Alcohol-derived DNA crosslinks are repaired by two distinct mechanisms

Article abstract of DOI:10.1038/s41586-020-2059-5

Acetaldehyde is a highly reactive, DNA-damaging metabolite that is produced upon alcohol consumption¹. Impaired detoxification of acetaldehyde is common in the Asian population, and is associated with alcohol-related cancers^1,2. Cell

Full text of DOI:10.1038/s41586-020-2059-5

Article
Alcohol-derived DNA crosslinks are repaired
by two distinct mechanisms
https://doi.org/10.1038/s41586-020-2059-5
Received: 17 August 2018
Michael R. Hodskinson1,8, Alice Bolner2,8, Koichi Sato2, Ashley N. Kamimae-Lanning1,
Koos Rooijers2, Merlijn Witte2, Mohan Mahesh1,5, Jan Silhan1,6, Maya Petek1,7,
David M. Williams3, Jop Kind2, Jason W. Chin1, Ketan J. Patel1,4 & Puck Knipscheer
2
ꢀ✉
ꢀ✉
Accepted: 31 January 2020
Published online: xx xx xxxx
Check for updates
Acetaldehyde is a highly reactive, DNA-damaging metabolite that is produced upon
alcohol consumption1. Impaired detoxifcation ofacetaldehyde is common in the
Asian population, and is associated with alcohol-related cancers1,2. Cells are protected
against acetaldehyde-induced damage by DNA crosslink repair, which when impaired
causes Fanconi anaemia (FA), a disease resulting in failure to produce blood cells and a
predisposition to cancer3,4. The combined inactivation ofacetaldehyde detoxifcation
and the FA pathway induces mutation, accelerates malignancies and causes the rapid
attrition ofblood stem cells5–7. However, the nature ofthe DNA damage induced by
acetaldehyde and how this is repaired remains a key question. Here we generate
acetaldehyde-induced DNA interstrand crosslinks and determine their repair
mechanism in Xenopusegg extracts. We fnd that two replication-coupled pathways
repair these lesions. The frst is the FA pathway, which operates using excision—
analogous to the mechanism used to repair the interstrand crosslinks caused by the
chemotherapeutic agent cisplatin. However, the repair ofacetaldehyde-induced
crosslinks results in increased mutation frequency and an altered mutational
spectrum compared with the repair ofcisplatin-induced crosslinks. The second repair
mechanism requires replication fork convergence, but does not involve DNA
incisions—instead the acetaldehyde crosslink itselfis broken. The Y-family DNA
polymerase REV1 completes repair ofthe crosslink, culminating in a distinct
mutational spectrum. These results defne the repair pathways ofDNA interstrand
crosslinks caused by an endogenous and alcohol-derived metabolite, and identify an
excision-independent mechanism.
To study the repair ofalcohol-induced DNA damage, we generated an and has been comprehensively studied in Xenopus egg extracts11. To
acetaldehyde-crosslinked DNA substrate. Acetaldehyde reacts with examine the repair ofAANAT-ICLs using this system, the oligonucleotide
guanine to create a crosslink precursor, N2-propanoguanine (PdG)8 was ligated into a plasmid (pICL-AANAT). We also generated plasmids
(Fig. 1a). In a 5′-CpG sequence, PdG reacts with the N2-amine ofguanine containing a cisplatin interstrand crosslink (pICL-Pt) or PdG (pPdG),
on the opposite strand to create an acetaldehyde interstrand crosslink and unmodified control plasmids (pCon) (Extended Data Fig. 1g, h).
(AA-ICL); this crosslink exists in equilibrium between three states9. We Crosslinked vectors were stable in non-replicating Xenopusegg extracts
synthesized a site-specific native AA-ICL, denoted AANAT-ICL, within an (Extended Data Fig. 1i).
oligonucleotide duplex (Extended Data Fig. 1a, b, d, Supplementary
Fig. 1). A control reaction of PdG with deoxyinosine—which lacks an
N2-amine—did not crosslink, thus confirming the site-specificity ofthe
Acetaldehyde crosslinks are repaired by two routes
AANAT-ICL (Extended Data Fig. 1c, for gel source data see Supplementary Cisplatin ICLs are repaired by the replication-dependent FA pathway,
Fig. 2). AANAT-ICLs were stable at physiological pH and temperature, which involves unhooking of the ICL by endonucleases, translesion
with less than 10% reversal after 72 h at 37ꢀ°C (Extended Data Fig. 1e). synthesis (TLS) to bypass the adduct, and homologous recombina-
However, increased temperature (55ꢀ°C) or exposure to acidic condi- tion to resolve double-strand breaks12–14 (Fig. 1b, Extended Data
tions reversed AANAT-ICLs, which is consistent with Schiff-base hydroly- Fig. 2a). Replication of pICL-Pt in Xenopus egg extracts generates a
sis and protonation of the deoxyguanine (dG) N2-amine10 (Extended temporal pattern ofrepair intermediates, starting with converged forks
Data Fig. 1e, f). DNA crosslink repair is conserved among vertebrates (‘figure 8’ structure) and followed by low-mobility products that include
1MRC Laboratory of Molecular Biology, Cambridge, UK. 2Oncode Institute, Hubrecht Institute–KNAW and University Medical Center Utrecht, Utrecht, The Netherlands. 3Department of
Chemistry, The University of Sheffield, Sheffield, UK. 4MRC Weatherall Institute of Molecular Medicine, University of Oxford, John Radcliffe Hospital, Oxford, UK. 5Present address: Department
of Chemistry, Imperial College London, London, UK. 6Present address: Institute of Organic Chemistry and Biochemistry of the Czech Academy of Sciences, Prague, Czech Republic. 7Present
address: Department of Biochemistry, University of Cambridge, Cambridge, UK. 8These authors contributed equally: Michael R. Hodskinson, Alice Bolner. e-mail: kjp@mrc-lmb.cam.ac.uk; p.
✉
knipscheer@hubrecht.eu
Nature | www.nature.com | 1

Article
a
N
O
O
N
O
N
O
N
O
OH
H
O
2 ×
N
N
N
N
N
N
N
N
5′-CpG
NH
NH
OH HN
HN
NH
N
N
N
O
N
H
N
H
O
N
HN
NH
2
N
PdG
O
N
H
dG
N
N
N
O
O
AARED-ICL
N
N
N
N
N
H
N
N
N
NH
HN
HN
N
+
N
N
N
N
H
N
N
N
H
N
NaCNBH3
AANAT-ICL
H
RRI
Resolution
b
pICL
F8
OC/SC
pICL-AANAT
c
pCon
pPdG
pICL-Pt
Time
(min)
F8
RRI
OC
SC
Native agarose
pICL-Pt
pICL-AANAT
d
e
pICL
Time (min)
Replication/
repair
Backbone
NotI
5′
5′
ICL (88 nt)
44 nt
Linear (44 nt)
2 × 44 nt
ICL (88 nt)
NotI
NotI
*
*
44 nt linear
Quant 28 nt
Denaturing PAGE
Denaturing PAGE
f
g
h
30
20
10
0
30
30
pICL-AANAT
mock
pICL-AANAT
20
10
20
10
0
pICL-AANAT
ΔFANCD2
pICL-AANAT
+ p97i
pICL-Pt
mock
pICL-AARED
pICL-AARED
+ p97i
pICL-AANAT
pICL-Pt
0
pICL-Pt
ΔFANCD2
0
50
100 150 200
Time (min)
0
50
100 150 200
0
50 100 150 200 250
Time (min)
Time (min)
Fig. 1 | AA-ICL repair by Fanconi-dependent and Fanconi-independent
mechanisms. a, Reaction scheme of the formation of AANAT-ICL. Two
acetaldehyde molecules react with deoxyguanine (dG) to generate PdG; this
reacts with a 5′-CpG guanine on the opposite strand. The AANAT-ICL exists in an
equilibrium between three forms, shown in the dashed box. The crosslink can
be chemically reduced with sodium cyanoborohydride to form ICL-AARED (solid
box). b, Replication intermediates generated during ICL repair. c, Plasmids
were replicated in Xenopus egg extracts, and the reaction products were
resolved by electrophoresis and visualized by autoradiography. Figure 8
structures (F8), later RRIs, open circle (OC) and supercoiled (SC) products
(grey arrow) are indicated. Six independent experiments were performed. d,
Scheme for the NotI ICL repair assay. Wavy lines indicate the part of the strand
that is synthesized during repair. e, Plasmids were replicated in Xenopus egg
extracts, repair intermediates were isolated, digested by NotI, and resolved by
PAGE. Accumulation ofthe 44-nt product (white arrow) indicates replication/
repair. The asterisk marks a product that is probably generated from end-
joining activity in some extracts. Ten independent experiments were
performed. f, Quantification ofICL repair based on the gels in e, as described in
the Supplementary Methods. Ten independent experiments were performed.
g, Quantification ofrepair in mock-depleted or FANCD2-depleted
(∆FANCD2) extract, based on the gel in Extended Data Fig. 2e. Three
independent experiments were performed. h, Quantification ofrepair in the
presence or in the absence ofp97i, based on the gel in Extended Data Fig. 3c.
Three independent experiments were performed.
homologous recombination intermediates (termed replication/repair Before DNA replication (t = 0), this resulted in fragments of 88 nucle-
intermediates (RRI)) and resolved nicked and supercoiled products14 otides (nt) (2 × 44 nt, crosslinked), 44 nt (low-level background of
(Fig. 1b, c). Because the structures ofcisplatin crosslinks and acetalde- non-crosslinked plasmids) and the unresolved vector backbone. For
hyde crosslinks differ substantially (Extended Data Fig. 1j), we asked pICL-AANAT and pICL-Pt, the proportion ofthe 44-nt fragment increased
whether they were repaired by similar mechanisms. We replicated over time, confirming ICL repair (Fig. 1e). We quantified the repair
pICL-AANAT and pICL-Pt in Xenopus egg extracts, along with non- products and found that they accumulated in greater quantities for
crosslinked controls, and separated the products by electrophoresis pICL-AANAT compared with pICL-Pt (around 30% compared with around
(Fig. 1c). Nicked and supercoiled products accumulated rapidly in 20%, respectively, at 180 min) and, notably, pICL-AANAT displayed a
experiments with pCon and pPdG, indicating that there was little or faster rate ofrepair (11% compared with 1% at 50 min) (Fig. 1f, Extended
no impediment to replication. Experiments with pICL-AANAT resulted Data Fig. 1k, l). These data suggest that a proportion of pICL-AANAT
in RRI products similar to those seen with pICL-Pt; however, an earlier was processed in a similar manner to pICL-Pt, but that pICL-AANAT was
accumulation ofnicked and supercoiled products compared with pICL- additionally repaired by a second, faster mechanism.
Pt suggested that some pICL-AANAT were repaired quickly.
Because cisplatin ICLs are repaired by the FA pathway (Extended Data
We developed an assay, termed the ‘NotI assay’, to determine whether Fig. 2a), we examined whether the same pathway also repaired AANAT
-
the nicked and supercoiled products that we observed were indeed ICLs. In Xenopusegg extracts, both pICL-Pt and pICL-AANAT stimulated
AANAT-ICL repair products. Repair intermediates were digested using monoubiquitination of FANCD2—the activation step of the FA path-
NotI, labelled at the 3′ end and resolved by denaturing PAGE (Fig. 1d). way (Extended Data Fig. 2b). FANCD2 depletion causes a defect in the
2 | Nature | www.nature.com

a
b
30
20
10
0
d
On native agarose
X-structure
pICL-AANAT
HincII
HincII
pICL-Pt
pICL
pICL-AA + geminin
pICL-Pt + geminin
Replication
0
50 100 150 200
Time (min)
(1) Base excision
(2) Nucleotide excision
Linear
or
Linear
(5.6 kb)
EcoRI
LacR
AꢀIII
(5.6 kb)
Replication
pICL
-lacO
Rightward
fork
HincII/
APE1
–1
Arms
Arms
HincII
Leftward fork
–20
LacR
Long Short
AP
Long Short
(3.3 kb) (2.3 kb)
(3.3 kb) (2.3 kb)
AANAT
or
AANATrev
H
H
N
N
OH
dG
dG
dG
dG
e
O H
N
H
N
H
pICL-Pso
HincII/APEI
pICL-AANAT
HincII
HincII
HincII/APEI
c
pICL-AANAT-lacO
–LacR +LacR
pICL-AANATrev-lacO
–LacR +LacR
Time (min)
X-structure
Time
(min)
Ext.
–1
–1
–1
–1
Linear
Quant
Long arm
Short arm
Native agarose
pICL-AANAT
f
pICL-Pt
–
+
–
+
p97i
–20
Time (min)
Quant
Long arm
–20
Short arm
Native agarose
AꢀIII EcoRI digest
Fig. 2 | The alternative AA-ICL repair route requires DNA replication and fork
convergence but not DNA excision. a, Quantification of repair in the presence
or in the absence of geminin, based on the gel in Extended Data Fig. 4a. Three
independent experiments were performed. b, Schematic representation of
AA-ICL lacO plasmids, showing replication intermediate and digestion sites.
c, Plasmids were replicated in Xenopus egg extracts in the presence or in the
absence of LacR. Repair intermediates were digested and separated on a
sequencing gel. Grey arrows, −1 product; white arrows, −20 product. Two
independent experiments were performed. d, Scheme of the assays used to
detect base excision (left) and nucleotide excision (right) pathways in e and f,
respectively. DNA fragments formed after HincII/APE1 digestion (left box) or
HincII digestion (top and right box). e, Plasmids were replicated in Xenopus egg
extracts with p97i. Repair intermediates were digested and separated by
electrophoresis. Red arrows indicate the arms from APE1 incisions. The
quantification ofthese species is shown in Extended Data Fig. 5e. Four
independent experiments were performed. f, Plasmids were replicated in
Xenopus egg extracts with or without p97i. Repair intermediates were digested
and separated by electrophoresis. White arrows indicate the arms from
backbone incisions. The quantification ofthese species is shown in Extended
Data Fig. 5h. Six independent experiments were performed.
unhooking ofcisplatin ICLs, which results in the persistence ofnascent Data Figs. 1f, 3a). Replication ofa plasmid containing an AARED-ICL (pICL-
leading strands one nucleotide before the ICL12 (−1 position, Extended AARED) in Xenopus egg extracts yielded RRIs that resembled those of
Data Fig. 2a). To visualize this, we replicated plasmids in mock and the FA pathway, and there was very little accumulation of nicked and
FANCD2-depleted extracts, digested the intermediates, and separated supercoiled products (Extended Data Fig. 3b). The NotI assay revealed
the products on a sequencing gel (Extended Data Fig. 2c, d). For both that repair ofpICL-AARED was slower than that ofpICL-AANAT, and it was
pICL-Pt and pICL-AANAT, FANCD2 depletion resulted in persisting −1 abolished by p97i (Fig. 1h, Extended Data Fig. 3c–e). AARED-ICL is there-
products (Extended Data Fig. 2d) and fewer extension products, indicat- fore repaired exclusively by the FA pathway, indicating that the alterna-
ing that the repair ofboth pICL-Pt and pICL-AANAT involves the FA path- tive repair route is restricted to the native AA-ICL.
way. We then questioned whether the second route ofAANAT-ICL repair
required the FA pathway. In NotI assays, FANCD2-depleted extracts did
not support pICL–Pt repair, but pICL-AANAT was still partially repaired
Fork convergence is required for AA-ICL repair
(Fig. 1g, Extended Data Fig. 2e–h). FA-dependent pICL-Pt repair requires To further characterize the second, faster mechanism of AANAT-ICL
unloading ofthe CMG helicase by the p97 segragase15. Consistent with repair, we first asked whether it was replication-dependent. The addi-
published results, a p97 segregase inhibitor (referred to as p97i) halted tion ofrecombinant geminin, which inhibits DNA replication, blocked
pICL-Pt repair (Extended Data Fig. 2i–m). By contrast, p97i blocked all repair ofboth pICL-Pt and pICL-AANAT (Fig. 2a, Extended Data Fig. 4a–c).
only a proportion of AANAT-ICL repair, while the faster route of repair Two replication forks must converge in order for pICL-Pt repair to ini-
was unaffected by this treatment (Extended Data Fig. 2i–m). Together, tiate, promoting the ubiquitination and unloading of CMG helicase.
these results indicate that pICL-AANAT repair proceeds through both an Although CMG unloading is not required for the second route ofAANAT
FA-dependent and an FA-independent mechanism. ICL repair, we questioned whether fork convergence was necessary.
Chemical reduction of AANAT-ICL by sodium cyanoborohydride We generated an AANAT-ICL plasmid containing a lacOarray that, when
results in a single, stable form of the acetaldehyde crosslink (AARED
bound by Lac repressor (LacR), blocks the replication fork16,17. Because
-
-
ICL), as confirmed by high-performance liquid chromatography (HPLC) the AANAT-ICL is non-symmetrical we generated two versions of this
analysis (not shown), resistance to hydrolysis, and matrix-assisted laser plasmid, with the ICL in either orientation with respect to the rightward
desorption/ionization (MALDI) mass spectrometry (Fig. 1a, Extended fork (pICL-AA-lacO and pICL-AAreverse-lacO, respectively, Fig. 2b).
Nature | www.nature.com | 3

Article
a
c
AꢀIII
AꢀIII
~150 nt
~550 nt
BamHI
b
Undepleted
100 25 10
Insertion defect
–1
Extension defect
~80 nt
5
100 100 Rel. vol.
Right-
–20
ward fork
Mock
Leftward fork
*
REV1
0
pICL-Pt
pICL-AANAT
6REV1
pICL-AARED
6REV1
Mock
Mock
6REV1
Time (min)
A
T G C
Ext.
–1/0
0
–1/0
–1/0
–1
–1
AꢀIII digest
0
–1
–1
AꢀIII BamHI digest
T
NA
T
A
N
Primer extension
A
t
i
AA
P
-
n
G
d
97
p
o
d
e
CL
CL-A
I
I
ICL-
Adduct on
+
p
pC
pP
p
p
top strand?
76 nt
Replication/
repair
220 nt
141 nt
Digestion
220 nt
PrimerBOT
PrimerTOP
pICL
52/54 nt
76 nt
141 nt
*
Adduct on
bottom strand?
Denaturing
PAGE
52/54 nt
Fig. 3 | AA-ICL repair is mediated by REV1. a, Scheme for the formation of
products detected on the sequencing gel in c. b, Western blot analysing REV1 in
the mock-depleted and the REV1-depleted (∆REV1) extract, compared with a
titration of undepleted extract. The asterisk indicates a background band. Nine
independent experiments were performed. c, Plasmids were replicated in
mock-depleted or REV1-depleted extracts and repair intermediates were
digested and separated on a sequencing gel alongside a sequencing ladder.
White arrows, 0 products; grey arrows, −1 products; open arrows, 0/−1
products. Three independent experiments were performed. d, Scheme ofthe
primer extension assay. Late repair products were digested and used as
template for primer extension from labelled TOP or BOT (bottom) primers.
e, Primer extension products were separated by denaturing PAGE. Four
independent experiments were performed.
We replicated these plasmids and separated the digested repair inter- role in AANAT-ICL repair, we complemented NEIL3-depleted extract with
mediates on a sequencing gel (Fig. 2b, c). In the absence ofLacR, both recombinant wild-type or catalytically inactive NEIL3 (Extended Data
leftward and rightward leading strands arrived at the −20 position, Fig. 5a). Unlike those of the psoralen-ICL plasmid (pICL-Pso)15, pICL-
approached the −1 position, and were extended past the lesion over AANAT repair intermediates were unaffected by the lack ofNEIL3 activ-
time. In the presence ofLacR, arrival ofthe leftward fork was inhibited ity (Extended Data Fig. 5b). Furthermore, human HAP1 cells in which
(as indicated by the absence of–1 and –20 products), showing that the NEIL3 was knocked out (∆NEIL3) were not hypersensitive to acetalde-
LacR block was functional. The rightward fork persisted at the –20 hyde, and NEIL3 deficiency did not further sensitize a FANCL-deficient
position, which suggests a failure in repair progression. Moreover, the strain (Extended Data Fig. 5c, d). It is plausible that another glycosylase
formation of extension products was impeded upon incubation with unhooks the AANAT-ICL, so we examined the accumulation of abasic
LacR, and this was similar for both orientations ofAANAT-ICL (Fig. 2c) and sites—the product ofglycosylase cleavage. Recombinant APE1 cleaves
pICL-Pt-lacO (Extended Data Fig. 4d). This indicates that the unhook- abasic sites resulting in the generation ofarm fragments upon lineari-
ing of AANAT-ICL requires replication fork convergence. Consistent zation ofreplication intermediates. Such arms were generated during
with this, in the presence of LacR, repair of pICL-AA-lacO and pICL- pICL-Pso repair, but they did not form during AANAT-ICL repair15 (route
AAreverse-lacOwas greatly reduced (Extended Data Fig. 4e–h). These 1, Fig. 2d, e, Extended Data Fig. 5e–g). This indicates that no abasic site
data show that the collision of one replication fork with an AANAT-ICL is formed and that AANAT-ICL repair does not cut the N-glycosyl bond.
is insufficient for repair. Although it is plausible that FA-independent
Next, we tested whether nucleotide excision products are formed
AANAT-ICL unhooking is a consequence oftwo forks colliding, we think when an AANAT-ICL is processed in Xenopusegg extract. Backbone inci-
that the mechanism is more likely to be enzymatic, as is the case for all sions should generate arm fragments when repair intermediates are
known ICL repair pathways.
linearized, as shown for repair of pICL-Pt (route 2, Fig. 2d, f). These
arms were also formed during pICL-AANAT repair, which is consistent
with FA pathway activity. Notably, when p97i was added—blocking
the FA repair route—no incision products formed, indicating that the
New repair route does not involve DNA excision
Whereas cisplatin ICLs are unhooked by nucleolytic incisions, pso- second ICL-AANAT repair pathway does not create backbone incisions
ralen and abasic site ICLs are preferentially cleaved at a glycosidic (Fig. 2f, Extended Data Fig. 5h–j). To further confirm this, we examined
bond by NEIL3 glycosylase12,15,18. To investigate whether NEIL3 has a the formation of adducts. In the FA pathway, incisions of one strand
4 | Nature | www.nature.com

a
pPdG
pICL-Pt
pICL-AANAT
FA pathway
FA pathway
Second pathway
ꢀ
ꢁ
ꢂ
ꢁ
ꢀ
ꢂ
ꢁ
ꢀ
ꢂ
ꢀ ꢁ ꢂ
&7&77&&*&7&77&7
*$*$$**&*$*$$*$
&7&77&&*&7&77&7
*$*$$**&*$*$$*$
&7&77&&*&7&77&7
*$*$$**&*$*$$*$
&7&77&&*&7&77&7
*$*$$**&*$*$$*$
Unhooking
and TLS
Unhooking
and TLS
Unhooking
and TLS
Replication
and TLS
Bottom strand
unhooked
Bottom strand
unhooked
Adduct on
top strand?
ꢁ
ꢀ
*
ꢁ
*
ꢀ
&7&77&&*&7&77&7
&7&77&&*&7&77&7
&7&77&&*&7&77&7
&7&77&&*&7&77&7
*$*$$**&*$*$$*$
*$*$$**11$*$$*$
*$*$$*11*$*$$*$
*$*$$*11*$*$$*$
TLS
TLS
TLS
&7&77&&117&77&7
TLS
&7&77&11&7&77&7
TLS
&7&77&11&7&77&7
&7&77&1&7&77&7
TLS
TLS
*$*$$**&*$*$$*$
*$*$$**&*$*$$*$
*$*$$**&*$*$$*$
*$*$$**&*$*$$*$
*
ꢂ
*
ꢀ
Top strand
unhooked
Top strand
unhooked
ꢀ
ꢀ
Adduct on
bottom strand?
Adduct on
bottom strand
b
pICL-AANAT
pICL-AANAT + p97i
pICL-Pt
pPdG
6
4
2
6
4
2
0
6
4
2
0
6
4
2
0
T
G
C
A
1.0
0.5
0
77&&*&7&7
0
&7&77&&*&7&77&7
&7&77&&*&7&77&7
10 15
Nucleotide
&7&77&&*&7&77&7
&7&77&&*&7&77&7
1
5
1
1
1
5
10
15
5
10
15
5
10
15
Nucleotide
Nucleotide
Nucleotide
Fig. 4 | AA-ICL repair causes point mutations. a, Scheme showing the
positions of DNA lesions and potential adducts (shown in red) generated
during repair. Mutations are generated during translesion synthesis (blue Ns
15-bp region flanking the lesions. Strand specificity is lost because sample
preparation involves PCR amplification; as such, only the top sequence is
indicated below the graphs. Mutation frequency is corrected for mutations
found in pCon. Two independent experiments were performed.
and arrow). b, Distribution and frequency of nucleotide misincorporation in a
result in an unhooked adduct on the other (Extended Data Fig. 5k). is encoded by FANCV, also known as MAD2L2), had a very similar effect
Therefore, late repair intermediates ofpICL-Pt contain adducts on the on AA-ICL repair to that of REV1 depletion (Extended Data Fig. 6d). It
top or the bottom strand, because either can be incised. Repair ofpICL- also caused an insertion defect in the FA pathway during AARED-ICL
AANAT also generated adducts, but they were largely restricted to the repair, as well as persistent stalling—especially ofthe rightward fork—in
top stand—it is possible that adducts on the bottom strand either were AANAT-ICL repair. As Pol ζ is primarily known for its extension activity20
,
present in amounts below the detection limit or were processed in the this defect could be due to co-depletion of the REV1–Pol ζ complex19
extract (Extended Data Fig. 5k, right). However, after the addition of (Extended Data Fig. 6b, c). In summary, the second pathway of AANAT
-
p97i no adducts were detected, indicating that they depend on the FA ICL repair generates an adduct on the bottom strand that requires
pathway. In conclusion, the second, faster repair route for AANAT-ICLs REV1 and Pol ζ for bypass. The top strand, however, is readily extended
does not involve a DNA excision step. This route must therefore operate without these TLS factors.
by cutting within the crosslink itself.
To further examine adduct formation, we isolated late repair prod-
ucts and subjected them to primer extension reactions at the ICL region
using a high-fidelity polymerase (Fig. 3d). As expected, pICL-Pt repair
products—which contain adducts on either parental strand—induced
Role of REV1 in acetaldehyde crosslink repair
Such a repair pathway would create an adduct on one or both strands, the formation ofstalled extension products on both strands (Fig. 3e).
and should require TLS for nucleotide insertion opposite the adduct For AANAT-ICL, stalling also occurred on both strands but was more
and extension beyond it. The TLS factor REV1 is critical for pICL-Pt extensive on the bottom strand. Moreover, treatment with p97i—that
repair, and so we tested whether it also operates in AANAT-ICL repair. is, blocking the FA pathway—almost entirely eliminated stalling on
Plasmids were replicated in mock and in REV1-depleted extracts, and the top strand (Fig. 3e). These results suggest that the second repair
digested intermediates were analysed on a sequencing gel (Fig. 3a–c). pathway regenerates dG on the top strand, but creates a dG adduct on
For pICL-Pt, REV1 depletion led to the accumulation of insertion the bottom strand.
products (0 products) and reduced amounts of extension products19
(Fig. 3c). By contrast, for pICL-AANAT, REV1 depletion led to the accumu-
lation of –1 products, which is indicative of REV1-mediated insertion
Acetaldehyde crosslink repair is mutagenic
opposite the adduct (Fig. 3c, Extended Data Fig. 6a). Notably, this was Finally, we examined the fidelity ofAANAT-ICL repair. To achieve this, we
also the case for pICL-AARED repair, indicating that there are different replicated plasmids in Xenopusegg extracts, recovered late repair prod-
TLS mechanisms within the FA pathway. Furthermore, REV1 depletion ucts, and subjected them to high-throughput sequencing (Extended
caused more extensive leading-strand stalling at the rightward than at Data Table 1, Fig. 4a, b). Consistent with a previous report19, we con-
the leftward fork for AA-ICLNAT (Fig. 3c). This suggests that unhooking by firmed that pICL-Pt is repaired with a low error rate—less than 2% of
the second pathway may create an adduct on the bottom strand, which the products carried a mutation at sites corresponding to crosslinked
is bypassed by REV1. To test this, we inhibited the FA pathway using guanines (Fig. 4b, Extended Data Fig. 7a). The most common substitu
-
p97i and examined lesion bypass in the second pathway of AANAT-ICL tion was G>T transversion. Repaired pICL-AANAT products show two
repair. As expected, repair of pICL-Pt in the presence of p97i resulted differences: first, around 10% carry mutations at the crosslinked sites;
in persistent stalling at the –20 to –40 position, due to defective CMG and second, the mutational spectrum differs, with C>G, C>A, G>C and
unloading15 (Extended Data Fig. 6a). Similar stalled products were G>T transversions all observed (Fig. 4b, Extended Data Fig. 7a–c). The
observed for AANAT-ICL repair, indicating that the FA pathway repair was frequency ofconsecutive mutations (at least 2 nt) around the ICL was
inhibited (Extended Data Fig. 6a). However, we observed an accumula- around 100-fold lower than that ofsingle mutations, in agreement with
tion ofrightward –1 products, indicating that REV1 depletion prevents a report suggesting that ICLs do not drive tandem mutations caused by
lesion bypass by the second pathway. By contrast, the leftward fork was acetaldehyde21. AANAT-ICL repair products generated in the presence of
extended without hindrance as no stalled products accumulated and p97i—blocking the FA route—lost most of their mutations at position
a considerable amount ofextension products were formed (Extended G8. This indicates that FA-dependent bypass ofthe top strand adduct
Data Fig. 6a). Depletion ofREV7 (the regulatory subunit ofPol ζ, which is the predominant source ofmutation at G8. Under these conditions,
Nature | www.nature.com | 5

Article
3.
Langevin, F., Crossan, G. P., Rosado, I. V., Arends, M. J. & Patel, K. J. Fancd2 counteracts
the toxic effects of naturally produced aldehydes in mice. Nature 475, 53–58 (2011).
mutations are almost entirely restricted to C7 and the mutational spec-
trum is almost identical to that obtained from pPdG repair (Fig. 4b,
Extended Data Fig. 7b). These data strongly suggest that the second
route ofAANAT-ICL repair reverses this crosslink to yield a monoadduct
that is similar or identical to the original propanoguanine. Consistent
with this model, we found that the insertion step of TLS past a PdG
adduct is mediated by REV1 (Extended Data Fig. 6e).
4. Garaycoechea, J. I. & Patel, K. J. Why does the bone marrow fail in Fanconi anemia? Blood
123, 26–34 (2014).
5.
6.
7.
Hira, A. et al. Variant ALDH2 is associated with accelerated progression of bone marrow
failure in Japanese Fanconi anemia patients. Blood 122, 3206–3209 (2013).
Garaycoechea, J. I. et al. Genotoxic consequences of endogenous aldehydes on mouse
haematopoietic stem cell function. Nature 489, 571–575 (2012).
Garaycoechea, J. I. et al. Alcohol and endogenous aldehydes damage chromosomes and
mutate stem cells. Nature 553, 171–177 (2018).
8.
9.
Wang, M. et al. Identification of DNA adducts of acetaldehyde. Chem. Res. Toxicol. 13,
1149–1157 (2000).
Cho, Y. J. et al. Stereospecific formation of interstrand carbinolamine DNA cross-links by
crotonaldehyde- and acetaldehyde-derived α-CH3-γ-OH-1,N2-propano-2′-deoxyguanosine
adducts in the 5′-CpG-3′ sequence. Chem. Res. Toxicol. 19, 195–208 (2006).
Conclusion
In summary, we have determined the repair mechanisms ofan impor-
tant class of endogenous DNA crosslinks—those caused by acetal-
dehyde. The central role of the FA repair pathway in removing such
crosslinks agrees with the strong genetic evidence underpinning two-
tier protection against this aldehyde3,6,7. However, we also uncovered
a new pathway ofDNA crosslink repair that operates without excision
repair. This mechanism requires replication fork convergence and is
unique in that it unhooks the ICL by cutting within the crosslink itself.
Repair ofAANAT-ICLs by both pathways is error-prone and requires the
TLS polymerases REV1 and Pol ζ. However, this new repair modality
has an obvious advantage in that it avoids the creation of DNA strand
breaks or abasic sites, both ofwhich can promote large-scale genome
instability.
10. Loudon, G. M. Organic Chemistry 874–878 (Oxford Univ. Press, 2002).
11. Duxin, J. P. & Walter, J. C. What is the DNA repair defect underlying Fanconi anemia? Curr.
Opin. Cell Biol. 37, 49–60 (2015).
12. Knipscheer, P. et al. The Fanconi anemia pathway promotes replication-dependent DNA
interstrand cross-link repair. Science 326, 1698–1701 (2009).
13. Long, D. T., Räschle, M., Joukov, V. & Walter, J. C. Mechanism of RAD51-dependent DNA
interstrand cross-link repair. Science 333, 84–87 (2011).
14. Räschle, M. et al. Mechanism of replication-coupled DNA interstrand crosslink repair. Cell
134, 969–980 (2008).
15. Semlow, D. R., Zhang, J., Budzowska, M., Drohat, A. C. & Walter, J. C. Replication-
dependent unhooking of DNA interstrand cross-links by the NEIL3 glycosylase. Cell 167,
498–511 (2016).
16. Duxin, J. P., Dewar, J. M., Yardimci, H. & Walter, J. C. Repair of a DNA-protein crosslink by
replication-coupled proteolysis. Cell 159, 346–357 (2014).
17. Zhang, J. et al. DNA interstrand cross-link repair requires replication-fork convergence.
Nat. Struct. Mol. Biol. 22, 242–247 (2015).
18. Klein Douwel, D. et al. XPF-ERCC1 acts in unhooking DNA interstrand crosslinks in
cooperation with FANCD2 and FANCP/SLX4. Mol. Cell 54, 460–471 (2014).
19. Budzowska, M., Graham, T. G. W., Sobeck, A., Waga, S. & Walter, J. C. Regulation of the
Rev1–Pol ζ complex during bypass of a DNA interstrand cross-link. EMBO J. 34, 1971–1985
(2015).
20. Johnson, R. E., Washington, M. T., Haracska, L., Prakash, S. & Prakash, L. Eukaryotic
polymerases ι and ζ act sequentially to bypass DNA lesions. Nature 406, 1015–1019
(2000).
Online content
Any methods, additional references, Nature Research reporting
summaries, source data, extended data, supplementary informa-
tion, acknowledgements, peer review information; details of author
contributions and competing interests are available at https://doi.
org/10.1038/s41586-020-2059-5.
21. Matsuda, T., Kawanishi, M., Yagi, T., Matsui, S. & Takebe, H. Specific tandem GG to TT base
substitutions induced by acetaldehyde are due to intra-strand crosslinks between
adjacent guanine bases. Nucleic Acids Res. 26, 1769–1774 (1998).
1.
Brooks, P. J., Enoch, M. A., Goldman, D., Li, T. K. & Yokoyama, A. The alcohol flushing
response: an unrecognized risk factor for esophageal cancer from alcohol consumption.
PLoS Med. 6, e1000050 (2009).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
2.
Lai, C. L. et al. Dominance of the inactive Asian variant over activity and protein contents
of mitochondrial aldehyde dehydrogenase 2 in human liver. Alcohol. Clin. Exp. Res. 38,
44–50 (2014).
© The Author(s), under exclusive licence to Springer Nature Limited 2020
6 | Nature | www.nature.com

spectrometry analysis. This work was supported by a project grant from the Dutch Cancer
Society (KWF HUBR 2015-7736 to P.K.), the Gravitation program CancerGenomiCs.nl from the
Netherlands Organisation for Scientific Research (NWO), part of the Oncode Institute, which is
partly financed by the Dutch Cancer Society. K.S. was supported by the Uehara Memorial
Foundation, the Mochida Memorial Foundation for Medical and Pharmaceutical Research, and
the JSPS Postdoctoral Fellowship for Research Abroad. A.N.K.-L. was supported by the
Wellcome Trust. J.S. was supported by Cancer Research UK. This work was supported by a
European Research Council Starting Grant (ERC-STG 678423-EpiID) to J.K. and K.R. and an
NWO Veni grant (016.Veni.181.013) to K.R.
Reporting summary
Further information on research design is available in the Nature
Research Reporting Summary linked to this paper.
Data availability
Datasets generated during the current study are available from the
corresponding authors upon reasonable request.
Author contributions K.J.P. and P.K. initiated and supervised the study. M.R.H. co-ordinated key
aspects of the project. K.S., A.N.K.-L. and K.R. contributed equally. M.R.H., M.M., J.S., M.P.,
D.M.W. and J.W.C. designed the strategy and synthesized the AA-ICL and PdG adducts.
A.N.K.-L. performed cell toxicity experiments. A.B., M.W. and P.K. designed biochemical assays
in Xenopus egg extracts. A.B., K.S. and M.W. conducted Xenopus extract assays. K.R. and J.K.
performed bioinformatic analysis. K.J.P. and P.K. wrote the manuscript and M.R.H. designed
and created the figures.
Code availability
Code used in the current study is available from the corresponding
authors upon reasonable request.
Competing interests The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020-
2059-5.
Correspondence and requests for materials should be addressed to K.J.P. or P.K.
Reprints and permissions information is available at http://www.nature.com/reprints.
Acknowledgements We thank G. P. Crossan, J. T. P. Yeeles and members of the Knipscheer and
Patel laboratories for critical reading of the manuscript; Hubrecht animal caretakers for animal
support; J. C. Walter and D. R. Semlow for providing us with recombinant X. laevis NEIL3
proteins, the AP-ICL and lacO plasmids; and S. Y. Peak-Chew, S. Maslen and M. Skehel for mass

Article
Extended Data Fig. 1 | See next page for caption.

Extended Data Fig. 1 | Acetaldehyde ICLs are stable and are repaired in
Xenopusegg extracts. a, Scheme for the synthesis of the precursor, 4-(R)-
aminopentane-1,2-diol. b, Site-specific synthesis of a PdG adduct in a DNA
oligonucleotide. c, Denaturing PAGE showing crosslink formation between dG
and PdG, but not between dG and inosine (ino), which lacks an N2 amine. Two
independent experiments were performed. d, Confirmation of AA-ICL
formation by MALDI MS. The peak at m/z 12,370.2 represents the imine or
pyrimidopurinone form. Two further peaks at m/z 5,979.41 and 6,409.74 equate
to masses for the two parent oligonucleotides, consistent with the mass ofthe
carbinolamine form after dissociation back to PdG and dG under the
desorption/ionization conditions. Three independent experiments were
performed. e, Stability of AANAT-ICL as a function of temperature and time, as
determined by radiolabelling ([α-32P]dCTP) and resolution by denaturing PAGE.
Error bars represent s.e.m. from three independent experiments. f, AANAT-ICL is
susceptible to hydrolysis in aqueous acid, whereas AARED-ICL is stable. Pre-
purification crosslink reactions were incubated with or without formic acid
and products were resolved by denaturing PAGE. Three independent
experiments were performed. g, Scheme depicting the type and position ofthe
DNA lesions used in this study. Duplex DNA with or without the indicated
lesions was annealed into a backbone vector to generate circular plasmids with
or without damage. h, To determine the percentage of crosslinks, the ICL-
containing plasmids were digested with NotI, labelled at the 3′-end by end
filling with [α-32P]dCTP, and separated by denaturing PAGE. Crosslinked DNA
(88 nt) shows slower mobility compared with non-crosslinked DNA (44 nt). The
percentage ofcrosslinks was calculated by comparing the 88-nt product with
the 44-nt products. Two independent experiments were performed. i, AANAT
-
ICL and Pt-ICL are stable in Xenopus egg extracts. Plasmids were incubated in a
high-speed supernatant (HSS) extract. DNA was extracted and analysed as
described in h. Three independent experiments were performed. j, Solution
structures ofa cisplatin ICL and a reduced form ofan acetaldehyde ICL (PDB:
1DDP22 and 2HMD23, cartoon representation generated in PyMOL). k, The
indicated plasmids were replicated in Xenopus egg extracts and repair
intermediates were digested with NotI, labelled at the 3′-end, and resolved by
denaturing PAGE. The increase in intensity ofthe 44-nt band over time
indicates ongoing replication and repair. A higher mobility band, probably
generated from end-joining activity in some extracts, is indicated by an
asterisk. This gel is the independent experimental duplicate ofthat in Fig. 1e. l,
Quantification ofrepair based on the intensity ofthe 44-nt product on the gel
in k, as described in the Supplementary Methods. This graph is the
independent experimental duplicate ofthat in Fig. 1f. Additional replicates of
these experiments are presented in Fig. 2a, Extended Data Figs. 2k–m, 4b.
22. Huang, H., Zhu, L., Reid, B. R., Drobny, G. P. & Hopkins, P. B. Solution structure of a
cisplatin-induced DNA interstrand cross-link. Science 270, 1842–1845 (1995).
23. Cho, Y. J., Kozekov, I. D., Harris, T. M., Rizzo, C. J. & Stone, M. P. Stereochemistry modulates
the stability of reduced interstrand cross-links arising from R- and S-α-CH3-γ-OH-1,N2-
propano-2′-deoxyguanosine in the 5′-CpG-3′ DNA sequence. Biochemistry 46, 2608–2621
(2007).

Article
Extended Data Fig. 2 | See next page for caption.

Extended Data Fig. 2 | Acetaldehyde ICLs are repaired by Fanconi-dependent
and Fanconi-independent mechanisms. a, Model for ICL repair by the FA
pathway. Upon convergence of two replication forks at the crosslink, the CMG
helicase is unloaded from the DNA to enable the approach of one replication
fork to the −1 position. Ubiquitylation of FANCD2 promotes the recruitment of
the XPF-ERCC1-SLX4 (XES) complex to the ICL, which enables nucleolytic
incisions that unhook the crosslink. This step could be preceded by fork
reversal of one of the stalled replication forks24. Incisions generate a broken
strand and a strand with an adduct; the latter is bypassed by TLS whereas the
broken strand is repaired by homologous recombination. In mammalian cells,
replicated in mock or FANCD2-depleted extracts and repair intermediates were
digested with NotI, labelled at the 3′-end, and resolved by denaturing PAGE.
Quantification ofrepair based on the intensity ofthe 44-nt product is shown in
Fig. 1g. f, The independent experimental duplicate ofFig. 1g. g, The
independent experimental triplicate ofFig. 1g, but using only pICL-AANAT
.
h, Plasmid pICL-AANAT was replicated in FANCD2-depleted extract, or FANCD2-
depleted extract supplemented with a recombinant FANCI–FANCD2 complex
(ID). Reaction samples were resolved by native agarose gel and visualized by
autoradiography. RRIs, open circle (OC) and supercoiled (SC) products are
indicated. The stalled repair product (grey arrow) is indicated. Two
independent experiments were performed. i, The indicated plasmids were
replicated in Xenopus egg extracts in the presence or in the absence ofp97i and
the intermediates were resolved by native agarose gel electrophoresis. The
stalled repair products (grey arrow) are indicated. Seven independent
experiments were performed. j, The indicated plasmids were replicated in
Xenopus egg extracts in the presence or in the absence ofp97i, and repair
intermediates were digested with NotI, labelled at the 3′-end, and resolved by
denaturing PAGE. The increase in intensity ofthe 44-nt band (white arrow) over
time indicates ongoing replication and repair. A higher mobility band,
probably generated from end-joining activity in some extracts, is indicated
with an asterisk. k, Quantification ofrepair based on the intensity ofthe 44-nt
product on the gel shown in j, as indicated in the Supplementary Methods.
l, Quantified independent experimental duplicate ofj. m, Quantified
independent experimental triplicate ofj.
it has been shown that a single fork can pass over the ICL without unhooking25
this ‘traverse’ gives rise to a structure that resembles the one generated after
fork convergence and CMG unloading and could follow the same steps
subsequently. b, The indicated plasmids were replicated in Xenopus egg
extracts and reaction samples were analysed by western blot with FANCD2
antibody. Two independent experiments were performed. c, Western blot of
FANCD2, showing a titration of Xenopus egg extracts compared to mock and
FANCD2-depleted extracts. Two independent experiments were performed.
d, The indicated plasmids were replicated in mock or in FANCD2-depleted
;
extracts in the presence of [α-32P]dCTP. Repair products were digested by AflIII,
separated on a sequencing gel alongside a ladder derived from extension
primer S, and visualized by autoradiography. The white arrow denotes the −1
product, which is 2 nt larger in pICL-Pt owing to the position of the ICL. Three
independent experiments were performed. e, The indicated plasmids were
24. Amunugama, R. et al. Replication fork reversal during DNA interstrand crosslink repair
requires CMG unloading. Cell Rep. 23, 3419–3428 (2018).
25. Huang, J. et al. The DNA translocase FANCM/MHF promotes replication traverse of DNA
interstrand crosslinks. Mol. Cell 52, 434–446 (2013).

Article
Extended Data Fig. 3 | Reduced acetaldehyde ICLs are repaired by the
Fanconi pathway. a, MALDI MS data confirms the identity and the stability of
the reduced ICL. Three independent experiments were performed. b, Plasmids
were replicated in extract and products were resolved on a native agarose gel.
Three independent experiments were performed. c, The indicated plasmids
were replicated in Xenopus egg extracts in the presence or in the absence of
p97i and the repair intermediates were analysed using the NotI digestion assay.
The increase in intensity ofthe 44-nt band indicates ongoing replication and
repair. A band with higher mobility, probably generated from end-joining
activity in some extracts, is indicated with an asterisk. Quantification ofrepair
based on this gel is shown in Fig. 1h. Three independent experiments were
performed. d, The independent experimental duplicate ofFig. 1h. e, The
independent experimental triplicate ofFig. 1h.

Extended Data Fig. 4 | Acetaldehyde ICL repair requires DNA replication and
replication fork convergence. a, The indicated plasmids were replicated in
Xenopus egg extracts in the presence or in the absence of geminin. Repair
intermediates were digested with NotI, labelled at the 3′-end, and resolved by
denaturing PAGE. Quantification of repair based on the intensity ofthe 44-nt
product is shown in Fig. 2a. b, The independent experimental duplicate of
Fig. 2a. c, The independent experimental triplicate of Fig. 2a but using only
pICL-A ANAT. d, pICL-P t-lacO was replicated in Xenopus egg extracts containing
[α-32P]dCTP in the presence or in the absence of LacR. The repair intermediates
were digested by AflIII and EcoRI, separated on a sequencing gel and visualized
by autoradiography. Two independent experiments were performed. e, The
indicated plasmids were replicated in extracts in the presence or in the absence
ofLacR. Repair products were digested with NotI, labelled at the 3′-end, and
resolved by denaturing PAGE. Three independent experiments were
performed. f, Quantification ofrepair based on the intensity ofthe 44-nt
product on the gel shown in e, as described in the Supplementary Methods.
Three independent experiments were performed. g, The independent
experimental duplicate of f. h, The independent experimental triplicate of f.

Article
Extended Data Fig. 5 | See next page for caption.

Extended Data Fig. 5 | The alternative route of AA-ICL repair does not involve
DNA excision. a, Western blot showing a titration of Xenopus egg extracts
compared to the NEIL3-depleted (∆NEIL3) extract and NEIL3-depleted extract
supplemented with recombinant wild-type NEIL3 (WT) or catalytically inactive
NEIL3 (MUT). Three independent experiments were performed. b, The
indicated plasmids were replicated in NEIL3-depleted extract containing
[α-32P]dCTP, supplemented with wild-type (WT) or catalytically inactive (MUT)
NEIL3. Replication intermediates were resolved by native agarose gel
electrophoresis and visualized by autoradiography. Three independent
experiments were performed. c, Clonogenic survival of wild-type, NEIL3-,
FANCL- or NEIL3/FANCL-deficient human HAP1 cells after a 2-h exposure to
acetaldehyde. Three independent experiments were performed. Data are
mean s.e.m. d, The median lethal dose (LD50) of acetaldehyde for the survival
ofwild-type and deficient HAP1 cells was calculated by regression analysis of
the curves presented in c. Data are mean s.e.m. Three independent
experiments were performed. e, Quantification of the arm fragments resulting
from APE1 treatment (AP sites) from the gel in Fig. 2e. f, Quantification ofthe
APE1 arms as in e, from an independent duplicate experiment without the
addition ofp97i. g, Quantification ofthe APE1 arms as in e, from an
independent triplicate experiment without the addition ofp97i. As a positive
control we used a plasmid containing an abasic-site-induced interstrand
crosslink (pICL-AP) that is also repaired via the glycosylase NEIL3.
h, Quantification ofthe HincII arm fragments from the gel in Fig. 2f. i,
Quantification ofthe HincII arms as in h, from an independent duplicate
experiment. j, Quantification ofthe HincII arms as in h, from an independent
triplicate experiment. k, Schematic representation ofthe formation ofDNA
adducts by unhooking incisions during ICL repair (left). These adducts are not
removed during ICL repair in Xenopus egg extracts14 and can therefore be
visualized. Plasmids were replicated in Xenopus egg extracts in the presence or
in the absence ofp97i. Late reaction samples were digested with AflIII and AseI
and separated on a sequencing gel. Adducts on either the top or the bottom
strand (white arrowheads) were detected by strand-specific Southern blotting
(right). Three independent experiments were performed.

Article
Extended Data Fig. 6 | See next page for caption.

Extended Data Fig. 6 | Both routes of AA-ICL repair are mediated by REV1
and REV7. a, The indicated plasmids were replicated in mock or REV1-depleted
extracts, in the presence or in the absence of p97i. Reaction intermediates were
digested by either AflIII or AflIII and BamHI, separated on a sequencing gel
alongside a ladder derived from extension of primer S, and visualized by
autoradiography. White arrows denote 0 products, dark grey arrows indicate
−1 products, and light grey arrows indicate 0/−1 products (not separated). Two
independent experiments were performed. b, Western blot detection ofREV7
in REV1- or mock-depleted Xenopus egg extracts compared to a titration of
undepleted extract. Two independent experiments were performed.
Two independent experiments were performed. d, The indicated plasmids
were replicated in mock- or REV7-depleted extracts and reaction intermediates
were digested by either AflIII or AflIII and BamHI, separated on a sequencing
gel and visualized by autoradiography. Grey arrows indicate −1 products. Two
independent experiments were performed. e, Indicated plasmids were
replicated in mock- or REV1-depleted extracts and reaction intermediates were
digested by AflIII and BamHI, separated on a sequencing gel alongside a ladder
derived from extension ofprimer S, and visualized by autoradiography. The
asterisk indicates a 121-nt background fragment caused by a second BamHI
restriction site in the leftward fork. Two independent experiments were
performed.
c, Western blot detection of REV7 and REV1 in REV7-depleted (∆REV7) or mock-
depleted Xenopus egg extracts compared to a titration of undepleted extract.

Article
Extended Data Fig. 7 | Mutagenic outcome of acetaldehyde crosslink repair.
a, Frequency of nucleotide misincorporation in a 15-bp region flanking the
lesions present in the indicated plasmids. The mutation frequencies for the
same plasmid that has not been replicated in Xenopus egg extracts (NR) and for
the control vector (pCon) are plotted together. Strand specificity is lost
because the sample preparation involves PCR amplification; as such, only the
top sequence is indicated below the graphs. See Fig. 4. b, Distribution and
frequency ofnucleotide misincorporations in a 15-bp region flanking the
lesions present in the indicated plasmids. Independent duplicate sequencing
experiment ofFig. 4b. The heights ofthe bars represent the mutation
frequency minus the baseline mutations found in pCon. c, Frequency of
nucleotide misincorporations in a 15-bp region flanking the lesions present in
the indicated plasmids (data from the same sequencing experiment as in b).
The mutation frequency for pCon is also plotted.

Extended Data Table 1 | Sequence and read numbers for high-throughput sequencing experiments
a, Amplicon sequence for sequencing experiments 1 and 2. b, Total and specific read numbers for sequencing experiment 1 (see Fig. 4). c, Total and specific read numbers for sequencing
experiment 2 (see Extended Data Fig. 7).