PNAS PLUS
Unreplicated DNA remaining from unperturbed S
phases passes through mitosis for resolution in
daughter cells
Alberto Morenoa,1, Jamie T. Carringtona,1, Luca Albergantea,b, Mohammed Al Mamuna,b, Emma J. Haagensena,2,
Eirini-Stavroula Komselic, Vassilis G. Gorgoulisc,d,e, Timothy J. Newmana,b, and J. Julian Blowa,3
a
School of Life Sciences, University of Dundee, Dundee DD1 5EH, United Kingdom; bSchool of Science and Engineering, University of Dundee, Dundee DD1
4HN, United Kingdom; cDepartment of Histology and Embryology, School of Medicine, University of Athens, GR-11527 Athens, Greece; dBiomedical
Research Foundation of the Academy of Athens, GR-11527 Athens, Greece; and eFaculty Institute of Cancer Sciences, Manchester Academic Health Science
Centre, University of Manchester, Manchester M20 4QL, United Kingdom
To prevent rereplication of genomic segments, the eukaryotic cell
cycle is divided into two nonoverlapping phases. During late mitosis
and G1 replication origins are “licensed” by loading MCM2-7 double
hexamers and during S phase licensed replication origins activate
to initiate bidirectional replication forks. Replication forks can stall
irreversibly, and if two converging forks stall with no intervening
licensed origin—a “double fork stall” (DFS)—replication cannot be
completed by conventional means. We previously showed how the
distribution of replication origins in yeasts promotes complete genome replication even in the presence of irreversible fork stalling.
This analysis predicts that DFSs are rare in yeasts but highly likely in
large mammalian genomes. Here we show that complementary
strand synthesis in early mitosis, ultrafine anaphase bridges, and
G1-specific p53-binding protein 1 (53BP1) nuclear bodies provide
a mechanism for resolving unreplicated DNA at DFSs in human cells.
When origin number was experimentally altered, the number of
these structures closely agreed with theoretical predictions of DFSs.
The 53BP1 is preferentially bound to larger replicons, where the probability of DFSs is higher. Loss of 53BP1 caused hypersensitivity to
licensing inhibition when replication origins were removed. These
results provide a striking convergence of experimental and theoretical evidence that unreplicated DNA can pass through mitosis
for resolution in the following cell cycle.
DNA replication
but capable of becoming active if necessary (5–9). We previously used mathematical analysis to show how the distribution
of replication origins in yeasts can be explained by the need for
complete genome replication in the presence of irreversible fork
stalling (4). Our theory predicts that organisms with significantly
larger genomes than yeast, such as those of mammals, will experience a much greater probability of replication failure genome-wide.
In this work, we provide evidence for a postreplicative mechanism that allows the resolution of these unreplicated segments of
DNA that involves segregation of template DNA strands during
mitosis by the creation of ultrafine anaphase bridges (UFBs) and
their recognition in the subsequent G1 phase by the DNA repair
protein p53-binding protein 1 (53BP1). We show that 53BP1 nuclear bodies correlate with the expected number of DFSs, both
when the number of replication origins is reduced and when the
number of replication origins is increased. We also show that 53BP1
preferentially associates with DNA in larger replicons, as predicted
Significance
We provide evidence that in organisms with gigabase-sized
genomes, such as humans, one or more stretches of DNA typically remain unreplicated when cells enter mitosis and are segregated to daughter cells via structures called ultrafine anaphase
bridges. p53-binding protein 1 (53BP1) accumulates at the subsequent DNA structures inherited by each daughter cell in the
following G1 phase to facilitate resolution in S phase. We show
that the abundance of these structures match theoretical predictions for the number of unreplicated DNA segments when the
number of replication origins is artificially increased or decreased.
We show that 53BP1 preferentially binds to chromosomal regions
with low numbers of replication origins. This work challenges the
prevailing view of how genome stability is maintained in proliferating cells.
| MCM | cell cycle | 53BP1 | UFB
D
uring the eukaryotic cell cycle, the genome must be precisely
duplicated with no sections left unreplicated and no sections
replicated more than once. To prevent rereplication, the process
is divided into two nonoverlapping phases: during late mitosis
and G1 replication origins are “licensed” for subsequent use by
loading MCM2-7 double hexamers, and during S phase DNAbound MCM2-7 is activated to form processive CMG (CDC45MCM-GINS) helicases that drive replication fork progression.
The prohibition of origin licensing during S phase and G2 ensures that rereplication of DNA cannot occur. However, the
inability to license new origins after the onset of S phase provides
a challenge for the cell to fully replicate the genome using its
finite supply of licensed origins. Replication forks can irreversibly stall when they encounter unusual structures on the DNA,
such as DNA damage or tightly bound protein–DNA complexes.
When replication initiation occurs at a licensed replication
origin the MCM2-7 double hexamer forms a pair of bidirectionally
orientated CMG helicases (1–3). If one fork irreversibly stalls, the
converging fork from a neighboring origin can compensate by
replicating all of the DNA up to the stalled fork. However, if two
converging forks both stall and there is no licensed origin between
them—a “double fork stall” (DFS)—new replicative machinery
cannot be recruited to replicate the intervening DNA (4). To
compensate for this potential for underreplication, origins are
licensed redundantly, with most (typically >70%) remaining dormant
www.pnas.org/cgi/doi/10.1073/pnas.1603252113
Author contributions: A.M. designed, performed, and optimized experiments on 53BP1
nuclear bodies, UFBs and immunofluorescence, and ChIP-Seq; J.T.C. performed experiments on 53BP1 nuclear bodies, RPA and γ-H2AX foci, mitotic EdU, and clonogenic assays
and performed the flow-cytometry experiments; L.A. performed the ChIP-Seq analysis;
L.A. and M.A.M. performed the mathematical and computational analyses; A.M. and E.J.H.
developed the 3D FACS protocol; E.-S.K. and V.G.G. made the HBEC cells; A.M. and J.T.C.
analyzed data; T.J.N. coordinated the theoretical work; J.J.B. led the project; and A.M., J.T.C.,
and J.J.B. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
1
A.M. and J.T.C. contributed equally to this work.
2
Present address: Northern Institute for Cancer Research, Medical School, Newcastle University, Newcastle upon Tyne NE2 4HH, United Kingdom.
3
To whom correspondence should be addressed. Email: [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1603252113/-/DCSupplemental.
PNAS | Published online August 11, 2016 | E5757–E5764
SYSTEMS BIOLOGY
Edited by James E. Cleaver, University of California, San Francisco, CA, and approved July 6, 2016 (received for review February 26, 2016)
A
3
Mean number of DFSs
Frequency x 103
4
3
2
1
0
wild-type
B
5
10
100
1000
10000
origins
added
2
1
0
1
origins
removed
Mathematical model
Computer model: IMR90
Computer model: HeLa
40
60
80
100
120
140
160
Replication Origins (%)
Inter−Origin Distances (kbp)
C
S/G2
mitosis
D
E
50
Frequency of cells (%)
12 μm
DAPI
53BP1
Untreated
S
G1
G1 Cells
Poisson Distr.
(λ = 0.95)
40
30
20
10
0
Aphidicolin (0.2 μM)
0
1
2
3
4
5
6
7
Number of 53BP1 bodies per cell
G
1
Mean 53BP1 bodies
Mean 53BP1 bodies
F
0.8
0.6
0.4
0.2
0
2
4
6
8
G1 (h after nocodazole release)
1.0
0.8
0.6
0.4
0.2
0
I
II
III
IV
V
S phase stage (EdU morphology)
Fig. 1. Potential mechanism for resolution of DFSs. (A) Distribution of replicon sizes in HeLa cells, based on data from ref. 13. The red bar represents the
average replicon size of ∼31 kb. (B) Mean number of DFSs predicted using a mathematical model (4), and a computational model that uses origin data from
HeLa and IMR-90 (13) when origins are added or depleted randomly. (C) Model for segregation of unreplicated DNA to daughter cells for resolution in the
next cell cycle. (D) The 53BP1 nuclear bodies in untreated and aphidicolin-treated cells. (E) Frequency of G1-specific 53BP1 nuclear bodies (n = 100, three
replicates). χ2 test for a fitted Poisson, P = 0.771. (F) Frequency of G1-specific 53BP1 nuclear bodies at the times indicated after nocodazole treatment and
mitotic shake-off (n = 150, three replicates, error bars are SEM). χ2 test, P = 0.924. (G) Frequency of 53BP1 nuclear bodies at different stages of the replication
timing program, as defined by O’Keefe et al. (32) (n = 150, three replicates, error bars are SEM). χ2 test, P = 4.998 × 10−4.
by the theoretical analysis. This experimental work strongly supports the theoretical analysis of DFSs in organisms of differing
genome size that we present in an accompanying paper (10).
Results
Refinements in technology have led to a convergence of origin
mapping data in mammalian tissue culture cells (11–13). Fig. 1A
shows the spacing between ∼90,000 replication origins (i.e., the
replicon sizes) in HeLa cells derived from the data of Picard
et al. (13). The average interorigin distance is ∼31 kb, consistent
with initiation events being ∼100 kb apart (11, 14, 15) and ∼30%
of origins being stochastically activated in any given S phase (6–9,
16, 17). Compared with yeast, human cells have an irregular
E5758 | www.pnas.org/cgi/doi/10.1073/pnas.1603252113
distribution of origins with an unexpectedly high number of very
large replicons (10). Using a mathematical approach that we
have previously derived and validated (4) we estimate that one
or two DFSs are expected to occur in every HeLa cell S phase
(Fig. 1B). Similar numbers were obtained when we performed
computational analyses based on the origin mapping data from
both HeLa and primary IMR90 cells (Fig. 1B). The predicted
number of DFSs increases when replication origins are removed
and decreases when they are added (Fig. 1B). Theoretical analysis
indicates that the distribution of origins in human cells is constrained to produce, on average, only a small number of DFSs and
therefore indicates that human cells should possess a postreplicative
mechanism capable of resolving these spontaneous events (10).
Moreno et al.
Mcm5
104
GAPDH
103
50k
20 50
D
1500
1000
500
0
32
48
64
3
2
1
Mcm5 RNAi
0
40
Total lysates
-
+
F
Dox
Tubulin
+
Dox
Chromatin
60
Frequency of cells (%)
Cdc6
-
Mcm5
Lamin B
origins
removed
origins
added
80
90
100
Non-induced
Cdc6 overexpression
40
Poisson distr. (λ = 0.95)
Poisson distr. (λ = 0.65)
30
20
10
0
1
2
3
4
5
1.0
I
***
1
0
Mathematical model
Computer model
HeLa: Mcm5 RNAi
HeLa: Cdt1 RNAi
HBEC: Cdc6
40
60
80
0.8
5
cm
s iC
on
2
tro
l
H
Mean 53BP1 bodies
Mean number of DFSs
70
Number of 53BP1 nuclear bodies per cell
wild-type
3
60
50
0
G
50
Replication Origins (% control)
Mcm5 RNAi (h)
E
100k
siM
Early S-phase Mcm2
2000
16
50k
DNA content
70
2500
Control
100k
DNA content
Mcm5 depletion (%)
C
104
103
102
LaminB
0
G2
G1
Mcm5
105
Mcm5
0.6
Tubulin
0.4
1.00
0.23
0.2
0
100
120
140
160
Replication Origins (%)
siControl
siMcm5
RNAi
Fig. 2. Origin number and the frequency of 53BP1 nuclear bodies. (A) Immunoblot of total and chromatin-bound MCM5 in HeLa cells after MCM5 RNAi.
(B) Three-dimensional FACS of HeLa cells labeled with EdU (Left) and MCM2 (Right). Red, G1 phase: EdU negative and G1 DNA content. Blue, early S phase:
incorporation of EdU without significant increase in total DNA content. Orange, late S phase: EdU positive cells with >G1 DNA content. Green, G2 phase: EdU
negative and G2 DNA content. (C) FACS of chromatin-associated MCM2 signal in early S phase HeLa cells with indicated periods of MCM5 RNAi. (D) Frequency
of G1-specific 53BP1 nuclear bodies (y-axis values) after MCM5 knockdown versus relative number of replication origins quantified by 3D FACS of DNA-bound
MCM2 (x-axis values). Each point represents the mean of 100 cells. (E and F) CDC6-inducible HBEC cells. Immunoblot of CDC6 and tubulin in whole-cell lysates
(E, Top) and MCM5 and Lamin B1 in chromatin samples (E, Bottom). Frequency distribution of 53BP1 nuclear bodies in HBEC cells (F) (n = 100, three replicates).
χ2 tests for fitted Poissons, P > 0.87. The two conditions are significantly different (Wilcoxon rank sum test, P = 2.843 × 10−6). (G) Compilation of the predicted
number of DFSs using the mathematical model and the computer simulation (from Fig. 1B) and the mean number of 53BP1 nuclear bodies in vivo (from D and
F and Fig. S2E). (H) Frequency of G1-specific 53BP1 nuclear bodies in control and MCM5 RNAi-treated IMR-90 cells (n = 150, three replicates, error bars are
SEM). t test, P = 1.79 × 10−4. (I) Immunoblot to show the depletion of MCM5 after RNAi. Quantification of band intensity is indicated below the blot.
Because a DFS will create a large segment of unreplicated
DNA, our analyses suggest that humans and metazoans in general with significantly larger genomes than yeasts have evolved
mechanisms to resolve them. Unreplicated or damaged DNA
may require topological unhooking for accurate segregation
during mitosis and the lesions generated by this process could
then be repaired in the following cell cycle (18, 19) (Fig. 1C).
Chromatid condensation during mitosis could provide a procMoreno et al.
essive unwinding activity to separate unreplicated DNA, leaving
single-stranded gaps that could then be partially filled in during
mitosis (20) or the following cell cycle. This mechanism depends
on the resolution and segregation of topologically intertwined
strands and could require a limited amount of DNA strand
cutting (19–21). We therefore predict that this mechanism might
be able to deal with only a small number of DFSs, as our theory
predicts (10).
PNAS | Published online August 11, 2016 | E5759
SYSTEMS BIOLOGY
Chromatin
20 40 50 80
Mcm5 depletion (%)
Mcm2 intensity
105
32 48 64
mid-late S
16
early S
32 48 64
EdU intensity
Total lysates
16
PNAS PLUS
B
Control RNAi (h) Mcm5 RNAi (h)
Mean number of 53BP1 bodies
A
B
DAPI
% 53BP1 bodies with RPA foci
A
53BP1
5 μm
RPA
40
***
30
20
10
0
Merge
siControl
siMcm5
RNAi
D
Normalised 53BP1 signal strength
Mean γ-H2AX foci
1.2
0.8
0.4
0
1.00
0.98
0.96
0.94
40
30
20
10
1
5
10 20
50 100 200 5001000
Replicon length (kbp)
F
10000
5000
2000
Number of Replicons
Replicon length (kbp)
1.02
siControl siMcm5
RNAi
E
50+
1.04
Number of replicons
C
500
200
50
20
53BP1− Replicons
53BP1+ Replicons
1000
100
10
53BP1−
53BP1+
Replicon type
1
6 -6
10 - 10
16 - 1
6
25 - 2
5
40 - 4
60 - 60
10 - 1 0
160 - 00
1
250 - 60
2
40 0 - 50
60 0 - 400
10 0 - 60
0
16 00 100
- 0
2500 160
4000 250 0
60 00 40 0
00 - 6 00
- 1 00
00 0
00
5
Replicon length (kbp)
Fig. 3. The 53BP1 is enriched at genomic loci that correspond to large replicons. (A) Representative image of the colocalization between G1-specific 53BP1
nuclear bodies and RPA foci. (B) Percentage of total cellular 53BP1 nuclear bodies that colocalize with RPA after treatment with control or MCM5 RNAi (n =
100, three replicates, error bars are SEM). t test, P = 3.01 × 10−3. (C) Mean frequency of G1-specific γ-H2AX foci in HeLa cells after MCM5 RNAi (n = 150, three
replicates, error bars are SEM). t test, P = 0.585. (D) Plot of the average 53BP1/IgG signal ratio per kilobase against replicon size. A strong and significant
correlation is observed (Spearman ρ = 0.91, P < 10−15). (E) Distribution of the size of 53BP1+ and 53BP1− replicons. t test, P < 10−15. (F) Frequency distribution
of 53BP1+ and 53BP1− replicons across different replicon sizes. χ2 test, P < 10−15.
Previous studies have suggested that 53BP1 recognizes these
aberrant structures in the cell cycle following underreplication.
The 53BP1 forms “nuclear bodies” in G1 phase that are symmetrically distributed between sister cells, possibly corresponding
to lesions generated by transmission of DFSs through mitosis in
the parent cell (18, 22). Consistent with this idea, the number of
53BP1 nuclear bodies increases in cells treated with replication
inhibitors (23–25) (Fig. 1D and Fig. S1 A and B), suggesting that
53BP1 can recognize structures resulting from defective DNA
replication. As our theory predicts (10), 53BP1 nuclear bodies in
normal G1 phase HeLa cells conform to a Poisson distribution
(suggesting that they are due to independent stochastic events)
with a mean close to the predicted number of DFSs (Fig. 1E
and Fig. S1 C and D). The number remains stable during G1
E5760 | www.pnas.org/cgi/doi/10.1073/pnas.1603252113
(Fig. 1F) but declines as cells progress through S phase, supporting the idea that they are resolved in a replication-dependent
manner (Fig. 1G and Fig. S1E).
To provide evidence for a link between DFSs and 53BP1
nuclear bodies, we depleted replication origins in HeLa cells
using RNAi against two components of the origin licensing system, MCM5 and CDT1 (6) (Fig. 2A and Fig. S2A). We then used
a 3D flow cytometry protocol, measuring DNA content, 5-ethynyl2′-deoxyuridine (EdU) incorporation, and chromatin-bound
MCM2 to determine the amount of chromatin-bound MCM2-7
in cells entering S phase (Fig. 2B). Because origins are only licensed for use if they are associated with MCM2-7, the amount of
DNA-bound MCM2 at the onset of S phase provides a measure
of the number of available origins. Depletion of MCM5 by RNAi
Moreno et al.
Moreno et al.
Discussion
In this work we present evidence that in unperturbed cell cycles
of human cells unreplicated DNA is frequently present at the
end of G2, is partially filled in during early mitosis, and is segregated during mitosis for resolution during the following cell
cycle. Our theoretical analysis (4, 10) suggests that in organisms
such as humans with gigabase-sized genomes DFSs will routinely
occur and create sections of unreplicated DNA that must be
resolved by a postreplicative mechanism. The symmetrical distribution of 53BP1 nuclear bodies between daughter cells and
their induction by replicative stresses means that they could mark
the products of unreplicated DNA segregated to daughter cells.
We show that when replication origins are deleted or added
there is a strong correlation between the number of 53BP1 nuclear bodies and our theoretical predictions of DFSs as presented in our accompanying paper (10). We show that 53BP1
preferentially associates with larger replicons, in line with our
theoretical predictions of DFS distribution. We also provide
evidence for a mechanism for the processing of the unreplicated
PNAS | Published online August 11, 2016 | E5761
PNAS PLUS
chromosome fragile sites, and late-replicating DNA (Fig. S6 D
and E). Taken together, these analyses show that 53BP1 is more
likely to bind to DNA in large replicons, as predicted if 53BP1
recognizes DNA structures resulting from DFSs.
The recent discovery that aphidicolin-treated cells exhibit
EdU incorporation during early mitosis indicates that replication
stress causes damage that is resolved postreplicatively by DNA
repair synthesis (20). If unreplicated DNA is unwound during
mitosis, ssDNA will be exposed, thereby providing a template for
complementary strand synthesis. Consistent with this idea, MCM5
RNAi caused a significant increase in early-mitotic EdU foci (Fig. 4
A–C). This result, when combined with the colocalization of 53BP1
and RPA and lack of increased γ-H2AX foci (Fig. 3A), implies that
the foci of EdU incorporation during early mitosis represent sites
of DNA synthesis of unreplicated DNA (dashed lines in Fig. 1C).
This postreplicative mechanism may not be able to complete
replication of all of the unreplicated DNA generated by DFSs,
which may be hundreds of kilobases in size (10). It has been
suggested that UFBs, which contain single-stranded DNA, might
represent a mechanism for resolving partially replicated stretches
of DNA (20, 26–29) (dashed lines in Fig. 1C). Consistent with
this idea, we observed that in untreated HeLa cells the number
and distribution of UFBs closely matched the numbers of 53BP1
nuclear bodies. Further, the number of UFBs increased in line
with 53BP1 nuclear bodies when MCM2-7 was partially depleted. This is consistent with the idea that UFBs provide a
mechanism by which unreplicated DNA generated by DFSs is
transmitted through mitosis to daughter cells to become ssDNA
lesions coated with 53BP1 that form nuclear bodies (Fig. 4 D–F).
Because we predict that DFSs occur frequently in normal
cells, 53BP1 is likely to be performing a function in binding to the
products of DFSs in G1 phase. The 53BP1 is known to protect
damaged DNA from undergoing homologous recombination (30)
and could perform this function at DFSs, which may allow the
structures to be resolved by an alternative pathway in S phase. To
explore this idea further, we examined a possible synthetic interaction between loss of 53BP1 and an increase in DFSs created
by partial knockdown of MCM2-7. RNAi transfected cells were
treated with increasing concentrations of hydroxyurea (HU) before a colony assay was performed. Cells partially depleted for
MCM2-7 were hypersensitive to HU, due to their inability to use
dormant replication origins (6–8). The 53BP1-depleted cells
showed a sensitivity similar to that of control cells. However, cells
depleted of 53BP1 showed a highly synergistic sensitivity to HU
when combined with partial knockdown of MCM5 (Fig. 4 G and
H and Fig. S6G). This shows that although 53BP1 is not essential
it works together with dormant origins to protect cells from the
consequences of replication fork failure.
SYSTEMS BIOLOGY
reduced the amount of DNA-bound MCM2 at S phase entry
(Fig. 2C). Similar results were obtained with RNAi against the
licensing factor CDT1 (Fig. S2 A and B). In both cases, overall
EdU incorporation was not affected (Fig. S2C). In line with our
theoretical predictions, the number of 53BP1 nuclear bodies increased in proportion to the reduction in DNA-bound MCM2
(Fig. 2D and Fig. S2 D and E). Similar results were obtained in
U2OS cells treated with MCM5 RNAi (Fig. S2 F and G).
Our theory also predicts that if cells have higher than normal
numbers of licensed origins, the number of DFSs should reduce
(Fig. 1B). To increase DNA-bound MCM2-7 we used a human
bronchial epithelial cell line overexpressing the licensing protein
CDC6 (Fig. 2E). The number of 53BP1 nuclear bodies in these
hyperlicensed cells was reduced 30% compared with noninduced
controls (Fig. 2F and Fig. S3 A and B), in line with our model.
Fig. 2G combines all our data on the number of 53BP1 nuclear
bodies (Fig. 2 D and F and Fig. S2E) to show that there is excellent agreement between our theoretical predictions (Fig. 1B)
and the experimental data from cells with reduced or increased
numbers of licensed origins. Fig. 2 H and I shows that there is
also an increase in the frequency of 53BP1 nuclear bodies after
MCM5 RNAi treatment of primary IMR-90 cells. This suggests a
similar relationship between the amount of DNA-bound MCM2-7
and the number of 53BP1 nuclear bodies in both normal cells
(IMR-90) and cancer cells (HeLa and U2OS). Taken together,
our data provide strong support for the idea that failures of DNA
replication caused by spontaneous DFSs cause the appearance of
53BP1 nuclear bodies in the subsequent G1.
We next investigated the nature of the lesions marked by
53BP1 nuclear bodies in G1 cells. Our theory suggests that these
structures represent single-stranded or partially single-stranded
regions of DNA rather than double-strand DNA breaks (Fig. 1C).
We therefore investigated the colocalization of 53BP1 nuclear
bodies with the ssDNA binding protein RPA (replication protein
A). Fig. 3 A and B shows that in untreated cells ∼7% 53BP1
nuclear bodies were associated with measurable levels of RPA
(18), but partial depletion of MCM2-7 caused a large increase in
colocalization to >30%. The increase of RPA in 53BP1 nuclear
bodies in cells with a reduced origin number might reflect the
larger distance between stalled forks, and hence longer stretches
of unreplicated DNA that bind RPA. To rule out the possibility
that G1-specific 53BP1 nuclear bodies mark double-strand breaks
generated by synthetic reduction of licensed origins in the preceding S phase, we also quantified the frequency of G1-specifc
γ-H2AX foci in response to MCM5 RNAi (Fig. 3C and Fig. S4).
No significant increase of G1-specifc γ-H2AX foci was observed
between the control and cells depleted of MCM5 (t test, P = 0.59).
Although DFSs can occur at any region of the genome, theoretical analysis predicts that they are more likely in larger
replicons rather than in smaller ones (4). To test this, we performed chromatin immunoprecipitation with anti-53BP1 antibodies and sequenced the precipitated DNA (Fig. S5). Cell
fractionation and immunoblotting revealed that a majority of
53BP1 (∼75%) is associated with chromatin (Fig. S6B), and
quantification of GFP-53BP1 intensity revealed that only ∼1% of
53BP1 signal originates from nuclear bodies (Fig. S6A). This
means that the majority of DNA bound by 53BP1 is not associated with 53BP1 nuclear bodies. Consistent with this, the total
genomic coverage from 53BP1 and IgG precipitations were
comparable (Fig. S6C). However, the 53BP1/IgG binding ratio
showed a highly significant correlation between replicon size and
the strength of 53BP1 association (Fig. 3D and Fig. S5D). We
then identified 1-kb regions of the genome with a high 53BP1/IgG
ratio (P < 10−3); replicons were defined as 53BP1+ when they
contained one or more 53BP1-enriched regions and 53BP1−
otherwise. 53BP1+ replicons were on average approximately three
times larger than 53BP1− replicons (Fig. 3 D and E and Fig. S6F).
There was also a weak correlation between 53BP1 binding,
DAPI
EdU
C
***
4
5
cm
5
siM
on
tro
l
6
siC
B
10 μm
Mean mitotic EdU foci
A
Mcm5
3
Tubulin
2
1.00
0.23
1
0
siControl
siMcm5
RNAi
E
Mean number of UFBs
3μm
DAPI
BLM
F
*
2.5
Frequency of cells (%)
D
2.0
1.5
1.0
0.5
siControl
siMcm5
H
1
s
si iMc
53 m
BP 5 +
1
si
53
BP
5
l
tro
cm
on
M
si
C
si
53BP1
Mcm5
Tubulin
Relative Number of Colonies (log)
RNAi
RNAi
RNAi Control
RNAi Mcm5
Poisson distr.
RNAi Control ( =1.4)
30
Poisson distr.
RNAi Mcm5 ( =2)
20
10
0
0
G
40
0
1
2
3
4
5
6
7
8
9 10
Number of UFBs per cell
100
10
RNAi Control
RNAi Mcm5
RNAi 53BP1
RNAi Mcm5 + 53BP1
1
0
0.2 0.4 0.6 0.8 1.0 1.3 1.7 2.0
HU (μM)
Fig. 4. MCM5 RNAi effects on mitosis. (A) Representative image of early-mitotic HeLa cell with foci of EdU incorporation. (B) Quantification of foci of EdU
incorporation during prophase and prometaphase HeLa cells after MCM5 RNAi (n = 100, three replicates, error bars are SEM). t test, P = 3.43 × 10−8.
(C) Immunoblot to show depletion of MCM5 after MCM5 RNAi. Quantification of band intensity is indicated below the blot. (D) Representative image of UFBs
stained with BLM in an anaphase HeLa cell. (E) Frequency of UFBs after 48-h treatment with MCM5 RNAi (n = 75, three replicates, error bars are SEM). t test,
P = 0.0473. (F) Frequency distribution of UFBs (n = 100 cells, four replicates). χ2 tests for Poissons, P > 0.85. A significant difference was observed. Wilcoxon
rank sum test, P = 5.095 × 10−3. (G) Immunoblot showing the knockdown of MCM5 and 53BP1 by RNAi in HeLa cells. (H) Clonogenic assay after treatment, as
seen in G, with increasing HU (three replicates, error bars are SEM).
DNA between DFSs, involving complementary strand synthesis
occurring in early mitosis, the resolution of partially replicated
DNA via UFBs, and their association with 53BP1 in G1.
A recent paper has shown that when DNA replication is
inhibited the condensation of chromosomes during early mitosis is
associated with the appearance of focal sites of DNA synthesis
(20). Our theoretical work is based on the idea that, from the end
of G1 through to the end of metaphase, MCM2-7 cannot be
loaded onto DNA even if DFSs have occurred (4, 10). MCM2-7
forms the core of the replicative CMG helicase that unwinds DNA
at the replication fork, and so the essential problem for completing
replication at DFSs is to provide an alternative DNA unwinding
activity. The chromosome condensation that resolves sister chromatids during early mitosis could provide such an alternative DNA
unwinding activity. Once ssDNA is exposed, DNA polymerases
will perform complementary strand synthesis to substantially fill in
the gaps. Consistent with this we show that the frequency of foci of
DNA synthesis in early mitosis is in line with our predictions of the
number of DFSs and increases when origin number is reduced.
We imagine that this complementary strand synthesis, which is
not driven by a processive helicase, might not always fully complete DNA replication but may leave small gaps or lesions on the
DNA. UFBs represent a potential intermediate that could allow
such partially unreplicated DNA segments arising from DFSs to
E5762 | www.pnas.org/cgi/doi/10.1073/pnas.1603252113
be segregated to daughter cells (18–20). We show that the number
of UFBs in untreated cells increases in line with our predictions
when origin number is reduced.
Consistent with our theoretical model, we show that 53BP1 is
preferentially bound to DNA in larger replicons, where DFSs are
more likely to occur. Under normal circumstances, there is only a
low colocalization of RPA with 53BP1 nuclear bodies, but this increases markedly when origin number is reduced. We imagine that
normally a significant proportion of the remaining unreplicated
DNA can undergo complementary strand synthesis in early mitosis,
leaving little ssDNA to bind RPA in the subsequent G1. However,
the distance between stalled forks in DFSs will increase following a
reduction in origin number, and this should result in more ssDNA
remaining after progression through mitosis, as we demonstrate.
We also show that the increase in 53BP1 nuclear bodies in G1 is not
significantly associated with an increase of γ-H2AX foci after partial depletion of origins, suggesting that 53BP1 nuclear bodies are
not simply sites of double-stranded DNA breaks. The 53BP1 nuclear bodies are ultimately dispersed as the DNA replicates during
S phase, suggesting that the unusual DNA structures that they mark
are not fully resolved until another round of replication has occurred.
Finally, we show that 53BP1 synergizes with dormant origins
to protect genome integrity in the presence of replicative stress,
as evidenced by its hypersensitivity to HU when the number of
Moreno et al.
Materials and Methods
Cell Culture. HeLa and IMR-90 cells were obtained from the American Type
Culture Collection and used at a population doubling level lower than 30 and
20, respectively, and maintained in DMEM (41966; Invitrogen), supplemented
with 10% FBS (10270106; Invitrogen) and penicillin and streptomycin at 37 °C
in 5% CO2. HBEC-Cdc6-Tet-On (human bronchial epithelial cells) were grown
in keratinocyte serum-free medium (17005-075; Invitrogen) supplemented
with 50 μg/mL bovine pituitary extract and 5 ng/mL hEGF (17005-075; Invitrogen).
The HBEC cell line was developed as described in ref. 31. Briefly, immortalized HBECs were infected with PLVX-Tet-On with blasticidin resistance (3 μg/mL)
and PLVX-TRE-Cdc6 with zeocin resistance (12.5 μg/mL). Clones with robust
doxycycline-dependent induction (5 μg/mL) were selected.
RNAi and Transfections. siRNA duplexes were obtained from Thermo Fisher
Scientific, and the sequences were as follows: control: 5′-UAGCGACUAAACACAUCAA -3′; MCM5: 5′-GGAUCUGGCCAGCUUUGAU -3′; CDT1: SMARTPool
M-003248-02; and 53BP1: 5′- GAAGGACGGAGUACUAAUA-3′.
Transfection was performed with Lipofectamine RNAiMAX (Invitrogen).
Fifty nanomolar siRNA was mixed with the Lipofectamine in Opti-MEM
medium (Invitrogen). The mixture was added to 50–60% confluent cells in
antibiotic-free DMEM (Invitrogen). Cells were subjected to different times of
transfection to obtain variable reductions in protein level.
Immunoblotting and Antibodies. Immunoblotting was performed as previously
described (8). Western blotting was performed according to standard procedures. Extraction of the chromatin-bound fraction was performed by treatment
with CSK extraction buffer (10 mM Hepes, pH 7.4, 300 mM sucrose, 100 mM
NaCl, 3 mM MgCl2, and 0.5% Triton-X-100) for 10 min on ice. The pellet, containing chromatin-associated proteins, was processed for Western blotting. The
antibodies used were MCM5 (sc-136366; Santa Cruz), CDT1 (ab183478; Abcam),
tubulin (T6199; Sigma-Aldrich), 53BP1 (A300-272A; Bethyl), Lamin B1 (16048;
Abcam), GAPDH (ab9484; Abcam), and CDC6 (05-550; Millipore).
Immunofluorescence. Cells were seeded in six-well plates containing glass
coverslips. At the required times for each experiment they were fixed with 4%
(vol/vol) formaldehyde, permeabilized with 0.1% Triton X-100 in TBS, and
blocked with 0.5% fish skin gelatin (G-7765; Sigma) for 1 h. Cells were then
incubated with the relevant antibodies overnight at 4 °C and washed with
0.1% TBS-Tween before incubation with Alexa secondary antibodies (Invitrogen). Cells were incubated with DAPI (D9542; Sigma) and mounted using
Vectashield mounting medium (H-1000; Vector Laboratories). The antibodies
used were 53BP1 (NB-100-904; Novus), Cyclin A (ab16726; Abcam), BLM (sc7790; Santa Cruz), RPA (ab2175; Abcam), γ-H2AX (2577L; Cell Signaling), and
phospho-histone H3 (9701S; Cell Signaling). For incorporation of EdU during
early mitosis, asynchronous HeLa cells were incubated in 40 μM EdU (Invitrogen) for 30 min before fixation. To visualize incorporated EdU the cells
were incubated in Click-iT EdU reaction, following the manufacturer’s pro-
1. Evrin C, et al. (2009) A double-hexameric MCM2-7 complex is loaded onto origin DNA
during licensing of eukaryotic DNA replication. Proc Natl Acad Sci USA 106(48):
20240–20245.
2. Gambus A, Khoudoli GA, Jones RC, Blow JJ (2011) MCM2-7 form double hexamers at
licensed origins in Xenopus egg extract. J Biol Chem 286(13):11855–11864.
Moreno et al.
PNAS PLUS
tocol (Thermo Fisher Scientific). Cells in prophase and prometaphase were
identified by phospho-H3 antibody signal and DAPI morphology.
Image Acquisition and Analysis. Microscopy images were acquired using an
Olympus IX70 deltavision deconvolution microscope. An Olympus 63× oil
immersion objective was used, and images were captured using a CCD
camera. Data from microscopy experiments was analyzed using Volocity 3D
analysis software (PerkinElmer). The nucleus was outlined as the region of
interest, and lower intensity threshold was set to a number that indicated
the intranuclear background.
Three-Dimensional Flow Cytometry. Cells were incubated with 40 μM EdU
(Invitrogen) for 30 min before trypsinization and collection. Cells were
preextracted with CSK extraction buffer for 10 min on ice and then fixed in
2% (vol/vol) formaldehyde for 15 min. For MCM2 labeling, cells were permeabilized in ice-cold 70% (vol/vol) ethanol for 10 min and incubated for 1 h
with anti-BM28 primary antibody (1:500). After staining with AlexaFluor
488-labeled secondary antibody (Invitrogen) cells were washed and Click-it
EdU reaction was performed for 30 min. Finally, cells were treated with
propidium iodide (PI) solution (50 μg/mL PI, 50 μg/mL RNaseA, and 0.1%
Triton-X-100) and transferred to FACS tubes for analysis. Samples were acquired using a BD FACSCanto and the results analyzed using FlowJo software.
ChIP Sequencing. Cells were cross-linked for 30 min using 1.5 mM ethylene
glycol-bis(succinimidyl succinate) followed by 10 min with 1% formaldehyde.
Reactions were stopped with 2 M glycine and cells were resuspended in CSK
buffer for 10 min. Cells were treated with 5 μL micrococcal nuclease (2,000 U/μL)
for 10 min at 37 °C, neutralized with 2× RIPA [100 mM Tris·HCl, pH 7.4, 300 mM
NaCl, 2% (vol/vol) IGE-Pal CA-630, 0.5% Na deoxycholate, and 1 mM EDTA] and
incubated on ice for 10 min. Samples were then precleared with Protein A
Dynabeads for 1 h at 4 °C and then incubated with 7 μg anti-53BP1 antibody
(A300-272A; Bethyl) rotating overnight at 4 °C. DNA was eluted (1% SDS, 0.1 M
NaHCO3, and 0.1% Tween-20) and used for library preparation using the NEBNext ChIP-seq kit and sequenced on an Illumina HiSEq 2500 by Edinburgh Genomics. The raw sequence data were assessed, aligned, and combined using
R version 3.2.2, Rsubread version 1.20.2, and SAMtools version 1.2. Aligned reads
were analyzed using a script based on R version 3.2.2 and Rsamtools 1.22. The
quality assessment and a detailed description of the analysis pipeline are available in SI Materials and Methods. Files containing the aligned reads are available
at the European Nucleotide Archive (accession no. PRJEB12222) and the R script
used for the analysis is available as Dataset S1.
Mathematical and Computational Analysis. The mean number of DFSs was
computed with the formula
logð2Þ
K
X
Ng
Ni
−
log 1 + logð2Þ
,
Ns
N
s
i=1
where Ng indicates the genome size, NS indicates the median stalling distance, and the Ni ði = 1..kÞ indicate the length of the K replicons. Replication
origin (RO) depletion and augmentation experiments were performed by
randomly removing or increasing the number of ROs. More details on the
mathematical model used are described in refs. 4 and 10 and an extended
summary of the approach used is available in SI Materials and Methods.
Clonogenic Assay. Cells were transfected in 10-cm dishes and replated into sixwell dishes before treatment. HU was added to cells for 48 h before medium
was replaced with fresh growth medium. After 10 d, cells were washed, fixed, and
stained with Crystal Violet. The number of colonies >1 mm were recorded. For
each genotype, cell viability of untreated cells was defined as 100%.
ACKNOWLEDGMENTS. This work was supported by Wellcome Trust Grants
WT096598MA and 097945/B/11/Z; Greek General Secretariat for Research
and Technology Program of Excellence II (Aristeia II) Grant 3020 (to V.G.G.);
the Dundee Imaging Facility, supported by Wellcome Trust Award 097945/B/
11/Z and Medical Research Council Award MR/K015869/1; and the Flow Cytometry and Cell Sorting Facility at the University of Dundee. V.G.G. and E.-S.K.
received an Experimental Research Center Elpen Scholarship and National
Scholarships Foundation-Siemens Aristeia Fellowship.
3. Remus D, et al. (2009) Concerted loading of Mcm2-7 double hexamers around DNA
during DNA replication origin licensing. Cell 139(4):719–730.
4. Newman TJ, Mamun MA, Nieduszynski CA, Blow JJ (2013) Replisome stall events have
shaped the distribution of replication origins in the genomes of yeasts. Nucleic Acids
Res 41(21):9705–9718.
PNAS | Published online August 11, 2016 | E5763
SYSTEMS BIOLOGY
dormant origins was reduced. The 53BP1 gene (TP53BP1) is not
essential and has been associated with protecting damaged DNA
from undergoing inappropriate homologous recombination (30).
The synthetic interaction that we show between loss of 53BP1
and partial MCM knockdown could therefore be a consequence
of unreplicated DNA undergoing inappropriate homologous
recombination during S phase.
In the accompanying paper (10) we provide a theoretical analysis
of origin distribution that leads to the conclusion that DFSs are
almost inevitable in the large genomes of human cells. The experimental work presented here provides a potential mechanism
by which DFSs can be processed, involving partial filling-in of
unreplicated segments during early mitosis, segregation to daughter cells via UFBs, their association with 53BP1 nuclear bodies
during G1, and their ultimate resolution during the next S phase.
Our work therefore provides both experimental and theoretical
evidence that structures resulting from replication failure can pass
through mitosis for resolution in the next cell cycle.
5. Woodward AM, et al. (2006) Excess Mcm2-7 license dormant origins of replication
that can be used under conditions of replicative stress. J Cell Biol 173(5):673–683.
6. Ge XQ, Jackson DA, Blow JJ (2007) Dormant origins licensed by excess Mcm2-7 are
required for human cells to survive replicative stress. Genes Dev 21(24):3331–3341.
7. Ibarra A, Schwob E, Méndez J (2008) Excess MCM proteins protect human cells from
replicative stress by licensing backup origins of replication. Proc Natl Acad Sci USA
105(26):8956–8961.
8. Ge XQ, Blow JJ (2010) Chk1 inhibits replication factory activation but allows dormant
origin firing in existing factories. J Cell Biol 191(7):1285–1297.
9. Blow JJ, Ge XQ, Jackson DA (2011) How dormant origins promote complete genome
replication. Trends Biochem Sci 36(8):405–414.
10. Mamun AM, et al. (2016) Inevitability and containment of replication errors for eukaryotic
genome lengths spanning megabase to gigabase. Proc Natl Acad Sci USA 113:E5765–5774.
11. Besnard E, et al. (2012) Unraveling cell type-specific and reprogrammable human
replication origin signatures associated with G-quadruplex consensus motifs. Nat
Struct Mol Biol 19(8):837–844.
12. Mesner LD, et al. (2013) Bubble-seq analysis of the human genome reveals distinct
chromatin-mediated mechanisms for regulating early- and late-firing origins.
Genome Res 23(11):1774–1788.
13. Picard F, et al. (2014) The spatiotemporal program of DNA replication is associated
with specific combinations of chromatin marks in human cells. PLoS Genet 10(5):
e1004282.
14. Jackson DA, Pombo A (1998) Replicon clusters are stable units of chromosome
structure: Evidence that nuclear organization contributes to the efficient activation
and propagation of S phase in human cells. J Cell Biol 140(6):1285–1295.
15. Conti C, et al. (2007) Replication fork velocities at adjacent replication origins are
coordinately modified during DNA replication in human cells. Mol Biol Cell 18(8):
3059–3067.
16. Blow JJ, Ge XQ (2009) A model for DNA replication showing how dormant origins
safeguard against replication fork failure. EMBO Rep 10(4):406–412.
17. Kunnev D, et al. (2015) Effect of minichromosome maintenance protein 2 deficiency
on the locations of DNA replication origins. Genome Res 25(4):558–569.
18. Lukas C, et al. (2011) 53BP1 nuclear bodies form around DNA lesions generated by
mitotic transmission of chromosomes under replication stress. Nat Cell Biol 13(3):
243–253.
19. Naim V, Wilhelm T, Debatisse M, Rosselli F (2013) ERCC1 and MUS81-EME1 promote
sister chromatid separation by processing late replication intermediates at common
fragile sites during mitosis. Nat Cell Biol 15(8):1008–1015.
E5764 | www.pnas.org/cgi/doi/10.1073/pnas.1603252113
20. Minocherhomji S, et al. (2015) Replication stress activates DNA repair synthesis in
mitosis. Nature 528(7581):286–290.
21. Ying S, et al. (2013) MUS81 promotes common fragile site expression. Nat Cell Biol
15(8):1001–1007.
22. Harrigan JA, et al. (2011) Replication stress induces 53BP1-containing OPT domains in
G1 cells. J Cell Biol 193(1):97–108.
23. Rappold I, Iwabuchi K, Date T, Chen J (2001) Tumor suppressor p53 binding protein
1 (53BP1) is involved in DNA damage-signaling pathways. J Cell Biol 153(3):613–620.
24. Anderson L, Henderson C, Adachi Y (2001) Phosphorylation and rapid relocalization
of 53BP1 to nuclear foci upon DNA damage. Mol Cell Biol 21(5):1719–1729.
25. Polo SE, Jackson SP (2011) Dynamics of DNA damage response proteins at DNA
breaks: a focus on protein modifications. Genes Dev 25(5):409–433.
26. Chan KL, North PS, Hickson ID (2007) BLM is required for faithful chromosome segregation and its localization defines a class of ultrafine anaphase bridges. EMBO J
26(14):3397–3409.
27. Chan KL, Hickson ID (2009) On the origins of ultra-fine anaphase bridges. Cell Cycle
8(19):3065–3066.
28. Barefield C, Karlseder J (2012) The BLM helicase contributes to telomere maintenance
through processing of late-replicating intermediate structures. Nucleic Acids Res
40(15):7358–7367.
29. Biebricher A, et al. (2013) PICH: A DNA translocase specially adapted for processing
anaphase bridge DNA. Mol Cell 51(5):691–701.
30. Bunting SF, et al. (2010) 53BP1 inhibits homologous recombination in Brca1-deficient
cells by blocking resection of DNA breaks. Cell 141(2):243–254.
31. Petrakis TG, et al. (2016) Exploring and exploiting the systemic effects of deregulated
replication licensing. Semin Cancer Biol 37-38:3–15.
32. O’Keefe RT, Henderson SC, Spector DL (1992) Dynamic organization of DNA replication in mammalian cell nuclei: Spatially and temporally defined replication of chromosome-specific alpha-satellite DNA sequences. J Cell Biol 116(5):1095–1110.
33. Morgan M, et al. (2009) ShortRead: A bioconductor package for input, quality assessment and exploration of high-throughput sequence data. Bioinformatics 25(19):
2607–2608.
34. Liao Y, Smyth GK, Shi W (2013) The Subread aligner: Fast, accurate and scalable read
mapping by seed-and-vote. Nucleic Acids Res 41(10):e108.
35. Weddington N, et al. (2008) ReplicationDomain: A visualization tool and comparative
database for genome-wide replication timing data. BMC Bioinformatics 9:530.
Moreno et al.
Supporting Information
Moreno et al. 10.1073/pnas.1603252113
SI Materials and Methods
RO Dataset. Genomic regions of HeLa cells associated with replication origins were obtained from the OriSeq project (13) using
the h19 human genome. Each genomic region was associated with
a single point on the genome, which was obtained by taking the
middle points of the genomic region under consideration. The
replication origin positions allowed us to divide each chromosome
into a set of replicons, defined as an area on a chromosome delimited on both sides by replication origins. The largest replicon of
each chromosome was removed from the analysis because these
replicons correspond to the centromeric areas, which are difficult
to study using sequencing experiments. When the replication
origin positions were used to compute the probability of DFSs, the
genome size was adjusted accordingly. These data were used in all
of the circumstances that required the positions of ROs.
Mathematical Model and Computational Simulations. The details of
the mathematical model used with an extensive analysis of its
theoretical consequences can be found in ref. 10 and only a
limited description will be presented here. In ref. 4 we derived a
simple model of DFS, using two key assumptions: that all of the
licensed ROs can be activated as necessary (therefore any time
constraint is excluded, a condition likely to be true for somatic
human cells) and that there is a constant small probability per
nucleotide for each individual replication fork to irreversibly stall.
In the same article (4), we showed that if all of the replicons are
significantly smaller than the median stalling distance and with a
distribution, characterized by a limited dispersion (conditions
compatible with experimentally derived replicon size distribution
in yeasts), the probability of at least one DFS in a given cell cycle is
Pðzero DFSÞ ≈ 1 −
ðlog 2Þ2 Nl Ng 1 + R2 ,
2
2
Ns
[S1]
where Nl is the average replicon size, Ng is the size of the genome, Ns is the median stalling distance, and R is the ratio of the
SD to the mean of the replicon size distribution.
Because in human cell lines a very large variation can be observed in replicon size and few replicons are as large as half of the
median stalling distance, the above approximation is inappropriate. Hence, we lifted this approximation and used an exact
approach (10). Under these new conditions, in a given genome
with K replicons Ni (with i = 1, . . ., K − 1) we have
!
K
X
Ng
Ni
Pðzero DFSÞ = exp −logð2Þ +
log 1 + logð2Þ
.
Ns
Ns
i=1
Computational experiments on RO depletion were performed by
randomly sampling a given percentage of ROs from the data for
each chromosome with the extremes excluded, and by computing
the mean number of DFSs using Eq. S2 and Eq. S4. Computational
experiments on RO augmentation were performed by inserting a
given percentage of new ROs in random positions on each chromosome. To bolster biological reality and for compatibility with
the RO dataset used, insertions that resulted in the creation of
replicons smaller than 4 kb were ignored. Each computational
experiment was repeated 100 times to account for stochastic variability. The SEM associated with the expected number of DFSs
was consistently lower than 0.1% and therefore was not reported.
ChIP Sequencing Experiments. Two technical replicas were performed for each of five ChIP experiments on 53BP1 and IgG
control. Each experiment was associated to a unique key (Fig.
S5A). Quality of the reads was assessed using the qa function of
the ShortRead R package (33). The analysis is based on a sample
of 1 million reads for each replica. Diagnostic plots (Fig. S5 B
and C) indicate a consistently good quality of the reads, and no
trimming was deemed necessary. Reads were aligned to the
human genome UCSC version h19, to ensure compatibility with
the replication origin dataset, using the align function of the
Rsubread package with default parameter values (34). Alignments containing indels were ignored. The number and percentage of aligned reads were consistently high, as indicated by
Fig. S5A. Because cells were preextracted using CSK treatment,
the aligned reads were tested for contamination from mitochondrial DNA. Contamination was detected in both technical
replicas of 53BP1 in Experiment 5, with an amount of mitochondrial DNA up to ∼100 times higher than other replicas. This
was considered an indication of potential problems in the experimental procedure, and both technical replicas were excluded
from further analysis. All of the remaining experiment and
technical replicas were combined into a single file containing
the 53BP1 and IgG reads (http://www.ebi.ac.uk/ena/data/view/
PRJEB12222; files ERX1269088 and ERX1269089, and contains
the control IgG and 53BP1 immunoprecipitates, respectively),
sorted, and indexed using SAMtools version 1.2.
Analysis of Binding Events. Aligned 53BP1 and IgG reads were associated with a unique replicon by considering the initial position of
the reads. Because the total number of reads was significantly different for the two antibodies, the number of reads for each replicon
was normalized by dividing it by the total number of chromosomal
reads of the antibody under consideration. For each replicon, the
ratio of the normalized number of reads was then considered. More
formally, for each replicon i, 53BP1 signal Sigi was computed as
[S2]
The distribution of DFSs can be well approximated by a Poisson
distribution, because in a Poisson distribution the probability of
zero events is
Pðzero eventsÞ = expð− λÞ,
[S3]
where λ is mean number of events. Then
λ = −logðPðzero eventsÞÞ.
[S4]
Combining Eq. S4 with Eq. S2 we can therefore calculate the
mean number of DFSs in human cell lines.
Moreno et al. www.pnas.org/cgi/content/short/1603252113
53BP1i
Sigi = P
j 53BP1j
!,
!
IgGi
P
,
j IgGj
where 53BP1i and IgGi are the numbers of 53BP1 and IgG
reads associated with replicon i, and the index j varies along all of
the replicons of the chromosome under consideration. To account
for the noise associated with the ChIP sequencing experimental
pipeline, the replicons were then grouped according to their size
and the mean of each group was computed. In the most populated
length group the mean was associated with limited errors; however,
very large and very small replicons are characterized by a much
larger uncertainty. Moreover, areas with limited mappability can be
1 of 6
frequently observed in large replicons, resulting in a weakened
53BP1 signal (Fig. S5D, which reports the same information of Fig.
3B with error bars indicating the SEM).
To assess 53BP1 enrichment, the genome was divided into
contiguous areas of 1 kb and for each area k, Sigk was computed
5
4
3
Frequency of cells (%)
EdU
CyclinA
Merge
1
0
Aphidicolin
D
100
G1 Phase Population
90
S Phase Population
40
80
G2 Phase Population
70
Poisson Distribution G1 Phase
60
50
40
30
20
10
0
0
1
2
3
4
5
6
7
Number of 53BP1 nuclear bodies per cell
E
% of cells
53BP1
2
Control
C
B
**
6
100
90
80
70
60
50
40
30
20
10
0
G1
T0
T2
S
T4
T6
Frequency of cells (%)
Mean 53BP1 nuclear bodies
A
as discussed above. For each chromosome, Sigk is approximately
log-normally distributed. It is therefore possible to use a fitted
log-normal distribution to associate, with each area k, a P value
that indicates the probability of observing by chance a value of
Sigk as large as observed or more.
Control
35
Aphidicolin
30
25
20
15
10
5
0
0
1
2
3
4
5
6
7
8
9 10 11 12
Number of 53BP1 nuclear bodies per cell
G2
T8 T10 T12
Time after Noc release (h)
Fig. S1. Dynamics of 53BP1 nuclear bodies. (A) Mean number of G1-specific 53BP1 nuclear bodies in untreated and aphidicolin-treated cells. Error bars are
SEM of three replicates. (B) Representative image of 53BP1 nuclear body identification at different stages of the cell cycle. Cells were labeled for 53BP1 (green),
EdU (orange), and Cyclin A (red). (C) Frequency of 53BP1 nuclear bodies at different stages of the cell cycle. Only G1-specific nuclear bodies fit a Poisson
distribution. (D) Frequency of G1-specific 53BP1 nuclear bodies in untreated and aphidicolin-treated cells. n = 100. (E) Frequency of cells at different stages of
the cell cycle after release from nocodazole arrest. Cells in G1 were assessed for the frequencies of 53BP1 nuclear bodies from T2–T8 (green area) and cells from
T8–T12 were used to identify the S-phase pattern (yellow area).
Moreno et al. www.pnas.org/cgi/content/short/1603252113
2 of 6
B
Cdt1 RNAi (h)
RNAi
Control 16
32
48
64
Early S-phase Mcm2
Total Lysates
A
Mcm5
Cdt1
*
GAPDH
5000
4000
3000
2000
1000
0
Control 16
32
48
64
Cdt1 RNAi (h)
EdU
C 10
Mcm5 RNAi (h)
5
Control
16
32
48
64
104
103
102
50k 100k
50k 100k
50k 100k
50k 100k
50k 100k
DNA Content
Cdt1 RNAi (h)
Control
16
50k 100k
50k 100k
32
48
64
104
103
102
50k 100k
40
40
RNAi Control
RNAi Control
20
15
10
5
25
20
15
10
5
0
0 1 2 3 4 5 6 7 8 9 10 11 12
0 1 2 3 4 5 6 7 8 9 10 11 12
Number of 53BP1 nuclear bodies
Number of 53BP1 nuclear bodies
E
Poisson distr. RNAi Cdt1 ( = 2.40)
3.0
2.5
2.0
1.5
1.0
0.5
Cdt1 RNAi
0
40 50
60
70
80
90 100
Replication Origin (%)
Mean 53BP1 nuclear bodies
F
3.5
G
3
***
5
25
Poisson distr. RNAi Control ( = 1.46)
30
cm
Poisson distr. RNAi Mcm5 ( = 2.68)
tro
l
30
siM
Poisson distr. RNAi Control ( = 1.46)
RNAi Cdt1 (50-60% origin deletion)
35
RNAi Mcm5 (50-60% origin deletion)
on
35
0
Mean 53BP1 nuclear bodies
50k 100k
siC
Number of cells (%)
D
50k 100k
DNA Content
Number of cells (%)
EdU
105
Mcm5
2
Tubulin
1.00
1
0
0.21
U2OS
siControl siMcm5
RNAi (U2OS)
Fig. S2. Quantification of origin reduction. (A) Western blot for replication origin depletion in HeLa cells using RNAi against Cdt1. Cells were transfected for
the indicated time before harvesting. (B) Representative FACS plot of chromatin-associated Mcm2 in early S-phase HeLa cells after Cdt1 depletion. (C) PI/EdU
FACS profiles for Mcm5 and Cdt1 depletion in HeLa cells. (D) Frequency distribution of 53BP1 nuclear bodies in Mcm5 RNAi-treated cells (Left) and Cdt1 RNAitreated cells (Right). (E) Mean number of G1-specific 53BP1 nuclear bodies in response to varying degrees of licensing knockdown as measured in B after Cdt1
RNAi. The 53BP1 nuclear bodies were quantified in triplicate in at least 100 cells. (F) Mean number of G1-specific 53BP1 nuclear bodies in U2OS cells after
treatment with control or MCM5 RNAi. Error bars are SEM of three replicates, P = 1.4 × 10−12. (G) Total U2OS cell extract after treatment with control or MCM5
RNAi, immunoblotted for MCM5.
Moreno et al. www.pnas.org/cgi/content/short/1603252113
3 of 6
B
**
- Dox
+ Dox
104
Mcm2
1
103
0.5
(30% RO addition)
105
0
EdU
Mean 53BP1 nuclear bocies
A
- Dox
+ Dox
104
103
102
50K
100K
50K
100K
DNA Content
Fig. S3. Cdc6 overexpression. (A) Mean number of 53BP1 nuclear bodies in uninduced and overexpression of Cdc6 in HBEC cells. Error bars are SEM of three
replicates, P = 0.00662. (B) FACS profiles of the distribution of chromatin-associated Mcm2 and the EdU incorporation of uninduced and Cdc6 overexpressing
HBEC cells.
DAPI
EdU/CyclinA
B
G1
5
Mcm5
S
Tubulin
1.00
γ-H2AX
s iM
cm
s iC
on
tro
l
A
0.27
Merge
10 μm
Fig. S4. G1-specific γ-H2AX foci. (A) Representative staining of γ-H2AX foci in asynchronous HeLa cells. Intense γ-H2AX foci were quantified in cells that
stained negatively for EdU and Cyclin A. (B) Transfected cells used for microscopy were tested for successful depletion by immunoblotting for Mcm5.
Quantification of Mcm5 reduction is indicated below the corresponding treatments.
Moreno et al. www.pnas.org/cgi/content/short/1603252113
4 of 6
A
Experiment
53BP1 Experiment 1 – Technical replica 1
53BP1 Experiment 1 – Technical replica 2
53BP1 Experiment 2 – Technical replica 1
53BP1 Experiment 2 – Technical replica 2
53BP1 Experiment 3 – Technical replica 1
53BP1 Experiment 3 – Technical replica 2
53BP1 Experiment 4 – Technical replica 1
53BP1 Experiment 4 – Technical replica 2
53BP1 Experiment 5 – Technical replica 1
53BP1 Experiment 5 – Technical replica 2
IgG Experiment 1 – Technical replica 1
IgG Experiment 1 – Technical replica 2
IgG Experiment 2 – Technical replica 1
IgG Experiment 2 – Technical replica 2
IgG Experiment 3 – Technical replica 1
IgG Experiment 3 – Technical replica 2
IgG Experiment 4 – Technical replica 1
IgG Experiment 4 – Technical replica 2
IgG Experiment 5 – Technical replica 1
IgG Experiment 5 – Technical replica 2
2.0
30
20
Quality Score
Reads
15’085’705
15’085’705
25’316’792
25’316’792
29’105’981
29’105’981
35’869’224
35’869’224
18’644’065
18’644’065
18’585’722
18’585’722
14’856’583
14’856’583
17’489’711
17’489’711
32’639’637
32’639’637
37’701’283
37’701’283
Proportion of reads
B
Key
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
10
9
10
30
20
10
5
6
7
8
30
20
10
1
2
3
10
30
20
10
7
8
30
20
10
2
1
2
3
4
5
0.5
0.0
2.0
1.5
1.0
0.5
102030
102030
102030
102030
3
6
7
8
9
10
1
2
3
4
5
2.0
1.5
1.0
0.5
0.0
2.0
1.5
1.0
0.5
0.0
4
0 50 100 0 50 100 0 50 100 0 50 100
Cycle
D
10
102030
Proportion of reads
Quality Score
10
1
9
Average (calibrated) base quality
20
6
8
0.0
4
30
5
7
1.0
Cycle
9
6
1.5
0 50 100 0 50 100 0 50 100 0 50 100
C
Aligned reads
13’431’878 (89.04 %)
13’289’715 (88.09 %)
22’624’95’5 (89.37 %)
22’427’05’5 (88.59 %)
25’570’708 (87.85 %)
25’320’996 (87.00 %)
31’837’763 (88.76 %)
31’546’518 (87.95 %)
16’632’221 (89.21 %)
16’468’113 (88.33 %)
16’502’266 (88.79 %)
16’368’731 (88.07 %)
13’250’134 (89.19 %)
13’130’364 (88.38 %)
15’158’814 (86.67 %)
15’002’940 (85.78%)
28’864’826 (88.43 %)
28’550’761 (87.47 %)
34’159’997 (90.61 %)
33’851’465 (89.79 %)
102030
102030
102030
102030
102030
Average (calibrated) base quality
Normalised signal strength
1.55
1.50
1.45
1.40
5
10
20
50 100 200
500 1000
Replicon length (kbp)
Fig. S5. Details of 53BP1 ChIP sequencing. (A) Experiment names, key, and number of reads. Details of the alignment can be found in SI Materials and
Methods. (B and C) ChIP sequencing quality analysis for 53BP1 (B) and IgG (C). Quality score with respect to the length of the reads (Left) and quality distribution (Right). (D) Plot of the average 53BP1/IgG signal ratio per kilobase against replicon size including SEM (error bars not reported when fewer than three
values are present).
Moreno et al. www.pnas.org/cgi/content/short/1603252113
5 of 6
B
10
9
C
***
10-1
108
107
CSK Soluble
106
53BP1
105
LaminB
104
Tubulin
103
102
101
% reads for 1kb regions
Sum intensity 53BP1 (a.u.)
A
53BP1
1.50
Replication Timing
Log2(Early/Late)
1.55
2
1
0
-1
10D
10E
10F
10G
11C
11D
11E
11F
11G
11H
12B
12E
13A
13C
13D
14B
14C
15A
16C
16D
17B
18A
18B
1A
1B
1D
1E
1F
1I 1G
1K
1L
20B
22B
2C
2D
2E
2F
2G
2H
2I
2J
3A
3B
3C
3D
4A
4C
4D
4F
5C
5D
5E
5F
6B
6C
6E
6F
6G
7B
7C
7D
7E
7F
7G
7H
7I
7J
8B
8C
8D
9B
9D
XB
XC
XD
-2
Fragile site ID
Genome wide
F
Density (x10-5)
E
Genome Wide
1.45
53BP1/IgG
10-5
10-7
Nuclear Nucleus
Bodies
D
10-3
Null
53BP1enriched
5
IgG
53BP1 - Replicons
53BP1 + Replicons
4
3
2
1
5
10
20
50 100 200
500
Replicon length (kbp)
Number of colonies
G
400
300
200
100
0
Control
Mcm5
53BP1
RNAi
Mcm5+
53BP1
Fig. S6. 53BP1 ChIP sequencing distributions. (A) Total intensity of 53BP1 in the nucleus and in nuclear bodies obtained by quantifying microscopy images of
cells expressing GFP-53BP1. (B) Immunoblot of 53BP1 in CSK-extracted pellets and CSK-solubilized fractions of HeLa cells. (C) Distribution of ChIP sequencing
reads for 53BP1 and control IgG across the entire human genome, with DNA grouped into 1-kb bins. (D) ChIP sequencing 53BP1/IgG ratio for common fragile
sites. (E) Mean replication timing computed for each 1-kb genomic region enriched in 53BP1 using timing data from Weddington et al. (35). The null distribution, using all of the values reported for HeLa, is plotted for comparison (Left). The distributions are significantly different (Wilcoxon signed rank test P <
10−10). (F) Frequency of 53BP1+ and 53BP1− replicons of different sizes (χ2 test, P = 5 × 10−4). (G) Plating efficiency of cells for the clonogenic assay. Error bars
are SEM of three independent experiments for untreated cells and for the four different genotypes used in the clonogenic assays.
Other Supporting Information Files
Dataset S1 (PDF)
Moreno et al. www.pnas.org/cgi/content/short/1603252113
6 of 6
# Clean memory ------------------------rm(list = ls())
# Load libraries ------------------------library("pastecs")
library("Rsamtools")
# Load replication origin distribution for Hela cells ------------------------#
# Data are available at http://pbil.univ-lyon1.fr/members/fpicard/research.html
#
# Set working directory
setwd("~")
# Load and process data
new.ori.Hela.u.2000.core.info.hg19 <- read.delim("new-ori-Hela-u-2000-core-infohg19.bed", stringsAsFactors=FALSE)
MIdPos <- rowMeans(new.ori.Hela.u.2000.core.info.hg19[,2:3])
ChrData <- split(MIdPos, new.ori.Hela.u.2000.core.info.hg19[,1])
for(i in 1: length(ChrData)){
ChrData[[i]] <- sort(ChrData[[i]])
}
ChrDist <- ChrData
for(i in 1: length(ChrDist)){
ChrDist[[i]] <- ChrDist[[i]][-1] - ChrDist[[i]][-length(ChrDist[[i]])]
}
get_distances <- function(Positions){
return(Positions[-1] - Positions[-length(Positions)])
}
ChrDistFil <- ChrDist
for(i in 1: length(ChrDistFil)){
ChrDistFil[[i]] <- ChrDistFil[[i]][-which.max(ChrDistFil[[i]])]
}
Prob.noStall <- function(N,G,S){
return(
exp( (-G/S)*log(2) + sum( log(1 + (N/S)*log(2)) ) )
)
}
ComputeStallProb <- function(x){
# Compute distances
if(length(x)<2){
return(NA)
}
x <- sort(unique(x))
Dists <- x[-1] - x[-length(x)]
# mean(Dists)
return(1-Prob.noStall(Dists, sum(Dists), 10^7))
}
ComputeMaxGap <- function(x){
# Compute distances
if(length(x)<2){
return(NA)
}
x <- sort(unique(x))
Dists <- x[-1] - x[-length(x)]
return(max(Dists))
}
# Load alignments for 53BP1 and IgG and compare with RO distribution
--------------#
# Processed data are saved to ReadVSRO.RData
#
AllRep <- NULL
AllIgG <- NULL
All53 <- NULL
ChrCovIgG <- NULL
ChrCov53 <- NULL
ClustersVect <- rep(1, length(names(ChrData)))
names(ClustersVect) <- names(ChrData)
for(ChrToCheck in names(ChrData)){
print(paste("Loading", Sys.time()))
print(ChrToCheck)
data.gr<-GRanges(seqnames=c(ChrToCheck), ranges = IRanges(start=0,
end=536870912))
which <- data.gr
params<-ScanBamParam(which=which, what = c("pos"))
Bam53BP1Pos <- scanBam("Sorted53BP1.bam", param = params)
Bam53BP1Pos <- unlist(Bam53BP1Pos)
print(paste("Loaded 53BP1", Sys.time()))
params<-ScanBamParam(which=which, what = c("pos"))
BamIgGPos <- scanBam("IgGFil1357.sorted.bam", param = params)
BamIgGPos <- unlist(BamIgGPos)
----------
print(paste("Loaded IgG", Sys.time()))
# Get the bin limits (RO and smallest/largest value for IgG and 53BP1)
ExtRoPos <- c(0, ChrData[[ChrToCheck]], max(max(max(BamIgGPos, Bam53BP1Pos)),
max(ChrData[[ChrToCheck]])))
BinnedIgG <- hist(BamIgGPos, breaks = ExtRoPos, plot = FALSE)$count
Binned53 <- hist(Bam53BP1Pos, breaks = ExtRoPos, plot = FALSE)$count
# Filter regions beyond the largest RO and before the smallest RO
FilRep <- get_distances(ExtRoPos)
FilBinIgG <- BinnedIgG
FilBin53 <- Binned53
RemIds <- c(1, length(FilRep))
FilRep <- FilRep[-RemIds]
FilBinIgG <- FilBinIgG[-RemIds]
FilBin53 <- FilBin53[-RemIds]
# Filter the centromeric area, determined by the largest replicon
RemIds <- which.max(FilRep)
FilRep <- FilRep[-RemIds]
FilBinIgG <- FilBinIgG[-RemIds]
FilBin53 <- FilBin53[-RemIds]
# Save the information on a genome-wide matrix
AllRep <- c(AllRep, FilRep)
AllIgG <- c(AllIgG, FilBinIgG)
All53 <- c(All53, FilBin53)
rm(BamIgGPos)
rm(Bam53BP1Pos)
}
# All53 contains the read count for 53BP1
# AllIgG contains the read count for IgG
# AllRep contains the replicon length
save(All53, AllIgG, AllRep, file = "ReadVSRO.RData")
# Load alignments for 53BP1 and IgG and perform enrichment analysis
-------------#
# Processed data are saved to Enrich.RData
#
ConIntList <- list()
-----------
ConIntVect <- NULL
OE53BP1Bins <- NULL
AllRep <- NULL
AllBottomsInROs <- NULL
AllTopsInROs <- NULL
RandTopsInROs <- NULL
Ro53OEStat <- NULL
library("pastecs")
library("Rsamtools")
ClustersVect <- rep(1, length(names(ChrData)))
names(ClustersVect) <- names(ChrData)
SummaryMat <- NULL
SummaryAreas <- NULL
# Look at each chromosome independently
# (this simplify the memory usage and account for potential ploidity problems)
#
for(ChrToCheck in names(ChrData)){
# Load data
print(paste("Loading", Sys.time()))
print(ChrToCheck)
data.gr<-GRanges(seqnames=c(ChrToCheck), ranges = IRanges(start=0,
end=536870912))
which <- data.gr
params<-ScanBamParam(which=which, what = c("pos"))
Bam53BP1Pos <- scanBam("Sorted53BP1.bam", param = params)
Bam53BP1Pos <- unlist(Bam53BP1Pos)
print(paste("Loaded 53BP1", Sys.time()))
params<-ScanBamParam(which=which, what = c("pos"))
BamIgGPos <- scanBam("IgGFil1357.sorted.bam", param = params)
BamIgGPos <- unlist(BamIgGPos)
print(paste("Loaded IgG", Sys.time()))
# Divide the data into blocks of 1kbp
print(paste("Computing histograms", Sys.time()))
BrkPos <- seq(from=0, to=max(BamIgGPos)+1e3, by=1e3)
HIgG <- hist(BamIgGPos, breaks = BrkPos, plot=FALSE)
H53BP1 <- hist(Bam53BP1Pos, breaks = BrkPos, plot=FALSE)
# For each bin, compute the normalised signal to noise ratio
SigRat <- (H53BP1$counts/sum(H53BP1$counts))/(HIgG$counts/sum(HIgG$counts))
# Look at the distribution of signal to noise
# mean(SigRat[!is.infinite(SigRat)], na.rm=TRUE)
#
# Approximate the distribution with a log-normal
LogVal <- log(SigRat[!is.infinite(SigRat) & !is.nan(SigRat) & !is.na(SigRat)])
LogSd <- sd(LogVal[!is.infinite(LogVal)], na.rm=TRUE)
# Obtain the pval for overexpression (OEPv) and underexpression (UEPv)
OEPv <- plnorm(q = SigRat, meanlog = 0, sdlog = LogSd, lower.tail = FALSE)
UEPv <- plnorm(q = SigRat, meanlog = 0, sdlog = LogSd, lower.tail = TRUE)
# Look at the number of ROs for each bin
HROs <- hist(ChrData[[ChrToCheck]], breaks = BrkPos, plot=FALSE)
# How does the strength of overexpression compare to the presence of RO?
# boxplot(OEPv ~ HROs$counts)
# TB <- table(OEPv<1e-3, HROs$counts)
# barplot(100*(t(TB)/colSums(TB))[,2], main = paste("53BP1 enrichment and RO",
ChrToCheck), xlab="Number of ROs",
# ylab="Percentage of genomic areas enriched in 53BP1")
#
#
#
#
the
#
Record the information collected
Code is rather slow.
Parallelization over different chromosomes could be implemented to speed up
analysis
CombMat <- NULL
for(ROdIdx in 2: length(ChrData[[ChrToCheck]])){
if(ROdIdx %% 100 ==0){
print(paste(ROdIdx-1, "of", length(ChrData[[ChrToCheck]]), "done!"))
}
SubIdx <- which(BrkPos < ChrData[[ChrToCheck]][ROdIdx] & BrkPos >
ChrData[[ChrToCheck]][ROdIdx-1])
tOEpV <- OEPv[SubIdx]
tUEpV <- UEPv[SubIdx]
tOEpV <- tOEpV[!is.na(tOEpV)]
tUEpV <- tUEpV[!is.na(tUEpV)]
tIgGCt <- HIgG$counts[SubIdx]
t53BCt <- H53BP1$counts[SubIdx]
CombMat <- rbind(CombMat, c(min(tOEpV), sum(tOEpV<1e-3), min(tUEpV),
sum(tUEpV<1e-3),
ChrData[[ChrToCheck]][ROdIdx],
ChrData[[ChrToCheck]][ROdIdx-1],
sum(tIgGCt==0), sum(t53BCt==0), sum(tIgGCt==0 &
t53BCt==0),
length(SubIdx)))
}
SummaryMat <- rbind(SummaryMat, cbind(CombMat, rep(ChrToCheck,
nrow(CombMat))))
SummaryAreas <- rbind(SummaryAreas, cbind(OEPv, UEPv, BrkPos[-1], BrkPos[length(BrkPos)], rep(ChrToCheck, length(OEPv)),
H53BP1$counts, HIgG$counts))
rm(BamIgGPos)
rm(Bam53BP1Pos)
}
colnames(SummaryMat) <- c("Min PV (OE)", "Nr PV<1e3 (OE)", "Min PV (UE)", "Nr
PV<1e3 (UE)", "Replicon End", "Replicon Start",
"0IgG count", "053BP1 count", "0IgG&53BP1 count",
"Area Count", "Chromosome")
# SummaryMat contains the summary statistics for each replicon
# SummaryAreas contains the summary statistics for each 1kb slice
save(SummaryMat, SummaryAreas, file = "Enrich.RData")