Annu. Rev. Genet. 1994.

28:237—61
Copyright © 1994 by Annual Reviews Inc. All rights reserved

HORIZONTAL GENE
TRANSFER: Evidence and
Possible Consequences

M. Syvanen
Department of Medical Microbiology and Immunology, School of Medicine, University
of California, Davis, California 95616

KEY WORDS: phylogeny, mosaics, angiosperm, molecular evolution, transgenic organisms

CONTENTS
INTRODUCTION 237
TRANSFERS INVOLVING EUKARYOTES 239
Phylogenetic Congruency Test 239
Pitfalls Involving Paralogies 241
Pitfalls Involving Unequal Razes 242
Convergent Replacements 243
Tests of Significance 244
MOBILE GENETIC ELEMENTS 244
CHROMOSOMAL GENES IN BACTERIA 245
Mosaic Nature of the E. coli chromosome 245
Salmonella and the Criteria of GC Content 247
Codon Bias 249
Other Bacteria 250
ORGANELLAR GENOMES 251
THE ANGIOSPERM PARADOX 251
Classification Problem 251
Temporal Problem 253
CONCLUSION 255

As careful observation is far harder work than generalization, and still harder than
speculation, do you think it very possible that it may be overvalued.
From C. Darwin to Asa Gray, 1857.

INTRODUCTION

The possibility that genetic information can move between remotely related
species is an idea that has been strenuously resisted within traditional schools
237

0066-4197/94/1215-0237$05.00
238 SYVANEN

of biology. This is especially true of evolutionists who have developed neo-

Darwinian theory. This resistance is natural, since superficially the concept seems
at odds with the explanatory power of the phylogenetic tree in taxonomy and the
important role of reproductive isolation as a mechanism of speciation. At the same
time, progress in the field of molecular genetics over the past 25 years has led to
many different observations consistent with horizontal gene transfers. It is not just
that there are dozens of examples of likely horizontal transfers, but that the
mechanisms which could accommodate horizontal transfer of DNA are observed
everywhere. Multiple mechanisms for the physical transfer of DNA from one
species to another are known. Recombination mechanisms that can absorb this
DNA are ubiquitous. Examples of cells and organisms that will express and
incorporate products from foreign genes are too numerous to list. Indeed, the
concept of genetic coadaptation, which was a seriously accepted idea 10 years
ago, is now rarely cited. Coadaptation posits that genes within a genome become
so specifically coadapted with one another that they will fail to function with
alternative alleles. The many cases of transgenic animals and cell lines created in
vitro renders this idea obsolete. The only remaining question is whether the
horizontal transfer of genes occurs at a rate that significantly influences evolution.
In 1982, when I first theorized that horizontal gene transfers are likely a major
evolutionary force (91), my reasons were twofold: First, it had recently been
established that the mechanisms for gene transfer were present at a molecular
level, and my intuition told me that if a mechanism so potentially useful existed,
nature would find a way to use it. Second, a general evolutionary theory
incorporating the idea of the migration of genetic information across taxonomic
boundaries seemed to provide a simple and satisfying answer to a question seldom
asked: Why is the molecular biology for all living organisms so unified? In the
face of speciation, biology has retained a unity so profound that transgenic animals
can be created in the laboratory—a phenomenon that few would have predicted in
1982. In fact, most evolutionists still believe that a theory of evolution
incorporating horizontal transfer conflicts with the useful notions of phylogenetic
trees and reproductive isolation. In the following discussion on the nature of the
evidence supporting horizontal transfers, I try to resolve this apparent
contradiction.
When I first discussed the theoretical implications of horizontal transfer, there
were less than half a dozen suspected examples involving eukaryotes in the
literature (90, 91); only one of these has held up under closer scrutiny. Yet reports
of suspected cases continue to grow, and many appear to be serious candidates.
Because so many examples have been offered and then retracted, and because
there remains considerable skepticism as to the occurrence of horizontal transfer, I
examine some of the criteria for identifying gene migrations when they have
occurred, and I consider the rigor of the various criteria
 HORIZONTAL GENE FLOW 239

for making this judgment. I then explore some old problems that concern
flowering plants. In this case I believe transfers may be too frequent to apply
criteria that depend on knowledge of species phylogeny. I try to document the
crisis in angiosperm systematics and paleontology between neoDarwinian theory
and its database—a crisis that could be resolved by a theory of evolution
incorporating horizontal gene transfer.
The topic of horizontal transfer overlaps with the endosymbiotic theory of
organelles. Endosymbiosis basically involves the fusion of the entire genomes of
two organisms. I am considering these to be one part of the larger phenomenon of
cross-species gene transfer, which involves, in addition to endosymbiotic fusion,
the insertion of smaller genetic regions, including single genes or even parts of
genes. The mechanisms of transfer will likely involve viruses (30), direct
transformation, conjugation, or other as yet unexplored means. This review does
not focus on the possible mechanisms for DNA penetration through species
boundaries.
With the growth of the DNA database, especially of large homologous sets, the
subject of horizontal gene transfer has grown quickly and has been recently
reviewed (38, 82, 86).

TRANSFERS INVOLVING EUKARYOTES

Identifying horizontal transfers that have occurred in the recent past, such as the
spread of antibiotic resistance plasmids, is not problematic. The difficult cases are
those that have occurred so far in the past that subsequent divergence of the gene
sequence may have obscured the original relationships. It is on these cases that I
concentrate.

Phylogenetic Congruency Test

The most rigorous criteria available for establishing the occurrence of ancient
horizontal gene transfer involve comparing the phylogenetic tree constructed with
regard to a specific protein or gene from a number distantly related organisms with
that of the known phylogeny for those species. If one member belongs to a
radically unexpected assemblage, i.e. if an incongruency is seen between the
“gene tree” and the “species tree”, then a case can be made for a horizontal
transfer. This test, which has been used with varying degrees of rigor since the
earliest claims of horizontal transfer, has been most explicitly defined by Smith et
al (82).
Ideally, application of the phylogenetic congruency test requires that a number
of conditions be fulfilled. First, the genes being considered must have
phylogenetic information. Second, the homology being compared must represent
orthologous, but not paralogous, genes (see below). And third, the incongruent
example should be a gene whose substitution rate is not radically
240 SYVANEN

Table 1 Some horizontal transfers involving eukaryotes

Protein Case Reference
1. Glucose phosphate iso- gpi from E. coil was inherited from a relative of the 82
merase dicot Clarkia ungulata
2. Fe superoxide dismutase The Fe SOD from the protist Entamoeba his- 82
tolytica has prokaryotic origins
3. Aldolase Class 11 aldolase from yeast shows affinity to that 82
from E. coil
4. Cytochrome c The protein from Arabidopsis thaliana is of fungal 37
affinity
5. Xylanase Rumoncoccus-like gene is found in a rumen fungus 23

6. Thioredoxin Thioredoxin-m in plants has a bacterial origin 31

7. Glyceraldehyde-3-phos- gapdhA from E. coli and Anabaena have an affinity 16. 54

phate dehydrogenase with eukaryotic gapC
8. Elongation factor Tu rufA in Arabidopsis is from its endosymbiont 4

9. Ribosomal proteins L21 L2l and L22 in some plants is from its en- 22, 55
and L22 dosymbiont

These examples all involve nuclear genes in eukaryotes. Examples 1—

7 were identified by the phylogenetic congruency test and 8 and 9 were
inferred from finding Tu. L21. and L22 in both nuclear and chloroplast
genomes among contemporary plants.

different from the other genes being compared. Before these three
conditions can be assured, the number of genes being compared
needs to be reasonably large, i.e. more that 5 and possibly more than
10. Table 1 summarizes examples that involve eukaryotes, including
those reviewed by Smith et al. The phylogenetic congruency test
supports the hypothesis of lateral transfer in seven of the nine cases.
All of these examples involve a complete nuclear gene. Examples
6—10 are probably bacteria-to-eukaryotic transfers that involved
mitochondrial or chloroplast intermediates, i.e. they were part of the
original endosymbiosis that gave rise to those organelles; there are
no clues as to the mechanism of transfer for the other six.
In some of the examples in Table 1, the direction of transfer is
given. Logically, directionality is difficult to infer from the
phylogenetic congruency test, especially for ancient transfers
involving lineages that gave rise to many extant taxa. An example of
this difficulty is illustrated by glyceraldehyde
 HORIZONTAL GENE FLOW 241

3-phosphate dehydrogenase (line 7, Table 1). Doolittle et al (16) concluded that

gapdhA in Escherichia coli came from a eukaryotic source, since it was the sole
prokaryote in a eukaryotic clade. But with the finding of an orthologous gapdhA in
Anabaena, it seems more likely that an ancestral bacterium donated gapC to the
eukaryotes (54).

Pitfalls involving Paralogies

Most of the premature reports of possible horizontal transfers resulted from
comparing paralogous genes and treating them as if they were orthologous. Two
genes are defined as paralogous if they diverged after a gene duplication, whereas
the two genes are defined as orthologous if they diverged after a speciation event.
Let us construct the following scenario to illustrate this problem. Consider a gene
in an ancestral strain that underwent a duplication to give rise to gene A and gene
B, which subsequently diverged from one another. Further consider the evolution
of that ancestral strain into two major branches. In the extant species, all
descendants of gene A will belong to one orthologous set, and descendants of
gene B will belong to another. Members within the gene A group are said to be
paralogous to members of the gene B group. If both sets can be identified in the
extant species, this does not pose a problem. However, consider what happens if,
during the evolution of these two major branches, the duplicated pair is reduced
back to a single copy, and A survives in some extant species while B survives in
others. This could result in both gene A and gene B being located in the two
major clades and their paralogous nature not being recognized. If a single tree is
constructed, then the A-B gene tree would be incongruent with the species tree
and hence a horizontal transfer incorrectly inferred. This scenario can also arise
from what is essentially a sampling error, when too few genes and their
incompletely characterized products are compared.
Comparisons of paralogous genes were responsible for the early report that the
Cu-Zn superoxide dismutase in the bacterium Photobacterium leiognathi came
from its eukaryotic host, and that the plant leghemoglobin came from vertebrates.
As more sequences were analyzed and orthologous genes identified, it was found
that the respective gene trees became reasonably congruent with their species
trees (6, 82). Another controversial example has been the possible eukaryotic
origin of glutamine synthetase II found in plant symbiotic bacteria (78). The most
recent analysis of this example (82) favors a very ancient horizontal transfer,
though this is likely not the last word.
Also, the recent suggestion that cytochrome p450 in Fusariuni oxysporum
transferred from a Streptomyces ancestor (40) suffers from the problem that
orthologous families of cytochrome P450s are extremely difficult to identify. The
extent of this problem is illustrated by the complete cytochrome P450 gene tree
constructed by Degtyarenko & Archakov (14) who, though they
 242 SYVANEN

conclude horizontal transfers and ”convergent” replacements likely occurred, were

unable to relate this tree to any underlying phylogeny. Though the multiple
horizontal transfer of cytochrome P450 is not an unreasonable hypothesis, we
must conclude for the present that its phylogenetic “tracks” are indecipherable.

Pitfalls Involving Unequal Rates

In small data sets, unequal rates of substitution may not be apparent. As noted
by Felsenstein (20a), comparing genes experiencing grossly unequal substitution
rates can result in aberrant affinities during phylogenetic reconstruction. This
problem was encountered in striated muscle calmodulin in the chicken—a second
calmodulin in the chicken that, at the time of its discovery, had no counterpart.
Gruskin et al (28) showed that this gene, designated calmodulinlike (cl), was
highly diverged from the other chicken calmodulin gene (cam), as well as from all
known vertebrate and mammalian genes. From this fact, and from third base
similarities between cl and the electric eel calmodulin, they suggested that cl
entered the chicken by a horizontal transfer, possibly by a viral mediated
retrotransposition (because it is an intronless gene).
The case for a migratory chicken calmodulin was inferred from the minimal
replacement tree based on the half dozen available sequences. Today, there are
over 60 different sequences (66). With this database, we can reanalyze the
likelihood that the chicken cl gene resulted from a horizontal transfer. Figure 1
gives the phylogeny of a set of 18 calmodulins from various eukaryotic sources.
The length of the branches in the tree in Figure 1 gives a rough measure of the
number of replacements since common ancestors. This tree

Figure 1 Phylogenetic tree based on the protein sequence of calmodulin from a variety of
eukaryotes. This tree is determined from a hybrid of procedures—the branching order was
determined from the minimal replacement method, and the length of the branches
corresponds roughly to an average distance between taxa on opposite sides of each node.
 HORIZONTAL GENE FLOW 243

reveals a number of points. One is that the rate of change of the chicken cl gene is
much higher than that of the other vertebrate calmodulins. Indeed, the branch
labeled human (cam) includes calmodulins from rat, mouse, bovine, xenopus, and
the salmon, indicating that the gene is completely functionally constrained among
all these vertebrates. In addition, a human analog to cl is now known that clusters
with the chicken cl. Hence, there are apparently two different calmodulins: one is
the highly functionally constrained member found in most eukaryotes; the other is
a less constrained member found in human and chicken. The chicken cl gene may
not be orthologous to the other members to which it was originally compared, and
almost certainly is evolving at a much higher rate than the others. Thus, the
chicken cl gene fails on possibly two of the three conditions that we set out and,
therefore, can no longer be considered to have experienced a horizontal transfer.

Convergent Replacements

Any discussion of horizontal gene transfer must deal with the possibility of
convergent evolution. Some scientists still do not accept horizontal gene transfer
as the explanation for an unexpected association that is revealed by the
phylogenetic congruency test; rather they attribute this association to convergent
replacements. For example, Kemmerer et al (37), in their paper showing the
affinity of cytochrome c in Arabidopsis with those of fungi, do not offer a
mechanistic explanation. In a recent review of this problem (15), Doolittle
concluded that although functional and mechanistic convergence is common and
enzyme structural convergence has probably occurred, no convincing case for
genuine sequence convergence has yet been made. In the most widely cited
example of sequence convergence, that of the langur lysozyme converging toward
true ruminants (89), Doolittle showed that the lysozyme tree is congruent with the
species tree. That is, the convergent amino acid replacements that occurred in the
lineage leading to ruminants and the langur were too few in the background of
neutral changes to hide the affinity of the langur’s lysozyme with that of the other
primates. In a similar example, we encountered convergent changes in a bacterial
neomycin phosphotransferase that may be the result of natural selection, but the
number of these changes was too low to distort gene phylogeny (48).
The lack of evidence that convergent evolution at the sequence level is
responsible for discrepancies between species trees and gene trees has produced a
subtle shift in evaluating evidence for horizontal transfers. Fifteen years ago,
because the notion of horizontal gene transfer was so completely unexpected and
apparently contrary to the theory of the biological species, evidence in its favor
would possibly have been attributed to convergent evolution. Today, with
numerous mechanisms now known for moving genes through species barriers, and
with the better documented cases, major in-
244 SYVANEN

congruencies between species and gene trees are increasingly being explained by
horizontal transfer.

Tests of Significance -

To say we have found an “unexpected” assemblage from the phylogenetic

congruence test implies that statistical expectations of outcome should be
computed, thereby allowing us to calculate a significance for our “unexpected”
result. Unfortunately, rigorous statistical tests are not generally available for
testing the significance of phylogenetic trees. The computational problems are
immense (32). Indeed, for data sets involving large numbers of taxa, it can be
exceedingly difficult to even find the shortest tree, let alone test the significance of
that tree against another. This problem has attracted much interest recently and
approaches for calculating the significance of four and five competing taxa trees
have been suggested. These include Felsenstein’s (21) and Kishino & Hasegawa’s
(39) maximum likelihood methods, and Lake’s (46) method of maximum
parsimony. The difficulty with the maximum likelihood procedures is that before a
significance test can be computed, a model of the evolutionary process must be
assumed. This model usually assumes that both replacements along lineages and
branching events follow a Markov process.. However, it is one thing to ask if a
given tree is consistent with a specific model, and quite another to ask if it can
discriminate between two competing models. These new procedures are being
used to test the significance of phylogenetic incongruencies, but as yet these
approaches have not been applied to testing the significance of horizontal
transfers.
Constructing phylogenetic trees from sequence data that contain “homoplasy”
can make the congruency test even more difficult. Homoplasy arises when
distantly related species share unique traits. The difficulty lies in distinguishing
whether these shared traits reflect inheritance from a common ancestor, or whether
they were derived independently. Traditionally, homoplasy is considered to result
from processes such as convergent and parallel evolution and reversion to
ancestral states; obviously, characters within a data set involved in horizontal
transfers would also contribute to homoplasy. The implications of this are
described in greater detail in the discussion of the angiosperms.

MOBILE GENETIC ELEMENTS

If the phylogenetic congruency test is applied in the absence of other evidence, it
can be too restrictive. It may preclude from consideration genes that are frequently
involved in horizontal transfers. This is likely the case for many transposable
elements—horizontal transfer is so frequent for some elements that the phylogeny
of the species carrying these elements is lost. Kidwell (38)
 HORIZONTAL GENE FLOW 245

recently reviewed this subject in detail, so I briefly touch on just a few examples.
In the examples of horizontal transfers of eukaryotic genes, the first and still one
of the most convincing cases is the P-factor in Drosophila melanogaster. This case
is convincing because the transfer event occurred in recent years and was therefore
observed as it happened. Observing the event in natural populations in real time
makes for a very convincing case, much as observing the spread of antibiotic
genes on plasmids among pathogenic bacteria has firmly established the role of
plasmids in the horizontal spread of traits in bacteria. The P-factor is convincing
for a second reason. Houck et al (34) showed that the mite Proctolaepus is a
possible donor of the P-factor to the Drosophilids. In addition, the phylogenetic
congruency tests convincingly support horizontal transfer (13). Since the
discovery of P-factors, numerous examples involving other transposable elements
in Drosophila, as well as elements in plants, yeasts, and other metazoans, have
been recorded. In some of these examples, many of the possible transfer events
occurred in the distant past. Subsequent sequence divergence has so obscured the
affinities that reconstruction of their evolutionary histories is either impossible or
highly conjectural. The phylogenetic history of many mobile elements looks much
like the phylogeny of viruses (as some are mixed in with viruses) in that their
history is relatively independent of the phylogeny of their respective hosts
(104).
The two classes of sequences of highly repetitive DNA found in vertebrates, the
SINEs and LINEs, have attracted considerable attention because they are abundant
in the human genome. Dispersion of both is likely due to retrotransposition events,
with the LINEs appearing like common, genomic transposable elements and the
SINEs having some retroviral affinity. On the basis of the phylogenetic
congruency test, both elements have probably transferred horizontally, at least into
some of the taxa that now carry them (65, 68).

CHROMOSOMAL GENES IN BACTERIA

The subject of bacterial plasmids and their associated accessory DNAs has been
extensively reviewed recently (1, 49, 60) and thus is not covered here. It has been
known since Griffith discovered transformation in 1928 that the mechanisms for
the horizontal transfer of information are in place. That this mechanism acts on
chromosomal genes has been known since the 1950s. The only remaining question
has been: Does this happen in nature? This has been a difficult question to answer.

Mosaic Nature of the E. coli Chromosome

Until very recently it was even questioned whether genetic information transferred
between different strains of the same E. coli species. It is useful to
246 SYVANEN

review this history because the logical structure of its discussion has mirrored the
larger discussion of cross-species gene flow. In particular, two issues have
emerged. First, the argument has been made that horizontal gene transfer cannot
be important in E. coli, because if it had occurred, strain differences would have
been obliterated. Second, the process of answering this question is leading to new
criteria for recognizing horizontal transfers between closely related groups.
Selander’s group focused on natural populations of E. coli and concluded that
its population structure was “clonal”. From this, he concluded that gene flow and
recombination between natural E. coli strains must not be important (67, 101).
This widely cited conclusion was based on the finding that native populations of
E. coli could be divided, based on a genetic distance derived from enzyme
polymorphisms, into at least three different groups in which members of a group
were more closely related to each other than to members of the other groups, i.e.
E. coil populations can be divided into demes.
The notion that horizontal transfer among E. coli was absent because of its
clonal population structure was dropped after a number of comparisons were made
between extensive DNA sequences derived from different strains. This analysis
was initiated by Sawyer (76), but Milkman & Bridges (63, 64) have presented the
most rigorous analysis. They determined the complete nucleotide sequence of a
4400-bp region of the trp operon from 36 strains of E. coli selected from the same
set of strains that Selander’s group used to determine the “clonal” population
structure. They confirmed the earlier finding that the 36 strains could be divided
into essentially the same groups identified by enzyme polymorphisms. In addition,
however, they found when comparing strains within a group that one of the
members may have a short section that differs from that of the other members.
That is, in the idiosyncratic individual regions of similarity are occasionally
interspersed with regions of dissimilarity. Moreover, the region of dissimilarity
could often be found in one of the other groups, as if this region of DNA had
transferred from one to the other. From the average size of the dissimilar regions,
it has been estimated that the average recombination event usually results in the
transfer of only a few hundred to 1000 base pairs. Pairs of chromosomes with this
pattern of similarity interspersed with regions of dissimilarity are said to have a
mosaic pattern. A conclusion of a mosaic pattern is based on models whose
significance can be rigorously tested (58, 64, 76). In summary, this work clearly
demonstrates the presence of subpopulations within E. coli that are genetically
distinctive, but that occasionally experience a gene transfer from one group into
another, i.e. small breaches of a “clone’s” reproductive isolation do not destroy its
identity.
In general, a comparison of any two homologous DNA regions that display an
abrupt change in similarity raises the possibility of a mosaic structure, and hence a
horizontal transfer. The case for E. coli is solid because the numbers
 HORIZONTAL GENE FLOW 247

are sufficiently high that the statistical significance of the discontinuities is high.
Furthermore, the donor strains were identified. I have used the criterion of
discontinuity in the degree of similarity to identify possible horizontal transfers
involving small regions of mammalian beta globin genes (90). In this case, the
discontinuity was statistically significant, but it remains to be seen whether it
represents horizontal transfer.

Salmonella and the Criterion of GC Content

Considerable progress has been made in characterizing different pathogenic strains

of Salmonella, and this process has led to numerous claims of horizontal transfers.
There are many good examples of horizontal movement of virulence factors.
Beltran et al (5) have shown that the common surface antigens used to classify
Salmonella serotypes are discontinuously distributed across distantly related
strains, as determined by enzyme polymorphisms, which indicates that these
serotype genes move horizontally. The virulence capsular antigen encoded by the
viaB gene found in isolates of Salmonella dublin is most likely a horizontal
transfer. Although the source organism is not known, it is likely another
Salmonella serovar (77, 83). Hoyer et at (35) found that the sialidase gene, nanH,
moves between different strains of Salmonella and may even have originated
within an assemblage that includes the Clostridia.
Groisman et al (27) have viewed the chromosome in S. typhimurium as a
widespread mosaic of remotely related parts. This conclusion for Salmonella is
based, in part, on the comparison of genes among enterics. Even though E. coli and
S. typhimurium share genomes of comparable size with the same gene order and
with 90% of their genes being closely related, about 10% of the genome in S.
typhimurium encodes functions not found in E. coil. In addition, the GC contents
in these unique genes are frequently significantly lower than the 50% average for
the entire genome, a finding that has led to the suggestion of remote origins for
these regions. Examples of low GC content include phoN (26), the rfb genes for 0-
antigen synthesis (75, 98), and an unknown transcriptional regulator (27). These
considerations support the notion of a mosaic chromosome for Salmonella. On the
other hand, the case for remote species donors for these genes is problematical in
that the only evidence for remote origins is deviation of the GC content.
Because various different factors probably influence GC content, I examined
this point in greater detail with respect to the hypothesis of remote origins. I
collected a distribution of the GC content found in 757 different DNA fragments
from E. coli (Figure 2A) and in 131 fragments from S. typhimurium (Figure 2B)
For E. coli, the distribution can be divided into two groups. Most fragments cluster
about a GC content of 0.509. In fact, this main distribution is close to Gaussian,
with a standard deviation determined by a stochastic process, as shown. The
second class of DNA fragments skew significantly to
248 SYVANEN

0.2600 0.3000 0.3400 0.3800 0.4200 0.4600 0.5000 0.5400 0.5800

%GC-CONTENT

Figure 2 A. Distribution of GC content in DNA fragments from the chromosome in E. coli. The
sample was selected from the gene bank by listing all genes with Ba:Eco in their entry name and then
deleting all entries of plasmids, duplicated entries, and sequences of less than 100 bp. The theoretical
curve is the normal, which is computed for a mean of 0.499, and σ = 0.032. B. The distribution of GC
content in DNA fragments from the chromosome in S. typhimurium. The sequences are from gene
bank listings with Ba:Sty in their entry name and edited as were the E. coli sequences. The theoretical
curve is the normal with a mean of 0.516 and a σ = 0.032.

lower GC content. By the GC-content criterion, these fragments become can-

didates for horizontal transfers. From the data in Figure 2a, this argument has one
problem: the skewing is only to low GC content. The variation to high GC content
is consistent with the variance predicted for random fluctuation in GC about a
mean of 0.509. Why do we not see genes of high GC-content that came from
remote sources?
The case for S. typhimurium is not too different. The main clustering near the
median of 0.516 is only approximately normal (possibly due to the smaller sample
size) and in addition, there may be some skewing to high GC content but, as with
E. coli, most of the skewing is to the lower GC content. Because the skewing is
primarily to low GC content, support for remote species origins for these genes
would seem unlikely. Functional selection for low GC content would appear more
likely. Obviously, such DNA would have a lowered melting temperature, and it
would be easy to imagine replication or recombination scenarios that would select
for these genomic regions. For example, a simpler explanation for these low GC
regions is that they are accessory DNAs, regions that transfer horizontally among
different strains of Salmonella and E. coli. This model for the Salmonella
chromosome is analogous to the cassette model for the lambdoid bacteriophages
(11), a model that involves recurrent transfer and recombination events of
different cassettes due to resulting episodic selection for these accessory DNAs.
The lower GC content could thereby be
 HORIZONTAL GENE FLOW 249

0.2900 0.3300 0.3700 0.4100 0.4500 0.4900 0.5300 0.5700

%GC-CONTENT

selected at the step of recombination either by easing the melting of DNA, or

possibly by reducing the barrier to recombination between heterologous DNAs by
reducing the probability of mismatched bases in heteroduplex overlap regions
(74). (The reduced probability of a mismatch at the extremes of GC content can be
deduced from a simple statistical argument. Let us compare two randomly
generated sequences that are of the same GC content (= g) and compute the
probability (P) that at any given site they misalign. Then it can be shown that P =
{1 over 2} + g - g2, which reaches a maximum value of 0.75 at g = 0.5 and
declines to 0.5 as g approaches 0 or 1.)
To conclude, before a remote origin for these regions from the Salmonella can
be accepted, a homologous gene from a member of the donor clade must be
identified.

Codon Bias
Another criterion for identifying horizontal transfers is to compare codon usage of
a gene with the codon bias of the host organism. This is a difficult criterion to
apply because codon usage for minor proteins deviates from the bias. Medigue et
al (61) examined 740 genes from E. coli with respect to codon usage. They found
three classes: (a) the heavily expressed proteins that define bias in E. coli; (b) the
proteins, expressed in trace amounts, that use the rare codons; and (c) a residual
group. This third group contained the genes that would have been predicted to be
migratory, such as the insertion sequences and other identified mobile elements,
and the genes determining surface antigens (as with Salmonella, perhaps E. coli
changes its coat from time to time because of episodic immune selection). But
curiously, group (c) also contained genes specifying faithful DNA replication, the
antimutator genes. This latter fact correlates with the observation that a number of
antimutator genes are
250 SYVANEN

plasmid-borne. Perhaps the evolution of bacterial lineages find episodic need for
antimutators or, possibly, periodic need for high mutator activities.
Though analysis of codon usage is an interesting exercise in sequence analysis,
possible explanations for any deviation from the species bias are sufficiently
numerous to preclude its use as a criterion for inferring a horizontal transfer. At
best, codon bias can be used to support more compelling evidence.

Other Bacteria

As recently as 1988, it was argued that horizontal transfer contributed little or

nothing to the genomes of bacteria. Today, few bacteriologists are eager to push
this point—too many examples have been reported during the past five years.
Some cases mentioned here are supported by small data sets, and may eventually
not hold up as additional sequences are determined. Also, some of these horizontal
transfers may involve shuffling cassettes between a closely related group of
strains, much as with the examples described above in Salmonella. These
examples include the celY gene in Erwinia chrysanthemi (29), fimbrae genes of
Bacteroides nodosus (33), the CO dehydrogenase from carboxydotrophic bacteria
(44), capsule genes from Haemophilus influenzae (45), the M12 serotype of
streptococci (81, 100), and an amidase from Brevibacterium (84). Genes for the
photosynthetic reaction center from Rhodocyclus gelatinosus (an α purple
bacteria) define a clade with its homologous genes from ß purple bacteria; this
case is interesting historically because these photosynthetic bacteria were once
classified, before the days of rRNA trees, into one group on the basis of similar
photosynthetic biochemistry. The nitrogenase Fe protein from Archaebacteria (85)
and a superoxide dismutase in Halobacterium (57) involve archaebacterium and
may represent horizontal transfers across great phylogenetic distances.
Mosaic patterns indicative of horizontal transfer have also been observed in
antibiotic resistance genes in the genomes of Neisseria and Streptococcus.
Naturally occurring resistance in some of their species is encoded by altered
penicillin-binding proteins (PBPs). The pathogenic strains S. pneumoniae (17), N.
meningitidis (9, 50), and N. gonorrhoeae (87) were originally sensitive, but have
recently become resistant. Comparison of the PBP genes from the resistant and
sensitive strains reveals a mosaic pattern—regions of similarity interspersed with
regions of dissimilarity. The genes from resistant pathogenic strains consist of a
hybrid sequence resembling parts from the sensitive strain interspersed with
blocks from the naturally occurring resistant strain (59). Spratt et al (87) have
shown that, in Neisseria, blocks of a few hundred base pairs were transferred.
Radstrom et al (73) have documented a similar mosaic pattern in the
dihydropteroate synthase gene found in sulfonamide-resistant N. meningitidis.
 HORIZONTAL GENE FLOW 251

ORGANELLAR GENOMES

The concept of horizontal transfer involving eukaryotic plastids has a long history.
In 1905, Mereschkowsky (62) argued that chloroplasts were bacterial
endosymbiotes, and in 1928, Walling (97) made a similar case for the mito-
chondria. Some 50 years later Margulis (51) reintroduced and developed this idea.
Today it is the most widely accepted form of horizontal movement of genetic
information across high phylogenetic barriers. The demonstration by Woese &
Fox (102) that rRNA from the mitochondria and chloroplasts are more closely
related to bacteria than to their eukaryotic nuclear counterparts has provided the
most convincing evidence for the endosymbiotic theory. Indeed, their paper
represents the first application of the phylogenetic incongruency test to establish a
horizontal transfer.
But beyond the original endosymbiotic event creating the mitochondria and
plastids, these genomes have likely been involved in additional horizontal
transfers. For example, the plastid phylogeny, which is based on the rubisco rbcL
genes, places the nonchlorophyll-B algae with the purple bacteria, whereas their
16S RNA sequences place them with the cyanobacteria and green plant
chloroplasts (2, 3, 80, 95). Such conflicting phylogenies have led to the hypothesis
that the plastids arose by more than one endosymbiotic event. However, to accept
this theory, one must posit many parallel mutations (80); a much simpler
explanation is to accept the monophyly of the plastids, but to add subsequent
horizontal transfers (3). The finding of multiple genes in paramecium
mitochondria that have previously only been seen in plant chloroplast raises the
possibility of a major migration across species boundaries (72). This example
must await a more complete application of the phylogenetic congruence test
before it can be accepted as a horizontal transfer, or at least a horizontal transfer
that occurred after the early symbiosis events. At least this finding implies a
transfer from chloroplast to mitochondrion (20).
There are numerous accounts of organellar DNA that have moved into the
nucleus. These examples are relevant to the current review because they define
part of the pathway for endosymbiotic-mediated horizontal transfer. These transfer
events are both ancient, as evidenced by thioredoxin-F and GAPDH in Table 1, as
well as more recent (reviewed in 10, 19, 25, 69).

THE ANGIOSPERM PARADOX

Classification Problem

Darwin’s theory of evolution with modification from common ancestors rapidly

accepted by zoologists in 1859, in large part because metazoan assemblages were
so easy to organize into trees, with ancestral branch points repre-
252 SYVANEN

sented by animals that, though primitive, could be still classified with their
descendants. Botanists, however, have long had difficulty in classifying the
flowering plants (the angiosperms) into taxonomic groups above the family level.
The traditional problem has been that morphological characters upon which trees
could be constructed give conflicting trees, depending upon how the characters are
weighted. Thus, botanists have never been able to reach agreement upon the shape
of a species tree (88). Even some of the simplest questions remain unresolved. For
example, it is not known whether the angiosperms evolved from a common
ancestor that could be classified as an angiosperm (i.e. monophyletic), or whether
they arose from more than one nonangiosperm lineage (polyphyletic). Initially,
those who thought about the evolution and classification of the angiosperms had
trouble accepting Darwinian evolution, at least as it is summarized by the
phylogenetic tree. This problem of angiosperm classification was apparent when
the analysis was based on morphology. With the introduction of sequence data, the
problem has only worsened.
The problem in plant taxonomy lies in the widespread existence of parallel
traits; thus, any tree that can be constructed depends on a judgment as to which
shared traits are ancestrally derived and which evolved “independently” in
different lineages. Indeed, both Vavilov, in 1922, and Went, in 1971, focused on
parallel variation as major factors in angiosperm evolution. An early goal of
molecular studies of angiosperm evolution was to use sequence information
contained in a number of genes in order to determine acceptable species trees.
This attempt has so far failed for the simple reason that different genes have
yielded different assemblages. The first extensive data were based on the protein
sequence determined for cytochrome c from 25 angiosperms by Boulter (7). Since
then, plastocyanin (52), the large rubisco subunit, rbcL (12, 53), stable RNA genes
(7, 18, 53, 94), and glyceraldehyde-3-phosphate dehydrogenase (56) have been
used to determine trees. The assemblages determined from one gene (or group of
genes and characters) conflict with those determined by another. In addition, it has
been very difficult to find a satisfying congruence between any of the gene trees
and those relationships agreed upon by systematic botanists. I would suggest that
horizontal gene transfer has been sufficiently common during the descent of the
angiosperms that different genes in the same species have likely followed different
phylogenetic histories. If so, then the phylogenetic congruency test (at least, as
outlined above) cannot be used, since finding an acceptable species tree may well
be impossible.
The problem in angiosperm classification goes even deeper than conflicting
gene trees. Internally robust trees based on single genes are also elusive: the
sequence for a single gene is difficult to organize into a tree-like pattern. This
problem of parallel variation within a single gene has been apparent since the
earliest introduction of molecular data for plants (70). Hartman, Stevens, and
 HORIZONTAL GENE FLOW 253

I (93) showed that the cytochrome c tree based on angiosperm sequences was
internally homoplastic as compared to that from vertebrates. In this study, we
showed that in many respects the plant cytochrome c appears to have evolved in a
well-behaved fashion, not differently from, for example, the vertebrate
cytochrome c. The plant sequences (with the exception of Arabidopsis) diagnose
their own kingdom, and their rate of change, that is their molecular clock, is
relatively uniform. In fact, the amount of divergence among the sequences from
angiosperm cytochrome c is comparable to that in vertebrates. While the gene tree
for the vertebrate cytochrome c is reasonably congruent with accepted phylogeny,
the plants, on the other hand, yielded four groups of minimal trees that were both
highly incongruent with each other and with any acceptable phylogeny. That is,
more homoplasy appears in plant cytochrome-c evolution than in vertebrate
cytochrome-c evolution. If this internal cytochrome-c homoplasy is due to
horizontal transfer events, the segments of transfer must be considerably smaller
than a single gene. This means that not only have DNA fragments moved
horizontally, but they must have experienced a gene conversion event, thereby
leaving only a block of new information inside a preexisting gene.

Temporal Problem

Martin et al (56) drew attention to a major discrepancy that has emerged between
molecular evolution and paleobotany. The molecular distance between
glyceraldehyde-3-phosphate dehydrogenase (gapdA) sequences suggests that
various angiosperm families diverged as early as the Permian period, possibly
even the Carboniferous, close to 300 million years ago (MYA), while the fossil
record supports an angiosperm radiation in the late Cretaceous period—only about
110 MYA. An early divergence of multiple angiosperm families is also supported
by the amount of divergence seen for other sequences. Wolfe et al (103) have
shown that this holds for the chloroplast rRNA and rubisco sequences. The same
result can be deduced from an analysis of the cytochrome-c sequences (see the
molecular distance tree given in Figure 3). Nine different angiosperm lineages
appear to have diverged at least in the Triassic period, if not the Permian period.
The time axis is calibrated using the usual assumptions of a relatively constant
molecular clock where the tree is rooted to the ginkgo and where the time of
separation of the lineage giving rise to the ginkgo from that giving rise to the
angiosperms is in the upper carboniferous (about 260 MYA, though a time of 300
MYA—320 MYA has been used by some authors). The problem is that the
earliest fossils resembling members of those nine lineages are no older than 110
MYA. As described above, it is not easy to attribute this discrepancy to aberrant
sequence evolution of cytochrome c. Thus, the cytochrome c sequences, in
addition to those from
254 SYVANEN

Figure 3 An evolutionary distance tree based on cytochrome-c sequences. Distance is the number of
sites that differ between two sequences under comparison divided by the number of sites compared and
MYA is million years ago. The position of each node is the numerical average of distance between the
respective families on each side of the node. The isolated dashes designate the specific values upon
which this average is based. Wheat and leek represent two monocot families and pumpkin through
love-in-the-mist are from the dicot families cucurbitaceae, brassicaceae, asteracese, capialeae,
chenopodinum, polygonacoae, and ranunculi, respectively.

the chloroplast genes and gapdhA, also support the notion of an angiosperm
radiation more than 200 MYA.
The discussion of the apparent discrepancy between the fossil record and the
molecular clock determinations has focused primarily on one of two possibilities.
First, it has been suggested that the paleobotanical information is misleading:
perhaps a long precretaceous angiosperm history was never captured in the fossil
record. This explanation is impossible to formally disprove, but it is considered by
those who study plant fossils to be highly unlikely (42). The second possibility is
that protein sequence evolution in angiosperms might obey rules different from
those of metazoans, and thus be useless for estimating divergence times. That is,
the molecular clock does not keep the right time. However, as Martin et al (54)
show for gapdhA, as Wolfe et al (103) show for the chloroplast sequences, and as
we have shown for cytochrome c (93), the rates of angiosperm sequence
divergence are comparable to that of the metazoans. In addition, in each case, the
relative rate test is satisfied, indicating that the various angiosperm families and
outgroup taxa display comparable rates of sequence divergence. These
measurements of rate do have error, so that small but significant differences in
rates in different lineages under comparison may be missed, but the test is accurate
enough to preclude the size of
 HORIZONTAL GENE FLOW 255

the discrepancy in divergence times that has emerged between the paleobotanical
record and the protein sequence record.

A third explanation--angiosperm evolution driven by horizontal transfer--allows a

straightforward interpretation of both the paleobotanical and molecular evidence.
This hypothesis easily incorporates the notion that angiosperms are probably
polyphyletic. Krassilov has advocated polyphyly (42, 43) to resolve conflicts
within the fossil record itself. Polyphyletic origins suggest that modern
angiosperms are derived from multiple proangiosperm lineages that first appeared
in the early Permian period (ca 260 MYA), and that these multiple lineages
underwent concerted conversion into angiosperms in the early Cenozoic era or, as
described by Krassilov (42), underwent "angio-spermization." It would seem that
a theory of angiosperm evolution that incorporates the lateral spread of genes
should be taken seriously. To its advantage, it accommodates the otherwise
irreconcilable divergence times based on paleobotanical and plant molecular
evolution; it resolves conflicts within the fossil record; it answers the appearance
of like-traits in distantly related plants that occupy similar ranges [as suggested by
Went (99)], and finally it provides a rational explanation for the failure to organize
extant angiosperm families into consistent phylogenetic trees.

CONCLUSION

As mentioned in the introduction, geneticists who think at the molecular and

cellular level could easily accommodate a theory of evolution incorporating
horizontal gene transfer. Other areas of biology would be more seriously affected,
though probably not as seriously as feared. Speciation theories that rely on
reproductive isolation are perfectly consistent with the notion of
macroevolutionary trends that are influenced by horizontal transfers, as illustrated
by the examples of E. coli and Salmonella.

As we have seen with populations of E. coli and S. typhimurium, horizontal

transfers are apparent among subpopulations and with even more distantly related
organisms. This has not prevented genetically identifiable subpopulations from
being formed, nor has it prevented Escherichia and Salmonella from being
taxonomically associated with the enterics, and the enterics from being identified
as members of the purple bacteria. That is, bacteria can be organized into tree-like
patterns of descent. As I have previously written (92), it is ironic that after the
discovery of transferable plasmids, early workers were criticized for attempting to
reconstruct bacterial phylogeny because, it was argued, the horizontal spread of
DNA caused by plasmid movement would so scatter traits that tree-like descent
patterns would be lost. Some microbiologists studying plasmids had early on
assumed cross-species gene transfer, using it to argue against the possibility of
phylogenetic classification of bacteria. How-
256 SYVANEN

ever, horizontal gene flow does not necessarily preclude phylogenetic classi-
fication but, as we have seen with the angiosperms, it can make it more difficult,
just as the existence of reproductive isolation (i.e. the fact that taxonomic barriers
do exist) and the fact that these taxonomic groups can be organized into tree-like
patterns of descent, does not establish, a priori, the absence of horizontal gene
transfer. Evolution favors those groups with an appropriate balance between
genetic variation and genetic stability. If genes or gene segments introduced from
other species are viewed as just another source of genetic variability, then cross-
species gene transfer could logically exist concomitantly with reproductive
isolation and other biological barriers that limit its employment and sustenance.
Bacteria also provide a model for another aspect of horizontal transfer. In general,
the discussion has focused on the transfer of whole genes (see Table 1), whereas
the transfer of much shorter elements seems to be common in E. coli,
Streptococci, and Neisseria. I believe that the transfer of parts of genes is also a
factor in eukaryotes, as I suggested above for the plant cytochromes c and for
mammalian beta-globins (91).
Some neoDarwinian concepts, however, would likely have to be modified if
horizontal gene transfer plays an important role in evolution. In particular, the idea
that variations that arise within populations are independent of macroevolutionary
trends (sometimes referred to as “Wright’s” rule) is probably unrealistic,
especially with regard to the angiosperms. This notion will likely be replaced with
the idea that useful variation, at least in rapidly evolving populations, is influenced
by traits carried in competing lineages. This idea was explicitly proposed by
Vavilov in 1922 (96) as his “law of homologous variation”, before botanists
accepted the neoDarwinian paradigm. Vavilov’s idea was premature because no
mechanism in its support was known. But with the discovery of mechanisms for
moving genes horizontally, Went in 1972, and more recently others (42, 43, 47,
71) have suggested that this is possibly the mechanism causing the widespread
parallelisms or “homologous variations”.
In paleontology, the influence of a theory of macroevolutionary trends
influenced by horizontal gene transfers will likely be profound (91). Much of the
fossil record is characterized by extensive parallelisms. At the turn of the last
century, many paleontologists focused on these observations and constructed
theories to accommodate them. These led to “orthogenetic” theories (see ref. 24,
for an historical account). With the dominance of neoDarwinian theories, the ideas
of orthogenesis have been widely criticized as teleological or even as vitalistic.
These criticisms are, for the most part, unwarranted since many orthogenetic
theories were offered as legitimate scientific hypotheses. However, because
mechanisms to accommodate parallel variations were not known, the theories
lapsed. Now, with known mechanisms, these old theories should be reanalyzed in
a new light—a process that has already begun by
 HORIZONTAL GENE FLOW 257

paleontologists (19,36). One result of this reanalysis should be the organization of

fossils by space and time, much as they were in the past, with less emphasis on
identifying uniquely derived characters and on organizing organisms according to
cladistic methodology.
Systematic zoological classification will doubtless be influenced by the
introduction of horizontal transfers, though the form this may take is difficult to
anticipate. There are a number of outstanding problems that current classification
schemes have yet to solve: These include, for example, the relationships among
the metazoan phyla, the vertebrate classes, and the mammalian orders. From the
fossil record, each of these examples arose as a radiation during a brief geological
period without clear ancestral-descendant lines. Analysis of both morphological
and molecular characters has yet to resolve these branching patterns. Should
horizontal gene transfer be recognized as a significant factor by systematists, this
will no doubt have a profound influence on future taxonomic theories.

ACKNOWLEDGMENTS

Thanks to my long time collaborator Hy Hartman for many helpful discussions

and suggestions, and to Sue Greenwald for ideas and editing.

Any Annual Review chapter, as well as any article cited in an Annual Review chapter,
may be purchased from the Annual Reviews Preprints and Reprints service.
1-800-347-8007; 415.259-5017; email: arpr@class.org

Literature Cited

1. Arber W. 1991. Elements in microbial teritidis, S. heidelberg, S. infantis, S.

evolution. J. Mol. Evol. 33:4—12 newport and S. typhimurium. Proc.
2. Assali NE, Mache R, Loiseaux-de Goer Natl. Acad. Sci. USA 85:7753—57
S. 1990. Evidence for a composite phy- 6. Bogusz D, Appleby CA, Landsmann J,
logenetic origin of the plastid genome Dennis ES, Trinick MJ, Peacock WJ.
of the brown alga Pylaiella littoralis (L. 1988. Functioning haemoglobin genes
KjelIm). Plant Mol. Biol. 15:307—15 in non-nodulating plants. Nature 331:
3. Assali NE, Martin WF, Sommerville 178—80
CC. Loiseaux-de Goer S. 199]. Evolu- 7. Boulter D. 1974. The evolution of plant
tion of the Rubisco operon from pro- proteins with special reference to higher
karyotes to algae: structure and analysis plant cytochrome c. Curr. Adv. Plant
of the rbcS gene of the brown alga Sci. 4:1—16
Pylaiella littoralis. Plant Mol. Biol. 17: 8. Boulter D, Gilroy JS. 1992. Partial se-
853—63 quences of 18s ribosomal RNA of 2
4. Baldauf SL, Manhart JR. Palmer JD. genera from each of 6 flowering plant
1990. Evolutionary transfer of the chlo- families. Phytochemistry 31:1243—46
roplast tufA gene to the nucleus. Nature 9. Bowler LD, Zhang QY, Riou JY, Spratt
344:262—65 BG. 1994. Interspecies recombination
5. Beltran P. Musser JM, Helmuth R, between the penA genes of Neisseria
Farmer ii 111, Frerichs WM. et al. 1988. meningitidis and commensal Neisseria
Toward a population genetic analysis of species during the emergence of peni
Salmonella: genetic diversity and rela- cillin resistance in N. meningitidis: nat-
tionships among strains of serotypes S. ural events and laboratory simulation.
choleraesuis, S. derby, S. dublin, S. en- J. Bacieriol. 176:333—37
258 SYVANEN

10. Brennicke A, Grohmann L, Hiesel R, 23. Gilbert HJ, Hazlewood GP, Laurie JI,
Knoop V. Schuster W. 1993. The mi- 0rpin CG. Xue GP. 1992. Homologous
tochondrial genome on its way to the catalytic domains in a rumen fungal
nucleus: different stages of gene transfer xylanase: evidence for gene duplication
in higher plants. FEBS Lett. 325:140-55 and prokaryotic origin. Mol. Microbiol.
11. Campbell A, Botstein D. 1983. Evolu- 6:2065—72
tion of the lambdoid phages. In 24. Gould SJ 1977. Ontogeny and Phylog-
LAMBDA 11, ed. RW Hendrix, [!** AU: eny Cambridge, MA: Harvard Univ.
need to give other authors**!] et al, pp. Press/Belknap
365—8O. New York: Cold Spring 25. Gray MW. 1992. The endosymbiont hy-
Harbor Lab. Press pothesis revisited. Int. Rev. Cytol. 141:
12. Chase MW, Soltis DE, Olmstead RG, 233—357
Morgan D, [!** AU: give first five 26. Groisman EA, Saier MH Jr. Ochman H.
authors please**!l, et al. 1993. Phylo 1992. Horizontal transfer of a phospha-
genetics of seed plants—an analysis of tase gene as evidence for mosaic struc-
nucleotide sequences from the plastid ture of the Salmonella genome. EMBO
gene rbcl. Ann. Mo. Bot.. Gard. 80:528- J.11:1309--16
80 27. Groisman EA, Sturmoski MA, Solomon
13. Clark JB, Maddison WD, Kidwell MG. FR, Lin R, Ochman H. 1993. Molecular,
1994. Phylogenetic analysis supports functional, and evolutionary analysis of
horizontal transfer of p transposable el- sequences specific to Salmonella. Proc.
ements. Mol. Biol. Evol. 11:40—50 Natl. Acad. Sci. USA 90:1033—37
14. Degtyarenko KN. Archakov Al. 1993. 28. Gruskin KD, Smith TF, Goodman M.
Molecular evolution of P450 superfam- 1987. Possible origin of a calmodulin
ily and P450-containing monooxygenase gene that lacks intervening sequences.
systems. FEBS Let:. 332:1—8 Proc. Natl. Acad. Sci. USA 84:1605—8
15. Doolittle RF. 1994. Convergent evolu- 29. Guiseppi A, Aymeric JL, Cami B, Barras
tion—the need to be explicit, Trends F, Creuzet N. 1991. Sequence analysis of
Biochem. Sci. 19:15—18 the cellulase-encoding celY gene of
16. Doolittle RF, Feng DF, Anderson KL, Erwinia chrvsanthemi: a possible case of
Alberro MR. 1990. A naturally occur interspecies gene transfer. Gene 106:
ring horizontal gene transfer from a eu- 109—14
karyote to a prokaryote. J. Mol. Evol. 30. Hartman, H. 1977. Speculation on vi-
31:383—88 ruses, cells and evolution. Evol. Theor.
17. Dowson CG, Coffey TJ, Kell C, Whiley 3:159—63
RA. 1993. Evolution of penicillin resis- 31. Hartman H, Syvanen M, Buchanan BB.
tance in Streptococcus pneumoniae; the 1990. Contrasting evolutionary histories
role of Streptococcus mitis in the for- of chloroplast thioredoxins f and m. Mol.
mation of a low affinity PBP2B in S. Biol. Evol. 7:247—54
pneumomae. Mol. Microbiol. [!** AU: 32. Hem J. 1990. Reconstructing evolution
please complete**!] of sequences subject to recombination
18. Doyle JA, Donoghue MJ, Zimmer EA. using parsimony. Math. Biosci. 98:185—
1994. Integration of morphological and 200
ribosomal RNA data on the origin of 33. Hobbs M, Dalrymple BP, Cox PT, Liv-
angiosperms. Ann. Mo. Bot. Gard. 81: ingstone SP, Delaney SF, Mattick JS.
In press [!** AU: POSSIBLE TO UP 1991. Organization of the fimbrial gene
DATE? **!] legion of Bacteroides nodosus: class I
19. Erwin DH. Valentine JW. 1984. “Hope- and class II strains. Mol. Microbiol. 5:
ful monsters”, transposons, and meta- 543—60
zoan radiation. Proc. Natl. Acad. Sci. 34. Houck MA, Clark JB. Peterson KR,
USA 81:5482—83 Kidwell MG. 1991. Possible horizontal
20. Fejes E, Masters BS, McCarty DM. transfer of Drosophila genes by the mite
Hauswirth WW. 1988. Sequence and Proctolaelaps regalis. Science 253:
transcriptional analysis of a chloroplast
insert in the mitochondrial genome of 1125 -- 28
Zea mays. Curr. Genet. 13:509—15 35. Hoyer LL, Hamilton AC, Steenbergen
21. Felsenstein J. 1981. Evolutionary trees SM, Vimr ER. 1992. Cloning, sequenc-
from DNA sequences: a maximum like- ing and distribution of the Salmonella
lihood approach. J. Mol. Evol. 17:368— typhimurium LT2 sialidase gene, nanH,
76 provides evidence for interspecies gene
22. Gantt JS, Baldauf SL, Calie PJ, Weeden transfer. Mol. Microbiol. 6:873—84
NF. Palmer JD. 1991. Transfer of rpl22 36. Jeppsson L. 1986. A possible mecha-
to the nucleus greatly preceded its loss
from the chloroplast and involved the
gain of an intron. EMBO J. 10:3073—
78
 HORIZONTAL GENE FLOW 259
52. Martin PG. Boulter D, Penny D. 1985.
Angiosperm phylogeny studied using
nism in convergent evolution. Paleobio- sequences of five macromolecules.
logy 12:337—44 Taxon 34:393—400
37. Kemmerer EC, Lei M, Wu R. 1991. 53. Martin PG. Dowd JM. 1991. A compar-
Structure and molecular evolutionary ison of 18s ribosomal RNA and rubisco
analysis of a plant cytochrome c gene: large subunit sequences for studying an-
surprising implications for Arabidopsis giosperm phylogeny. J. Mol. Evol. 33:
thaliana. J. Mol. Evol. 32:227—37
274 - 82
38. Kidwell M. 1993. Lateral transfer in 54. Martin W, Brinkmann H, Savonna C,
natural populations of eukaryotes. Annu. Cerff R. 1993. Evidence for a chimeric
Rev. Genet.. 27:235—56 nature of nuclear genomes: eubacterial
39. Kishino H, Hasegawa M. 1989. Evalu- origin of eukaryotic glyceraldehyde-3-
ation of the maximum likelihood esti- phosphate dehydrogenase genes. Proc.
mate of the evolutionary tree topologies Natl. Acad Sci. USA 90:8692—6
from DNA sequence data, and the 55. Martin W, Lagrange T, Li YF, Bisanz-
branching order in hominoidea. J. Mol. Seyer C, Mache R. 1990. Hypothesis for
Evol. 29:170—79 the evolutionary origin of the chloroplast
40. Kizawa H, Tomura D, Oda M, ribosomal protein L21 of spinach. Curr.
Fukamizu A, Hoshino T, et al. 1991. Genet. 18:553—6
Nucleotide sequence of the unique 56. Martin W, Lydiate D, Brinkmann H,
nitrate/nitrite-inducible cytochrome P- Forkmann G, Saedler H, Cerff R. 1993.
450 cDNA from Fusarium oxysporum. Molecular phylogenies in angiosperm
J. Biol. Chem. 266:10632—37 evolution. Mol. Biol. Evol. 10: 140—62
42. Krassilov VA. 1977. The origin of an-
giosperms. Bot. Rev. 43:143—76 57. May BP, Dennis PP. 1989. Evolution
43. Krassilov VA. 1991. The origin of an- and regulation of the gene encoding
giosperms—new and old problems. superoxide dismutase from the arch-
Trends Ecol. Evol. 6:215—20 aebacterium Halobacterium cutirubrum
44. Kraut M, Hugendieck I, Herwig S, J. Biol. Chem. 264:12253—58
Meyer O. 1989. Homology and distri- 58. Maynard-Smith JM. 1992. Analyzing
bution of CO dehydrogenase structural the mosaic structure of genes. J. Mol.
genes in carboxydotrophic bacteria. Evol. 34:126—29
Arch. Microbiol. 152:335—41 59. Maynard-Smith JM, Dowson CG, Spratt
45. Kroll JS, Moxon ER. 1990. Capsulation BG. 1991. Localized sex in bacteria.
in distantly related strains of Haemo- Nature 349:29—31
philus influenzae type b: genetic drift 60. Mazodier P. Davies J. 1991. Gene trans-
and gene transfer at the capsulation fer between distantly related bacteria.
locus. J. Bacteriol. 172:1374—79
Annu. Rev. Genet. 25:147—71
46. Lake JA. 1987. A rate-independent
61. Medigue C, Rouxel T, Vigier P. Henaut
technique for analysis of nucleic acid A, Danchin A. 1990. Evidence for hor-
sequences: evolutionary parsimony. izontal gene transfer in Escherichia coli
Mol. Biol. Evol. 4:1987 speciation. J. Mol. Biol. 222:851—6
47. Lamboy WT. 1984. Evolution of 62. Mereschkowsky C. 1905. Uber Natur
flowering plants by fungus-to-host und Ursprung der Chromatophoren in
horizontal gene transfer. Evol. Theor. Pflanzenteilen. Biol. Zentrabl. 25:593—
7:45—51 [!** AU: please complete**!]
48. Lee KY, Hopkins JD, Syvanen M. 63. Milkman R, Bridges MM. 1990. Molec-
1991. Evolved neomycin ular evolution of the Escherichia coli
phosphotransferase from an isolate of chromosome. HI. Clonal frames. Genet-
Klebsiella pneumoniae. Mol. Microbiol. ics 126:505—17
5:2039—46 64. Milkman R, Bridges MM. 1993. Molec-
49. Levy SB, Miller RV, eds. 1989. Gene ular evolution of the Escherichia coli
Transfer in the Environment. New chromosome. IV. Sequence compari-
York: McGraw-Hill sons. 133:455—68
50. Lujan R. Zhang QY, Saez Nieto JA, 65. Mouches C, Bensaadi N, Salvado JC.
Jones DM, Spratt BG. 1991. Penicillin- 1992. Characterization of a LINE retro-
resistant isolates of Neisseria lactamica poson dispersed in the genome of three
produce altered forms of penicillin- non-sibling Aedes mosquito species.
binding protein 2 that arose by Gene 120:183—90
interspecies horizontal gene transfer. 66. Nakayama S. Kretsinger RH. 1993. Evo-
Antimicrob. Agents Chemother. lution of EF-hand calcium-modulated
35:300—4 proteins. III. Exon sequences confirm
51. Margulis L. 1971. Symbiosis and evo- most dendrograms based on protein se-
lution. Sci. Am. 225:48—57
quences: calmodulin dendrograms show
260 SYVANEN
significant lack of parallelism. J. Mol. prokaryotes. J. Mol. Evol. 29:422—28
Evol. 36:458—76 79. Shay JW. Werbin H. 1992. New evi-
67. Ochman H, Selander RK. 1984. Evi- dence for the insertion of mitochondrial
dence for clonal population structure in DNA into the human genome: signifi-
Escherichia coli. Proc. Natl. Acad. Sci. cance for cancer and aging. Mutat. Res.
USA 81:198—201 275:227—35
68. Ohshima K, Koishi R, Matsuo M, 80. Shivji MS. Li N, Cattolico RA. 1992.
Okada N. 1993. Several short Structure and organization of rhodophyte
interspersed repetitive elements (SINEs) and chromophyte plastid genomes:
in distant species may have originated implications for the ancestry of plastids.
from a common ancestral retrovirus: Mol. Gen. Genet. 232:65— 73
characterization of a squid SINE and a 81. Simpson WJ, Musser JM, Cleary PP.
possible mechanism for generation of 1992. Evidence consistent with horizon-
tRNA-derived retroposons. Proc. Natl. tal transfer of the gene (emm12) encod-
Acad. Sci. USA 90 : 6260-64 ing serotype M12 protein between group
69. Palmer JD. 1985. Comparative organi- A and group G pathogenic streptococci.
zation of chloroplast genomes. Annu. Infect. Immun. 60:1890—93
Rev. Genet. 19:325—54 82. Smith MW, Feng DF, Doolittle RF.
70. Peacock D, Boulter D. 1975. Use of 1992. Evolution by acquisition: the case
amino acid sequence data in phylogeny for horizontal gene transfers. Trends
and evaluation of methods using com- Biochem. Sci. 17:489—93
puter simulation. J. Mol. Biol. 95:5 83. Smith NH. Beltran P. Selander RK.
13— 27 1990. Recombination of Salmonella
71. Pirozynski, KA. 1988. Coevolution by phase I flagellin genes generates new
horizontal gene transfer: a speculation serovars. J. Bacteriol. 172:2209—16
on the role of fungi. In Coevolution of 84. Soubrier F, Levy-Schil S, Mayaux JF,
Fungi with Plants and Animals, ed. KA Petre D, Arnaud A, Crouzet J. 1992.
Pirozynski, DL Hawksworth, 11:248— Cloning and primary structure of the
68 wide-spectrum amidase from Brevi-
72. Pritchard AE, Venuti SE, Ghalambor bacterium sp. R312: high homology to
MA. Sable CL, Cummings DJ. 1989. the amiE product from Pseudomonas
An unusual region of Paramecium mi- aeruginosa. Gene 116:99—104
tochondrial DNA containing chloro- 85. Souillard N, Magot M, Possot O, Sibold
plast-like genes. Gene 78:121—34 L. 1988. Nucleotide sequence of regions
73. Radstrom P, Fermer C, Kristiansen BE, homologous to nifH (nitrogenase Fe pro-
Jenkins A, Skold O. Swedberg G. 1992. tein) from the nitrogen-fixing
Transformational exchanges in the archaebacteria Methanococcus the-
dihydropteroate synthase gene of Neis- rmolithotrophicus and Methanobacter-
seria meningitidis: a novel mechanism ium ivanovii: evolutionary implications.
for acquisition of sulfonamide resis- J. Mol. Evol. 27: 65—71
tance. J. Bacieriol. 174:6386—93 86. Sprague GF Jr. 1991. Genetic exchange
74. Rayssiguier C, Thaler DS, Radman M. between kingdoms. Curr. Opin. Genet.
1989. The barrier to recombination be- Dev. 1:530—33
tween Escherichia coli and Salmonella 87. Spratt BG, Bowler LD, Zhang QY, Zhou
typhimurium is disrupted in mismatch- J, Smith JM. 1992. Role of interspecies
repair mutants. Nature 342:396—401 transfer of chromosomal genes in the
75. Reeves P. 1993. Evolution of evolution of penicillin resistance in
Salmonella O antigen variation by pathogenic and commensal Neisseria
interspecific gene species. J. Mol. Evol. 34:115—25
transfer on a large scale. Trends Genet. 88. Stevens PF. 1984. Metaphors and typol-
9: 17—22 ogy in the development of botanical
76. Sawyer S. 1989. Statistical tests for de- systematics 1690—1960 or the art of
tecting gene conversion. Mol. Biol. putting new wine in old bottles. Taxon
Evol. 6:526—38 33:169—211
77. Selander RK, Smith NH, Li J. Beltran 89. Stewart CB, Schilling JW, Wilson AC.
P. Ferris KE, et al. 1992. Molecular 1987. Adaptive evolution in the stomach
evolutionary genetics of the cattle- lysozymes of foregut fermenters. Nature
adapted serovar Salmonella dublin. J. 330:401—4
Bacteriol. 174:3587—92 90. Syvanen M. 1984. Conserved regions in
78. Shatters RG, Kahn ML. 1989. Gluta- mammalian beta -globins: could they
mine synthetase II in Rhizobium: reex- arise by cross-species gene exchange? J.
amination of the proposed horizontal Theor. Biol. 107:685—96
transfer of DNA from eukaryotes to 91. Syvanen M. 1985. Cross-species gene
 HORIZONTAL GENE FLOW 261

transfer; implications for a new theory Salmonella enterica group El rfb gene
of evolution. J. Theor. Biol. 112:333-43 cluster: 0 antigen and the genetic basis of
92. Syvanen M. 1989. Migrant DNA in the the major polymorphism. Genetics 130:
microbial world. Cell 60:7—8 429—43
93. Syvanen M, Hartman H, Stevens PF. 99. Went FW. 1971. Parallel evolution. Tax-
1989. Classical plant taxonomic ambi- on 20:197—226
guities extend to the molecular level. 100. Whatmore AM, Kehoe MA. 1994. Hor-
[!** AU: please give journal**!] 28: izontal gene transfer in the evolution of
536—44 group a streptococcal emm-like genes-
94. Troitsky AV, Melekhovets YuF, gene mosaics and variation in vir
- Rakhimova GM, Bobrova VK, Valiejo- regulons. Mol. Microbiol. 11:363—74
Roman KM. Antonov AS. 1991. Angio- 101. Whittam TS. Ochman H, Selander
sperm origin and early stages of seed RK. 1983. Multilocus genetic structure
plant evolution deduced from rRNA se- in natural populations of Escherichia
quence comparisons. I. Mol. Evol. 32: coli. Proc. Nail. Acad. Sci. USA 80:
253—61 1751—55
95. Valentin K, Zetsche K. 1990. 102. Woese CR, Fox GE. 1977. Phylogenetic
Nucleotide sequence of the gene for the structure of the prokaryotic domain: the
large subunit of Rubisco from primary kingdoms. Proc. Natl. Acad.
Cyanophora paradoxa—phylogenetic Sci. USA 74:5088—90
implications. Curr. Genet. 18:199—202 103. Wolfe KH, Gouy M, Yang YW, Sharp
96. Vavilov NI. 1922. The law of homolo- PM, Li WH. 1989. Date of the mono
gous series in variation. J. Genet. cot-dicot divergence estimated from
12:47—89 chloroplast DNA sequence data. Proc.
97. Wallin JE. 1927. Symbioticism and the Nail. Acad. Sci. USA 86:6201—5
Origin of the Species. London: 104. Xiong Y, Eickbush TH. 1990. Origin
Baillere, Tindall & Cox and evolution of retroelements based
98. Wang L, Romana LK, Reeves PR. upon their reverse transcriptase se
1992.Molecular analysis of a quences. EMBO J. 9:3353—62