Repair of shattered chromosomes in Deinococcus
radiodurans: a natural contig assembly
Ksenija Zahradka1, 4, Dea Slade1, Adriana Bailone2, Suzanne Sommer2,
Dietrich Averbeck3, Mirjana Petranovic4, Ariel B.Lindner1 & Miroslav Radman1, 5
1
Université de Paris-Descartes, Faculté de Médecine; INSERM Site Necker, U571, 156 rue de
Vaugirard, 75015 Paris, France;
2
Institut de Génétique et Microbiologie, Bat. 409, Université
Paris-Sud, 91405 Orsay Cedex, France;
3
Institut Curie, Section Recherche, UMR 2027 CNRS,
Centre Universitaire de Paris-Sud, Bat. 110, 91405 Orsay Cedex, France;
4
Division of
Molecular Biology, Ruder Boskovic Institute, P.O. Box 180, 10002 Zagreb, Croatia;
5
Mediterranean Institute for Life Sciences, Mestrovicevo setaliste bb, 21000 Split, Croatia
Deinococcus radiodurans, an extremophile bacterium, survives excessive desiccation and
ionizing radiation that shatter its genome into hundreds of short 20-30 kb fragments.
Remarkably, these fragments are reassembled into a functional 3.28 megabase genome. Here,
we describe the novel molecular mechanism accounting for this repair process: an “extended
synthesis dependent strand annealing” (ESDSA) followed by homologous recombination. At
least two genome copies and random DNA breakage are requirements for effective ESDSA.
In ESDSA, chromosomal fragments with overlapping homologies are used both as primers
and templates for a massive synthesis of single strand extensions, like in a single-round
multiplex PCR. This synthesis depends on DNA polymerase I and incorporates more
nucleotides than does normal replication in intact cells. Newly synthesized complementary
single strand extensions appear to anneal with high precision so that contiguous DNA
1
fragments are joined together forming long linear intermediates. These intermediates require
RecA-dependent homologous recombination to mature into circular chromosomes that are
double-stranded patchworks of numerous DNA blocks synthesized before radiation
connected by DNA blocks synthesized after radiation.
___________________________________________________________________________
Along with starvation, heat and cold, dehydration or desiccation is perhaps the most frequent and
severe challenge to living cells1. Bacterium Deinococcus radiodurans is the best-known
extremophile among a handful of organisms found to resist extremely high exposures to desiccation
and ionizing radiation2-5. Both desiccation and radiation cause extensive intracellular DNA double
strand breakage (DSB)4. Because a single unrepaired DSB is usually lethal6, DSBs are considered
to be the most severe form of genomic damage. Standard vegetative prokaryotic and eukaryotic
cells can repair less than a dozen simultaneous DSBs7. D. radiodurans survives ionizing radiation
breaking its genome into several hundred fragments (over a thousand DSBs per cell containing at
least two genome copies) due to a DNA repair process accomplishing an efficient and precise
fragment assembly2,3,8. This molecular transaction is formally akin to the computer-assisted contig
assembly of shotgun sequenced random genomic fragments2.
Because D. radiodurans recA and polA are its most radiation sensitive mutants2,3, the
RecA recombinase and DNA polymerase I (PolA) are essential for the repair of shattered
chromosomal DNA. However, the molecular mechanism of this repair remained a mystery9-15. This
study reveals a new form of DNA repair involving mutually dependent DNA replication and
recombination whereby DNA fragments acquire “sticky ends” – long single-strand extensions
synthesized in a process where the fragments are used as primers and templates. Complementary
extensions anneal connecting contiguous fragments into long linear intermediates that are matured
to unit-size circular chromosomes by homologous recombination.
2
Hypotheses and the experimental strategy
A priori, there are six known mechanisms that, alone or in some combination, could rejoin
hundreds of partially overlapping chromosomal fragments in D. radiodurans, and none has been
excluded so far:
(i)
Non-homologous end-joining (NHEJ) of DNA fragments16,17
(ii)
Homologous recombination (HR) via conservative crossovers involving ends of
overlapping fragments18-21
(iii)
Single strand annealing (SSA) via a strand-biased 5’-exonucleolytic erosion of the two
ends resulting from a DSB providing a 3’single strand overhang for eventual annealing
with (a) each other (intra-chromosomal SSA) via an internal micro-homology (usually a
tetranucleotide) resulting necessarily in the deletion of the sequence between two microhomologies19,21, or (b) a fully complementary single strand end of a contiguous fragment
belonging to another chromosomal copy (inter-chromosomal SSA)20 thereby avoiding
sequence deletions (SSA in Fig. 1)
(iv)
Synthesis dependent strand annealing (SDSA) can assemble DNA fragments by strand
invasion, i.e., annealing of both recessed single-stranded ends to the same intact region
of a homologous chromosome (D-loop formation; the bracketed intermediate in Fig.1).
D-loop provides the primer for synthetic elongation of the annealed 3’ end to produce a
moving D-loop, like in transcription. Once elongated on the same template sequence,
the two extended complementary ends dissociate from their template and anneal21,22.
(v)
Break-induced replication (BIR) is, in its early stage, mechanistically akin to an SDSA
event involving only one broken end19,21. Following the initial one-end strand invasion
(D-loop formation), either (a) a genuine replication fork is formed resulting in a semiconservative copying of the template, or (b) the displaced newly synthesized single
3
strand (like in SDSA, Fig. 1), is itself copied by a discontinuous synthesis resulting in a
conservative replication of the homologous DNA template. Either version of BIR could
link together many DNA fragments by iteration of the replication process at the growing
ends, which would leave little, if any, original double-stranded material in the repaired
chromosome.
(vi)
Copy choice (CC) is a DNA replication mechanism that involves sequential template
switching between double-stranded templates from different overlapping fragments until
a full-size chromosome is newly synthesized23. This mechanism is not clearly
distinguishable from iterative BIR. In its original version, it is unidirectional and does
not require DNA breakage for template switching.
Mechanisms involving extensive homologies, (ii), (iii b), (iv), (v) and (vi), can assure a high
precision of chromosomal reconstruction, others can be (i), or must be (iii a), mutagenic. The
requirement and pattern of DNA synthesis is a diagnostic feature for the listed recombination repair
mechanisms. Mechanisms (i) and (ii) involve no significant DNA synthesis, whereas (iii a and b)
and (iv) involve significant but limited synthesis, (v) requires extensive, close to total, synthesis,
and in (vi) all repaired DNA is newly synthesized.
DNA synthesis and repair of D. radiodurans chromosomes shattered by radiation
To discriminate among alternative recombination repair mechanisms on account of required or
associated DNA synthesis, we explored the kinetics of DNA fragment joining by pulsed-field gel
electrophoresis (PFGE) and measured the rate of DNA synthesis by labelling DNA with a 15 min
3
H-TdR pulse. Following 7 kGy gamma radiation, shattering chromosomal DNA to about 20-30 kb
fragments (Fig. 2a), we observed a clear coincidence of DNA fragment assembly and a massive
DNA synthesis (Fig. 2b). This repair synthesis occurs before cell division at a much higher rate
than DNA replication in the growing unirradiated cell culture (Figs. 2b) and is absent in a polA
4
strain (Fig. 2d) that shows no evidence of DNA repair (Fig. 2c). The repair process is highly
efficient: no repair or synthesis is visible for 1.5 h after irradiation, but in the following 1.5 h,
chromosomal DNA appears fully reassembled (Fig. 2a, b). These observations cannot be easily
accommodated by mechanisms (i) and (ii), but generally allow all mechanisms which may (e.g., iii
and iv) or must (e.g., v and vi) involve extensive DNA synthesis.
We confirm, as previously observed20, a largely RecA-independent fragment assembly
(Fig. 2e) that is clearly accompanied by substantial DNA synthesis (Fig. 2f), but does not produce
complete chromosomes even after 24 h (Fig. 2e). These experiments establish a correlation between
the PolA–but not the RecA–dependent DNA synthesis and the repair of shattered D. radiodurans
chromosomes. However, RecA is required for the appearance of full-size chromosomes (Fig. 2),
which suggests a critical role in the late-stage process of maturation of circular chromosomes (Fig.
3).
Density gradient analysis of the structure of repaired chromosomes
How can small DNA fragments sustain the rate of DNA synthesis much higher than during normal
DNA replication (Fig. 2a, b), and why is such synthesis correlated with fragment assembly? To
address this double question, the repair of the D. radiodurans DNA shattered by ionizing radiation
was analysed by an adaptation of the classical Meselson-Stahl experiment24, i.e., by DNA density
labelling and analysis of its buoyant density by ultracentrifugation in CsCl equilibrium density
gradients. The extent and pattern of DNA repair synthesis was monitored by the incorporation of a
non-radioactive heavy analogue of thymidine (5-bromo-deoxyuridine – 5-BrdU) present only after
irradiation, for 3 h, when the bulk of DNA repair appears to be completed (Fig. 2a). The
incorporation of 5-BrdU does not hinder DNA repair (see Fig. 5). The radioactive labelling of DNA
by 3H-thymidine incorporation was carried out under three different regimens: (a) “pre-labelling”,
continuous labelling only before irradiation, (b) “post-labelling”, labelling only after irradiation
5
during the 3 h period of DNA repair, and (c) “pre- and post-labelling”, i.e., combined (a) and (b).
The DNA density analysis from unirradiated D. radiodurans cultures (Fig. 4c) agrees with
the Meselson-Stahl experiment with E. coli, i.e., the replication of all genomic components appears
to be semi-conservative: the H/L (heavy/light) density appears after the first replication cycle, H/H
during the second. As expected, upon DNA denaturation and centrifugation in alkaline CsCl
gradients, all single-stranded material segregates into either H or L densities (Fig. 4, a1 - c1).
Because DNA is monitored by 3H-thymidine radioactivity, the unirradiated cultures show, in the
“pre-labelling” regimen, only the H/L* material (* denotes 3H-label) (Fig. 4a), and only L* in
denaturing gradients (Fig. 4a1). In the “post-labelling” regimen predominantly H*/L and – to the
extent of second replication round – also H*/H* are found (Fig. 4b) (and only *H in denaturing
gradients, Fig. 4b1). We shall call the light DNA synthesized before irradiation “old” and the heavy
DNA synthesized after irradiation “new”. Because only old DNA is damaged in radiation
experiments and the repair takes place in the heavy medium, the two strands of the DNA repaired
by NHEJ and HR are expected to be essentially old/old, those repaired by BIR substantially
old/new and/or new/new, and those made by CC fully old/new and/or new/new (see above). None
was observed. In neutral density gradients, the old “pre-labelled” DNA was found spread towards
intermediate densities due to 5-BrdU incorporation after irradiation (Fig. 4a). On the other hand, the
“post-labelled” new DNA synthesized after irradiation is essentially heavy with significant skewing
towards lighter densities (Fig. 4b). Thus, the density patterns of DNA repaired after gamma
radiation do not resemble either semi-conservative or conservative replication and therefore do not
support NHEJ, HR, CC or BIR as the predominant DNA repair mechanism (text above and Fig. 4ac). Yet the amount of DNA repair synthesis (Fig. 2) and the extent of density shifts (Fig. 4a-c) do
not support standard SSA or SDSA mechanisms either, but rather some kind of DNA repair
creating large patchworks of old and new DNA material. To test for an old/new patchwork structure
at the level of single strands, the density of denatured repaired DNA was analysed in alkaline CsCl
6
density gradients. The densities of pre-labelled single strands are shifted substantially from light
(old) towards heavy (new) (Fig. 4c1). Because the size of the DNA isolated by the described
method is in the range of 15 to 25 kb (not shown), i.e., generally smaller than the in vivo size of
fragments following 7 kGy irradiation (20-30 kb), the detection of fragments containing old and
new material is diminished.
What is the size of newly synthesized DNA patches? Fig. 4a1 shows that, in the “prelabelling” regimen, the majority of single stranded material is heavier than light (old) but lighter
than heavy (new). The inverse is not true. In the “post-labelling” regimen, the single-stranded
DNA is under the heavy peak and skewed towards lighter densities (Fig. 4b1). In other words,
during repair in the presence of 5-BrdU, the old light strands are made heavier (by association with
new heavy material) than are the new heavy strands made lighter (by association with old light
material). This suggests that the tract of newly synthesized (density and radioactively labelled)
material is often longer than the size of original light (radioactively but not density labelled)
material. This DNA density analysis suggests a “distributive” mode24 of DNA (repair) replication
that generates a patchwork structure of old light radiation-produced fragments interconnected by
newly synthesized heavy DNA blocks that are at least as long as the initial fragments. The old and
new DNA blocks are connected via old/new hybrid DNA regions detected as H/L fragments. These
results exclude all repair mechanisms involving no, or very limited, DNA synthesis, as well as
those requiring complete, or close to complete, DNA synthesis.
Distribution of newly synthesized DNA patches revealed by DNA photolysis
Unlike natural bases, 5-BrdU is highly photosensitive; therefore, the postulated old-Thy/new-5BrdU patchwork structure of the repaired chromosomes can be tested by intracellular DNA
photolysis. 5-BrdU-incorporation sensitizes DNA to strand breaks by photolysis with ultraviolet
and short wavelength visible light25-27. Only the 5-BrdU substituted strands are highly sensitized
7
such that one-strand substitution in DNA duplex causes single-strand breaks, and two strand
substitution causes double-strand breaks as well (both directly and via closely spaced single-strand
breaks in the two complementary strands25,27). Thus, when DNA repair synthesis takes place in the
presence of 5-BrdU, the patchwork structure predicted by SSA-like repair (newly synthesized DNA
only in one strand – Fig. 1) should produce DNA susceptible only to limited single-strand breakage,
and DNA repaired by SDSA-like mechanisms (newly synthesized DNA in both complementary
strands) to both single- and double-strand breakage (red coloured strand regions in Fig. 1). The
protocol was as in previous experiments (7 kGy radiation and 3 h repair time in 5-BrdUsupplemented medium, Fig. 4), but instead of DNA isolation, cells were additionally starved in
buffer, ice cooled, UV irradiated on ice (the latter steps are to minimize repair of UV
photoproducts) and embedded in agar plugs for PFGE analysis of DNA.
We confirm that, also under our experimental conditions, one-strand 5-BrdU substitution
(during 1.5 h, i.e., less than one generation time) of unirradiated D. radiodurans chromosomes does
not significantly sensitize DNA to double-strand breakage by UV induced photolysis (not shown).
Two-strand 5-BrdU substitution results in extensive DNA photolysis (Fig. 5b). The intracellular
photolysis degrades DNA by double-strand breakage–only when repaired in 5-BrdU-medium–to
the fragment sizes similar to those seen after the initial DNA breakage by gamma rays (Fig. 5a).
UV light-induced DNA fragmentation does not increase beyond the saturating dose of 250-500
J/m2, whereas it does for the fully substituted DNA (Fig. 5b). The observation that fragmented
chromosomes recomposed in 5-BrdU medium can be decomposed by UV light to initial fragment
sizes suggests that most, perhaps all, reassembled fragments are linked together via double-stranded
blocks of newly synthesized DNA (red coloured in Fig. 1). This is not predicted by any of the
single six listed known mechanisms for DSB repair.
The efficiency of UV light in reaching saturation of photolysis of DNA repaired in 5-BrdU
medium suggests that the blocks of newly synthesized double stranded DNA patches in
8
reconstituted chromosomes are large, which is in agreement with density gradient analysis (Fig. 4ac). The fragmented chromosomes are probably not homologously paired because in such a case
(e.g., yeast meiosis) the SDSA repair synthesis would be limited to the short region between the
recessed ends of individual DSBs19,21. The proposed old/new DNA patchwork structure of
chromosomes repaired by an SDSA-like mechanism is drawn in Fig. 1.
Immunofluorescent microscopy reveals synthesis of single stranded DNA during DNA repair
Of the considered repair mechanisms, only the SDSA-like repair mechanism in Fig.1 predicts that
the massive DNA repair synthesis in Fig. 2b produces initially single-stranded extensions of the
numerous double-stranded fragments prior to their annealing via sequence complementarity. To test
this prediction, we have developed an immunofluorescent microscopy method to detect and
measure, on single cell level, the newly synthesized DNA in its single- and double-stranded form.
The method is based on the specificity of monoclonal anti-5-BrdU antibodies that bind the 5-BrdU
moiety in single-stranded but not in double-stranded DNA (see Methods). D. radiodurans cell
suspensions exposed to 5-BrdU were fixed and rendered permeable for antibodies, treated with 5BrdU-specific primary and fluorescence-tagged secondary antibodies, and the fluorescence was
analysed by microscopy. Similar to the experiment in Fig. 2b, newly synthesized DNA was labelled
by 10 min pulses of 5-BrdU. Under native condition, which reveals only the newly synthesized
single-stranded DNA, unirradiated cells show a low-level fluorescent background (Fig. 6a, c). After
DNA denaturation the antibody detects all newly synthesized DNA, and the same cell samples
show bright fluorescent foci whose intensity reflects the amount of incorporation of 5-BrdU into
DNA (Fig. 6b, d). This pattern of fluorescence is expected from a semiconservative DNA synthesis.
In 7 kGy -irradiated cell cultures, under native condition, 5-BrdU pulses at different time
periods cause the emergence of fluorescent foci after at least 1.5 h incubation in rich medium,
corresponding to the onset of global DNA synthesis (Fig. 2b). Their frequency and intensity peaks
9
at 3 h (Fig. 6a, c), then both start decreasing and the foci disappear by 6 h (Fig. 6a). Under
denaturing (Fig. 6b, d) conditions, the foci peak, as expected, at the same time as under native
conditions, when the rate starts decreasing to the level of normal DNA replication. This pattern was
seen in Fig. 2b. At 3 h, the intensity of fluorescence under native conditions (newly synthesized
single-stranded DNA) is about 15% of that under denaturing conditions (all newly synthesized
DNA). Thus, it appears that a significant fraction, and perhaps all, of DNA synthesis triggered by
DNA fragmentation produces initially single-stranded DNA that is rapidly converted to doublestranded DNA. This apparent rapidity of synthesis-dependent strand annealing may signal that the
complementary single strands are often synthesized almost coincidently in time and space.
Discussion
D. radiodurans is a small non-sporulating and non-pathogenic bacterium whose sequenced genome
is composed of two circular chromosomes (2.65 and 0.41 mega bp) and two circular plasmids
(0.177 and 0.046 mega bp)9. Its extreme radiation resistance appears as a by-product of natural
selection for its desiccation resistance4. Each of the 41 tested D. radiodurans radiation sensitive
mutants are also desiccation sensitive proportionally to their radiation sensitivity4 and selection for
desiccation resistance in other bacteria co-selects radiation resistance5. This correlation implies that,
although excessive desiccation and ionizing radiation cause damage to all key cellular components,
genome reconstitution rescues cellular life.
We interpret the results of our experiments as evidence that the capacity of D. radiodurans
to repair hundreds or thousands of double strand breaks in its genome is due to a mechanism that
Matthew Meselson suggested we call extended synthesis dependent strand annealing, ESDSA.
ESDSA differs from the standard “limited” SDSA in that it involves strand invasion (the bracketed
intermediate in Fig. 1) between dispersed fragments belonging to different chromosomal copies and
10
sharing overlapping sequence homology. The synthetic extension of the priming 3’ termini proceed
probably to the end of the invaded fragments, followed by the annealing of complementary newly
synthesized strand extensions of contiguous fragments from other chromosomal copies (Fig. 1).
The apparent rapidity by which single strands are converted to double strands (Fig. 6a, b) prompts
the hypothesis that the synthesis of complementary single strands often coincides in space and time.
Such a double coincidence would occur when two priming fragments separated by a large gap are
bridged by a templating fragment (e.g., via homologous pairing of their sequence overlaps)
providing for a synthetic filling-in of the missing sequence. If the sequences of the two priming
fragments were ABCD and GHIJ, that of the bridging and templating fragment could be DEFG, and
the sequence of the assembled contig is ABCDEFGHIJ.
The distinct features of the ESDSA model are that (i) it requires at least two chromosomal
copies broken at different positions, and (ii) it involves a single-round multiplex PCR-like step
(steps 2 and 3 in Fig. 1) resulting in long, newly synthesized single strand overhangs that allow for
an accurate annealing process. For consistency, the standard SSA mechanism (Fig. 1) should be
called synthesis-independent strand annealing, or SISA, because no synthesis is required for the
annealing step, only for the eventual gap filling after annealing (Fig. 1). Such mechanism was
proposed by Daly and Minton20. A SISA mechanism with an extensive nick translation synthesis is
a conceivable alternative to the ESDSA. However, the E. coli PolA Klenow fragment complements
D. radiodurans polA mutants for resistance to -radiation28, showing no requirement of the PolA
nick translation activity for DNA repair. Furthermore, this alternative “cycling” SISA mechanism
would produce almost completely newly synthesized DNA that is incompatible with our results
(Figs. 4 and 5b). Similarly, iterative BIR would also produce only semi-conservative (v, a) or
conservative (v, b) DNA replication leaving little, if any, original DNA duplex within the repaired
chromosome. A combination of the (v, b) version of BIR and HR is conceivable, yet would not
explain the massive growth of DNA in the recA mutant (Fig. 2e). The apparent regularity of the
11
alternation of old and new DNA blocks within the repaired chromosomes (Fig. 5) can be readily
explained by the ESDSA model. Although many details within the framework of the ESDSA
mechanism have not been clarified, e.g., the priming step in DNA strand elongation, the major
simple alternatives have been ruled out.
The pattern of DNA repair seen in PFGE (Fig. 2e), and particularly in optical mapping
experiments29, shows that the majority of DNA fragments “grow” progressively and that most
fragments are used up in the chromosomal assembly process (Fig. 2). One can therefore imagine
ESDSA “chain reactions” involving numerous “nucleation” points in the cell producing growing
linear double-stranded DNA repair intermediates29. Such long linear DNA, either shorter or longer
than the respective chromosome, would require homologous, typically RecA-dependent,
intermolecular or intramolecular crossovers to produce mature unit size circular chromosomes (Fig.
3). This scenario is consistent with PFGE experiments with a recA mutant showing the lack of a
complete set of chromosomal restriction fragments even after 24 h (Fig. 2e). However, both DNA
synthesis and repair are less efficient in the recA mutant (Fig. 2e, f) than in the wild type.
Therefore, it is conceivable that deinococcal RecA, through its preferential double-strand binding
activity12, brings together the recessed DNA fragments with overlapping homology, thereby
facilitating an accurate priming of strand extension in ESDSA (Fig 1 and the text above). However,
the key involvement of RecA is clearly in the chromosomal maturation process (Fig. 3).
A polA, but not recA, mutant is fully deficient in ESDSA repair (Fig. 2). However, DNA
base and sugar damages by radiation-induced oxygen free radicals, which can cause single-strand
breaks directly and by proximal base excision repair (BER) events, are at least 10 times more
frequent than DSBs30 and typically require PolA for the completion of repair31. Therefore, it is
expected that PolA protects the integrity of the DNA fragments produced by -radiation (see the
disappearance of DNA fragments in the polA mutant, Fig. 2c) by repairing single nucleotide gaps
created by the BER enzymes, and that it participates directly in strand extension synthesis (Fig. 1).
12
However, before the isolation of a conditional mutant of DNA polymerase III, we cannot exclude
the possibility that PolA only initiates a DNA polymerase III-catalysed single strand elongation or
even contributes only to the maintenance of fragments.
Unlike other bacteria, all tested radiation resistant bacterial species show condensed
nucleoids even after ionizing radiation15. It was suggested that a special ring-like chromatin holds
broken DNA ends in register and facilitates correct repair by NHEJ14. However, a high
concentration of DNA fragments in the condensed nucleoid is expected to facilitate any bimolecular
homologous interaction required for DNA repair, including ESDSA. Perhaps, in addition to the
peculiarity of the RecA12, the condensed chromatin is another evolutionary innovation of D.
radiodurans assuring its radiation resistance through a high efficiency ESDSA repair.
Can ESDSA account for the apparent fidelity of DNA contig assembly in D. radiodurans?
Generally, the larger the number of DNA fragments, the higher the precision required for avoiding
their incorrect assembly and the longer the required homology must be. If there is an unusual
amount of DSBs, an unusually long homology, and hence single strand exposure, is required. Such
unusually long overhangs are newly synthesized in the course of ESDSA (Fig. 1). However, all
homology-based repair of shattered DNA raises the question of the accuracy of the assembly of
chromosomal fragments containing repetitive sequences. The problem of incorrect fragment
assembly via repetitive sequences present in single strand tails could be circumvented in ESDSA if
the synthesis of complementary single strands were coincident in space and time, as discussed
above. But even for solitary strand extensions (Fig.1), the accuracy can be maintained if the
fragments’ tails were much longer than the longest repetitive sequences. In that case, annealing
only a limited repeated sequence block within a long non-complementary single-stranded overhang
could not readily link the two fragments. Partially annealed structures are the preferred substrates
for DNA helicases, which act to dissociate such aberrant DNA associations. Our experiments
suggest that the size of newly synthesized overhangs is similar to the fragment size, i.e., about 20 to
13
30 kb – much longer than the longest D. radiodurans’ repetitive sequences (IS of about 1kb)9,32.
The avoidance of a lethal assembly of non-contiguous DNA fragments may have provided
sufficient selective pressure for the evolution of the ESDSA mechanism (with long single stranded
overhangs) rather than a mechanistically more simple inter-chromosomal SISA (with shorter
overhangs) (Fig. 1).
The high fidelity ESDSA process requires that all fragment’s overhangs be extended by
copying a fragment that is contiguous in the intact chromosome. Each mispriming of strand
elongation, e.g., within a repeated sequence of a wrong fragment, will later cause the annealing of
two non-contiguous fragments generating a gross chromosomal rearrangement. We can think of
four strategies for assuring the fidelity of both priming of the single strand synthesis and of strand
annealing: (1) homologous pairing of recessed double-stranded DNA fragments before the
initiation of D-loops, perhaps by the peculiar deinococcal RecA12, (2) editing the pairing process by
mismatch repair proteins33,34, (3) repeat-binding proteins9 preventing sequence repeats from
becoming single-stranded or from annealing, and (4) stable secondary structures (hairpins) of
repetitive sequences32 preventing their annealing with another partner molecule. Strategy (1)
predicts that DNA assembled by fragment concatenation in the dead recA mutant cells (Fig. 2e)
could be riddled with rearrangements.
Ramifications. The extraordinary diversity of genetic lineages represented in the D. radiodurans
genome (Bacillus, Thermus and multiple eukaryote-like genes)9,32 might be a hallmark of an
efficient incorporation of foreign genes in its evolutionary past, probably due to the combination of
its high transformability and recombination repair. Reproducing in vitro the D. radiodurans’ DNA
recombination repair, perhaps under low fidelity conditions, could provide a potential tool for the
shuffling of chromosomal fragments from the entire biosphere.
The robustness of D. radiodurans has presumably evolved under harsh non-reproductive
14
conditions such as those in arid desert environments35. Extreme desiccation accompanied by
extensive DNA breakdown4, akin to an exaggerated apoptosis, leads eventually to the “clinical
death” of deinococcal cells. Yet upon supply of water and some ions, the dead bacterium
“resurrects” by reconstituting its shattered genome36. If we consider growing E. coli as a bacterial
paradigm of renewing epithelia (including the evolution of carcinomas), then D. radiodurans can
be a bacterial paradigm of long-lived, non-dividing neurons. Therefore, exploring mechanisms of
deinococcal robustness could inspire new approaches in regenerative medicine. Moreover, evolving
highly robust bacteria could provide an option for spreading life on other planets by directed
panspermia37.
Methods
Bacterial strains, growth conditions, and gamma irradiation. The following D. radiodurans
strains were used: R1 (ATCC 13939) wild type38, GY10922 (cinA-recA)1::kan39, and IRS501 polA
(J. R. Battista). A thymine-requiring (thy-) derivative of the R1 strain was isolated by selection on a
solid minimal medium containing thymine (50 g/ml) and trimethoprim (100 g/ml)40.
Bacteria were grown in TGY broth (0.5 % tryptone, 0.1 % glucose, 0.15 % yeast extract) at
30oC to the late exponential phase (OD650=0.6 - 0.8). Cultures were washed in 10 mM sodium
phosphate buffer, concentrated 10 times in the same buffer, and irradiated on ice with a
60
Co  ray
source at a dose rate of 11 Gy/s. A dose of 7 kGy was applied to the cells in all irradiation
experiments resulting in 80 to 90% survival. The number of viable cells was estimated by plating
serial dilutions onto TGY plates and the colonies were counted after 3-4 days incubation at 30oC.
Kinetics of DNA repair measured by pulsed-field gel electrophoresis. Irradiated cultures were
diluted in TGY to an OD650=0.2 and incubated at 30oC. At indicated intervals, 5-ml samples were
taken to prepare DNA plugs as described by Mattimore and Battista4. The DNA contained in the
plugs was digested with 60 units of NotI restriction enzyme (Roche) for 16 h at 37oC. After
15
digestion, the plugs were subjected to pulsed-field gel electrophoresis in 0.5xTBE using a CHEFDR III electrophoresis system (Bio-Rad) at 6 V/cm2 for 20 h at 14°C, with a linear pulse ramp of
50-90 s and a switching angle of 120o.
Rate of DNA synthesis measured by DNA pulse labelling. Unirradiated and irradiated
exponentially growing cultures were incubated and 0.5-ml samples taken and mixed with a 0.1 ml
pre-warmed TGY medium containing 6 Ci 3H-thymidine (Amersham; specific activity 86
Ci/mmol). Radioactive pulses were terminated after 15 min by addition of 2 ml ice-cold 10% TCA.
Samples were kept on ice for at least 1 h, and then collected by suction onto Whatman GF/C filters
followed by washing with 5% TCA and 96% ethanol. Filters were dried overnight at room
temperature, and placed in 5 ml scintillation liquid. The precipitated counts were measured in a
liquid scintillation counter (Wallac).
Radioactive and density DNA labelling. D. radiodurans thy- cells were radioactively and density
labelled during growth in the presence of 3H-thymidine and 5-bromo-2'-deoxyuridine (5-BrdU),
respectively. Density labelling was performed in all experiments by adding 5-BrdU to the medium
only after irradiation. The radioactive labelling was performed in three different regimens. (a) “Prelabelling”: cells were grown overnight in TGY supplemented with 5 Ci/ml 3H-thymidine. They
were collected by centrifugation, washed twice in the phosphate buffer, concentrated 10 times in
the same buffer, and exposed to 7 kGy  radiation. Both irradiated and unirradiated cultures were
diluted to an OD650=0.2 in TGY containing 20 g/ml 5-BrdU. The unirradiated culture was grown
in 5-BrdU-supplemented TGY for 2.5 h (corresponding to one mass-doubling), whereas the
irradiated culture was grown for 3 h (the time required for DSB repair to be completed) at 30 oC. (b)
“Post-labelling”: cells were grown overnight in non-radioactive TGY, and radioactivity (20 Ci/ml
3
H-thymidine) was added to the 5-BrdU-supplemented TGY only after irradiation. The
experimental procedure was otherwise the same as described for “pre-labelling”. (c) “Pre- and postlabelling”: the procedure was a combination of (a) and (b), i.e. the cells were radioactively labelled
16
before and after irradiation.
DNA preparation and density gradient analysis. Radioactively and density labelled DNA was
isolated from D. radiodurans by the use of the Qiagen DNeasy Tissue Kit according to the
supplier's instructions. The DNA was centrifuged to equilibrium in a neutral or alkaline cesium
chloride (CsCl) solution (1.7246 g/ml) in a VTi90 rotor (Beckman) for 24 h at 40.000 rpm and
20oC. To obtain the desired CsCl concentration, the refractive index of CsCl solution was adjusted
(by adding water) to 1.4030 for neutral gradients, and to 1.4050 for alkaline gradients. For alkaline
gradients, the DNA was denatured by 10 min heating in boiling water, followed by chilling in ice
water, and the CsCl solution was adjusted to pH 11.8. Gradients were collected from the bottom of
pierced tubes (OptiSeal 4.9 ml, Beckman) in about 25 (12-drop) fractions. 100-l aliquots of
fractions were applied on round filters, dried under the lamp for several hours, and the radioactivity
was measured in a scintillation counter.
UV-induced photolysis of BU-substituted DNA. D. radiodurans thy- culture was grown and
irradiated with 7 kGy  radiation as described above. The irradiated culture was diluted to an
OD650=0.2 and grown in 5-BrdU-supplemented TGY for 3 h. The cells were collected by
centrifugation, resuspended in the phosphate buffer and incubated (starved) in the buffer for one
hour at 30oC. Cell suspension was cooled in ice and exposed in a thin layer to indicated doses of
254-nm UV light. Both UV-irradiated and unirradiated cells were embedded in agarose plugs for
DNA analysis by PFGE (see above).
Immunofluorescent microscopy of cellular DNA synthesis. Exponentially grown D. radiodurans
thy- culture was harvested by centrifugation (6000 g), concentrated 10 times in a phosphate buffer
(66.7 mM, pH 6.8), and irradiated with 7 kGy  rays. Irradiated and unirradiated cells were diluted
to OD650=0.2 in TGY and incubated at 30oC. At different post-irradiation time points cells were
pulse-labelled with 5-BrdU (20 μg/ml; Sigma) for 10 min and fixed with 75% methanol at 4oC
overnight. Fixed cells were washed with PBSTE (PBS with 10 mM Tris-HCl and 1 mM EDTA;
17
wash buffer) and treated with lysozyme (2 mg/ml; Sigma) for 10 min. Cells were then washed
twice and blocked with 2% w/v bovine serum albumine (Sigma) in PBSTE (PBSTE-BSA) for 15
min. Half of the cells were treated with 4 M HCl for 1 h at 37 oC and washed twice (denatured
samples). Detection of 5-BrdU was subsequently performed via indirect immunofluorescence. Cells
were incubated in anti-mouse anti-BrdU monoclonal antibody (Becton Dickinson) diluted in
PBSTE-BSA (4 μg/ml) for 1 h at room temperature with gentle shaking, washed twice and
incubated in goat anti-mouse IgG-FITC (Sigma) diluted in PBSTE-BSA (12 μg/ml) for 1 h at room
temperature with gentle shaking. Cells were then washed twice and mounted onto an agarose film
on a glass slide. Native and denatured cell samples were examined with the Zeiss Axioplan 2
fluorescence microscope. Phase contrast and fluorescence images were taken at 100 x
magnification. Image analysis was performed using Metamorph software by measuring average
foci fluorescence per cell for 10 000 to 35 000 cells for each condition.
References
1. Potts, M. Desiccation tolerance: a simple process? Trends Microbiol. 9, 553-559 (2001).
2. Minton, K. W. DNA repair in the extremely radioresistant bacterium Deinococcus
radiodurans. Mol. Microbiol. 13, 9-15 (1994).
3. Battista, J. R., Earl, A. M. & Park, M. J. Why is Deinococcus radiodurans so resistant to
ionizing radiation? Trends Microbiol. 7, 362-365 (1999).
4. Mattimore, V. & Battista, J. R. Radioresistance of Deinococcus radiodurans: functions
necessary to survive ionizing radiation are also necessary to survive prolonged desiccation. J.
Bacteriol. 178, 633-637 (1996).
5. Sanders, S. W. & Maxcy, R. B. Isolation of radiation-resistant bacteria without exposure to
radiation. Appl. Environ. Microbiol. 38, 436-439 (1979).
18
6. Krasin, F. & Hutchinson, F. Repair of DNA double-strand breaks in Escherichia coli which
requires recA function and the presence of duplicate genome. J. Mol. Biol. 116, 81-98 (1977).
7. Fujimori, A. et al. Rad52 partially substitutes for Rad51 paralog XRCC3 in maintaining
chromosomal integrity in vertebrate cells. EMBO J. 20, 5513-5520 (2001).
8. Dean, C. J., Feldschreiber, P. & Lett, J. T. Repair of X-ray damage to the deoxyribonucleic
acid in Micrococcus radiodurans. Nature 209, 49-52 (1966).
9. White, O. et al. Genome sequence of the radioresistant bacterium Deinococcus radiodurans.
Science 286, 1571-1577 (1999).
10. Ferreira, A. C. et al. Characterization and radiation resistance of new isolates of Rubrobacter
radiotolerans and Rubrobacter xylanophilus. Extremophiles 3, 235-238 (1999).
11. Narumi, I. Unlocking radiation resistance mechanisms: still a long way to go. Trends
Microbiol. 11, 422-425 (2003).
12. Cox, M. M. & Battista, J. R. Deinococcus radiodurans – the consummate survivor. Nat. Rev.
Microbiol. 3, 882-892 (2005).
13. Harsojo, Kitayama, S. & Matsuyama, A. Genome multiplicity and radiation resistance in
Micrococcus radiodurans. J. Biochem. (Tokyo) 90, 877-880 (1981).
14. Levin-Zaidman S. et al. Ringlike structure of the Deinococcus radiodurans genome: a key to
radioresistance? Science 299, 254-256 (2003).
15. Zimmerman, J. M. & Battista, J. R. A ring-like nucleoid is not necessary for radioresistance in
Deinococcaceae. BMC Microbiol. 5, 17-27 (2005).
16. Lees-Miller, S. P. & Meek, K. Repair of DNA double strand breaks by non-homologous end
joining. Biochimie 85, 1161-1173 (2003).
17. Wilson, T. E., Topper, L. M. & Palbos, P. L. Non-homologous end-joining: bacteria join the
chromosome breakdance. Trends Biochem. Sci. 28, 62-66 (2003).
19
18. Kowalczykowski, S. C., Dixon, D. A., Eggleston, A. K., Lauder, S. D. & Rehrauer, W. M.
Biochemistry of homologous recombination in Escherichia coli. Microbiol. Rev. 58, 401-465
(1994).
19. Haber, J. E. Partners and pathways repairing a double-strand break. Trends Genet. 16, 259264 (2000).
20. Daly, M. J. & Minton, K. W. An alternative pathway of recombination of chromosomal
fragments precedes recA-dependent recombination in the radioresistant bacterium
Deinococcus radiodurans. J. Bacteriol. 178, 4461-4471 (1996).
21. Paques, F. & Haber, J. E. Multiple pathways of recombination induced by double-strand
breaks in Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. 63, 349-404 (1999).
22. Gloor, G. B. et al. Targeted gene replacement in Drosophila via P element-induced gap
repair. Science 253, 1110-1117 (1991).
23. Lederberg, J. Recombination mechanisms in bacteria. J. Cell Comp. Physiol. 45, 75-107
(1955).
24. Meselson, M. & Stahl, F. W. The replication of DNA. Cold Spring Harb. Symp. Quant. Biol.
23, 9-12 (1958).
25. Lion, M. B. Search for a mechanism for the increased sensitivity of 5-bromouracil-substituted
DNA to ultraviolet radiation. II. Single-strand breaks in the DNA of irradiated 5-bromouracilsubstituted T3 coliphage. Biochim. Biophys. Acta 209, 24-33 (1970).
26. Radman, M., Roller, A. & Errera, M. Protection and host cell repair of irradiated
bacteriophage lambda: III Ultraviolet irradiation of 5-bromouracil-substituted phage. Mol.
Gen. Genet. 104, 152-156 (1969).
27. Lohman, P. H. M., Bootsma, D. & Hey, A. H. The influence of 5-bromodeoxyuridine on the
induction of breaks in deoxyribonucleic acid of cultivated human cells by X-irradiation and
ultraviolet light. Radiat. Res. 52, 627-641 (1972).
20
28. Gutman, P. D., Fuchs, P. & Minton, K. W. Restoration of the DNA damage resistance of
Deinococcus radiodurans DNA polymerase mutants by Escherichia coli DNA polymerase I
and Klenow fragment. Mutat. Res. 314, 87-97 (1994).
29. Lin, J. et. al. Whole-genome shotgun optical mapping of Deinococcus radiodurans. Science
285, 1558-1562 (1999).
30. Milligan, J. R. et al. DNA strand-break yield after incubation with base excision repair
endonucleases implicate hydroxyl radical pairs in double-strand break formation. Int. J.
Radiat. Biol. 76, 1475-1483 (2000).
31. Lindahl, T. DNA repair enzymes. Annu. Rev. Biochem. 51, 61-87 (1982).
32. Makarova, K. S. et al. Genome of the extremely radiation-resistant bacterium Deinococcus
radiodurans viewed from the perspective of comparative genomics. Microbiol. Mol. Biol.
Rev. 65, 44-79 (2001).
33. Rayssiguier, C., Thaler, D. S. & Radman, M. The barrier to recombination between
Escherichia coli and Salmonella typhimurium is disrupted in mismatch-repair mutants. Nature
342, 396-401 (1989).
34. Mennecier, S., Coste, G., Servant, P., Bailone, A. & Sommer, S. Mismatch repair ensures
fidelity of replication and recombination in the radioresistant organism Deinococcus
radiodurans. Mol. Genet. Genomics 272, 460-469 (2004).
35. Rainey, F. A. et al. Extensive diversity of ionizing-radiation-resistant bacteria recovered from
Sonoran Desert soil and description of nine new species of the genus Deinococcus from a
single soil sample. Appl. Environ. Microbiol. 71, 5225-5235 (2005).
36. Harris, D. R. et al. Preserving genome integrity: the DdrA protein of Deinococcus
radiodurans R1. PloS Biol. 2, 1629-1639 (2004).
37. Crick, F. H. C. & Orgel, L. E. Directed panspermia. Icarus 19, 341-346 (1973).
21
38. Anderson, A. W., Nordon, H. C., Cain, R. F., Parrish, G. & Duggan, D. Studies on a radioresistant micrococcus. I. Isolation, morphology, cultural characteristics and resistance to
gamma radiation. Food. Technol. 10, 575-578 (1956).
39. Bonacossa de Almeida, C., Coste, G., Sommer, S. & Bailone, A. Quantification of RecA
protein in Deinococcus radiodurans reveals involvement of RecA, but not LexA, in its
regulation. Mol. Genet. Genomics 268, 28-41 (2002).
40. Little, J. G. & Hanawalt, P. C. Thymineless death and ultraviolet sensitivity in Micrococcus
radiodurans. J. Bacteriol. 113, 233-240 (1973).
Acknowledgements We thank John Battista for teaching us deinococcology and for the gift of the polA mutant, Matt
Meselson, Bob Wagner, Davor Zahradka, Avi Stark and members of the M.R. lab for reading this manuscript and/or
providing advice during this work, and Mikula Radman for the illustrations. K.Z. held a fellowship from the Necker
Institute during the work in M.R.’s laboratory. D.S. holds a Boehringer Ingelheim Foundation predoctoral fellowship.
The laboratory (1) was funded by INSERM and the Necker Institute (Mixis/PLIVA contract), (2) by EDF-France and
CNRS (GEOMEX), (3) by EDF-France and (4) by the Croatian Ministry of Science, Education and Sports.
Authors' contributions Experiments in Figs. 2, 4 and 5 were carried out by K.Z., those in Fig. 6 by D.S., and the
global experimental design was by M.R. Results from Fig. 2b and f were obtained in, and with the scientific and
material support of, laboratory (2), those from Fig. 2a and e in laboratory (3), those from Figs. 2c and d, 4 and 5 in
laboratory (4) and those from Fig. 6 in laboratory (1).
Correspondence and requests for materials should be addressed to M.R. (radman@necker.fr).
Legends to the Figures
Figure 1. Two selected strand annealing models for the assembly of small
chromosomal fragments into long linear intermediates. Several genomic copies of
22
exponentially growing D. radiodurans cells are each randomly broken into about 100
fragments. Following 5' to 3' end recession (1) diffused fragments can rejoin directly via
annealing of complementary single stranded overhangs of overlapping fragments
belonging to different chromosomal copies (step 4, inter-chromosomal SSA pathway; the
standard, deletion-prone intra-chromosomal SSA is not shown); the gaps are repaired by
synthesis and strand excess trimmed by nucleases (5). In the ESDSA pathway, the endrecessed fragments (1) prime synthesis templated on the homologous regions of other
partially overlapping fragments (2), presumably via a moving D-loop shown in the
bracketed intermediate. The strand extension can run to the end of the template producing
fragments with a 3’ single strand overhang extended by de novo synthesis (3). All
fragments can anneal under the constraints of their single-stranded homology. The
extended single strands can a priori engage in multiple rounds of extension (like in a
single round multiplex PCR with an “ends-out” growing primer) until their extended ends
find a complementary partner for SSA (4); these linear intermediates are trimmed and/or
gap-filled and finally ligated (5). The de novo strand extension synthesis is in red.
Hydrogen bonds are indicated only for inter-fragment associations. All data support the
ESDSA pathway for producing long linear DNA.
Figure 2. DNA repair and synthesis following gamma irradiation of D. radiodurans.
a, c, e show the kinetics of DNA repair in wild type, polA and recA strains respectively. b,
d, f show, on the log scale, the respective rates of DNA synthesis. a, (C) the line shows
the PFGE NotI restriction pattern of DNA from unirradiated wild type cells, (0) immediately
after 7 kGy gamma irradiation (DNA fragments are about 20-30 kb) and from the irradiated
cells grown in rich medium for 1.5, 3 and 4.5 h. (S) shows the S. cerevisiae chromosomes
as molecular markers. b, Incorporation of radioactive 3H-thymidine during 15 min pulse-
23
labelling measures the global rate of DNA synthesis in irradiated (red line) and nonirradiated (black line) cultures. c, The same experiment as in a for the polA mutant. d, The
same experiment as in b for the polA mutant. e, The same as in a and c for the recA
mutant. f, The same as in b and d for the recA mutant.
Figure 3. Maturation of circular chromosomes by homologous recombination.
Whereas RecA-independent ESDSA can produce long linear intermediates (Fig. 1),
closing the structure of the D. radiodurans genome (consisting of four circular elements)
by ESDSA, or any SSA mechanism, is highly unlikely. To mature into the unit size circular
chromosomes, the overlapping long linear fragments - or linear intermediates longer than
chromosome - require crossovers (X) via homologous recombination (HR). This step
requires RecA (Fig. 2).
Figure 4. Analysis of repaired D. radiodurans DNA by 5-BrdU density labelling. D.
radiodurans thy- cells were radioactively and density labelled by the growth in the
presence of 3H-thymidine and 5-BrdU, respectively. Density labelling was performed by
adding 5-BrdU to the medium only after irradiation. Genomic DNA was extracted and
analysed by CsCl equilibrium density centrifugation. HH, HL and LL refer to doublestranded heavy/heavy, heavy/light and light/light molecules. H and L refer to singlestranded heavy and light molecules. a, a1, 3H-thymidine «pre-labelled» DNA (labelled
before irradiation). b, b1, 3H-thymidine «post-labelled» DNA (labelled after irradiation). c,
c1, 3H-thymidine «pre- and post-labelled» DNA. a, b, c, Native DNA in neutral CsCl
gradients. a1, b1, c1, Denatured DNA in alkaline CsCl gradients. Open black circles: DNA
density profile from unirradiated cell culture grown for 2.5 h (one mass-doubling) in 5-
24
BrdU-supplemented medium. Closed red circles: DNA density profile from 7 kGy-irradiated
cells grown for 3 h in 5-BrdU-supplemented medium.
Figure 5. Photolysis of D. radiodurans DNA shattered by ionizing radiation and
repaired in the presence of 5-BrdU. a, PFGE patterns of NotI fragments of D.
radiodurans DNA (1-6) and the molecular weight standards of S. cerevisiae chromosomes
(S). DNA from: unirradiated cells (1); cells irradiated with 7 kGy gamma radiation (2);
irradiated cells incubated for 3 h in the standard medium before (3) and after (4) exposure
to 1000 J/m2 UV light; irradiated cells repaired in 5-BrdU containing medium before (5) and
after (6) exposure to 1000 J/m2 UV light. b,
Unlimited photolysis of fully 5-BrdU
substituted DNA (1-5) versus limited photolysis of DNA repaired in 5-BrdU (6-10). DNA
from: cells grown for 4.5 generations in 5-BrdU-supplemented medium before UV
irradiation (1) and after exposure to 100 J/m2 (2), 250 J/m2 (3), 500 J/m2 (4) and 1000 J/m2
(5); cells irradiated with 7 kGy gamma radiation and repaired in 5-BrdU-supplemented
medium (6) and then irradiated with 100 J/m 2 (7), 250 J/m2 (8), 500 J/m2 (9) and 1000
J/m2 (10) UV light.
Figure 6. Detection at single cell level of repair-associated newly synthesized
single-stranded DNA by immunofluorescent microscopy. D. radiodurans thyunirradiated (UNIR) and irradiated (IR; 7 kGy) cells were pulse-labelled with 5-BrdU at
different time points and revealed by immunofluorescence under native (a, c) and
denatured (b, d) conditions. Unirradiated cells were labelled in exponential growth phase.
Error bars represent standard errors. Asterisks denote a statistically significant difference
when compared to unirradiated exponential cells according to T-test (p<10-2). c, d,
25
Representative images (fluorescence images, adjusted to the same intensity scale, and
phase-contrast) of cells pre- (UNIR) and post-irradiation (IR).
26
radman_fig1
SSA
ESDSA
27
28
29
radman_fig4
30
31
radman_fig6
32