in nature. Here I describe
research that tests the resource concentration tradeoff
hypothesis, first by a systematic comparison of the growth

rates of several bacterial strains, and then by evolution

2

experiments designed to select for competitive fitness at

high and low resource concentrations.

The microbial system

The tradeoff hypothesis in question has relevance both
to theoretical issues in microbial ecology and evolution and
to more applied questions in bioremediation and industrial
processes. Attractive organisms for testing resource
concentration tradeoff theory are microbes that degrade
anthropogenic compounds. Many such substances, such as 2,4—
dichlorophenoxyacetic acid (2,4—D), are known to be toxic to
humans and potentially disruptive to ecological processes.
Effective bioremediation of toxic substrates "requires the
selection of inoculant strains able to effectively degrade
environmental pollutants over a wide range of concentrations"
(Greer et a1. 1992). Determining whether there is tradeoff
in bacterial fitness at low versus high resource
concentrations is therefore important for identifying the
most useful inoculant strains.

Current bioremediation practice focuses primarily on
enhancing growth conditions for indigenous microflora that
are capable of growth on an unwanted substrate. If, however,
a very clear tradeoff exists, such that some organisms are
superior at high substrate concentrations and others at low
concentrations, then it may be possible to improve the
effectiveness of bioremediation by applying microbes that are

appropriate for a particular site based on their kinetic

3

parameters. This strategy may thus enable "rapid degradation
at high pollutant concentrations or thorough degradation at
low pollutant concentrations" (Greer et al. 1992). 2,4-D is
a herbicide used worldwide that has been subject to much
research regarding its toxic effects (Brueggeman 1993, Ebasco
and Herra 1993). Research on the ecology, metabolism and
genetics of 2,4-D degrading bacteria is among the most
extensive for microbes used in bioremediation. Therefore,
2,4-D degrading bacteria are an excellent system to use for
testing the substrate concentration tradeoff hypothesis.
Knowledge of the genetics underlying 2,4—D degradative
functions may allow future molecular analyses of adaptive
genetic changes that have occurred during the experimental
evolution portion of the studies reported in this
dissertation. Bacterial strains vary in their complement of
genes encoding the 2,4-D degradative pathway, which can be
borne either on the chromosome or on plasmids. The
degradative pathways for 2,4-D metabolism and the underlying
genes have been thoroughly reviewed by Calabrese (1994).
Moreover, 2,4-D degradative functions exhibit considerable
diversity at the levels of gene organization, pathways,
physiological traits, and bacterial classification. (Don and
Pemberton 1981; Calabrese et al. 1992). Horizontal transfer
and the subsequent evolution of a relatively few ancestral
2,4—D metabolic genes are believed to account for the current

diversity (Don and Pemberton 1981; McGowan 1995).

The comparative study

Numerous previous studies in evolutionary biology have
employed either a comparative or an experimental approach to
test hypothetical tradeoffs between different performance
measures. Unlike most of these previous studies, I have
brought both comparative and experimental tests to bear on
the substrate concentration tradeoff hypothesis, within the
context of a single biological system. The comparative
approach involved testing for statistical correlations
between growth rates measured at high and low resource
conditions using seven strains of 2,4-D degrading bacteria
isolated from nature (Chapter 1).

The tradeoff hypothesis predicts a negative correlation
between growth rates at high and low resource concentrations.
I tested this prediction by measuring the growth rates of the
seven strains at 5 ug/ml and 500 ug/ml of each of two
substrates. The analysis of growth rates was performed using
two different methods, one traditional and the other a modern
approach that incorporates phylogenetic information. The
traditional method treats each strain as an independent
observation, and it is the method that has been typically
employed by microbial ecologists. The modern methods
incorporates information on the phylogenetic relationships of
the strains whose properties were measured, thus avoiding

spurious correlations that may reflect unique historical

5

events rather than general evolutionary trends (Harvey and

Pagel 1991).

The experimental study

Evolutionary experimentation using bacteria is a
powerful method for testing hypotheses in population biology.
Such laboratory evolution has given important insight into
numerous biological phenomena at multiple levels. At the
genetic level, such studies have addressed the evolution of
pleiotropy and epistasis (Lenski 1988) and the effects of
mutation rate differences on fitness (Chao and Cox 1983).
Experimental evolution has also addressed more ecological
aspects of evolution, including the coevolution of bacteria
and viruses (Lenski and Levin 1985) and the effects of
antibiotic production on competitive interactions (Chao and
Levin 1981). Physiologically oriented studies have addressed
the relationship between metabolic flux and fitness
(Dykhuizen et al. 1987; Dykhuizen and Dean 1990). Basic
theoretical issues in evolutionary biology such as the roles
of adaptation, history and chance in evolutionary change
(Travisano et al. 1995), as well as applied corcerns such as
the enhancement of competitive fitness in the biodegradation
of toxic compounds (Korona et a1. 1994), have been tested by
the use of experimental evolution.

Rapid bacterial growth during experimental evolution
allows changes to be tracked over evolutionarily significant

time scales. Changes in population composition can be

6

accurately monitored using genetic and phenotypic markers
that can be easily scored. Environmental parameters can be
precisely controlled, allowing strong inferences about the
causal effects of natural selection. Large population sizes
ensure that a great number of spontaneous mutations occur
every generation. Other important features of bacteria
include the ability to freeze ancestral and experimental
lines for later analyses, and to measure relative fitness in
any environment by competing derived lines against their
ancestors. Moreover, the general ease of culture, storage,
and analysis make it feasible to obtain replicate lines for
each experimental treatment. Replication of lines allows one
to assess whether genetic responses to imposed selection are
uniform or heterogeneous.

To complement the comparative study, I conducted a two—
stage evolution experiment that allowed evolving populations
of 2,4—D degrading bacteria to adapt to either high or low
substrate conditions (Chapter 2). To test the tradeoff
hypothesis, evolved lines were placed in competition with
their ancestors to assess their performance under regimes of
both resource scarcity and abundance. The tradeoff
hypothesis predicts that lines that adapt significantly to
one regime must lose competitive fitness in the alternative

regime.

7

Effects of adaptation to a single substrate
on performance on a different substrate
In addition to testing the tradeoff hypothesis described
above, I have used the same experimentally evolved lines to
test the consequences of adaptation to one substrate for
competitive performance on an alternative substrate (Chapter
3). Such consequences are relevant to concerns about the
evolution of genetically modified organisms following their
release into the environment. In particular, it is often
assumed that genetically modified organisms will perform
their intended function and then “fade away," thus minimizing
concerns about any unintended effects they may have.
However, modified organisms used in bioremediation can be
expected to adapt evolutionarily to growth on the
anthropogenic substrate that they are intended to degrade.
If such adaptation results in improved competitiveness for
alternative, naturally occurring substrates, then this will
increase the likelihood that the modified organisms will
persist in the environment. Therefore, it becomes important
to ask how adaptation to growth on an anthropogenic
substrate, such as 2,4-D, affects competitiveness for a

natural substrate, such as succinate.

Chapter 1

APPLICATION'OP TRADITIONAL AND PHYLOGENETIC COMPARATIVE
METHODS TO TEST FOR.A.TRADEOFP IN BACTERIAL GROWTH RATES AT
LOW VERSUS’HIGH SUBSTRATE CONCENTRATION

INTRODUCTION

It is widely assumed that bacterial species cannot be
successful competitors at both low and high resource
concentrations, but instead they must pursue one strategy or
the other (Matin and Veldkamp 1978; Gerson and Chet 1981;
Greer et al. 1992). This view is closely related to the
distinction between r— and K—strategists in animals and
plants (Pianka 1970; Mueller and Ayala 1981). When applied
to microorganisms, Andrews and Harris (1986) expect r—
strategists to have high maximum growth rates and to require
abundant resources to support their rapid growth, whereas K;
strategists should have lower maximum growth rates and
require fewer resources to support their slower growth. In
other words, maximum growth rate (”max) is viewed as the
primary determinant of its competitiveness when resources are
abundant, whereas substrate affinity (Kg) becomes more
important when resources are scarce.

Understanding the nature and generality of tradeoffs in
bacterial competitiveness at low versus high resource
concentrations may have practical application for

bioremediation of soils and groundwater contaminated by

8

9

xenobiotic compounds (Greer et al. 1992). That is,
successful bioremediation may depend on identifying bacterial
strains whose growth parameters are well suited to site—
specific conditions in terms of resource concentration.
However, prior studies that have looked for possible
tradeoffs in growth rates as a function of resource
concentration have yielded mixed results (Greer et al. 1992;
Luckinbill 1984; Vasi et a1. 1994; Zhou and Tiedje ms). Some
of these studies employed an experimental approach, in which
bacteria were allowed to evolve in the laboratory; the
tradeoff hypothesis was then tested by comparing the growth
properties of ancestral and derived genotypes. Other studies
have used a comparative approach, in which growth parameters
were estimated in the laboratory for bacterial strains that
had been isolated from nature and correlations between these
parameters tested.

In recent years, evolutionary biologists have recognized
a serious problem with basing inferences about evolutionary
adaptations and tradeoffs upon the traditional comparative
method (Harvey and Pagel 1991). The traditional comparative
method treats each strain (or species) as an independent
observation for purposes of statistical analysis, but in fact
certain different pairs of strains (or species) are more
closely related to one another than are other pairs.
Consequently, not all observations are truly independent,
creating the opportunity for spurious correlations between

various organismal traits. In other words, phylogenetic

10

relatedness tends to produce phenotypic similarity simply
because of common ancestry, whether or not the resulting
pattern has any particular adaptive significance. For
example, consider the universe of all warm-blooded
terrestrial vertebrates (i.e., birds and mammals). Treating
each species as an independent observation, there is a very
strong association between viviparity (as opposed to
oviparity) and the possession of hairs (as opposed to
feathers). But viviparity and hairiness may each have
evolved only once in these combined groups, and so the
association between these two traits may reflect a sort of
historical accident rather than having any adaptive
significance. Fortunately, new methods for comparative
analysis have been developed that incorporate phylogenetic
relationships between species and thereby allow reliable
statistical inferences about associations between traits of
interest (Clutton-Brock and Harvey 1977; Stearns 1983;
Cheverud 1985; Lynch 1991; Felsenstein 1985; Grafen 1989;

Harvey and Pagel 1991).

For this study, I performed comparative analyses to test
the hypothesized tradeoff in bacteria between relative
performance at high and low resource concentrations, using
both the traditional approach (in which each strain is
treated as an independent observation) and newer methods that
depend on phylogenetically independent contrasts. Seven
strains of bacteria isolated from nature were examined, all

of which can grow on either 2,4-dichlorophenoxyacetate (2,4-

11

D) or succinate as a sole carbon source. Growth rates of all
the strains were measured at both low and high concentrations
of each substrate. I will show that the two comparative
methods yield different results, although neither analysis
provides compelling support for the tradeoff hypothesis. I
conclude by discussing possible explanations for these
results as well as considering an alternative approach to

study this issue.

MATERIALS AND METHODS
Bacterial strains
All of the bacteria used in this study were isolated

from either soil or sludge, and all are able to catabolize
2,4-D (Tonso et al. 1995). Designations and geographical
origins of the seven strains are as follows: TFD2 (Michigan),
TFD3 (Oregon), TFD6 (Michigan), TFD13 (Michigan), TFD20
(Michigan), TFD41 (Michigan), and TFD43 (Australia). Based
on 16S ribosomal DNA sequences (see below), these seven

strains all belong to the B subgroup of the Proteobacteria

(McGowan 1995). I also sought to include in our study
several 2,4—D degrading strains from the a and.y subgroups of
the Proteobacteria in order to increase phylogenetic
diversity, but these other strains did not grow reliably in

our standard culture medium.

12

Culture medium
Growth rates were measured in acid—washed flasks using

the same base mineral medium for all strains and experiments.

The base medium contained, per liter, 1.71 g K2HPO4, 0.3 g
Na2PO4, 0.33 g (NH4)2$O4, 0.246 g MgSO4O7H20, 0.12 g
Na2EDTA02H20, 20 mg NaOH, 4 mg ZnSO4-7H20, 3 mg MnSO40H20, 1
mg CuSO405HZO, 30 mg FeSO407H20, 52 mg Na2S04, and 1 mg
NaMoO402H20. Succinate or 2,4-D was added to the base medium

at a concentration of either 5 or 500 ug/ml.

EStimation of growth rates

I initially sought to estimate two growth parameters for

each strain: “max (maximum growth rate) and Ks (resource
concentration that supports a growth rate of ”max/2)-

However, I was unable to obtain reliable estimates of KS due

to various technical difficulties (Lenski et al. 1997). For
example, certain strains grew somewhat even when we added no
carbon to the medium (using acid-washed flasks and double—
distilled water); the cells presumably used some airborne
organic material (Geller 1983). Other strains exhibited
maximum growth rates at intermediate concentrations of
substrate, in contrast to the hyperbolic form of the Monod
model (G.J.V., unpublished observations). Therefore, I opted
instead to measure each strain's growth rate at two substrate

concentrations, one low (5 ug/ml) and one high (500 ug/ml),

in order to test the tradeoff hypothesis.

13

Strains were removed from storage at -80°C and
inoculated into 10 ml of medium (in 50-ml flasks) containing
either 2,4-D or succinate (500 ug/ml). These initial

cultures were allowed to grow for three days at 25°C while

being shaken at 120 rpm. Cells from each initial culture
were then transferred into fresh medium containing either 5
or 500 ug/ml of the corresponding substrate and allowed to
grow for four days; this constituted a preconditioning step
to ensure that cells were physiologically acclimated to the
relevant medium. Cells from each preconditioning culture
were transferred again into fresh medium with the same
substrate and initial concentration to begin the experiment
proper. Each combination of strain, substrate, and
concentration was replicated three—fold. Bacterial cultures
were sampled repeatedly at 3-5 hour intervals over a period
of 24-30 hours. At each time point, the total biovolume
(cell number x average cell size) of a culture was measured
using a Coulter particle counter. Biovolume measurements
were log—transformed, and a growth rate was calculated as the
rate of change in biovolume during the period of exponential
increase. The mean of the three independently estimated
growth rates (for each combination of strain, substrate, and
concentration) was used in the comparative analyses.
Examples of log—transformed population volume data for
strains TFD2 and TFD20 during growth on 500 Hg/ml succinate

are shown in Figure 1.

14

18.5— a

16.5—
16— I

Ln [population volume (fl)]

15.5~

 

l l l l l l T l l l l

Hours

185— b

—L
m
L

17.5—

—L
\l
l

16.5j

—L
a)
1
ID

Ln [population volume (fl)]
<O|I

15.5—

 

15 l l l I l I T l l l

Hours

Figure 1. Exponential growth of strains TFD2 and TFD20.
Log-transformed population biovolume data (in femtoliters)
are shown for (a) a relatively fast growing strain (TFD2),
and (b) a slow growing strain (TFD20) during growth on 500
ug/ml succinate. Population biovolume equals the average
cell size multiplied by the number of cells. Open circles,
closed diamonds and closed squares represent three
independent replicate populations, respectively, for both
TFD2 and TFD20.

15

16S ribosomal DNA sequencing

For six of the strains used in this study, 380 basepairs
(bp) of 16S ribosomal DNA sequence were available from an
earlier study (McGowan 1995). Using the Macromolecular
Sequencing Facility at Michigan State University, I also
sequenced the seventh strain, TFD13, using the same PCR and
sequencing primers that were used previously for the other
strains. All of the sequences are downstream of the primer
that begins at ESCherichia coli position 519 in the 16S
ribosomal DNA gene. The 16S rDNA gene was first amplified by
PCR primers fDl (5'ccgaattcgtcgacaacAGAGTTTGATCCTGGCTCAG 3')
and rDl (5' cccgggatccaagcttAAGGAGGTGATCCAGCC 3') (Weisburg
et al. 1991). A sample (1 ml) of stationary phase TFD13
culture (grown in Luria broth) was centrifuged at 14,000 rpm
for 10 minutes and resuspended in 1 ml of sterile double-
distilled H20. PCR was conducted in multiple reaction vials,
with each containing 5 pl of resuspended bacterial culture,
0.5 pl of each primer (10 pmol/ml), 10 pl of 10X reaction
buffer with MgClZ, 2 pl of mixed dNTP, 1 p1 of Taq DNA
polymerase, and 81 pl of distilled H20. Reaction solutions
were overlaid with mineral oil. Amplifications were
performed in a Perkin-Elmer DNA Thermal Cycler 480 using a
sequence of 6.5 minutes at 95°C; 35 cycles of 60 seconds at
94°C, 60 seconds at 40°C, and 60 seconds at 72%:; 10 minutes
at 72°C; and maintenance at 4°C. PCR products were purified

using a Qiagen Qiaquick-spin PCR kit, eluted into 50 ml of TE

16

buffer, and stored at —20°C. Then, 9 pl of the purified PCR
product solution was added to 9 pl of H20 and 2 pl of 10

pmol/ml primer 519R (5' GTATTACCGCGGCTGCTGG 3').

Construction and scaling of the phylogeny

The PAUP (Phylogenetic Analysis Using Parsimony)
computer program (Swofford 1991) was used to infer the
phylogeny of the bacterial strains included in this study,
using all 380 bp of 16S ribosomal DNA sequence (Figure 2). A
bootstrap analysis (“branch and bound" method, 100
iterations) was performed to ascertain confidence in the
resulting phylogenetic tree, and the percentage of bootstrap
samples supporting each grouping is so indicated (Swofford
1991). The estimated branch lengths between ancestral nodes
were lengthened slightly using a method that seeks to correct
systematic biases in estimating ancestral character states
(see Table 1, Felsenstein 1985). Growth rates corresponding
to each inferred ancestor were calculated as the average of
the growth rates for the extant strains derived from the
ancestor, weighted by the estimated branch lengths between

ancestors and derived strains (Felsenstein 1985).

RESULTS

Traditional comparative.method: application
The tradeoff hypothesis predicts a negative correlation
between growth rates at high and low substrate

concentrations. I first tested this hypothesis using the

17

Figure 2. Phylogenetic relationships. The phylogenetic
relationships of the seven strains used in this study were
inferred from 380 bp of their 16S ribosomal DNA sequences
using the PAUP program (Swofford 1991). The letters at each
node indicate ancestral nodes. A “branch and bound"
bootstrap analysis (Swofford 1991) was performed to ascertain
confidence in the inferred relationships. The percentage of
bootstrap samples supporting each grouping are indicated on
the branch below each node. See Table 1 for branch lengths.

18

 

 

 

 

 

 

 

 

 

 

 

TFD2
‘°° B TFD3
57
"—0 rTFD13
97 D
A TFD20
96
E
TFD6
»— TFD41
100 F
TFD43

 

 

Figure 2. Phylogenetic relationships.

19

Table 1. Phylogenetic tree branch lengths. Scaled and
unscaled branch lengths for the phylogenetic tree of seven
strains of 2,4—D degrading bacteria shown in Figure 2.
Unscaled branch lengths reflect the number of base pair (bp)
differences between strains over 380 bp of 16S rDNA sequence.
The estimated branch lengths between ancestral nodes were
lengthened slightly by a method that seeks to correct

systematic biases in estimating ancestral character states
(Felsenstein 1985).

20

Table 1. Phylogenetic tree branch lengths.

Branch Segment Scaled Length unsealed Length

A to B 14.55 11
B to 2 5 5
B to C 12.27 7
C to 3 13 13
C to D 8.87 8
D to 13 1 1
D to E 6.5 6
E to 6 1 1
E to 20 1 1
A to F 19.33 16
F to 41 S 5
F to 43 10 10

21

traditional comparative method, which treats each strain as
an independent observation. Both succinate (Figure 3a) and
2,4—D (Figure 3b) gave positive correlations between growth
rates measured at 5 and 500 pg/ml. For succinate, this
correlation was significant (r = 0.7747, p = 0.0408), whereas
for 2,4—D the correlation was not quite significant (r =

0.7431, p = 0.0556).

This traditional analysis clearly does not indicate
any tradeoff between performance at high and low substrate
concentrations. Rather, it suggests the opposite trend,
whereby some strains are superior at both low and high
substrate concentrations. However, there are two important
considerations that may render this conclusion invalid.
First, it is inappropriate to treat each strain as an
independent observation, unless one uses phylogenetic
information to scale the data in order to meet the
assumptions of standard statistical tests. Failure to do so
may dramatically alter the apparent correlation between
traits. For example, growth rates at high and low substrate
concentrations may be positively correlated for one pair of
distantly related taxonomic groups, even though several pairs
of more closely related strains within each group show the
predicted negative correlation (Figure 4). In the next
section, I will reanalyze this data using comparative methods
that incorporate information on phylogenetic relationships.
Second, some strains may be fortuitously preadapted to the

laboratory environment for reasons that have nothing to do

Figure 3. Correlations using the older comparative method.
Correlations between growth rates at low and high substrate
concentrations for (a) succinate and (b) 2,4-D are shown.

The traditional method uses all strains as independent
observations, regardless of their phylogenetic relationships
to one another. The correlation is significant for succinate
(r = 0.7747, p = 0.0408). The correlation is not quite
significant for 2,4-D (r = 0.7431, p = 0.0556).

23

Growth rate (per hour) on 5 pg/ml succinate
o
a
l

 

 

l I a r I l I I I
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45

Growth rate (per hour) on 500 pg/ml

 

Growth rate (per hour) on 5 pg/ml 2,4-D
o
T

 

 

0.08—

0.06— I

0'04 l I l l l l
o 0.05 0.1 0.15 0.2 0.25 0.3

Growth rate (per hour) on 500 pg/ml 2,4-D

Figure 3. Correlations using the older comparative method.

24

Figure 4. Phylogenetic effects on correlation analyses.
Growth rates schematically illustrated at low and high
substrate concentrations are positively correlated across two
distantly related taxonomic groups (circles and squares),
even though several pairs of more closely related strains
within each group exhibit the negative correlation predicted
by the tradeoff hypothesis. Lines connect the most closely
related pairs. Circles correspond to species A - F in Figure
5b, while squares represent species G - L. The overall
correlation would appear to be positive using the
traditional comparative method that treats each species as an
independent observation. Newer comparative methods take into
account phylogenetic relationships among strains and could
reveal the preponderance of negative associations between the
two traits.

25

 

 

0.1—
.9
c0 0.
*3 0.09- 0-..
'8 0 .....
a) O O
E 0.08— "'o
BI
:1.
In
a 0.07—
"3
O
f 0.06—
% l...
o °I I
E 0.05- I. .......... I
e _
0 0.04
0.1 0.15 0.2 0.25 0.3

Growth rate (per hour) at 500 pg/ml substrate

Figure 4. Phylogenetic effects on correlation analyses.

26

with the competitive relationships between strains in the
soil or other natural environments. For example, one strain
may grow much faster than another — at both low and high
substrate concentrations — simply because the first strain is
better suited to the particular combination of base mineral
medium, temperature, and other variables used in the
laboratory experiments. If one could analyze performance in
an environment more similar to that in which the strains had
evolved, then one might see the predicted tradeoff. In a
later section, I examine evidence that certain strains are
fortuitously better adapted to the laboratory environment

than are other strains.

Phylogenetically independent contrasts: theory'

The purpose of using phylogenetic information in
comparative studies is to obtain the truly independent data
that are required for robust statistical analysis, data that
reflect independent evolutionary events. If all of the
species in a study shared a common ancestor from which they
had evolved for the same length of time, then these species
could be appropriately treated as statistically independent
observations (Figure 5a). But if the species are
hierarchically related (as in most phylogenies), then
statistical problems arise because of the relative lack of
independence between the states of more recently diverged

lineages (Figure 5b). Statistical problems may also arise if

27

Figure 5. Equidistant and hierarchical strain phylogenies.
Strains (or species) are properly treated as independent
observations only if they shared a common ancestor from which
they had all evolved for the same length of time. This would
require a "star" phylogeny, in which all species pairs are
equidistant (a). More generally, pairs of strains (or
species) are hierarchically related (b), and newer
comparative methods have been developed that take into
account these relationships.

28

 

 

Ll % L1 L1 LHlLH

 

 

 

 

 

Figure 5. Equidistant and hierarchical strain phylogenies.

29

evolutionary rates vary among the lineages (Harvey and Pagel
1991).

Several different methods have been developed in recent
years to incorporate phylogenetic information into
comparative analyses. Some of the earliest methods relied on
making contrasts only between independent pairs of extant
species (Clutton—Brock and Harvey 1977; Stearns 1983;
Cheverud 1985; Lynch 1991), but these methods lacked
statistical power because they discarded potentially useful
information. More recent methods are preferable because they
make use of all the relevant information that is available in
the data (Felsenstein 1985; Grafen 1989; Harvey and Pagel
1991).

Felsenstein's (1985) method of independent comparisons
makes use of the fact that any two extant species (say, 1 and
2) at adjacent tips on a phylogenetic tree share a unique
common ancestor. Consequently, any evolutionary change in
the traits of interest (say, X and Y) since 1 and 2 split
from their most recent common ancestor is independent of
changes in X and Y at all other locations in the phylogenetic
tree. Hence, the differences in traits X and Y between
species 1 and 2 are statistically independent from the
differences in X and Y between other such pairs of adjacent
species in the tree. For example, (X1 — X2) and (Y1 - Y2)
are independent of (X3 - X4) and (Y3 - Y4). Moreover, this
independence holds for adjacent pairs of nodes at higher

levels (i.e., earlier in time) in the tree. By using this

30

method, one can extract n — 1 independent contrasts from n
extant species, when all of the differences between adjacent
tips and nodes are calculated. Statistical tests of the
correlation between traits of interest are then made using
these n - 1 independent contrasts (weighted for expected
variance based on branch lengths). This method assumes that
the true branching pattern and branch lengths of the
phylogeny are known, or at least that they can be estimated
with some confidence.

Of course, one cannot know the actual values of the
traits at the ancestral nodes, but these can be estimated by
assuming a particular model of evolutionary change.
Felsenstein's method assumes a model of evolutionary change
that is equivalent to random Brownian motion. In such a
process, each trait has an equal likelihood of a positive or
negative change of equal magnitude over any given period of
time. Hence, the expected mean change is zero, and the
expected variance around this mean is directly proportional
to the elapsed time (estimated from the branch lengths). The
value of any trait at each ancestral node is therefore
estimated simply as the midpoint of its descendant lineages
scaled by the distances of those lineages from the node.

Various modifications of these methods have been
developed that assume different models of evolutionary change
(Grafen 1989; Harvey and Pagel 1991). Martins and Garland
(1991) ran computer simulations to examine the effect of

these different assumptions on the statistical validity of

31

the resulting evolutionary inference. Although they found
that these assumptions had some effect on statistical
validity, fortunately their effects were usually minor. More
importantly, all of the methods that incorporate phylogenetic
relatedness (but which assume different models of
evolutionary change) performed much better than the
traditional approach (which ignores phylogeny and treats each
species as an independent observation), even when the assumed
model of evolution is incorrect. Therefore, in practice,
methods that use phylogenetically independent contrasts will
yield more reliable inferences than traditional comparative
analyses, even when the assumptions that underlie the

phylogenetic method are violated.

Phylogenetically independent contrasts: application

I used Felsenstein's method of independent contrasts
(see above) to re-examine the correlation between bacterial
growth rates at high and low substrate concentrations, using
the phylogenetic relationships among the seven strains
depicted in Figure 2. Table 2 gives the growth rates that
were measured for each strain as well as the growth rates
inferred for all of the ancestral nodes. The branch lengths
of the phylogenetic tree, both scaled and unsealed, are given
in Table 1.

In succinate, the significant positive correlation
between growth rates at high and low concentrations, based on

the traditional comparative method, is not supported by the

32

Table 2. Growth rates of strains and anestral nodes.
Growth rates were measured for each strain, and growth rates
for each ancestral node in the strain phylogeny were also
inferred (see Figure 2). The values shown for each strain
are averages based on three replicate determinations. The
values inferred for each ancestral node used the method of

Felsenstein (1985). The growth rates have units of hr‘l.

Table 2. Growth rates of strains and ancestral nodes.

Strain or Succinate 2,4-D
Ancestral Node 5 pg/ml 500 pg/ml 5 pg/ml 500 pg/ml

TFD2 .1341 .3729 .0905 .1661
TFD3 .2790 .4199 .1359 .2073
TFD6 .0800 .0539 .0606 .0941
TFD13 .0899 .0454 .1078 .0799
TFD20 .1011 .0524 .0994 .0605
TFD41 .0737 .2283 .0909 .1128
TFD43 .2942 .3737 .1498 .2630
Node A .1451 .3027 .1035 .1590
Node B .1435 .3222 .0982 .1560
Node C .1666 .1979 .1170 .1314
Node D .0900 .0464 .1041 .0796
Node E .0906 .0532 .0800 .0773
Node F .1472 .2768 .1105 .1629

34
method of phylogenetically independent contrasts (Figure 6a:

r = 0.5401; p = 0.2108). In 2,4-D, the phylogenetic method
again gives no indication of a significant correlation
(Figure 6b: r = 0.5373; p = 0.2136), whereas the traditional
method suggested a positive correlation. Thus, by employing
phylogenetically independent contrasts, comparative data that
seemed to provide compelling evidence that flatly
contradicted the tradeoff hypothesis now appears to be merely
inconclusive. These changed results, although inconclusive
with respect to the tradeoff hypothesis, suffice to show that
phylogenetic considerations are important for comparative

studies in microbial ecology.

An indirect test of the preadaptation hypothesis

In short, the comparative approach provides no support
for the hypothesized tradeoff between relative growth rates
at high and low substrate concentrations. A possible
explanation for the failure of this hypothesis is that
certain strains may be fortuitously preadapted to growth in
the laboratory. Thus, one strain may grow faster than
another simply because the first strain prefers the
particular combination of mineral medium, temperature, and
other variables used in laboratory experiments at both high
and low substrate concentrations. A direct test of this
hypothesis is beyond the scope of this study and would

require the ability to perform competition experiments

Figure 6. Correlations using new comparative methods.
Correlations between growth rates at low and high substrate
concentrations for (a) succinate and (b) 2,4-D are shown.
These new methods take into account phylogenetic
relationships among strains. The correlations are
insignificant on both succinate (r = 0.5401, p = 0.2108) and
2,4-D (r = 0.5373, p = 0.2136), whereas the traditional
comparative method indicated stronger positive correlations
(Figure 3).

0.06 - a
0.05 -
0.04 3 I
0.03 -
0.02 1

0.01 l'

fl

 

0"""""'I I I I I I I n
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

 

Scaled growth rate contrasts - 5 pg/ml succinate

Scaled growth rate contrasts - 500 pg/ml succinate

0.03 —

5 pg/ml 2,4-1)

.0

o
N
l

0.01 3 I

 

Scaled growth rate contrasts

0

 

I I I I I I I I
O 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Scaled growth rate contrasts - 500 pg/ml 2,4—D

Figure 6. Correlations using new comparative methods.

37

between strains at high and low substrate concentrations in
situ in the environments from which the strains were
isolated. However, an indirect test can be performed by
asking whether growth rates on succinate and 2,4-D are
positively correlated. That is, all other aspects of the
growth conditions (medium composition, temperature, etc.) are
held constant, so that those strains that are fortuitously
preadapted to these conditions should benefit regardless of
the substrate that is provided.

To that end, I used the method of phylogenetically
independent contrasts to calculate the correlation between
growth rates on succinate and 2,4-D. At low substrate
concentrations (5 pg/ml), there is a marginally
nonsignificant positive correlation between growth rates on
succinate and 2,4-D (r = 0.7487; p = 0.0528). At high
substrate concentrations (500 pg/ml), the correlation between
growth rates on the two substrates is even weaker (r =
0.4932; p = 0.2607). The positive correlations across
substrates are consistent with the preadaptation hypothesis,
but the test is an indirect one and the statistics are again

inconclusive.

DISCUSSION
This study had two overlapping goals. The first of
these goals was to examine whether bacterial strains that are
good competitors at high substrate concentrations tend to be

inferior competitors at low substrate concentrations, and

38

vice versa. This tradeoff hypothesis, if valid, has
important implications for understanding the structure of
microbial communities as well as for identifying bacterial
strains that are most likely to be useful in applications
such as bioremediation. Previous tests of this hypothesis
have yielded mixed results (Luckinbill 1978, 1979, 1984;
Greer et al. 1992; Lenski et al. 1997; Zhou and Tiedje ms),
and the results of my study must be added to those that do
not support this hypothesis. However, there are important
caveats to this interpretation that will be considered below.
The second goal of our study was to illustrate the
application of recent advances in the comparative method to
microbial ecology (Felsenstein 1985; Harvey and Pagel 1989;
Harvey and Pagel 1991). The traditional comparative approach
has been to view each species or strain as an independent
observation when one calculates correlations between two
traits, such as growth rates at high and low substrate
concentrations. This approach implicitly ignores the
phylogenetic relationships among species or strains.
However, there is a general lack of statistical independence
of observations with respect to evolutionary hypotheses
because of these phylogenetic relationships. As a
consequence, the traditional comparative method may often
generate spurious correlations (Martins and Garland 1991). A
newer comparative method explicitly takes into account these
phylogenetic relationships and thereby avoids this problem.

In this study, I found that correlations between growth rates

39

at high and low substrate concentrations were weaker when the
phylogenetic relationships were taken into account than when
they were ignored (contrast Figures 3 and 6).

These recent advances solve certain statistical problems
with the comparative approach as it was formerly applied.
Nonetheless, there remain vexing problems in determining the
validity of inferences based on the comparative method. I
will now discuss two such problems.

It is typically assumed that phylogenetic relationships
among organisms can be determined using essentially any set
of molecular or other data capable of resolving evolutionary
relatedness. The presumption is that any one gene has the
same genealogy (at the level of species and higher taxa) as
do all other genes. But with bacteria, this assumption may
be violated, especially for those traits that are encoded by
plasmids or other mobile genetic elements. Phylogenetic
relationships among the strains used in this study were
inferred from chromosomal sequences encoding the 16S
ribosomal subunit. However, tfdA and other genes that encode
the ability to catabolize 2,4-D are often plasmid-borne (Don
and Pemberton 1981), so that the phylogenetic relationships
used for independent contrasts may have been inappropriate
for testing the correlation between growth rates at high and
low concentrations of 2,4-D.

Indeed, McGowan (1995) recently showed that the
phylogeny of bacteria that degrade 2,4-D obtained using the

16S ribosomal gene is different from the phylogeny of the

4o

tfdA genes themselves, which implies horizontal gene transfer
(see also Dykhuizen and Green 1991). When evaluating the
potential tradeoff between growth rates at high and low
concentrations of 2,4-D, it is unclear whether the tfdA or
168 ribosomal gene phylogeny is more appropriate, because it
is unclear a priori which genes mediate this hypothetical
tradeoff. Therefore, in addition to scaling data based on
the 16S ribosomal gene phylogeny, I did the same scaling
based on the phylogeny of gene tfdA for five of the seven
strains used in the ribosomal gene analysis for growth on
2,4-D (McGowan 1995). The qualitative patterns from the tfdA
based analysis are similar to those from the ribosomal based
analysis. Using unscaled growth rate estimates for the five
strains, there is a significant positive correlation between
growth rates at 5 and 500 pg/ml 2,4—D (r = 0.9466, 2—tailed P
= 0.0147). Scaling based on the tfdA phylogeny for these
five lines reduced the correlation coeffecient of this
relationship and made it statistically insignificant (r =
0.8104, 2-tailed P = 0.0962). Because of their qualitative
similarity to the results based on the 16S ribosomal gene,
these results using the tfdA gene suggest that the failure of
the tradeoff hypothesis is most likely not due to effects of
horizontal gene transfer.

Also, horizontal gene transfer is probably irrelevant to
the catabolism of succinate, which is an intermediate
compound in central metabolism. The fact that I failed to

find any evidence of a tradeoff between growth rates at high

41

and low succinate concentrations also indicates that possible
complications due to horizontal gene transfer were not
generally responsible for the failure of the tradeoff
hypothesis.

The second caveat arises because the hypothesized
tradeoff between performances at high and low substrate
concentrations may be masked by fortuitous differences among
strains in the extent of their preadaptation to laboratory
growth conditions. Imagine, for example, that some strains
are adapted to habitats that are similar in temperature,
mineral composition, and so on to the conditions used in my
laboratory experiments, whereas other strains prefer
dissimilar conditions. Certain strains would have a
consistent advantage over other strains regardless of
substrate concentration, and these differences would
therefore promote a positive correlation between growth rates
measured at low and high substrate concentrations. This
effect might therefore obscure or even override any negative
correlation due to the hypothesized tradeoff.

Several evolutionary studies with higher organisms have
shown this systematic bias against detecting tradeoffs.

These studies have relied on contrasting the results of
comparative studies in which the performance capabilities of
natural isolates are measured under artificial conditions
dissimilar to those under which they evolved with the results
of experimental evolutionary studies. In these experimental

studies, populations are allowed to evolve in the laboratory,

42

so that the organisms are adapted evolutionarily to the same
artificial conditions in which their performance capabilities
will be measured. For example, to test the predictions of
the theory of r— and K-selection, Mueller and Ayala (1981)
allowed replicated experimental populations of Drosephila
melanogaster to evolve for eight generations in the
laboratory at either high or low population density. They
then estimated the per capita growth rates of the derived
populations at different population densities, and they
compared these responses across the selected populations.
They observed the predicted tradeoff, such that the
populations selected at low density grew more slowly at high
densities than did those selected at high density, and vice
versa. However, this same tradeoff was not detected when
Mueller and Ayala (1981) performed a comparative study using
25 strains of D. melanogaster isolated from nature,
presumably because of fortuitous differences among these
strains in their preadaptation to the laboratory environment.
A direct demonstration of the confounding effect of
fortuitous preadaptation to environments that are novel (from
the standpoint of an organism's evolutionary history) was
made by Service and Rose (1985), also using D. melanogaster.
By means of experimental evolution, the authors demonstrated
a life-history tradeoff in the flies between early fecundity
and longevity, when the assays of fecundity and longevity
were performed in the same culture medium that had been used

during the experimental evolution. However, when these same

43

traits were measured in a novel culture medium, the tradeoff
was largely obscured.

It is clear from these studies that the power of the
comparative method suffers when performance traits are
measured under conditions that differ substantially from
those that prevailed during an organism's evolutionary
history. At the same time, it is also clear that evolution
experiments performed in the laboratory allow one to
circumvent this problem not by eliminating the differences
between nature and the laboratory, but rather by using
populations that have a defined history of adaptation to the
laboratory environment in which the performance assays are
conducted. Thus, I have also performed evolution experiments
(Chapter 2) to re-examine the tradeoff between growth rates
at low and high substrate concentrations. My experiments
proceeded in two stages. The first stage allowed bacterial
strains isolated from nature to become adapted for many
generations to one of two laboratory growth conditions:
chemostat culture (wherein cells perpetually experience
resource limitation) and serial batch culture (wherein cells
are periodically given a surfeit of resources). Then, in a
second stage, populations selected under one regime were
propagated for many more generations under the alternative
regime, and vice versa, with control populations continuing
under their former regimes. I then measured the competitive
performance of the derived strains at both low and high

substrate concentrations to ascertain whether or not

44

tradeoffs have occurred during evolution under the two
regimes.

These evolution experiments are similar in outlook to
those performed by Luckinbill (1978, 1984) to test the
predictions of r- and Keselection theory. In Luckinbill's
experiment, replicated populations of E. coli were propagated
by serial batch culture under either r- or Keselection
regimes, with the only difference being that under the r—
selection regime the bacteria were transferred to fresh
medium before they reached their stationary-phase population
density. This experiment failed to show any tradeoff in
adaptation to these two environments. However, Vasi et al.
(1994) have suggested that Luckinbill's putative Keselection
regime may have selected for essentially the same growth
parameters as his r—selection regime. Their suggestion was
based on theoretical and empirical analyses of changes in
growth parameters in populations of E. coli evolving under a
serial batch regime; these analyses indicated that maximum
growth rate was the primary target of selection, just as it
would be under perpetual exponential growth. Therefore, in
my experiment a chemostat culture regime is employed to
generate more extreme selection for faster growth rate under
conditions of limiting substrate concentrations.

In conclusion, the hypothesis of a tradeoff in bacterial
growth rates between low and high substrate concentrations is
difficult to test because both comparative and experimental

approaches are fraught with subtle inferential problems. The

45

hypothesis is not supported by the results of this study.
Nonetheless, the tradeoff hypothesis is intuitively appealing
and may have considerable importance for the field of
microbial ecology. Therefore, efforts should continue to
develop new ways to test this hypothesis that avoid these

problems.

Chapter 2

EXPERIMENTAL TESTS FOR A TRADEOFF IN BACTERIAL FITNESS AT
W VERSUS HIGH SUBSTRATE CONCENTRATION

INTRODUCTION

Pleiotropic tradeoffs involving organismal fitness in
different environments have long been recognized as an
important dynamic of adaptive evolution (Reznick 1985; Bell
and Koufopanou 1986). Such tradeoffs arise from genetic,
structural, and physiological constraints on the simultaneous
optimization of different traits. Hypothesized tradeoffs
include r- vs..K- traits (MacArthur and Wilson 1967; Pianka
1970), adaptation to high vs. low temperatures (Mongold et
al. 1996), longevity vs. fecundity (Service and Rose 1985),
and growth vs. reproduction (Warner 1984).

r- vs. K- selection is among the most studied tradeoff
theories in ecology. r- selection occurs when the primary
component of fitness is the intrinsic rate of population
increase under low density conditions, while K- selection
optimizes population carrying capacity under high density
conditions (Roughgarden 1971). Characteristic r- traits
include high growth and death rates, short life spans, and
success in low density and fluctuating conditions.

Reciprocal K? traits include slow population growth and lower

mortality rates, increased longevity, and success in stable

45

47

and high density environments. Numerous investigations have
shown a tradeoff pattern between r- and K? type traits (Cody
1966; Mueller and Ayala 1981; Andrews and Rouse 1982), while
others have failed to reveal predicted patterns (Luckinbill
1978, 1979, 1984). One Specific prediction of r- and K—
selection theory is that organisms which compete well in
abundant resource environments should perform relatively
poorly under resource—limiting conditions (Andrews and Harris
1986).

Microbial ecologists commonly assume that tradeoffs
between competitive abilities under abundance vs. scarcity
are pervasive throughout the microbial world (Matin and
Veldkamp 1978; Konings and Veldkamp 1980; Kuenen and Harder
1982; Veldkamp et al. 1984; Greer et al. 1992). The rapid
growth of bacteria, their large population sizes, and their
ease of manipulation make them excellent organisms for
studying tradeoffs such as this one and for the study of
resource—based competition theory in general (Levin et al.
1977; Hansen and Hubbell 1980). The hypothesized tradeoff
between competitive ability at high vs. low resource levels
is not only important for ecological life-history theory, but
also has applied significance for the selection of microbes
useful for bioremediation (National Research Council 1993).

This study experimentally tests the substrate
concentration tradeoff hypothesis, as it shall be referred to
here. This hypothesis is most commonly expressed in terms of

kinetic parameters that are important components of fitness

43

at high and low substrate concentrations. At high substrate

concentrations, the maximum exponential growth rate (pmax) is

the major fitness component, whereas under conditions of

scarcity, “max/Ks becomes important, where K5 is the
concentration of substrate that allows a growth rate of 1/2
Hmax- (”max/Ks is a better indicator of competitive ability
than is Ks alone because an increase in Ks does not
necessarily reflect a decrease in fitness at low substrate
concentrations if accompanied by a parallel increase in
Mmax). There is some evidence in favor of the substrate
concentration tradeoff hypothesis from studies that compare
the kinetic parameters of small numbers of bacterial species
(Matin and Veldkamp 1978; Kuenen and Harder 1982; Veldkamp et
al. 1984). These studies, however, are not statistically
powerful and other more systematic studies have yielded
results that do not support the hypothesis (Greer et al.
1992, Lenski et al. 1997, Chapter 1).

The hypothesized tradeoff between the improvement of
flmax and pmax/Ks assumes a more general tradeoff involving
gross competitive fitness in abundant vs. scarce resource
conditions. Populations that evolve under abundant resource
conditions should simultaneously improve their performance at
high substrate levels and lose competitive ability under
scarce resource conditions, and vice versa. Therefore, I
have tested for a tradeoff between overall competitive
ability at high and low resource concentrations. Bacterial

populations were allowed to evolve under both types of

49

conditions and evolutionary changes in the competitive
ability of derived strains were measured in both the
environment in which they evolved and the alternative
environment. My results show that a tradeoff is not
compulsory because bacteria were capable of simultaneously
improving competitive ability at both high and low resource
concentrations. The substrate concentration tradeoff
hypothesis is not accurate as a general rule for bacteria,
but a significantly modified form of the hypothesis may still

be true.

EXperimental evolution overview

Evolution experiments were conducted in two temporally
distinct stages and in two different environments (Figure 7).
In the first environment - serial batch culture — bacterial
populations are transferred daily from stationary phase
cultures into fresh medium and allowed to grow back to high
density; hence populations experience high concentrations of
carbon source at regular intervals. Natural selection favors
mutants with improved competitive ability under high resource
conditions. In the second selective regime - chemostat
culture — evolving populations experience high density and
low substrate concentrations continuously as fresh media
flows into the cultures at a slow, constant rate and the high
equilibrium population size holds resource concentrations
very low. Here natural selection favors those mutants that

are best able to scavenge for scarce resources.

50

Figure 7. Derivation of bacterial strains. Horizontal lines
indicate periods of evolution with the selective regime and
sole carbon source indicated above and below, respectively.
Each horizontal line represents two independently evolved
replicate populations. See MATERIALS AND METHODS for a detailed
description of strains and nomenclature as well as the number
of evolutionary generations for each line.

51

 

 

 

Batch >FB <FBr
F< Chemostat >FC <Fcr .1“
Mt
Batch >SB <:: 331

”t

.<
Chemostat >80 <SC1

SCI
Stage I

(succinate)

Figure 7.

 

Batch 9 FEB

 

9 FBBr

 

>FCB

 

> FCBr

 

> SBB

 

> SBBr

 

>808

 

> SCBI

Chemostat > FBC
> FBCr

 

 

>FCXS

 

'9 FCCr

 

>SBC

 

9 SBCr

 

>800

 

> SCCr

Stage II
(2,4-0)

Derivation of bacterial strains.

52
A problem of testing tradeoff hypotheses by comparing

the traits of extant species is the different degrees of
“preadaptation" among natural isolates to the environment in
which traits are measured. For example, one bacterial strain
may perform relatively better than other natural strains at
both high and low resource concentrations not because it has
optimized performance under both conditions, but rather
because it happens to be better fit than are other strains to
lab variables other than substrate concentration. Such pre—
adaptation has been shown to obscure a tradeoff between early
fecundity and longevity when Drosqphila melanogaster traits
are measured in an environment other than that in which they
evolved (Service and Rose 1985).

To reduce the effect of any such differences between the
natural isolates used here on the rate and extent of
adaptation in my experiments, I conducted two stages of
evolution. The first stage was designed to allow adaptation
to general laboratory culture conditions so that subsequent
adaptation on a relatively uncommon carbon source would be
largely specific to that substrate. In Stage I of the
evolution experiments, two bacterial strains capable of
degrading the herbicide 2,4-dichlorophenoxyacetic acid (2,4-
D) were used to found populations that underwent 75 days of
evolution in either the batch or chemostat regime (Figure 7).
I was interested in observing whether bacterial strains that
had apparently been shaped by different selective forces in

nature responded to the laboratory selection regimes in

53

similar or different manners. Therefore, I selected one
strain that grew fast under laboratory conditions (on both
succinate and 2,4-D) and one that grew relatively slowly (on
both substrates) to start the evolving lines.

Separate lines were established from each natural
isolate in both the batch and chemostat environments for the
first stage (Stage I) of evolution, in which succinate (a
central substrate in bacterial metabolism) was the sole
carbon source. In the second stage (Stage II), populations
founded from Stage I lines were propagated in both batch and
chemostat environments in a complete factorial design with
2,4—D, which is catabolized into succinate, as the sole
carbon source. This design allows examination of the
specificity of adaptation not only with respect to culture

regime (this chapter), but also substrate type (Chapter 3).

MATERIALS AND METHODS
Bacterial strains and culture media

The two ancestral bacterial strains used in this study,
TFD3 and TFD13, are both capable of catabolizing 2,4-D and
were isolated from sludge in Oregon and soil in Michigan,
respectively (Tonso et al. 1995). These two strains, which
both belong to the genus Burkholderia in the B subgroup of
Proteobacteria, were selected because one of them (TFD3)
grows relatively fast in 2,4-D medium (”max = 0.21 hr'l),

while the other (TFD13) grows relatively slowly (Hmax = 0.08

hr‘l) (Chapter 1). Hereafter, TFD3 shall be designated "F"

54
(fast), and TFD13 will be referred to as strain "S" (slow).

All laboratory evolution and competition experiments were
conducted in the same base mineral salts base medium (AN

medium, Chapter 1), with either succinate or 2,4—D added to

the base medium at a concentration of 500 pg/ml.

Stage I batch evolution

Two replicate lines of both strain F and strain S were
established by inoculating 10 ml flasks of succinate medium
from frozen clonal cultures. Upon reaching stationary phase,
the lines were transferred into fresh medium. The new
cultures were then grown to stationary phase and the cycle
repeated daily. The transfer ratios were 1/100 (0.1 ml
transferred to 9.9 ml) for the F lines and 18/100 (1.8 ml
transferred to 8.2 ml) for the S lines. These ratios allowed
approximately 6.6 and 2.4 generations of growth per day,
respectively. Lines derived from F and S underwent 75 daily
transfer cycles, allowing about 500 generations of evolution
for F descendants and 180 generations for S lines. Upon
completion of Stage I evolution, each line was stored as a
whole population in 10% glycerol at -80° C. Single colonies
of Stage I lines evolved from F and S were isolated on agar
plates, cultures from which were also stored frozen. These
clonal cultures, hereafter designated FBI, FB2, SBl and SB2
(B = batch), were used for the competition assays comparing

ancestral with Stage I evolved strains. Two of these clones,

55

FBl and SBl, also were used as ancestors to found lines for

the second stage of evolution.

Stage I chemostat evolution

Two evolutionary lines each of strains F and S were
established by inoculating succinate medium into chemostat
vessels from the frozen, clonal ancestral cultures. Strain F
vessels were maintained at 18 ml and strain S vessels at 50
ml. Chemostat flow was maintained at 2.5 ml/hr, setting
growth rates at 0.139 hr"l for the F lines and 0.050 hr‘1 for
the S lines. These rates are one third and slightly over one
half (0.56) the maximum growth rates of F and S,
respectively, on 500 pg/ml succinate as estimated from cell
density data for replicate cultures during exponential phase
growth. (Chemostat dilution rates relative to ancestral
maximum growth rates are shown in Table 3.) Strain F
cultures underwent 75 days of chemostat evolution (about 360
generations), while strain S cultures were stored after only
73 days (about 130 generations) due to clogging of the
nutrient input tube from bacterial accumulation. Wall growth
was allowed to proliferate undisturbed throughout the
experiment. This had the effect of increasing cell density
and reducing substrate concentrations beyond theoretical
expectations and thereby amplifying the selective difference
between the batch and chemostat regimes. Chemostat vessels

were vortexed prior to withdrawing a sample for freezer

Table 3. Chemostat dilution rates relative to ancestral
maximum growth rates. The dilution rates for both stages of
chemostat evolution, maximum growth rates (pmax) of proximate
ancestors, and the ratios of dilution rate to flmax are shown
for each type of evolutionary line. Dilution rate is the
flow rate of media divided by chemostat volume.

Line Type Dilution Rate llmax of proximate Dilution Rate!
ancestor llmax
FC 0.139 0.420 0.331
SC 0.050 0.090 0.556
FBC 0.111 0.205 0.542
FBCr 0.111 0.217 0.511
FCC 0.111 0.204 0.545
FCCr 0.056 0.130 0.429
SBC 0.042 0.100 0.419
SBCr 0.042 0.104 0.402
SCCr 0.042 * *

* A reliable growth rate estimate for strain SCr (proximate
ancestor of SCCr1) was not obtained due to extensive clumping
of cells during exponential growth.

57

storage to allow a random mixture of cells from both the
liquid and wall populations of the chemostat.

Samples of each evolved culture were stored frozen in
10% glycerol. Single colonies of evolved chemostat cultures
were isolated on agar plates, cultures from which were stored
frozen. These clonal cultures, hereafter designated FCl,
FC2, SC1 and SC2 (C = chemostat), were used for the
competition assays comparing ancestral with evolved strains.
FCl and SC1 also became ancestors for the next stage of
evolution. All Stage I (and Stage II) evolved clones were
confirmed to be true descendants of strains F and S by
comparing electrophoretic gel patterns of REP PCR
amplification products of the ancestors and descendants
(versalovic et al. 1991). Also, single streptomycin-
resistant colonies of FBl, FCl, 881, and SC1 were selected
and stored frozen as clonal cultures. These cultures were
designated FBr, FCr, SBr, and SCr, and were used to found one

half of the Stage II evolution lines.

Stage II batch evolution
Two evolution lines each from FBl, FBr, FCl, FCr, SBl,
SBr, SC1 and SCr were established by inoculating flasks of 10
ml AN medium containing 500 pg/ml 2,4-D with single colonies
isolated from the frozen clonal cultures. Transfer cycles
were carried out on 2,4-D medium as during Stage I evolution
for 75 days. FBl, FBr, and FCl descendant lines were

transferred daily at 100-fold dilutions into fresh medium

58

(6.6 generations/day, 500 total generations), while FCr
descendants were transferred with 13-fold dilutions due to a
significantly slower growth rate (3.7 generations/day, 280
total generations). SBl and SBr descendants were transferred
at 5-fold dilutions (2.25 generations/day, 170 total
generations), while SC1 and SCr descendants were transferred
at 3-fold dilutions (1.5 generations/day, 115 total
generations). Mixed population samples were stored frozen at
25 day intervals and after 75 days. Single colonies from
each evolved culture were selected, grown up and stored
frozen. These clonal samples were used for competition
assays and were designated FBBl, FBB2, FBBrl, FBBrZ, FCBl,
FCB2 , FCBrl , FCBr2 , SBBl , SBB2 , SBBrl , SBBrZ , SCBl , SCB2 ,
SCBrl and SCBr2. In this nomenclature, the first letter
reflects the original ancestor, the second and third letters
represent the environments in which the strain evolved during
Stages I and II, respectively, the letter ‘r’ indicates that
the line is derived from a streptomycin-resistant clone at
the start of Stage II, and the numerals distinguish replicate

lineages.

Stage II chemostat evolution
Two replicate chemostat cultures each of FBl, FBr, FCl,
FCr, SBl, SBr, SC1 and SCr were established by inoculating
vessels of AN medium containing 500 pg/ml 2,4-D with Single
colonies isolated from the frozen clonal cultures. These

cultures were grown to turbidity before initiating chemostat

59

medium flow. A flow rate of 2.0 ml/hr was maintained for 75
days, after which both mixed population and clonal cultures
were stored frozen. Vessels containing descendants of
strains FBl, FBr, and FCl contained 18 ml of culture allowing
growth at 0.111 per hr (3.8 generations/day, 290 total
generations), while vessels containing descendants of FCr
held 36 ml, allowing 145 generations of evolution.
(Chemostat dilution rates relative to the maximum growth
rates of Stage II proximate ancestors are shown in Table 3.)
All vessels with descendants of strain S held 48 ml of
culture allowing growth at 0.042 per hour (about 1.4
generations per day, 110 total generations). (Stage II
chemostat growth rates were lower than Stage I rates to keep
chemostat lines near or below 1/2 of their “max in 500 pg/ml
2,4-D.) Evolved clonal cultures were designated as FBC1,
FBC2, FBCrl, FBCr2, FCCl, FCC2, FCCrl, FCCr2, SBCl, SBC2,
SBCrl, SCCrl and SCCr2. Three of the chemostat lines, SBCr2,
SCCl and SCC2 were lost due to contamination of chemostat
vessels. All Stage II derived lines were confirmed to have
maintained their ancestral marker state, either resistant or
sensitive to streptomycin, and their ancestral REP PCR gel

pattern.

Batch competition experiments
All competition experiments were performed with clonal
samples of evolved lines in the same culture conditions used

to propagate the evolving populations. Two competing

60

genotypes were mixed together in the same flask and allowed
to compete for a common pool of limiting nutrient. In Stage
I competitions, the ancestors and descendants all competed
against a streptomycin-resistant clone isolated from FBl or
SBl for the F and S lines, respectively. (These resistant
clones, FBr and SBr, were also used to establish 8 of the 32
Stage II lines). For Stage II descendants, one competitor
was one of the 28 derived genotypes (four were lost to
contamination) and the other was the proximate ancestral
genotype carrying the reciprocal streptomycin marker,
allowing the competitors to be easily distinguished.
(Proximate ancestors of Stage II genotypes are the Stage I
derived clones used to begin the Stage II evolution lines).
For example, streptomycin-sensitive lines FBBl and FBB2
competed against streptomycin-resistant FBr, while FBBrl and
FBBr2 competed against FB1.

Both strains in a competition were grown separately for
24 hr (one batch transfer growth cycle) to allow acclimation
to competition conditions. Batch competitions lasted either
1 or 2 days for F—derived lines and either 4 or 6 days for S—
derived lines, depending on the expected relative fitnesses
of the competitors. Competitors were initially mixed in
fresh medium (mixing ratios are given below) and Shaken
thoroughly to allow even dispersion of the two competitors.
Densities of the evolved and ancestral strains at the
beginning and end of a competition were obtained by diluting

the culture onto selective and non—selective agar. The

61

density of the resistant competitor could be estimated
directly from the colony count and dilution factor, while the
density of the sensitive competitor was estimated as the
difference between the non-selective plate count and the
selective plate count adjusted for the dilution factor.

The initial mixing ratios, daily transfer dilutions, and
length of duration for each set of batch competition
experiments are listed below. These were varied to obtain
good resolution of each strain at both initial and final time
points. Dilution rates for batch competitions are listed in
Table 4.

F lines— 0.05 ml samples of Stage I F-derived lines (F,
FB1, FB2, FC1, and FC2) were mixed with equal volumes of the
FBr competitor in 9.9 ml of medium. The FCl and FC2 cultures
were plated at 0 and 24 hours, whereas the F, FB1 and FB2
cultures were diluted 100-fold into 9.9 ml of fresh medium at
24 hours and plated at 48 hours. All Stage II lines
descended from strain F (sets FBB, FBC, FCB, and FCC; Stage
II sets consist of lines that share Stage I and Stage II
selective regimes in common) were mixed in equal 0.05 ml
volumes with their proximate reciprocal ancestor (FB, FBr,
PC, or FCr) and plated at 0 and either 24 or 48 hours, with
48 hour competitions being diluted 100-fold at 24 hours.

S lines- 1 ml samples of S, SB1, and SB2 were mixed
with an equal volume the SBr competitor in 8 ml of medium,
whereas 1.67 ml samples of SC1 and SC2 were mixed with equal

volume of SBr in 6.67 ml of medium. S, SB1 and SB2

62

Table 4. Dilution rates of competition experiments.
Dilution rates of both batch [= ln(daily dilution factor)]
and chemostat (= flow rate/chemostat volume) competition
experiments are Shown for each set of replicate lines.

Line Set Batch.Di1ution. Chemostat Dilution
Rate (per day) Rate (per day)
FB 4.605 2.667
FC 4.605 2.667
SB 1.609 1.000
SC 1.099 1.000
FBB 4.605 2.667
FCB 4.605 1.333
FCC 4.605 1.333
FBC 4.605 2.667
SBB 1.609 1.000
SCB 1.099 1.000
SCC 1.099 1.000

SBC 1.609 1.000

63

competition cultures were diluted 5—fold daily and plated at
0 and 96 hours (4 days), while SC1 and SC2 flasks were
diluted 3-fold daily and plated at O and 144 hours (6 days).
Stage II lines from sets SBB and SEC (and the control 881)
were competed with SB1 or SBr in the same manner as the S set
competitions described above. The lines of sets SCB and SCC
(and the control SC1) competed against SC1 or SCr as per the

Stage I SC1 and SC2 competitions described above.

.Measurement of selection rate constant for batch competitions
Relative performance between two competitors is

expressed as the selection rate constant (SRC), rij, and is

given by

rij = (lnlNi(t l/Ni(0)] - 1n[Nj(t)/Nj(0)l)/t

where Nfi(0) and Nj(0) are the starting densities of the two
competing genotypes, and.N1(t ) and Nj(t ) are their
corresponding densities after t days of competition
(Travisano and Lenski 1996). The selection rate constant
equals the difference between the genotypes' realized
Malthusian parameters during direct competition for the same
pool of limiting resource. (If there is a difference in the
plating efficiencies of the genotypes, this does not affect
the selection rate constant, granted that plating
efficiencies remain the same between initial and final

samples.) The selection rate constant, a difference between

64

Malthusian parameters, was chosen to represent performance
rather than the more commonly used ratio of Malthusian
parameters (Lenski et al. 1991), because the former is less
sensitive to sampling error when competitors have very
different Malthusian parameters, which was sometimes the case
in this study. Marker effects Of streptomycin—resistant
lines have been factored out of the SRC values presented in
Tables 5 and 6, so that a positive SRC reflects superior
performance than the relevant ancestor, and a negative value
shows decreased performance. This was done to standardize
SRC values relative to an ancestral value of zero. For
example, to obtain the SRC for strain FBBl relative to its
proximate ancestor FB1, both strains were independently
competed directly against strain FBr. Then the SRC of FB1
relative to FBr was subtracted from that of FBBl relative to

FBr to obtain the SRC of FBBl relative to FB1.

EXperimental design of batch competitions

In both the 2,4-D and succinate media, usually five
estimates, and a minimum of three, of the selection rate
constant for each competitor were obtained. Competitions
were performed in blocks of five pairwise matches, with five
replicates per match, for a total of 25 separate cultures per
block. For the Stage I lines, one block consisted of F, FB1,
FB2, FC1, and FC2 each being competed against FBr, and a
second with S, SB1, SB2, SC1 and SC2 each competing against

SBr. Each block of competitions was performed in both 500

65

pg/ml 2,4-D and 500 pg/ml succinate AN medium. For Stage II
lines, these blocks consisted of one control competition
between the two reciprocal ancestors (e.g., FB1 vs. FBr) and
a set of four lines against their reciprocal ancestors. For
example, one block had pairings of FB1 vs. FBr, FBBl vs. FBr,
FBB2 vs. FBr, FBBrl vs. FB1, and FBBr2 vs. FB1. (The only
exception to this was for the SBC and SCC sets, which both

had fewer than four lines due to contamination.)

Protocol and SRC.measurement for chemostat competitions
Chemostat competitions were also performed under the

same conditions as the chemostat evolution experiments. The
genotype pairings and block designs were the same as for the
batch competitions, with the exception that two replicates
(rather than five) were obtained for each pair of competing
genotypes. For each competition, both competing genotypes
were first grown to stationary phase in the competition
medium. They were then mixed at a ratio of 100 (sensitive
competitor) to 1 (resistant competitor) in the chemostat
vessels, at which time media flow through the vessels was
initiated. To allow a period of physiological acclimation to
competition conditions, cultures were allowed to grow
overnight (12—24 hours) before being sampled for the first
time. Each culture was sampled and plated several times on
selective and non-selective agar over a 6—10 day period to
monitor changes in density of the two genotypes. Chemostat

competition dilution rates are listed in Table 4.

66

The selection rate constant for each replicate was
calculated as the slope of the value ln(R/S) over time (in
days) where R is the density of the resistant competitor and
S is the density of the sensitive competitor. This
calculation using several density measurements over time is
equivalent to that for the batch competitions, which uses
density data from only the start and finish of the

competition period.

RESULTS
Adaptation to selective regime

Of the 36 evolved lines (Stages I and II) analyzed in
this study, 22 Of them show Significant adaptation to their
selective regime based on an SRC > 0 in competition with the
ancestor (Tables 5 and 6). In four sets of lines (FC, SB,
FBB, and SCB), all lines in each set adapted significantly to
their selection regime, while other sets contain lines (total
of 13) that show a non—significant increase in competitive
performance. Only one strain, FCBr2, actually indicates a
decrease in performance relative to its ancestor, but this
decrease is not significant.

Of the Stage I lines derived from F, the fast ancestor,
neither batch lineage (FB1 or FB2) shows a significant
competitive improvement in their batch selective regime, but
both lines selected in the chemostat regime (FCl and FC2) did
improve significantly in chemostat. The opposite trend

occurred among the Stage I descendants of the slow ancestor

67

Table 5. Results of competition experiments for evolved
lines from Stage I selection in succinate. Selection rate
constants (SRC) of Stage I evolved lines (relative to their
appropriate ancestors) for both the regime in which evolution
occurred and the alternative regime are presented. Lines
with SRC values that are significantly different from the

appropriate ancestor are presented in bold (t-test, 2-tailed
P < 0.05).

Table 5. Results of competition experiments for evolved

lines from Stage I selection in succinate.

Evolved Line

FB1
FB2
FCl

FC2

SB1
SB2
SC1

SC2

Selected in

Batch
Batch
Chemostat

Chemostat

Batch
Batch
Chemostat

Chemostat

SRC (per day)

Selective

Regine
0.0260
0.0910
2.8721

2.0268

0.2531
0.3759
0.1729

0.2521

Alternative
Regime

0.0472
-0.0374
-0.3093

0.8125

0.3715
0.6026
0.2379

0.0897

69

Table 6. Results of competition experiments for evolved
lines from Stage II selection in 2,4-D. Selection rate
constants (SRC) of Stage I evolved lines (relative to their
appropriate ancestors) for both the regime in which evolution
occurred and the alternative regime are presented. Lines
with SRC values that are significantly different from the

appropriate ancestor are presented in bold (t-test, 2-tailed
P < 0.05).

Table 6.

Results of competition experiments for evolved

lines from Stage II selection in 2,4-D.

Evolved Line

FBB1
FBB2
FBBrl

FBBr2

FCBl
FCB2
FCBrl

FCBr2

FCCl
FCC2
FCCrl

FCCr2

FBCl
FBC2
FBCrl

FBCr2

Selected in

Batch
Batch
Batch

Batch

Batch
Batch
Batch

Batch

Chemostat
Chemostat
Chemostat

Chemostat

Chemostat
Chemostat
Chemostat

Chemostat

SRC (per day)

Selective Alternative
Regime Regime
0.4026 0.1821
0.9675 0.2162
0.3116 0.2268
0.3439 -0.4448
0.4171 0.3177
0.6390 -0.2586
0.6788 0.0150

—0.6600 -0.3296
0.2341 0.3291
1.2510 0.6807
0.9161 -0.4631
0.7944 -0.0517
0.1432 —O.l468
0.2944 —0.2777
0.3265 —0.3362
0.6090 -0.3798

71

Table 6 (cont’d).

Evolved Line Selected in SRC (per day)
Selective Alternative
Regime Regime
SBBl Batch 0.0379 —0.0154
SBB2 Batch 0.1073 0.1013
SBBrl Batch 0.2230 —0.0151
SBBr2 Batch 0.5401 0.0098
SCBl Batch 0.6924 0.2494
SCB2 Batch 0.7670 0.0758
SCBrl Batch 0.4899 -0.2048
SCBr2 Batch 0.3680 —0.4172
SCCrl Chemostat 0.0013 0.0925
SBC1 Chemostat 0.0701 0.1207
SBC2 Chemostat 0.0026 0.0031
SBCrl Chemostat 0.0763 0.0845

72
(S). SB1 and SB2 both improved significantly in batch, but

neither SC1 nor SC2 improved significantly in chemostat.
Among the eight Stage II F lines that continued evolving
in the same regime as that of their Stage I ancestral lineage
(sets FBB and FCC), seven Show significant adaptation (Table
6). In the two sets that switched selective regimes (FBC and
FCB), however, only four out of eight improved significantly
in their Stage II regime. Those Stage II strains that share
strain S as a common ancestor, however, showed the opposite
overall trend of adaptation between those lines that did and
did not switch selective regimes. In the SBB and SCC sets,
only one of five lines show significant improvement in their
selective regime, whereas in the sets that switched regimes

(SBC and SCB), six of seven lines adapted significantly.

Performance in alternative regime

Of the 36 total lines, ten show significant changes in
the alternative regime to that in which they were selected
(Tables 5 and 6). Seven of these ten lines (FC2, SB2, FBB1,
FBBrl, FBBr2, FCC2, and FCCrl) are also significantly
different than their ancestor in their selective regime. The
evolutionary changes of these seven lines, therefore, are of
greatest consequence for the substrate concentration tradeoff
hypothesis. Only two lines (FBBr2 and FCCrl) show the
pattern of evolutionary changes predicted by the tradeoff
hypothesis - significantly improved performance in its

selective environment (chemostat) associated with

73

significantly diminished competitiveness the alternative
regime (batch) (Figure 8a). The other five lines all
improved significantly both in their selective and
alternative regimes, contrary to expectations of the tradeoff
hypothesis (Figure 8b). Among the 22 lines that adapted
significantly to their own selective regime, two lines (FBBr2
and FCCrl) show a significant tradeoff pattern, eight show
non-significant tradeoff patterns, seven Show non-significant
improvements in their alternative regime, and five lines
(FC2, SB2, FBB1, FBBrl, and FCC2) improved significantly in

both regimes.

.HOmogeneous and heterogeneous responses to selection

Only one of the eight Stage II sets with multiple lines
(FBC set) of lines shows a consistent performance response
among all lines in both their selective and alternative
environments. The competition results for the four FBC lines
(Figure 9, open squares) suggest that they all may have
improved in their selective regime while worsening in the
alternative batch regime. Most of the changes in FBC lines
are not significant when individual lines are compared to the
ancestor separately. However, when data from separate lines
are pooled, the set as a whole shows a statistically
significant performance decrease in the non-selective (batch)
regime (t = 2.512, 21 degrees of freedom, 2-tailed P =
0.0202) and a not quite significant improvement (t = 2.234, 8

degrees of freedom, 2-tailed P = 0.0560) in the selective

74

 

   

 

 

1.75 - FCCr1 a 1.75 — FBB1 b
1 .5 — 1 ,5 _
1.25 — 1 .25 -
B 1" B 1“
a a
8 0.75 -l 8 0.75 -
o o
g 0 5 — g 0.5 —
C E
.g 0.25 — g 0.25 —
O O
2 2
a 0- a 0—
-0.25 - -0.25 —
-0.5 — -0.5 —
-0.75 — -0.75 —
Chemostat Batch Batch Chemostat

Figure 8. Tradeoff and non-tradeoff patterns of adaptation.
(a) Strain FCCrl shows the pattern of performance in its
selective and non-selective regimes that is expected by the
substrate concentration tradeoff hypothesis. (b) Strain FBB1
and five other strains (not shown) exhibit the opposite
pattern, improving their performance in both batch and
chemostat regimes. Error bars indicate 95% confidence
intervals.

75

_L
N
l

—l
I

F’
m
I

.'

.0
.n
I

(124
E]

 

 

SRC in selective regime
o
O)
l
E]

l l l

l I
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
SRC in alternative regime

Figure 9. Heterogeneous and homogeneous selection responses.
Most Stage II sets show a heterogeneous pattern of
performance change in their alternative regime as exemplied
by the FCC lines (filled circles), where two lines improved
and two became worse. Alternatively, two sets have a
homogeneous pattern, as exemplified by the FBC lines (open
squares), which all appear to have slightly decreased
performance in their alternative (batch) regime.

76

(chemostat) regime, suggesting an overall tradeoff tendency
for the set.

For most Stage II sets, however, there is no consistent
improvement or detriment in their non-selective regime. For
all six sets showing such variation, one or two lines appear
to be worse in the non-selective regime while the other two
or three appear to have improved. This is exemplified in
Figure 9 by the FCC set (filled circles), in which two lines
Show improvement in the alternative (batch) regime and two
either show or suggest performance decreases. Again, many of
the individual changes among all lines in these six sets are
not significant, but the overall pattern strongly suggests
that improved performance in a selective regime (which
occurred or appeared to have occurred in all lines but one)
can be associated with either improved or worsened

performance in the alternative regime.

Interpretation of chemostat competition data

While analysis of data from the batch competitions is
straightforward, there are some features of the chemostat
competition data that make its interpretation more
challenging and warrant mention. The ideal results of
chemostat competitions are log—linear slopes of the ratio of
competitors’ densities over time (Figure 10a) that are,
moreover, very similar for the two independent replicates.
This, however, is not always the case. My data can be

classified into four basiq.types, corresponding to the four

Figure 10. Four types of chemostat competition data. The
natural log ratios of competing genotypes over time in the
chemostat are shown. Circles and diamonds distinguish data
points between two independent replicate competition
experiments and solid lines represent the two corresponding
linear regressions. (a) Both replicates of FCCrl behave
similarly, are linear, and have a mean SRC that is
significantly different from the ancestral controls.

(b) FBB1 replicates are similar and significantly different
from the ancestral controls, but they do not appear to follow

a simple linear trajectory. (c) Replicates of SCB2 are
similar and linear, but they are not significantly different
from the ancestral controls. (d) Replicates of FBCl are not

similar and their mean is not significantly different from
the ancestral controls.

78

 

 

 

 

b l9
l8
I7
l6
#5
I4
l3
l2
9 l1
_ _ _ _. _ I no
A.
Seam”: 5
I00
a
Inu
e I7
16
IR.
I4
l3
l2
l1
_ _ a I _ _ no
1..

$0.550”: 5

Days

Days

6

 

 

_
024.6

:muatome 5

 

 

_
0246

209809 5

 

Days

Days

Four types of chemostat competition data.

Figure 10.

79

graphs in Figure 10. In the first case (Figure 10a, strain
FCCrl), the data points in each replicate are approximately
log linear over time, the two replicate trajectories are very
similar, and their mean slope is significantly different from
that of the ancestral control competition. A second type of
result (Figure 10b, strain FBB1) again exhibits similar
replicate patterns and a significantly different slope than
the ancestral control, but the data points do not seem to
follow a simple linear trajectory over time. Figure 10c
(strain SCB2) shows a third case where the replicates are
similar and relatively linear, but their mean slope does not
differ significantly from the ancestral competition.

Finally, there are also instances (Figure 10d, strain FBC1)
where the replicate populations seem to Show different
trajectories and the mean slope is not significantly
different from the ancestor.

My results exhibit multiple instances of each of these
four types of data, and I will briefly discuss the
interpretation of the types that deviate from similarity
between replicates (Figure 10d) or from simple linearity
(Figure 10b). The causes of these dissimilar or non-linear
results are difficult to specify, especially for the former
case. Dissimilarity of replicates reduces the statistical
significance of any difference in mean slope between derived
and ancestral competitions because it increases variance. In
all cases where we observe a significant evolutionary change

by a derived line relative to its ancestor in the chemostat,

80

the two replicates Of both the derived competition and the
ancestral control behave similarly. Because dissimilar
replicates tend to mask any actual evolutionary change that
took place in a derived line, the effect may be a reduction
in the number of statistically significant cases of
evolutionary change from which to draw conclusions bearing on
the tradeoff hypothesis. Nonetheless, those cases of
significant and clear—cut evolutionary change revealed by
these competition studies are sufficient to draw robust
conclusions about the substrate concentration tradeoff
hypothesis.

The other unexpected pattern in chemostat competition
data occurs in cases where the slope of the log ratio of
competing genotypes changes significantly over time (Figure
10b). Possible causes of this pattern include different
rates of chemostat wall accumulation by the two genotypes
(Chao and Ramsdell 1985); frequency—dependent selection,
where the fitness of one genotype relative to the other
changes as the relative frequency of the genotypes changes
(Levin 1988); or insufficient time for both competitors to
reach equilibrium because one competitor has a longer
transition from stationary phase to equilibrium growth than
the other (Jannasch 1969; Veldkamp et al. 1984).

Experimentally testing these possibilities was not
feasible within this study, given the large number of
ancestral and derived strains that were studied. However,

there are several considerations which legitimize the use of

81

the mean slope over the entire time-course of the
competitions (rather than selecting only a sub-period of the
experiments, in which trajectory appears log—linear, over
which to measure performance). First, there may be different
causes for this pattern in different competitions. One case
may be due to frequency dependent selection and another due
to gradual accumulation of wall growth. Hence, it is not
clear whether to focus on flatter or steeper parts of the
competition trajectory. Second, not all cases of changing
slopes follow the same pattern. In some cases, the slope
decreases over time, while in others it increases. Third,
and most importantly, among those derived strains that show a
significant change from their ancestor in chemostat
competitiveness, none of them show a change in sign of the
slope during the competition, but only a modification in the
same direction. All of these considerations suggest that
taking a simple log ratio slope over all time points is a
feasible, reasonable, and consistent approach to measuring
competitiveness for these experiments. However, even if I
limited my analysis to those derived strains that are
Significantly different than their ancestor in chemostat and
showed constant slopes, this would not alter the fundamental

conclusions of this investigation.

82
DISCUSSION

Implications for the tradeoff hypothesis

Systematic comparative tests of the substrate
concentration tradeoff hypothesis (Greer et al. 1992, Lenski

et al. 1997, Chapter 1) have failed to support the

hypothesis. Greer et al. (1992) and Lenski et al. (1997)

both measured the maximum growth rate (pmax) and substrate

affinity (”max/Ks) values for several strains of 2,4-D

degrading bacteria. These parameters determine growth rates

at high and low substrate levels respectively. Neither study

found a significant negative correlation between Hmax and

Hmax/Ks as was expected from the tradeoff hypothesis. I also

measured growth rates at high and low substrate
concentrations on two different substrates for seven strains
of 2,4—D degrading bacteria isolated from a variety of
locations and conditions (Chapter 1). While the tradeoff
hypothesis predicts a negative correlation between high and
low substrate concentration growth rates, the opposite trend
was observed. When each data point was treated as an
independent observation, there was a significant positive
correlation between growth rates at 5 and 500 pg/ml of
succinate, and an almost significant positive correlation
between growth rates at the same concentrations of 2,4—D.
These correlations remained positive but decreased to
insignificance after data was scaled to account for

phylogenetic relationships among the strains.

83

A related comparative study was conducted with sixteen
strains of benzene degrading bacteria (Zhou and Tiedje, ms).
The maximum rates (Vmax) and half—velocity coefficients (Km)
of benzene utilization were measured for each strain. The
strains were then categorized into three basic kinetic types:
high vmax with high Km, low vmax with low Km, and high vmax
with low Km, The first two categories are expected by the
traditional substrate concentration tradeoff hypothesis, but
the third category (which included two of the sixteen
strains) is not. These two unexpected strains offer
additional comparative evidence that the tradeoff hypothesis
is in need of modification, even though most strains
conformed to the predicted pattern. In summary, systematic
comparative studies have either not strongly supported or
have contradicted the tradeoff hypothesis.

Experimental evolution has been previously used by
Luckinbill (1984) to test predictions of r— and K— selection
theory that are similar to predictions of the substrate
concentration hypothesis tested here. In Luckinbill’s study,
replicate evolving lines of EScherichia coli were maintained
by serial batch culture using transfer protocols designed to
impose either r- or K; type selection On evolving
populations. Cultures under the r- selection regime were
transferred prior to reaching stationary phase, whereas to
impose K; selection, populations were allowed to reach and
experience a prolonged stationary phase before being

transferred to fresh medium. Although no tradeoff was shown

34

by this experiment, the selective regime intended to impose
significant K; selection may not have done so. Vasi et al.
(1994) conducted theoretical and empirical studies of growth
parameter changes during evolution of E. coli populations in
a batch transfer regime and concluded that even in
Luckinbill’s putatively K- selection regime, maximum
exponential growth rate was probably the primary target of
selection. This evaluation renders the absence of tradeoff
in Luckinbill’s experiment inconclusive regarding the
relationship of competitive abilities at high vs. low
nutrient concentrations.

Vasi et al. (1994) measured “max and Ks values for
twelve independently derived lines of E. coli before and
after 2000 generations of evolution in a serial batch

transfer selective regime. Both Hmax and Ks increased

significantly overall among the twelve lines, but the ratio
“max/Ks did not Change significantly. Their study,
therefore, also failed to support the tradeoff hypothesis.
In this study, I tested directly for changes in

competitive ability at high and low resource concentrations
following evolution in regimes that clearly favored
adaptation to one condition or the other. The results bear
upon the tradeoff hypothesis more clearly and strongly than
do those of previous studies. My overall results not only
fail to support the hypothesis, but several specific results
directly contradict it. Only two strains (FBBr2 and FCCrl)

Show significant tradeoff patterns, having lost

85

competitiveness at high 2,4-D concentrations after having
evolved and improved competitively under low 2,4-D
conditions. Evolutionary changes in the FBC set as a whole
are also suggestive of the tradeoff pattern (Figure 9).
However, five lines (FC2, SB2, FBB1, FBBrl, and FCC2) show
results that flatly contradict the tradeoff hypothesis,
improving significantly both in their selective regime and in
the alternative regime.

Overall, 14 (12 non-significant) lines show the tradeoff
pattern, and 21 (16 non-significant) show the opposite
pattern (Table 7). These bi-directional responses to
selection indicate no propensity for tradeoffs in competitive
performance under the two selective regimes (Table 7). It is
likely that at least some evolution has been overlooked among
lines not showing significant changes by my competition
assays. This could be due to noisy data, insufficient
replication, and/or insufficiently long periods of
competition. If this is the case, then one expects that more
precise data would tend to move cases of masked evolution
from insignificance to significance in the direction
suggested by my actual data. The overall pattern of
evolutionary changes among these lines strongly suggests that
there are multiple evolutionary pathways by which bacteria
can adapt to high or low substrate conditions and, moreover,
that these pathways can be either beneficial or detrimental

to fitness under the alternative conditions. These data also

86

Table 7. Frequency of tradeoff and non-tradeoff patterns.
The number of lines that show a statistically significant
tradeoff (ST), non-significant tradeoff (NST), non-
significant correlated improvements in both batch and
chemostat (NSCI), and significant correlated improvements
(SCI) are shown. These categories are shown for the 22 lines
that adapted significantly to their selective regime and then
for the 35 total lines that either Show or suggest such
adaptation.

ST NST NSCI SCI

Significantly
Adapted 2 8 7 5

"RNaI 2 12 16 5

87

suggest that such changes in the alternative regime may occur

in positive and negative directions at a similar frequency.

Strain history and evolutionary change

Also of interest in the design of these experiments were
the effects of strains' selective history on subsequent
patterns of adaptation (cf. Travisano et al. 1995). Two
patterns in the data are suggestive of such effects. First,
the selective histories of F and S in the wild may have
influenced the rate and extent of their adaptation to batch
and chemostat environments during Stage I evolution. Among
the four F-derived lines, neither batch-selected line
improved significantly in batch, but both chemostat-selected
lines did improve significantly in chemostat. In contrast,
S-derived lines show the opposite pattern, with both batch-
selected lines becoming significantly better in batch, but
neither chemostat-selected line having significantly improved
in chemostat. Prior selection in nature had rendered strain
F able to grow fast when resources were plentiful, and strain
F therefore had much more room for improvement under a
condition of resource scarcity (chemostat) than one of
abundance (batch culture). The opposite appears to be true
for strain S, which prior natural selection left with
relatively slow growth in substrate rich conditions and thus
with considerable room for improvement in the batch culture

regime.

88

In the second pattern, F-derived Stage II lines Show the
opposite adaptive trend between lines that did and did not
switch regimes from Stage I to II than do S—derived Stage II
lines. Seven of the eight F—derived lines that shared a
common selective regime for both stages of evolution (FBB and
FCC sets) adapted significantly to their Stage II regime,
while only four of the eight lines that switched regimes (FBC
and PCB sets) are significantly better in their Stage II
environment. Among those S-derived lines that switched
regimes, however, six of seven improved significantly during
Stage II, whereas only one of five among those that did not
switch regimes is significantly improved. These results
suggest that the selective history of strain S in the wild
may have left it with more potential for reversible evolution
in alternating environments than did the natural history of
F. Because this explanation is ad hoc, confirmation would

require further experimentation.

Potential mechanisms of adaptation

Here I first discuss the population genetic processes
that caused phenotypic changes in these evolving lines and
then I speculate on potential demographic mechanisms
underlying their adaptation. Seven lines underwent
significant fitness changes in both their selective regime
and the alternative regime. Were the fitness changes in the
alternative regime due to pleiotropic effects of beneficial

mutations in the adaptive regime? Or were they due to the

89

random fixation of mutations that were effectively neutral in
the selective regime (Kimura 1983)? The vast majority of
mutations that affect fitness in a given environment are
expected to be harmful rather than beneficial, because it is
generally much easier to disrupt than to improve the
performance of a well-functioning entity. Therefore, if one
considers mutations that are selectively neutral in one
regime (say, batch culture), but affect fitness in an
alternative environment (say, chemostat), then one would
expect most of them to reduce, rather than improve, fitness
in the alternative regime. In this study, however, six of
the eight lines mentioned above show significant improvements
in both types of regime. These data imply that improvements
in the alternative regimes were not due to substitution of
random mutations by drift, but instead were caused by the
same mutation that were responsible for improvements in the
selective regimes. In other words, selected mutations were
generally beneficial in both regimes, indicating positive
pleiotropy (and contrary to the antagonistic pleiotropy
assumed by the tradeoff hypothesis).

The parameters “max and ”max/Ks are central components
of fitness in the batch and chemostat regimes, respectively.
They are also the parameters in which the substrate
concentration tradeoff hypothesis is most often expressed.
The dual improvement of performance by several evolved lines

in both regimes suggests that adaptive mutations can occur

that simultaneously increase both Hmax and “max/Ks- For

90

lines evolved in batch culture, the lag and stationary phases

of growth are potential objects of selection as well as

exponential growth rate. However, a simple increase in Hmax

would also increase the value “max/Ks: improving fitness at

both high and low resource concentrations. This one-step
path to improving both kinetic parameters seems more likely

than an increase in flmax/Ks being correlated with an

improvement in lag or stationary phase fitness.

The strengths and limitations of this study

The greatest strength of this investigation relative to
previous tests of the substrate concentration tradeoff
hypothesis is that the evolution relevant to the conclusions
of the study occurred under well-defined, experimental
laboratory conditions, rather than in unknown conditions in
the wild. Instead of basing conclusions on correlative
patterns of data for traits with unknown selective histories
in the wild, I have been able to compare the competitive
performance of derived lines with that of their clonal
ancestors after evolution in known selective conditions.
This means that differences among evolved lines relative to
each other and to the ancestor are due to the selective
forces inherent in the lab evolution regimes rather than to
phylogenetic relationships, differential preadaptation to
laboratory conditions, or unknown selective forces in nature.

Also strengthening this study is the experimental

design, which includes multiple starting lines, two stages of

91

evolution with different resources, independently evolving
replicate populations, and two distinct selective regimes.
Multiple starting lines allow observation of the influence of
strain history, if any, on subsequent evolutionary patterns.
Two evolutionary stages allowed adaptation to lab conditions
during Stage I and allowed Stage II lines to either continue
in the same selective regime or alternate to the other. Most
importantly, the two selective regimes used, batch and
chemostat culture, are clearly distinct in the selective
forces they impose on their evolving populations.

In this test of the substrate concentration tradeoff
hypothesis, evolutionary changes in competitive fitness in
both selective and alternative regimes were estimated by
direct competition experiments between evolved and ancestral
strains. This approach tests the hypothesis more directly
than would comparison of the kinetic parameters (between
evolved and ancestral strains) that are thought to underlie
fitness in abundant and scarce resource conditions. While
measurement of these parameters would prove useful in
clarifying which components of fitness changed most during
evolution, evolutionary dynamics (whether in the lab or in
the wild) are ultimately a function of overall competitive
ability.

Additionally, these experiments were performed with
naturally isolated strains of bacteria capable of degrading a
common herbicide rather than by using strains with long

laboratory histories and no particular applied significance.

92

The substrates used are potential resources for the strains
in nature, increasing the relevance of this study for applied
bioremediation issues. The components of this study have
resulted in conclusions that almost unambiguously contradict
the hypothesis in question.

These experiments provide little or no evidence for a
tradeoff in competitive performance under conditions of
resource abundance versus scarcity. However, the maximum
number of evolutionary generations undergone by any of my
derived lines was 1000 (FBB set, 500 in batch-succinate, 500
in batch-2,4-D). Although I have shown short—term
simultaneous fitness increases in alternative environments to
be possible and even probable, the long-term potential for
this simultaneous Optimization of traits may be limited. For
example, a particular strain evolving in the batch regime for
2000 generations may increase its SRC in batch relative to
its ancestor from 0 to 1.0 while simultaneously increasing
its SRC in the chemostat environment from 0 to 0.5. Further
evolution (say, another 2000 generations) in batch, however,
may increase its batch SRC from 1.0 to 1.5 while
Simultaneously reducing its chemostat SRC from 0.5 to 0.25
(Figure 11). Hence, the putative tradeoff may in fact exist,
but only beyond a certain threshold level of adaptation to a
particular environment. Below this threshold, improvement of
fitness in the non-selective environment is possible. The
only way to conclusively test whether such a tradeoff

threshold exists would be to conduct experimental evolution

93

Figure 11. Hypothetical adaptive threshold for tradeoff.

The competitive ability of an evolving population is shown to
increase (with a decreasing rate of increase) over 4000
generations of evolution in its selective regime (say,
abundant resource conditions; solid curve). The correlated
performance of that strain in its alternative regime (scarce
resource conditions; dashed curve) also increases, but only
up to about 2000 generations, after which time it begins to
decrease. In this case, the substrate concentration tradeoff
hypothesis would be false when applied to the first 2000
generations of adaptation, but accurate for subsequent
adaptation to the selective regime.

'L5—

'L25—

(175-

0.5 -

Selection Rate Constant

0.25 —

 

 

 

l l l l
0 1000 2000 3000 4000
Generations

Figure 11. Hypothetical adaptive threshold for tradeoff.

95

for a long enough period to allow the slope of fitness
increase over time to approach zero as the potential for
further adaptation to a given environment decreases over time
(Lenski and Travisano 1994). Despite this limitation, my
results clearly show that a tradeoff between competitive
ability at high vs. low resource conditions does not exist
over the entire trajectory of adaptation to abundant and

scarce resource regimes.

Chapter 3

PLEIOTROPIC EFFECTS OF.ADAPTATION TO A.SINGLE CARBON SOURCE
FOR GROWTH ON.ALTERNATIVE SUBSTRATES

INTRODUCTION

Biological traits that are not important components of
fitness in a particular environment may nonetheless evglye
by a variety of mechanisms. Such non—selective mechanisms of
evolution include pleiotropic side-effects of adaptive
mutations on unselected traits (Wright 1977) and the random
fixation of effectively neutral alleles by genetic drift
(Kimura 1983). Along with adaptation by natural selection,
these more indirect mechanisms of evolution are relevant to
such issues as the genetic divergence of populations within a
species (Cohan 1984; Travisano et al. 1995) and the fate of
genetically modified organisms in natural environments
(Tiedje et al. 1989; Lenski 1993; Regal 1993).

Several studies have shown that isolated populations of
the same species may evolve different adaptations even to the
same selective conditions (Gould and Lewontin 1979; Cohan
1984). Some Of clearest evidence for this phenomenon comes
from laboratory experiments using fruitflies (Cohan and Graf
1985; Hoffman and Cohan 1987) and bacteria (Travisano et al.
1995; Travisano and Lenski 1996). These results appear to

conform with Sewall Wright's concept of an “adaptive

96

97

landscape” that has several fitness peaks, each representing
a distinct genetic solution to a particular selective
environment.

Alternative pathways of adaptation to a particular
selective environment may lead to heterogeneous changes in
traits that are not important for fitness in that
environment. For example, several populations of organisms
that are under selective pressure to utilize a particular
substrate more efficiently may each find a different
physiological mechanism to do so. The effects of these
different adaptations on the ability of organisms to utilize
some alternative substrate may be positive in some cases,
neutral in others, and negative in yet other populations. As
a consequence of these different correlated responses, some
populations may be fortuitously well adapted, and others
poorly adapted to environments that contain this alternative
substrate.

In this study, I have used evolving populations of
bacteria to investigate the effect of adaptation to one
carbon source on competitive performance on an alternative
substrate. The bacterial strains employed are two natural
isolates of the B—proteobacteria, each capable of degrading
the herbicide 2,4-dichlorophenoxyacetic acid (2,4-D).
Experimental evolution occurred in two separate stages, one
in which succinate was the sole carbon source, and a second

in which 2,4-D was the limiting substrate (Figure 12).

98

Batch

Batch FB > FBB
> FE <
FBr > FBBr

Chemostat > FC < FC > FCB
FCr > FCBI

 

 

 

 

 

 

Batch

8 I), SB 9 SBB
30 > 88 <
SBr > SBBr

Chemostat SC > SCB
> so <
SCr > SCBI

 

 

 

 

 

 

Stage I

(succinate) Stage II
(2,4-0)

Figure 12. Derivation of bacterial strains. Horizontal
lines indicate periods of evolution with the selective regime-
and sole carbon source indicated above and below,
respectively. Each horizontal line represents two
independently evolved replicate populations. See MATERIALS AND
MEmmms (Chapters 2 and 3) for detailed strain and
nomenclature descriptions as well as the number of
evolutionary generations for each line.

99

The use of environmentally important substrates
increases the relevance of my results to concerns about the
evolutionary adaptation of genetically modified organisms
subsequent to their release in the environment. One concern
is that certain environmental conditions may cause bacteria
that are degrading a toxic substrate to produce an even more
toxic metabolite in the process. For example, bacteria may
produce highly toxic vinyl chloride during the biodegradation
of tetrachloroethylene under methanogenic conditions (Vogel
and McCarty 1985). In a similar vein, genetically modified
bacteria that are used to degrade some toxic compound may
evolve novel degradative pathways that yield a more toxic
metabolite even as they continue to degrade the intended
substrate. Concerns over these and other potential adverse
effects are magnified if there is a significant likelihood
that the released organisms may persist indefinitely in the
environment (Tiedje et al. 1989; Lenski 1993; Regal 1993).
And as the organisms adapt to the natural environment in
which they were released, the likelihood that they can
persist will increase.

Of particular relevance to these concerns is the effect
that evolutionary adaptation by bioremediative organisms to
their intended anthropogenic substrate has on their abilities
to grow on, and compete for, naturally occurring substrates.
If adaptation to an anthropogenic substrate always leads to a
reduction in fitness on natural substrates, then this

provides a measure Of safety for the release of organisms

100

modified for bioremediation. If, however, such adaptation
may indirectly enhance fitness on natural substrates, then
this increases the likelihood that the released organisms
(and any adverse effects they may cause) will persist in the
environment.

In Stage I of the experimental evolution, two naturally
occurring strains of 2,4-D degrading bacteria were used to
found replicate populations that were propagated in batch
culture with succinate (a naturally occurring substrate) as
the sole carbon source (see Figure 12). One of these strains
(F) grows relatively fast under laboratory conditions,
whereas the other (S) grows more slowly. In Stage II of the
evolution experiment, clonal isolates derived from the Stage
I batch culture lines were used to found replicate
populations that were further propagated in batch culture,
except that 2,4-D was provided as the sole carbon source
(Figure 12). Additional Stage II batch lines were founded
using clones from Stage I lines that had undergone evolution
in a different selective regime - chemostat culture. I then
measured the competitive performance of all of these derived
lines relative to their immediate ancestors on both succinate
and 2,4-D. The results of these experiments demonstrate that
adaptation to either substrate can be associated with both
improvements and losses in competitive performance on the

alternative substrate.

101
MATERIALS AND METHODS

All strains, evolution experiments, and batch
competition protocols are the same as described in the
MATERIALS AND METHODS section of Chapter 2, except that batch
competition assays were conducted in each strain’s
alternative substrate (succinate or 2,4-D) as well its
selective substrate. Dilution factors and the duration of
competition experiments were always the same for both
substrates. This study focuses on a subset of the strains
described in Chapter 2 (Figure 12). Batch competition
experiments were performed on both selective and alternative
substrates for strains that had been propagated in the batch
regime during their most recent stage of experimental
evolution. These strains are FB1, FB2, SB1, and SB2, all of
which grew on succinate in batch culture during Stage I
evolution; and sixteen strains that grew on 2,4-D in batch
culture during Stage II evolution. Among the latter group,
the FBB and SBB sets had been previously grown on succinate
in batch culture during Stage I; whereas sets PCB and SCB had
been previously grown on succinate in chemostat culture
during Stage I. Competition dilution rates are listed in

Table 8.

102

Table 8. Dilution factors and rates for batch competition
experiments. Daily dilution factors and corresponding
dilution rates (ln(daily dilution factor) per day), are shown
for each set of replicate lines.

Line Set Daily Dilution. Dilution Rate
Factor (per day)

FB 1/100 4.605

SB 1/5 1.609

FBB 1/100 4.605

FCB 1/100 4.605

SBB 1/5 1.609

SCB 1/3 1.099

103
RESULTS

.Adaptation to selective substrate

In this section, I report the extent of adaptation by
the various lines to the substrate in which they were most
recently selected. In the next section, I describe the
effect of this adaptation to one substrate on the performance
of each line on the alternative substrate.

Of the Stage I lines propagated on succinate in batch
culture, both SB1 and SB2 significantly improved their
competitive performance in that environment, whereas neither
FB1 nor FB2 showed significant improvement (Table 9 and
Figure 13). The greater extent Of adaptation by S-derived
lines (which underwent fewer than half the number of
generations during Stage I as did the F-derived lines)
suggests that the slower growing strain S had more room for
improvement in the batch regime than did strain F (see
Chapter 2).

Six of the eight Stage II F—derived lines significantly
improved their competitiveness on 2,4—D, which was the
substrate supplied during Stage II. All four lines in the
FBB set demonstrably improved, as did two of the four FCB
lines (Table 9, Figure 14). Similarly, five of the eight S—
derived Stage II lines Show significant fitness increases,
but only one of these five belongs to the SBB set of lines
(Table 9, Figure 15). All four SCB lines, which were

subjected to changes in both culture regime and growth

104

Table 9. Results of competition experiments for evolved
lines from Stage I selection in succinate and Stage II
selection in 2,4-D. Selection rate constants (SRC) of
evolved lines (relative to their appropriate ancestors) for
both the substrate in which evolution occurred and the
alternative substrate are presented. Lines with SRC values
that are significantly different from the appropriate
ancestor are presented in bold (t-test, 2-tailed P < 0.05).

105

Table 9. Results of competition experiments for evolved
lines from Stage I selection in succinate and Stage II
selection in 2,4-D.

Line Selected in SRC (per day)
Selective Alternative
Substrate Substrate
FBl Succinate 0.0260 0.0236
FB2 Succinate 0.0910 -4.6052*
SB1 Succinate 0.2531 0.0366
SB2 Succinate 0.3759 0.1582
FBB1 2,4-D 0.4026 —0.0929
FBBZ 2,4-D 0.9675 0.0147
FBBrl 2,4—D 0.3116 0.3932
FBBr2 2,4-D 0.3439 —0.8023
FCBl 2,4-D 0.4171 -2.7843
FCB2 2,4-D 0.6390 -1.1000
FCBrl 2,4-D 0.6788 1.9561
FCBr2 2,4-D -0.6600 —0.5012

* The asterisk marking strain FB2 for performance on 2,4—D
indicates that the maximum SRC for FB2 on 2,4-D was inferred,

rather than measured directly.

The value is based on its

competitor's (FBr) ability to grow at least 100-fold per day

on 2,4-D, while FB2 is unable to grow on 2,4—D.

Table 9 (cont’d).

SBBl
SBB2
SBBrl

SBBr2

SCBl
SCB2
SCBrl

SCBr2

2,4-D
2,4-D
2,4-D

2,4-D

2,4-D
2,4—D
2,4-D

2,4—D

106

0.0379
0.1073
0.2230

0.4092

0.6924
0.7670
0.4899

0.3680

-0.0287
0.1892
-0.0880

-0.0762

0.2455
0.0291
0.4913

-1.0028

 

 

 

 

 

 

 

 

107

 

   

 

 

 

 

 

 

 

a b
(15-
(13—
01-
411-
-0.3- -0.34
-0.5— -05-
Succinate Succinate 2,4-D
F8 153
. FB1 I SB1
D FB2 SB2
Figure 13. Performance changes of Stage I lines. (a) FB

lines; (b) SB lines.

95% confidence intervals,

Mean SRC values are shown, along with
for both the selective substrate

(succinate, left column) and the alternative substrate (2,4-
D, right column). * See Table 9.

108

   

-2sé

 

 

 

 

 

 

 

 

 

-353
2.4-D Succinate 2.4-D Succinate
FBB FCB
I FBB1 I FCB1
[3 F332 [3 F032
m FBBr1 fl FCBr1
FBBr2 fl FCBr2

 

 

 

 

 

 

Figure 14. Performance changes of Stage II F-derived lines.
(a) FBB lines; (b) FCB lines. Mean SRC values are shown,
along with 95% confidence intervals, for both the selective
substrate (2,4-D, left column) and the alternative substrate
(succinate, right column).

109

 

as.

 

 

 

 

 

 

 

 

 

2.4-D Succinate 2.4-D Succinate
SBB SCB
I 8881 - 8081
m 3332 3032
SBBrl I SCBr1
El SBBr2 SCBr2

 

 

 

 

 

 

Figure 15. Performance changes of Stage II S-derived lines.
(a) SBB lines; (b) SCB lines. Mean SRC values are shown,
along with 95% confidence intervals, for both the selective
substrate (2,4-D, left column) and the alternative substrate
(succinate, right column).

110

substrate between Stages I and II, improved Significantly

during Stage II.

Performance on alternative substrate

Among the thirteen lines that showed significant
improvement on their selective substrate, five of them (SB2,
FBBr1, FCBrl, SCBl and SCBr1) also improved significantly
(Table 10) on the alternative substrate (2,4—D for Stage I
and succinate for Stage II). Three additional lines (SB1,
FBB1, and SCB2) had non-significant performance increases on
the alternative substrate. In contrast, three of the
thirteen significantly adapted lines (FBBr2, FCB2, and SCBr2)
experienced significant losses of competitive ability on the
alternative substrate, and two others (FBB1 and SBBrl) were
suggestive of such losses (Table 10). Overall, ten lines
either show or suggest correlated improvements in both
substrates, while nine show or suggest fitness losses in the
alternative substrate (Table 10). One line (FCBr2) suggests
unexpected losses of performance on both the selective and
alternative substrates, although neither change was
statistically significant. Among the six sets of parallel
lines, only the SB set exhibited a homogeneous pattern of
performance for all its replicate lines (in this case only
two) on both selective and alternative substrates. The other
five sets showed (or suggested) both performance losses and

gains among their lines on the alternative substrate.

111

Table 10. Frequency of four types of correlated response to
selection. The number of lines that show statistically
significant correlated improvements (SCI) for both 2,4-D and
succinate, non-significant correlated improvements (NSCI) for
both substrates, non—significant tradeoffs between substrates
(NST), and significant tradeoffs between substrates (ST) are
shown. These categories are shown for the thirteen lines
that adapted significantly to their selective substrate and
then for the 19 total lines that either showed or suggested
such adaptation.

SCI NSCI NST ST

Significantly
Adapted 5 3 2 3

'RNal 5 5 6 3

112

Three lines where improvements on the selective
substrate were not significant nonetheless have significantly
changed performance on the alternative substrate. For
example, the data for SBB2 only suggest a small performance
increase on 2,4-D (SRC = 0.1073), its selective substrate,
yet this line shows a significant improvement in succinate
(SRC = 0.1892). FB2 and FCBl data also merely suggest
adaptation to their selective substrates of succinate and
2,4—D, respectively. These two lines, however, show
significant and very large performance losses on their
respective alternative substrates. Indeed, line FB2 appears
to have entirely lost its ability to grow on 2,4-D; pure
inocula of FB2 into 2,4-D medium never become turbid. Strain
FB2 was shown to be a true descendent of strain F by
comparison of their REP PCR “fingerprints" (Chapter 2). This
genetic test confirms that the inability of FB2 to grow on
2,4-D represents an evolutionary loss of that function
(rather than the accidental introduction of a non-2,4-D
degrading contaminant strain). Similarly, strain FCBl shows
the second largest fitness loss (SRC = —2.7843) on its
alternative substrate (succinate) of the twenty experimental

lines.

DISCUSSION
Population dynamics of adaptation
The evolutionary changes in competitive performance on

non—selective substrates could be caused either by

113
pleiotropic effects of adaptive mutations (Wright 1977) or by

the random drift of effectively neutral alleles (Kimura
1983). The latter possibility can be readily dismissed with
respect to my experiments. Only a very small minority of
mutations that affect fitness in a given environment are
expected to be beneficial rather than harmful, because it is
usually much easier to disrupt than to improve the
performance of a complex entity. That is, if one considers
mutations that are selectively neutral in one environment
(say, 2,4-D medium), but that affect fitness in an
alternative environment (say, succinate medium), then one
would expect most of them to reduce, rather than improve,
fitness in the alternative environment.

In this study, however, five of the eight evolved lines
that showed significant performance changes in both 2,4-D and
succinate media underwent improvement on both substrates.
These data therefore imply that improvements in the
alternative substrates were not due to substitution of random
mutations by drift, but instead were caused by the same
mutations responsible for improvements in the selective
regimes. In other words, the selected mutations were often
beneficial in both regimes, indicating positive pleiotropy.
This being the case, it seems likely that the three cases of
significant performance loss on the alternative substrate
(associated with significant performance improvement on the
selective substrate) are also due to pleiotropy, in this case

negative pleiotropy.

114

It is also likely that each line’s improvement in its
selective environment was due to only one or a few adaptive
mutations. This small number is because the duration of
experimental evolution in this study was relatively short
(maximum of 500 generations during either stage of
evolution). Mathematical models indicate that many
generations (often several hundred) are required for each
beneficial mutation to sweep sequentially through a large,
clonal population (Lenski et al. 1991). The results of
previous evolution experiments with E. coli support the
predictions of these models (Lenski and Travisano 1994; Elena

et al. 1996).

MUltiple adaptive pathways

My results clearly show that replicate evolving
populations often achieved adaptation to their selective
substrate by different mechanisms. This inference follows
from the fact that in three of the four Stage II sets (each
containing four independent replicate lines), at least one
line shows significant correlated improvements on both
selective and alternative substrates, whereas at least one
other line shows a significant loss of performance on its
alternative substrate after adaptation to its selective
substrate (Table 9, Figures 14 and 15). Another way to
describe these patterns is that the variance in competitive
fitness among replicate populations on alternative substrates

is much greater than the corresponding variance on selective

115

substrates (Travisano et al. 1995). Thus, among the strains
that showed significant improvement on their selective
substrate, the corresponding SRC values ranged from 0.25 to
0.97; whereas among the strains that changed significantly in
their performance on the alternative substrate, the SRC
values ranged from -4.61 to 1.96 (Table 9).

Even for replicate lines within a set that showed the
same pattern of performance changes on both selective and
alternative substrates, there is sometimes additional
evidence that these changes were caused by different
underlying physiological mechanisms. For example, lines SCBl
and SCBr1 both adapted significantly to their selective
substrate (2,4-D), and they both showed significant
correlated improvements on their alternative substrate
(succinate). However, data presented in Chapter 2 of this
dissertation suggest that one of these lines, SCBl, had
improved performance in the chemostat culture regime during
adaptation to the batch regime, whereas the other strain,
SCBr1, appears to have reduced performance in chemostat
culture. [The difference in chemostat performance between
each strain and its proximate ancestor was not significant,
but the resulting difference between the two evolved strains
was highly significant (t = 8.298, 8 degrees of freedom, 2-
tailed P < 0.0001).] Assuming that these changes in
Chemostat performance reflect the pleiotropic effects of

mutations that were adaptative in the batch regime, then

116

these data would imply that even dual improvements of SCBl
and SCBr1 had occurred by different underlying mechanisms.
The replicate lines that indicate different mechanisms
of adaptation to 2,4-D in batch culture were founded from
base populations that were initially isogenic. This
corroborates the results of Travisano et al. (1995), which
showed that natural selection can cause significant
divergence of populations within a species even when the
populations are founded from the same progenitor and
experience identical environments. Such divergence is
presumably due to the fact that the replicate populations
incur different sequences of random mutations during
evolution, thus providing distinct patterns of genetic
variation across populations upon which natural selection may

act.

Potential physiological mechanisms of adaptation

This study focuses on the evolution of competitive
performance of bacterial strains at the population level, and
the results demonstrate a variety of correlated responses for
competitive ability on alternative substrates. For these
reasons, detailed analyses of the specific mechanisms of
genetic adaptation for the experimental lines was not
feasible within the scope of this study. Nonetheless, I will
briefly review physiological functions that are involved in
growth on succinate and 2,4—D, and then discuss a few

alternative physiological mechanisms that might account for

117

the performance changes observed in evolved lines. It will
become evident, however, that the performance changes alone
are insufficient to distinguish between alternative
scenarios.

Succinate is a key metabolite in the TCA cycle, which
has presumably been fine-tuned by natural selection over many
millions of years. It is conceivable that some regulatory
changes to the TCA cycle might be advantageous if the
organism is given a steady diet of nothing but succinate. It
seems more likely, however, that transport of succinate into
the cell would be the physiological step with the greatest
room for improvement. By contrast, growth on 2,4-D requires
several more specialized enzymes to take it through the
various steps by which it is degraded into succinate and
enters the central metabolic circuitry. Hence, improvements
in growth on 2,4-D might be expected to involve not only
transport but also the regulation and expression of any of
these enzymes.

The genes responsible for 2,4-D degradation are often
plasmid-borne but in some strains they are located on the
chromosome (Don and Pemberton 1981, Amy et al. 1985). The
most extensively studied plasmid that encodes 2,4—D
degradative functions is pJP4 from Alcaligenes eutrophus
strain JMP134 (Don et al. 1985). This plasmid carries seven
genes that are responsible for the breakdown of 2,4—D into
succinate, and possibly one or more genes that code for the

proteins responsible for 2,4—D transport. The first three

118
steps in this pathway (encoded by genes tfdA, tde, and

tde), result in cleavage of the aromatic ring. The rate-
limiting step for 2,4-D catabolism is widely thought to be
one of these first three steps (L. Forney, personal
communication). However, circumstantial evidence suggests
that permeability of the cell to 2,4-D may also sometimes
limit a strain’s overall growth rate on 2,4-D (T. Sassanella,
personal communication).

spage I selegtign 9n sucginate, Line SB2 adapted
significantly to growth on succinate, and it simultaneously
showed significant improvement in its performance on 2,4-D.
Plausible physiological foci of adaptation that could explain
correlated improvement on these two substrates include
succinate transport system, the TCA cycle, or some other
function that is not directly involved in succinate
metabolism. For example, an acceleration of a rate-limiting
step in the TCA cycle could benefit growth on both
substrates, as could an adaptation to some variable other
than the type of substrate, such as pH or temperature. And
while it seems likely that succinate and 2,4-D are
transported into the cell by different mechanisms (given
their very different biophysical characteristics), one can
nonetheless imagine that a change in some general feature of
the cell surface, such as loss of a capsule, could benefit
growth on both substrates if transport is limiting (T. Marsh,

personal communication).

119
Another Stage I line, FB2, completely lost its ability

to grow on 2,4—D. The 2,4-D degradative pathway of strain F
is chromosomally encoded (Amy et al. 1985; strain F is named
'RASC' in this reference), and so the loss of the ability to
grow on 2,4-D was not simply due to plasmid segregation, as
might have occurred in a strain with a plasmid-borne 2,4—D
degradative pathway. Instead, the loss must be due to a
mutation in a chromosomal gene that is critical for 2,4—D
degradation, but it is not clear whether that mutation was
somehow beneficial in succinate or whether it might be an
instance of random genetic drift.

Stage II selegtign gn 2.4-D. Among the thirteen Stage
II lines that adapted to 2,4—D, some became better
competitors for succinate whereas others became worse. Their
improved performance on 2,4-D could be due to modifications
in 2,4-D transport, 2,4-D specific catabolism, central
metabolism, or some function not directly involved in
nutrient acquisition, such as pH regulation. Improvements in
the last two categories would seem most likely to benefit
growth on both 2,4-D and succinate. However, as noted above,
any change in the cell envelope that increased its overall
permeability could also have this effect, even if the
specific transport systems for 2,4-D and succinate were
completely different. Hence, it seems safe to exclude only
changes that are specific to 2,4-D catabolic functions in the
case of those lines that showed correlated improvements on

succinate.

120
The physiological bases of adaptation to 2,4-D for those

strains that became worse on succinate is similarly unclear.
Improvements in 2,4-D transport or catabolism could prove
detrimental for growth on succinate. For example, the
duplication of one or more genes specific for growth on 2,4-D
may enhance growth rate on 2,4—D. That same duplication,
however, may impose a significant metabolic cost that reduces
the rate of growth on other substrates such as succinate.
Alternatively, a mutation that repressed a membrane protein
involved in succinate transport might benefit growth on 2,4-D
either by allowing a higher level of expression of an
alternative 2,4-D transport protein or by reducing the cost
associated with an unused function.

There are undoubtedly many more scenarios that could be
put forward to explain any given pattern of evolutionary
change with respect to the selective and alternative
substrates. I have presented these scenarios to illustrate
the difficulties of invoking overly specific relationships
between selective factors in the environment and an
organism’s evolutionary response to those factors. At the
same time, the diversity of plausible mechanisms suggests
that the physiological heterogeneity among replicate
populations may be far greater than even their divergent

ecological performances would indicate.

121

Possible relevance to the adaptation of
genetically modified organisms in natural habitats

In the discussion of relative benefits and risks of
releasing genetically modified organisms into nature, it has
often been argued that genetically altered microorganisms
will not persist because they are inherently less fit than
indigenous competitors (Brill 1985; Davis 1987; Davies 1988).
Lenski (1993) reviewed various arguments for why genetically
modified organisms might theoretically be less fit than their
natural counterparts, and be summarized some experiments with
bacteria that bear on those arguments.

Most of the studies reviewed by Lenski (1993) support
the view that modified microorganisms are competitively
inferior to the Strains from which they are derived.
Mutations that alter basic metabolic functions (Lenski 1988),
the expression of additional functions (Andrews and Hegeman
1976; Dykhuizen 1978; Koch 1983), the carriage of accessory
genetic elements (Lenski and Bouma 1987), and the
“domestication" of bacteria as they adapt to laboratory
conditions (Davis 1987; Regal 1988) all tend to decrease the
fitness of modified organisms relative to their progenitors,
although important exceptions to this generality exist (Biel
and Hartl, 1983; Edlin et al. 1984; Hartl et al. 1983; Blot
et al. 1994). Beyond the immediate fitness effects of
genetic manipulations that occur prior to release, there is
the further possibility that genetically modified organisms

may adapt evolutionarily to the local ecosystem in which they

122

are released. Such “post—release” adaptation seems more
likely than fortuitous preadaptation prior to release, and it
could lead to the indefinite persistence of modified
organisms in the environment (along with any adverse effects
they might cause).

Several studies, including three with bacteria in the
laboratory (Bouma and Lenski 1988; Lenski 1988; Modi and
Adams 1991) and one with insects in the field (McKenzie et
al. 1982), indicate that organisms may evolve so as to
mitigate the fitness costs associated with other genetic
changes. In one case, a population of E. coli carried a
plasmid, pACYC184, that encoded resistance to two
antibiotics, but which reduced the bacteria’s fitness in the
absence of antibiotic (Bouma and Lenski 1988). These
plasmid-bearing cells were then propagated in medium that
contained antibiotic, in order to prevent spontaneous
plasmid—free segregants from out-competing the plasmid—
bearing cells and taking over the population. After only 500
generations, the plasmid—bearing cells were, suprisingly,
more fit than both their plasmid-free progenitor and their
isogenic plasmid-free derivatives even in the absence of
antibiotic (R. Lenski, personal communication; Bouma and
Lenski 1988). In other words, evolution of a genetically
modified organism in one environment (with antibiotic) led to
a correlated improvement in fitness in an alternative

environment (without antibiotic).

123

The above results point towards a second concern over
the evolution of microorganisms that have been genetically
modified for bioremediation in the environment. What is the
effect of adaptation to an anthropogenic substrate on their
competitive fitness for naturally occurring substrates? If
such adaptation may have pleiotropic benefits for growth on
natural substrates, then this increases the likelihood that
these genetically modified organisms will persist in the
environment. The results of this study clearly show that
evolutionary adaptation of bacteria to growth on a common
toxic herbicide (2,4-D) sometimes improved their competitive
fitness on a natural substrate (succinate). The strains
employed in this investigation were already capable of
degrading 2,4-D, without genetic modification in the
laboratory. However, there is no obvious reason that
qualitatively similar results might be obtained for strains
that are genetically modified to degrade chlorinated aromatic
compounds (or perform some other bioremediation function)
more efficiently than natural strains (Ramos et al. 1987;
Chaudhry and Chapalamadugu 1991). In fact, such modified
organisms may have even more opportunities for general
improvements in their overall vigor, owing to the fitness
costs associated with their modification (see above).
Therefore, when considering the fate of genetically modified
organisms released into the environment for bioremediation of
anthropogenic substrates, it seems unrealistic to assume that

these organisms will simply do their job and then fade away.

Reference List

REFERENCE LIST

Any, P. 8., J. W. Schulke, L. M. Frazier, and R. J. Seidler.
1985. Characterization of aquatic bacteria and cloning of
genes Specifiying partial degradation of 2,4—
dichlorophenoxyacetic acid. Applied and Environmental
.Microbiology'49:1237—1245.

Andrews, J. E., and R. F. Harris. 1986. r- and K- selection
in microbial ecology. AdVances in Microbial Ecology 9:99—147.

Andrews, J. E., and D. I. Rouse. 1982. Plant pathogens and
the theory of r— and K- selection. American Naturalist
120:283-296.

Andrews, K. J., and G. D. Hegeman. 1976. Selective
disadvantage of nonfunctional protein synthesis in
EScherichia coli. JOurnal of.Molecular EVolution 8:317-328.

Bell, 6., and'V. Koufopanou. 1986. The cost of reproduction.
Oxford Surveys in EVolutionary Biology 3:83—131.

Bennett, Am F., and R. E. Lenski. 1993. Evolutionary
adaptation to temperature. II. Thermal niches of experimental
lines of EScherichia coli. EVolution 47:1-12.

Biel, 8. W}, and D. L. Hartl. 1983. Evolution of transposons:
natural selection for Tn5 in EScherichia coli K12. Genetics
103:581-592.

Bouma, J. E., and R. E. Lenski. 1988. Evolution of a
bacteria/plasmid association. Nature 335:351-352.

Brill,‘W. J. Safety concerns and genetic engineering in
agriculture. Science 227:381-384.

Brueggenan, J. 1993. Element G. Chemical methods only. Human
health effects of 2,4-D. Ebasco Environmental, Bellevue,
Washington.

Calabrese, V} Gm M. 1994. Ph.D. dissertation. Michigan State
University, East Lansing, MI.

Calabrese, V; G. E., N. L. Tonso, A, J. Sexstone, and‘W. E.
Holben. 1992. Characterization and classification Of diverse
2,4-dichlorophenoxyacetic acid (2,4-D) degrading populations
using whole plasmid probes in colony hybridization analyses.
Abstracts of the 92nd General Meeting of the American Society

124

125

for.Microbiologyu American Society for Microbiology,
Washington, D. C.

Chao, L., and E. C. Cox. 1983. Competition between high and
low mutating strains of EScherichia coli. EVolution 37:125-
134.

Chao, L., and B. R. Levin. 1981. Structured habitats and the
evolution of anticompetitor toxins in bacteria. Proceedings
of the National Academy of Sciences USA 78:6324-6328.

Chao, L., and G. Ramsdell. 1985. The effects of wall
populations on coexistence of bacteria in the liquid phase of

chemostat cultures. JOurnal of General.Microbiology 131:1229-
1236.

Chaudhry, G. R. and S. Chapalamadugu. 1991. Biodegradation of
halogenated organic compounds..Microbiological Reviews 55:59-
79.

Chaudhry, G. R., and G. H. Huang. 1988. Isolation and
characterization of a new plasmid from a Flavobacterium sp.
which carries the genes for degradation of 2,4-
dichlorophenoxyacetic acid. JOurnal of Bacteriology'170:3897-
3902.

Cheverud, J. E., M. M; Dow, and'w. Leutenegger. 1985. The
quantitative assessment of phylogenetic constraints in
comparative analyses: sexual dimorphism in body weight among
primates. EVolution 39:1335-1351.

Clutton-Brock, T. H. and P. H. Harvey. 1977. Primate ecology
and social organization. JOurnal of Zoology 183: 1—33.

Cody, M. L. 1966. A general theory of clutch size. Eyolution
20:174-184.

Cohan, F. M. 1984. Can uniform selection retard random
genetic divergence between isolated conspecific populations?
EVolution 38:495-504.

Cohan, F. E., and J.-D.Gra£. 1985. Latitudinal cline in
Drosophila melanogaster for knockdown resistance to ethanol
fumes and for rates of response to selection for further
resistance. EVolution 39:278-293.

Davies, J. 1988. Genetic engineering: processes and products.
Trends in Ecology and Evolution 3:87-511.

Davis, R. D. Bacterial domestication: underlying assumptions.
Science 235:1329-1335.

126

Don, R. E., and J. M. Pemberton. 1981. Properties of six
pesticide degradation plasmids isolated from Alcaligenes

[paradoxus and Alcaligenes eutrophus. JOurnal of Bacteriology
145:681-686.

Don, R. R., A. J. Weightman, R. J. Rnaclnnuss, and K. N.
Timmis. 1985. Transposon mutagenesis and cloning analysis of
the pathyways for degradation of 2,4-dichlorophenoxyacetic
acid and 3-Chlorobenzoate in Alcaligenes eutrophus JMP134
(pJP4). JOurnal of Bacteriology 161:85—90.

Dykhuizen, D. E., and A. M. Dean. 1990. Enzyme activity and
fitness: evolution in solution. Trends in Ecology'and
EVolution 5:257-262.

Dykhuizen, D. E., A. E. Dean, and D. L. Hartl. 1987.
Metabolic flux and fitness. Genetics 115:25-31.

Dykhuizen, D. E., and L. Green. 1991. Recombination in
EScherichia coli and the definition of biological species.
JOurnal of Bacteriology 173:7257—7268.

Ebasco and Herra. 1993. Element E. Chemical methods only.
Environmental effects of glyphosates and 2,4-D. Ebasco
Environmental, Bellevue, Washington and Herra Environmental
Consultants, Inc., Seattle, Washington.

Edlin, G., R. C. Tait, and R. L. Rodriguez. 1984. A
bacteriophage Lambda cohesive ends (cos) DNA fragment
enhances the fitness of plasmid—containing bacteria growing
in energy-limited Chemostats. Bio/Technology 2:251-254.

Elena, S. F.,‘V. 8. Cooper, and R. E. Lenski. 1996.
Punctuated evolution caused by selection of rare beneficial
mutations. Science 272:1802-1804.

Felsenstein, J. 1985. Phylogenies and the comparative method.
American Naturalist 125:1-15.

Geller, A, 1983. Growth of bacteria in inorganic medium at
different levels of airborne organic substances. Applied and
Environmental Microbiology 46:1258—1262.

Gerson, U. and I. Chet. 1981. Are allochthonous and
autochthonous soil microorganisms r— and K— selected? Rev.
Ecol. Biol. Sol. 18:285-289.

Gould, S. J., and R. C. Lewontin. 1979. The spandrels of San
Marco and the panglossian paradigm: a critique of the
adaptationist program. Proceedings of the Royal Society of
London B 205:581-598.

Grafen, A. 1989. The phylogenetic regression. Philosophical
Transactions of the Royal Society 326:119-156.

127

Greer, L. E., J..A. Robinson, and D. R. Shelton. 1992.
Kinetic comparison of seven strains of 2,4-dichlorophenoxy—
acetic acid-degrading bacteria. Applied and Environmental
Microbiology'38:1027-1030.

Hansen, 8. R., and S. P. Hubbell. 1980. Single-nutrient
microbial competition: qualitative agreement between

experimental and theoretically forecast outcomes. Science
207:1491-1493.

Hartl, D. L., D. E. Dykhuizen, R. D. Miller, L. Green, and J.
deFramond. 1983. Transposable element I850 improves growth
rate of E. coli cells without transposition. Cell 35:503—510.

Harvey, P. H. and.M. D. Pagel. 1989. Comparative studies in
evolutionary ecology: using the database. In waard a more
exact ecology (ed. P. J. Grubb and J. Whittaker), pp. 209-30,
Blackwell, Oxford.

Harvey, P. H. and.M. D. Pagel. 1991. The Comparative.Method
in Evolutionary'Biology; Oxford University Press, Oxford.

Healey, F. P. 1980. Slope of the Monod equation as an
indicator of advantage in nutrient competition..Microbial
Ecology 5:281-286.

Hoffman,.A. A., and F. M. Cohan. 1987. Genetic divergence
under uniform selection. III. Selection for knockdown
resistance to ethanol in Drosophila pseudoobscura populations
and their replicate lines. Heredity 58:425-433.

Jannasch, H.‘W. 1969. Estimations of bacterial growth rates
in natural waters. JOurnal of Bacteriology 99:156-160.

Kim, J., L. R. Ginsburg, and D. E. Dykhuizen. 1991.
Quantifying the risks of invasion by genetically engineered
organisms. In L. R. Ginzburg (ed.). Assessing Ecological
Risks of Biotechnology. Butterworth-Heinemann, Boston. pp.
193-214.

Rimura, M. 1983. The.Neutral Theory of.Mo1ecular Evolution.
Cambridge University Press, Cambridge.

Roch,.A. L. 1983. The protein burden of lac operon products.
JOurnal of Mblecular EVolution 19:455-462.

Konings, Wk N. and H. veldkamp. 1980. Phenotypic responses to
environmental change. In D. C. Ellwood, J. N. Hedger, M. J.
Lathan, J. M. Lynch, and J. H. Slater (eds.). Contemporary
.Microbial Ecology; Academic Press, New York. pp. 161-191.

128

Korona, R., C. H. Hakatsu, L. J. Forney, and R. E. Lenski.
1994. Evidence for multiple adaptive peaks from populations
of bacteria evolving in a structured habitat. Proceedings of
the National Academy of Sciences USA 91:9037-9041.

Kuenen, J. G., and Harder, W. 1982. Microbial competition in
continuous culture. In R. G. Burns and J. H. Slater (eds.).
Experimental.Microbial Ecology. Blackwell Scientific
Publications, Oxford, England. pp. 342-367.

Lenski, R. E. 1988. Experimental studies of pleiotropy and
epistasis in Eacherichia coli. I. Variation in competitive
fitness among mutants resistant to virus T4. Evolution
42:425—432.

Lenski, R. E. 1993. Evaluating the fate of genetically
modified microorganisms in the environment: Are they
inherently less fit? Experientia 49:201—209.

Lenski, R. E., and B. R. Levin. 1985. Constraints on the
coevolution of bacteria and virulent phage: a model, some
experiments, and predictions for natural communities.
American Naturalist 125:585—602.

Lenski, R. E., M. R. Rose, 8. C. Simpson and S. C. Tadler.
1991. Long-term experimental evolution in Escherichia coli.
I. Adaptation and divergence during 2,000 generations.
American Naturalist 138:1315-1341.

Lenski, R. E., and M. Travisano. 1994. Dynamics of adaptation
and diversification: a 10,000 generation experiment with
bacterial populations. Proceedings of the National Academy of
Sciences USA 91:6808-6814.

Lenski, R. E., G. J. velicer and R. Korona. 1997.
Experimental studies on the evolution of bacteria that
degrade 2,4—dichlorophenoxyacetic acid. In Horikoshi, K., M.
Fukuda, and T. Kudo (eds.). Microbial Diversity and Genetics
of Biodegradation. Japan Scientific Societies Press, Tokyo.
In press.

Leroi, A. E., A, R. Chippindale, and M. R. Rose. 1994. Long—
term laboratory evolution of a genetic life-history trade-off
in Drosophila melanogaster. 1. The role of genotype-by—
environment interaction. EVolution 48:1244-1257.

Levin, B. R. 1988. Frequency-dependent selection in bacterial
populations. Philosophical Transactions of the Royal Society
London B 319:459-472.

Levin, B. R., F. M. Stewart, and L. Chao. 1977. Resource—
limited growth, competition, and predation: a model and
experimental studies with bacteria and bacteriophage.
American Naturalist 111:3-24.

129

Luckinbill, L. S. 1978. r- and K- selection in experimental
populations of Escherichia coli. Science 202:1201—1203.

Luckinbill, L. S. 1979. Selection and the r/K continuum in

experimental populations of protozoa. American Naturalist
113:427—437.

Luckinbill, L. S. 1984. An experimental analysis of a life
history theory. Ecology 65:1170-1184.

Lynch, M. 1991. Methods for the analysis of comparative data
in evolutionary biology. EVolution 45:1065-1080.

thArthur, R. H., and E. O. Wilson. 1967. The Theory of
Island Biogeography. Princeton University Press, Princeton,
N.J.

Martins, E. P. and T. H. Garland. 1991. Phylogenetic analyses
of the correlated evolution of continuous characters: A
simulation study. Evolution 45:534-557.

Eatin,.A., and H. veldkamp. 1978. Physiological basis of the
selective advantage of a Spirillum species in a carbon-
limited environment. Journal of General.Microbiology 105:187-
197.

EcGowan, C. 1995. Ph.D. dissertation. Michigan State
University, East Lansing, MI.

McKenzie, J. A., M. J. Whitten, and M. A. Adena. 1982. The
effect of genetic background on the fitness of diazinon
resistance genotypes of the Australian sheep blowfly, Lucilia
cuprina. Heredity'49zl—9.

Modi, R. I., and.Adams, J. 1991. Coevolution in bacteria-
plasmid populations. Evolution 45:656-667.

Mbngold, J. A.,.A. F. Bennett, and R. E. Lenski. 1996.
Evolutionary adaptation to temperature. IV. Adaptation of
Eacherichia coli at a niche boundary. EVolution 50:35—43.

iflueller, L. D. and F. J..Ayala. 1981. Trade-off between r—
selection and K—selection in Drosophila populations.
Proceedings of the National Academy of Sciences USA, 78:1303—
1305.

National Research Council. 1993. In situ bioremediation: When
does it work? National Academy Press, Washington, D. C.

Pianka, E. R. 1970. On r- and K-selection. American
Naturalist 104:592-596.

130

Ramos, J. L., A.‘Wasserfallen, K. Rose, and R. N. Timmis.
1987. Redesigning metabolic routes: manipulation of TOL

plasmid pathway for catabolism of alkylbenzoates. Science
235:593—596.

Regal, P. J. 1988. The adaptive potential of genetically
engineered organisms. Trends in ECology and EVolution 3:536—
538.

Regal, P. J. 1993. The true meaning of ‘exotic species’ as a
model for genetically engineered organisms. EXperientia
49:225-234.

Reznick, D. 1985. Costs of reproduction: an evaluation of the
empirical evidence. Oikos 44:257-267.

Roughgarden, J. 1971. Density-dependent natural selection.
Ecology'52:453-468.

Service, P. M. and.M. R. Rose. 1985. Genetic covariation
among life history components: the effect of novel
environments. EVolution 36:1276-1282.

Stearns, S. 1983. The influence of size and phylogeny of
patterns of covariation among life-history traits in mammals.
Oikos 41:173-187.

Swofford, D. L. 1991. PAUP: Phylogenetic Analysis Using
Parsimony, version 3.05. Computer program distributed by the
Illinois Natural History Survey, Champaign, Illinois.

Tiedje, J. H., R. R. Colwell, Y. L. Grossman, R. Hodson, R.
E. Lenski, R. N. Rack and P. J. Regal. 1989. The planned
introduction of genetically engineered organisms: ecological
considerations and recommendations. Ecology 70:298—315.

Tonso, N. L., V; G» Natheson, and‘w. E. Holben. 1995.
Polyphasic Characterization of a suite of bacterial isolates
capable of degrading 2,4-D. Microbial Boology 30:3-24.

Travisano, H., J. A. angold, A- F. Bennett, and R. E.
Lenski. 1995. Experimental tests of the roles of adaptation,
chance, and history in evolution. Science 267:87-90.

Travisano, E., and R. E. Lenski. 1996. Long—term experimental
evolution in Escherichia coli. IV. Targets of selection and
the specificity of adaptation. Genetics 143:15-26.

Travisano, E., F. vasi, and R. E. Lenski. 1995. Long-term
experimental evolution in Eacherichia coli. III. Variation
among replicate populations in correlated responses to novel
environments. EVolution 49:189—200.

131

vasi, F., N. Travisano, and R. E. Lenski. 1994. Long-term
experimental evolution in Eacherichia coli. II. Changes in
life-history traits during adaptation to a seasonal
environment. American Naturalist 144:432-456.

veldkamp, H., H. van Gemerden,'w; Harder, and H. J.
Laanbroek. 1984. Competition among bacteria: An overview. In
M. J. Klug and C. A. Reddy (eds.). Current Perspectives in
.Microbial Ecology: American Society for Microbiology,
Washington, D. C. pp. 279—290.

verealovic, J., T. Roeuth, and J. R. Lupski. 1991.
Distribution of repetitive DNA sequences in eubacteria and
application to fingerprinting of bacterial genomes. NUcleic
Acids Research 19:6823-6831.

vogel, T. N. and P. L. EcCarty. 1985. Biotransformation of
tetrachloroethylene to trichloroethylene, dichloroethylene,
vinyl chloride, and carbon dioxide under methanogenic
conditions. Applied and Environmental.Microbiology'49zlo80-
1083.

warner, R. R. 1984. Deferred reproduction as a response to
sexual selection in a coral reef fish: a test of the life
historical consequences. EVolution 38:148-162.

weisburg, W; G., S. H. Barns, D..A. Pelletier, and D. J.
Lane. 1991. 16S ribosomal DNA amplification for phylogenetic
study. JOurnal of Bacteriology 173:697-703.

wright, S. 1970. “Surfaces" of selective value. Proceedings
of the National Academy of Sciences USA 58:165-172.

Zhou, J., and J. N; Tiedje. Manuscript. Kinetic
specialization of benzene-degrading bacteria for parts-per-
billion concentrations.

”lllllllllllllllllllls