Reconstructing evolution: gene
transfer from plastids to the nucleus
Ralph Bock
1
* and Jeremy N. Timmis
2
Summary
During evolution, the genomes of eukaryotic cells have
undergone major restructuring to meet the new regu-
latory challenges associated with compartmentalization
of the genetic material in the nucleus and the organelles
acquired by endosymbiosis (mitochondria and plastids).
Restructuring involved the loss of dispensable or redun-
dant genes and the massive translocation of genes from
the ancestral organelles to the nucleus. Genomics and
bioinformatic data suggest that the process of DNA
transfer from organelles to the nucleus still continues,
providing raw material for evolutionary tinkering in the
nuclear genome. Recent reconstruction of these events
in the laboratory has provided a unique tool to observe
genome evolution in real time and to study the molecular
mechanisms by which plastid genes are converted into
functional nuclear genes. Here, we summarize current
knowledge about plastid-to-nuclear gene transfer in the
context of genome evolution and discuss new insights
gained from experiments that recapitulate endosymbiotic
gene transfer in the laboratory. BioEssays 30:556–566,
2008. ß 2008 Wiley Periodicals, Inc.
Introduction
Eukaryotic cells arose more than a billion years ago through
endosymbiotic engulfment of free-living eubacteria. Subse-
quently, the enslaved bacterial endosymbionts were gradually
converted into two types of DNA-containing cell organelles:
the mitochondria, which stem from an a-proteobacterium, and
the plastids (chloroplasts), which are derived from a cyano-
bacterium (Fig. 1). Early genetic and biochemical studies
revealed that the genomes of plastids have been greatly
diminished compared with any possible free-living ancestor.
The plastid genome was shown to be far too small to encode
the proteome of the organelle and it was deduced that the
control of plastid biogenesis and function is massively
dependent upon nuclear genes. This conclusion is fully
confirmed by the 122 complete plastid genomes that have
been sequenced to date (http://www.bch.umontreal.ca/ogmp/
projects/other/cp_list.html). Soon after this realization, ele-
gant early experiments
(1,2,3)
paved the way for the current
understanding of peptide import from the nucleo-cytoplasmic
genetic compartment into the organelle.
(4)
An explanation of the observed nuclear control over the
organelle within the framework of the equally convincing
evidence for the endosymbiotic origin of mitochondria and
chloroplasts (Fig. 1) required answers to several fundamental
questions.
(5,6)
Firstly, what is the evolutionary relationship
between the nuclear genes that encode organellar proteins
and the ancestral prokaryote genes of similar function that
have not persisted in the plastid DNA (ptDNA)? Secondly, how
did the mutual dependence of the separate organellar and
nucleo-cytoplasmic systems arise and what are, or were, the
selective forces involved in its evolution? One answer to these
questions envisaged that redundancy was removed during
the early stages of endosymbiont evolution. In this scenario,
many ancestral host genes adapted to produce the host
product plus another of similar function that was active within
the endosymbiont. The second possibility was that genes of
the symbiont were physically transferred to the evolving
nucleus where they were functionally activated to service the
cytoplasmic compartment. In both hypotheses, genes were
rapidly deleted from the endosymbiont genome. The diminu-
tion of the organelle genomes has been attributed to savings
in DNA synthesis for the highly polyhaploid genomes and,
though it has not been observed under experimental con-
ditions, this process is clearly rapid in evolution as the plastids
of non-photosynthetic parasitic plants, such as Epifagus
virginiana, contain highly reduced genomes that must have
lost many genes relatively recently.
(7)
The nuclear control of plastid form and function results in
complex regulation of activity of genes in the two compart-
ments. The situation that exists in cells of the majority of sexually
1
Max-Planck-Institut fu
¨
r Molekulare Pflanzenphysiologie, Potsdam-
Golm, Germany.
2
School of Molecular and Biomedical Science (Genetics), The
University of Adelaide, Adelaide, Australia.
Funding agencies: Research on gene transfer in the authors’
laboratory is supported by the Max Planck Society (R.B.) and the
Australian Research Council (J.N.T.).
*Correspondence to: Ralph Bock, Max-Planck-Institut fu
¨
r Molekulare
Pflanzenphysiologie, Am Mu
¨
hlenberg 1, D-14476 Potsdam-Golm,
Germany. E-mail: rbock@mpimp-golm.mpg.de
DOI 10.1002/bies.20761
Published online in Wiley InterScience (www.interscience.wiley.com).
556 BioEssays 30.6 BioEssays 30:556–566, ß 2008 Wiley Periodicals, Inc.
Abbreviations: BAC, bacterial artificial chromosome; bp, basepair;
cDNA, copy DNA (or complementary DNA); cpDNA, chloroplast DNA;
DAPI, 4
0
,6-diamidino-2-phenylindole; EST, expressed sequence tag;
kb, kilobase pair; mt, mating type; norgDNA, nuclear organellar DNA;
NUMT, nuclear mitochondrial DNA; NUPT, nuclear plastid DNA;
ptDNA, plastid DNA; UTR, untranslated region.
Review articles
reproducing organisms is that a diploid nucleus contains
thousands of paired alleles on multiple linear disomic chromo-
somes. These chromosome pairs are separated to produce
haploid gametes carrying only a single allele after separation of
bivalents during meiosis. Random fertilisation then produces
offspring that show characteristic Mendelian ratios when individ-
ual alleles are tracked between generations. The contrasting
genetic situation in organelles is that they contain multiple haploid
genomes, each with a far smaller set of genes than the nucleus
and, in plants for example, there may be hundreds of polyhaploid
chloroplasts in each leaf cell (Fig. 2). Therefore, although plastid
genomes are genetically simple compared with the nucleus, they
may comprise a large proportion of total cellular DNA in the leaf
cells of some angiosperms, such as spinach (23% of total DNA,
13,000 plastid genomes per cell; Refs 8,9; Fig. 2). However, this
proportion is very variable
(9,10)
and is smaller in many other
species (e.g. 1,000 to 1,700 plastid genomes per cell in
Arabidopsis thaliana; Ref. 11), being dependent upon the nuclear
genome size, the degree of nuclear endopolyploidy, the number of
plastids per cell and the number of genomes per plastid. Thus the
subunit peptides of some plastid protein complexes may be
encoded by both nuclear and organellar genes that differ greatly
in number within the cell. For example, the large subunit of the
most abundant protein on the planet, ribulose bisphosphate
carboxylase/oxygenase (RuBisCO), is encoded by thousands of
identical genes in the chloroplasts of each tobacco leaf cell,
whereas its small subunit is encoded by low-copy-number genes
in the single nucleus (Fig. 3).
Figure 1. Intracellular gene transfer between genomes in the evolution of eukaryotic cells.
(62)
Arrows indicate the direction of gene
transfer, arrow colors correspond to the color of the compartment from which the transferred genetic material originated (red,
mitochondrion; blue, nucleus; green, plastid). Arrow thickness is roughly proportional to the amount of genetic information transferred at a
given evolutionary stage. Mitochondria arose through endosymbiotic uptake of an a-proteobacterium by a pre-eukaryotic cell. Conversion of
the a-proteobacterium into a mitochondrion was accompanied by massive translocation of genetic information into the nuclear genome. In
a second endosymbiosis, a cyanobacterium was engulfed and gradually converted into a plastid (chloroplast). This second endosymbiosis
event was again followed by large-scale information transfer out of the endosymbiont genome into the nuclear genome of the host cell. At the
same time, gene transfer from the mitochondrion to the nucleus continued on a small scale and, in addition, some nuclear and plastid nucleic
acid sequences invaded the mitochondrial genome. While some plastid DNA sequences transferred into the mitochondrion gave rise to
functional tRNA genes,
(32)
no example of formation of a functional mitochondrial gene from nuclear DNA has been identified to date. The
plastid genome seems to be remarkably immune to the invasion of foreign DNA sequences and, to date, no evidence of gene transfer
from either the nucleus or the mitochondrion into the plastid has been documented.
Figure 2. A leaf mesophyll cell of Spinacea oleracea stained
with DAPI. This picture of a mature spinach leaf cell illustrates
the disposition of cellular DNA that is concentrated in the
genetically complex diploid nucleus (n) but DNA is also seen in
the chloroplasts (cp). Multiple plastid genome copies form
nucleoids, of which there are several within each chloroplast. In
some species, the plastid genomes, though genetically very
simple compared with the nucleus, comprise almost as much of
the total cellular DNA as the nucleus because they are present
in multiple copies in each organelle and because there are
many chloroplasts in each leaf cell (usually more than in the cell
shown here). However there is wide species variation in the
proportion of cpDNA.
(9,11)
The chlorophyll molecules within the
plastids show red fluorescence. Mitochondria are not visible in
this preparation. The bar represents 10 mm.
Review articles
BioEssays 30.6 557
Ancestral organelle genes can be recognized
in the nucleus
Gene transfer from the plastid ancestor to the nucleus has
played a major role in the evolution of eukaryotes (Fig. 1).
Genome-wide phylogenetic comparisons of individual nuclear
genes of Arabidopsis thaliana
(12)
with representative prokar-
yote and eukaryotic genomes revealed that 866 of 9,368
nucleus-encoded proteins that were sufficiently conserved to
allow valid comparisons showed closest similarity to proteins
of the cyanobacteria. A further 834 proteins generated
phylogenetic trees that contained cyanobacterial branches.
Extrapolating these data to include genes that could no longer
be readily recognized because of their higher rate of
divergence suggested that about 4,500 of the 25,000 of
A. thaliana nuclear genes (18%) were acquired from a cyano-
bacterial ancestor of the plastid. Only some of the nuclear
genes derived from cyanobacteria were clearly recognizable
as functional in chloroplast biogenesis—others had taken on
novel functions elsewhere in, or outside, the cell.
(5)
These
results not only explain the dependence of the chloroplast on
nuclear genes but also implicate endosymbiotic gene transfers
as a provider of DNA for successful nuclear experimentation
on a massive scale and they bring into stark focus the
unexpected conclusion that the endosymbiont has made large
contributions to the genetic complexity of eukaryotic nuclei.
The importance of the endosymbiotic contribution to the
evolution of nuclear genes is confirmed by ESTanalyses in the
glaucophyte Cyanophora paradoxa, which showed that 10.8%
of nuclear genes were of cyanobacterial origin, though rather
fewer of them had non-plastid functions.
(13)
Sato et al.
(14)
also
made estimates of endosymbiont-derived nuclear genes for
A. thaliana (4.7%) and a red alga, Cyanidioschyzon merolae
(12.7%) and assumed that all those revealed were plastid
directed. Whether a significant proportion of genes that origi-
nated in the ancestral cyanobacterium has been reorganized
to take on non-plastid roles remains to be fully clarified.
Ancient or modern transfer, or both?
Much functional plastid gene transfer must have occurred
soon after the first conglomerate cells were formed but,
remarkably, the process continues, particularly in angio-
sperms. Millen et al.
(15)
studied the genomic location of the
chloroplast translation initiation factor 1 gene (infA) and
evidence of recurrent mobility emerged. In these experiments,
cases where nuclear relocations were inferred were accom-
panied by various degrees of decay of the corresponding
chloroplast sequence, from minor (but incapacitating) muta-
tions to major sequence decay or comprehensive deletion.
Characterization of nuclear infA genes provided evidence of
the acquisition of de novo transit peptides rather than the
hijacking of pre-existing nuclear genes encoding proteins with
expedient cellular targets (Fig. 4). However, clear cases of
gene hijacking exist. For example, Cusack and Wolfe
(16)
show
how the plastid gene rpl32 invaded an intron in the nuclear
Figure 3. Contribution of nuclear and plastid genes to the
plastid proteome. The plastid genome in angiosperms contains
a low number of genes (80) that contribute some polypeptides
to the multisubunit complexes that support photosynthesis and
translation. Several thousands of proteins required for chlor-
oplast biogenesis are not encoded in the ptDNA and are
provided via the nucleocytoplasmic genetic compartment.
Precursors of the required polypeptides are directed to the
chloroplast by N-terminal transit peptides that are removed on
import into the organelle.
Figure 4. Evolutionary changes required to acclimatize
plastid-derived DNA in the nucleus. A plastid gene in
a section of the prokaryotic-type plastome is usually nonfunc-
tional when it has simply been transposed to the nucleus as a
NUPT (a). It must somehow acquire a nuclear, eukaryotic-type
promoter and terminator, a polyadenylation signal and, if the
gene product is to find its way back to the organelle, it must
incorporate a sequence that encodes a transit peptide (b). Only
then will the plastid-like DNA be transcribed (c) and translated
(d) and its protein product become functional after importation
into the plastid (e).
Review articles
558 BioEssays 30.6
gene SODcp in an ancestor of mangrove and poplar trees to
give a chimeric SODcp-RPL32 gene. In mangrove, differ-
entially spliced transcripts encoding either the native SOD
precursor or a truncated version of plastid-bound SOD, which
also encodes the plastid RPL32, allow continued viability of the
species. Both proteins use the SOD transit peptide for protein
import into plastids. In poplar, a further evolutionary refinement
occurred following gene duplication and divergence of
SODcp-RPL32, which then allowed each of the two genes to
dedicate themselves to different products. The apparent high
frequency of functional gene transfer in angiosperms may
merely reflect that the group is a strong focus for genomic
analyses. An example outside angiosperms is found in a moss,
Physcomitrella patens, where the rpoA gene has relocated
to the nucleus compared with its chloroplast location in
Marchantia polymorpha.
(17)
Another contribution of ptDNA to nuclear genes was
revealed when bioinformatic analyses discovered nuclear
exons in Arabidopsis and rice that are derived from insertions
of ptDNA.
(18)
Although nucleotide sequences were used to
expose their origins, the amino acid sequences encoded by
these novel nuclear exons show little resemblance to those of
the organellar counterparts from which they originated. For
example, the Arabidopsis gene (At2g28820) encodes a protein
with similarity to alanine aminotransferases. Its carboxytermi-
nal domain is clearly derived from 477 bp of the plastid gene
rpl16 to which it is 84% similar. The 73 nucleotide differences
that distinguish the domain have a larger effect on the encoded
amino acid sequence which is only 74% similar to the
corresponding plastid protein, providing for an entirely new
function.
These data suggest that gene mobility is far from rare in the
plant kingdom. The regularity with which some genes that are
amenable to nuclear relocation have completed the process
implies that many of those that remain in the organellar
genome are forced to do so by strong selection,
(19)
although
some recalcitrant ones can be transferred experimentally.
(20)
The corollary of this conclusion is a paradox. Why did the
transposition of genes predisposed to nuclear transfer not
occur much earlier in evolution? It is as though novel
mechanisms of transfer have evolved that were at first absent
from the cell and the advent of their availability resulted in a
recent flush of independent relocations. Such evolutionary
developments could include a new system of protein import
into organelles, consistent with the variety of sequences that
lead to efficient targeting.
DNA per se or are RNA intermediates involved?
DNA molecules themselves could move between genetic
compartments, but some studies have reported evidence for
the involvement of RNA intermediates. The possibility of direct
DNA transfer is founded on evidence from experiments in
yeast
(21,22)
and on comparisons between organellar DNA and
nuclear DNA of the same plant species, which show that
intergenic spacers and other non-coding regions of organellar
DNA are found in nuclear-transferred copies as often as are
coding sequences.
(23)
When plant nuclear genome sequen-
ces reveal uninterrupted organelle-like DNA tracts that contain
the organellar introns, tRNAs and hundreds of kilobases of
organellar non-coding regions, DNA transfer seems most
likely to have been involved. The possibility of RNA inter-
mediates is based on observations of certain functional
mitochondrial protein-encoding genes that are present in the
mitochondrial DNA (mtDNA) in some species of angiosperms,
but are located in the nucleus in others.
(24,25)
Mitochondrial
protein-coding genes often have introns and their mRNAs are
post-transcriptionally modified by RNA editing, but the nuclear
copies seem to lack these organelle-specific marks suggest-
ing theyare derived from a fully processed mRNA.
(26)
Although
the possibility that cDNA intermediates might be involved in the
transfer of genes from mitochondria to the nucleus in flowering
plants cannot be excluded, there are other interpretations of
the same data.
(27)
The predominance of C-to-T (and the
associated G-to-A) substitutions caused by the hypermut-
ability of 5-methylcytosine that is abundant in plant nuclear
DNAs,
(28)
might be misinterpreted as signs of mitochondrial
editing. This class of substitution is by far the most-common
mutation in long tracts of mtDNA or ptDNA in the nucleus
(dubbed NUMTs and NUPTs, respectively; see below), whose
formation must have been directly from organellar DNA.
(28)
On
a broader scale, evidence that implicates RNA intermediates
in nuclear transfers in eukaryotic groups other than flowering
plants is so far lacking,
(22)
as is evidence that implicates cDNA
in the transfer of plastid genes.
(23)
Non-functional gene transfer: promiscuous
plastid DNA in the nucleus
During evolution, most ptDNA transfer events do not result in
functional nuclear genes because of the differences between
the nuclear and plastid genetic environments. The prerequisite
and first step in any scenario in which organellar genes
relocate to, and function in, the nucleus necessarily involves
migration and integration of nucleic acid molecules. After or
during nucleic acid transit, additional complex modifications
are required to overcome the ‘‘culture shock’’ for an expatriate
gene as it moves from a prokaryotic to a eukaryotic compart-
ment (Fig. 4). Experiments demonstrated fragments of extant
ptDNA integrated in the maize mitochondrial genome
(29)
and
into both mitochondrial and nuclear DNA of spinach.
(30)
In
tobacco, it was demonstrated that very long tracts of ptDNA,
some inferred to be as large in size as the entire plastid
genome (plastome), were abundant in the nucleus.
(31)
More-recent genome sequencing confirmed that plastid-
like DNA sequences (NUPTs) are common in the nuclear
genomes of photosynthetic species and other species that
contain a relict plastid.
(23,32,33)
The diminutive nuclear genome
Review articles
BioEssays 30.6 559
of Arabidopsis contains relatively little plastome-like DNA, but
the rice nuclear genome, which is also small, contains many
NUPTs totaling >800 kb (Refs 34–36). Chromosome 10 of rice
alone contains 28 tracts of ptDNA greater than 80 bp in length
including two large insertions of 131 and 33 kb (Ref. 37).
These integrants showed greater than 95% sequence identity
to the bona fide organellar genomes and they are assumed
to reveal recent transposition events. Shotgun sequencing
projects are prone to underestimate organellar DNA insertions
because they may be discarded as contaminants during
assembly. The BAC-by-BAC approach that first characterized
the 33 kb ptDNA insert on rice chromosome 10 (Ref. 37) must
be used to determine efficiently the presence or absence of
organelle DNA in the nucleus. This is important as initially it
was thought that norgDNAs were not present in the honeybee
genome,
(33)
whereas recent analysis indicates that they are
present, and in high copy numbers.
(38)
The data pertaining to NUPTs are currently sparse as the
plant genomes available for analysis were chosen specifically
for their small size. Evidence suggests that norgDNAs are
much more abundant in other species. For example, although
we now know that NUPTs are present in Arabidopsis, Ayliffe
et al.
(39)
were unable to detect any by the same Southern
blotting method that revealed a plethora of such sequences in
all other species tested. It is clear that the Nicotiana tabacum
nucleus has a very large representation of NUPTs and we
speculate that the species whose genomes are an order of
magnitude larger still, have correspondingly more NUPTs.
Genomic analysis suggests that, though they may be
initially integrated into the nucleus as long continuous tracts,
norgDNAs are far from static and evolve rapidly via multiple
substitutions, insertions, deletions and duplications.
(28)
Alter-
natively, they may have been fragmented and reassembled
during insertion or there may be mixtures of existing and
incoming DNA involving baroque interchromosome rear-
rangements that culminate in the formation of complex
mosaics containing both mitochondrial and plastid DNA from
many disjoined parts of the two original organelle genomes.
(33)
The apparent paucity of more diverged copies is surprising
and may be due to a dynamic equilibrium between ingress and
egress (i.e. insertion of new copies versus decay of old copies;
Sheppard and Timmis, unpublished results), or to preserva-
tion of the sequence by continuous copy correction (gene
conversion) with ptDNA and mtDNA.
(40)
In the absence of such
mechanisms, the sequence of some promiscuous DNAs may
diverge to the extent that their origin cannot be recognized.
Reconstruction of gene transfer processes in
the laboratory
To be able to decipher the precise mechanisms of gene
transfer from the plastid to the nuclear genome, an exper-
imental system is needed that facilitates the observation of
gene transfer processes in real time. The phylogenetic
evidence discussed above suggested that natural transfer
events can be monitored only over large evolutionary time
scales. This made it clear that strong selective pressure would
be needed to identify gene transfer events in laboratory
experiments. The development of techniques to alter the
genetic information of the plastid and, most importantly, the
ability to integrate foreign genes into the plastid genome by
transformation
(41,42)
have made it possible to design such
experiments. A key experiment demonstrating the ongoing
transfer of genetic material from plastids into the nucleus is
illustrated in Fig. 5. By chloroplast transformation, two genes
for antibiotic resistances are introduced into the plastid
genome of tobacco (N. tabacum) plants: a spectinomycin-
resistance gene (aadA) with plastid expression signals (a
plastid promoter and 5
0
untranslated region (5
0
UTR) and a
plastid 3
0
UTR conferring mRNA stability) and a kanamycin-
resistance gene (nptII) fused to nuclear expression signals
(promoter and terminator from the 35S gene of the cauliflower
mosaic virus, CaMV; Fig. 5A). Transplastomic plants contain-
ing these marker genes in their plastid genomes are resistant
to spectinomycin, but not to kanamycin, because the nuclear
expression signals of the nptII gene do not function efficiently
in the organelle. Background expression can be further
suppressed by insertion of a nuclear spliceosomal-type intron
into the nptII gene which cannot be spliced out in the plastid
(Fig. 5A). However, if the kanamycin-resistance gene were to
jump into the nucleus, it is immediately capable of expression,
allowing cell division and plant growth in the presence of
kanamycin (Fig. 5B).
Using these transplastomic lines with a eukaryotic-type
kanamycin-resistance gene in the plastid genome, two
alternative strategies were pursued to select for DNA transfer
from the plastid to the nucleus. In one approach, seeds
produced by fertilization of wild-type plants with pollen from
transplastomic plants were sown on synthetic medium
containing kanamycin.
(43)
In an alternative approach, leaf
explants from the transplastomic plants were subjected to
kanamycin selection on a plant regeneration medium.
(44)
Both
screens resulted in kanamycin-resistant plants at unexpect-
edly high frequencies. The seedling selection produced 16
antibiotic-resistant lines in a sample of 250,000 seeds (equal-
ing one transfer event in 16,000 pollen grains). The leaf
selection yielded 12 antibiotic-resistant lines from 1,200 tissue
explants subjected to regeneration in the presence of
kanamycin. If the data from the leaf selection are roughly
converted into an estimate for the gene transfer frequency at
the cellular level, this frequency is lower (approximately one
transfer event per 5 million somatic cells) than in pollen grains.
It should not be formally excluded that this discrepancy can be
attributed to differences in the experimental design and/or the
in vitro selection procedures used. For example, the trans-
plastomic lines used in the pollen experiment have two copies
of the nptII per plastome, the lines used in the leaf experiment
Review articles
560 BioEssays 30.6