ABSTRACT
In eukaryotic cells, a 5
'
flap DNA endonuclease activity and a ds DNA 5
'
-exonuclease activity exist within a single enzyme called FEN-1 [
f
lap
e
ndo-
n
uclease and 5(
f
ive)
'
-
e
xo-
n
uclease]. This 42 kDa endo-/exonuclease, FEN-1, is highly homologous to human XP-G,
Saccharomyces cerevisiae
RAD2 and
S.cerevisiae
RTH1. These structure-specific nucleases recognize and cleave a branched DNA structure called a
DNA flap, and its derivative called a pseudo Y-structure. FEN-1 is essential for lagging strand DNA synthesis in Okazaki fragment
joining. FEN-1 also appears to be important in mismatch repair. Here we find that human
PCNA, the processivity factor for eukaryotic polymerases, physically associates
with human FEN-1 and stimulates its endonucleolytic activity at branched DNA structures
and its exonucleolytic activity at nick and gap structures. Structural
requirements for FEN-1 and PCNA loading provide an interesting picture of this stimulation.
PCNA loads on to substrates at double-stranded DNA ends. In contrast, FEN-1 requires a free single-stranded 5'
terminus and appears to load by tracking along the single-stranded DNA branch. These physical constraints define the range of DNA
replication, recombination and repair processes in which this family of
structure-specific nucleases participate. A model explaining the exonucleolytic
activity of FEN-1 in terms of its endonucleolytic activity is proposed based on these
observations.
In all eukaryotic cells, an enzyme called FEN-1 [
The yeast proliferating cell nuclear antigen (PCNA) is the processivity factor
for DNA polymerases [delta] and [epsilon]. Like FEN-1, PCNA is one of the 10 essential proteins for DNA
replication (
7
). PCNA is also important in nucleotide excision repair (
11
). It is a homotrimer with a subunit molecular weight of 29 kDa and is highly
conserved from yeast to mammalian cells. Based on the crystal structure,
trimeric yeast PCNA forms a closed ring which appears to encircle double-stranded DNA (
12
). Processivity in DNA synthesis is achieved by PCNA binding to the polymerase,
thereby tethering the DNA polymerase at the primer terminus (
13
). In addition to this structural function in DNA replication, mammalian PCNA,
through its interactions with the cyclin-dependent protein kinase inhibitor p21 (CIP1/WAF1/SDI1), has also been
implicated in cell cycle control (
14
).
Recently, we found that yeast PCNA and yeast FEN-1 interact (
15
). We were interested in verifying this interaction in the most distant
multicellular eukaryote, human. Here we report that human PCNA binds to FEN-1 and stimulates the endonucleolytic cleavage of FEN-1 at flap structures and its exonucleolytic activity at nicks. In
this ternary interaction between DNA substrate, FEN-1 and PCNA, it is interesting to consider how these three components
assemble and interact. Does the mode of assembly influence the range of
substrates cleaved? Here, we describe studies using recombinant human FEN-1 confirming that it requires a fully single-stranded 5' terminus and flap for cleavage at the single- to double-stranded DNA junction. This is despite the
insensitivity of FEN-1 to the phosphorylation state of the 5' terminus and to the base in the 5' terminal nucleotide in endonucleolytic assays. Deviations
from single-stranded character anywhere between the 5' terminus and the cleavage junction prevent cutting. Hence,
heterologous loops and DNA bubbles, potentially important intermediates in a
variety of processes in DNA metabolism, are not recognized. These observations
have important implications for the physiologic role of FEN-1 in DNA replication, DNA recombination, and DNA repair. These studies
indicate that FEN-1 and PCNA may load onto different portions of branched DNA structures in
assembly of the ternary complex of these two proteins with DNA. PCNA does not
alter the substrate specificity at all. Based on the insensitivity of the
endonucleolytic activity of FEN-1 to the 5' phosphorylation status of the ss-DNA flap, but its sensitivity to the 5' phosphorylation status at a DNA nick (where FEN-1 previously has been considered to act
exonucleolytically), we describe a unified model that explains the
exonucleolytic activity of FEN-1 in terms of its endonucleolytic activity.
Protein G (Pharmacia) beads (10 [mu]l) were washed twice 0.4 ml Buffer A (40 mM HEPES pH 7.4; 2 mM MgCl
2
) containing 150 mM NaCl (designated buffer B when it includes the 150 mM NaCl).
Anti-human c-myc (anti-myc) or anti-human PCNA (2 [mu]g) monoclonal antibodies (mouse IgG1) were added to
the beads in 100 [mu]l buffer B and incubated at 4oC for 14 h. The loaded beads were washed three times with 400 [mu]l buffer B. Human FEN-1 (60 ng) was added to the beads in buffer B and incubated 2
h at 4oC. The beads were washed with 400 [mu]l buffer B three times. Any bound protein (FEN-1) was eluted with two washes of 10 [mu]l buffer A containing either 300 or 600 mM NaCl. This 20 [mu]l was mixed with 20 [mu]l 2* SDS loading buffer and fractionated on an 8%
PAGE. Immunoblotting used rabbit polyclonal anti-human FEN-1 anti-sera at a 1:100 dilution. The secondary antibody was goat anti-rabbit coupled to horse radish peroxidase at a 1:400
dilution. Enhanced chemiluminescence (ECL) was used for detection.
Oligonucleotides are SC1, CAGCAACGCAAGCTTG (strand adjacent to the flap strand);
SC3, GTCGACCTGCAGCCCAAGCTTGCGTTGCTG (bridge strand, which is annealed to the
flap and the adjacent strand); and SC5, ATGTGGAAAATCTCTAGCAGGCTGCAGGTCGAC (flap strand, which is the one cleaved by FEN-1). Additional flap substrates were composed of HJ41 (bridge strand), 43
(flap adjacent strand), and 86 (flap strand). HJ41, GGACTCTGCCTCAAGACGGTAGTCAACGTG (30mer). HJ43, CACGTTGACTACCGTC (16mer). HJ86, GCCGCCGCCGCCGCCTTTTTTTTTTTTTTTTTGAGGCAGAGTCC (44mer); note that upon Klenow fill-in, this 44mer becomes 46 nt long. The blocking oligo HJ87 is used in some
experiments to anneal to the flap strand, HJ86. HJ87, GGCGGCGGCGGCGGC (15mer).
Some structures use a different blocking oligo in place of HJ87. This is HJ88.
HJ88, AAAAAAAGGCGGCGG (15mer). In cases where we wanted to label the flap
strand at the 3' end, we needed to make the bridge strand longer than HJ41. This longer version is HJ89 [TTGGACTCTGCCTCAAGACGGTAGTCAACGTG (32mer)].
FEN-1 activity on loop substrates. For heterologous loops, we paired MY2 with
HJ90. MY2, GTATCTGCCGAAACTGATCCAGTTACAAGGCTGTGTCCTCAGAGGATC (48mer). HJ90,
GATCCTCTGAGGACACAGATCAGTTTCGGCAGATAC (36mer). For a bubble structure, we paired
MY2 with HJ91 [GATCCTCTGAGGACACTTTTTTCAGTTTCGGCAGATAC (38 mer)].
The endonuclease assay was done in a 15 [mu]l total volume containing 50 mM Tris-HCl (pH 8.0), 10 mM MgCl
2
, 25 mM NaCl, 0.5 mM [beta]-mercaptoethanol, 500 [mu]g/ml BSA, 10 fmol of flap substrate, FEN-1 at specified amounts (10 fmol = 0.4 ng = 20 U as defined
in ref.
1
), and, if present, PCNA trimer at specified amounts.
Assays were as described for the flap substrates, except for adjustment to 10 mM
Tris (pH 8), 5 mM MgCl
2
and 8 mM NaCl. Oligonucleotides are CLH2 (24mer) GTAGGAGATGTCCCTTGATGAATT; CLH3
(16mer) CGAACCCAGATACGGC and AI4 (41mer) GGCCGTATCTGGGTTCGAATTCATCAAGGGACATCTCCTAC. In cases where we wanted to 3' label the oligonucleotide downstream of a nick, we used HJ95, which is
identical to CLH3 except that it is one nucleotide shorter at its 3' end, permiting fill-in with the Klenow fragment of DNA pol I using labeled [[alpha]-
32
P]dCTP. Control experiments showed no nuclease activity in all analogous
experiments lacking FEN-1 (data not shown).
To prove that the human FEN1 specifically interacts with human PCNA, we
transformed pJG-hFEN1 into the YPH499 strain and mated with human PCNA bait and other
strains expressing unrelated baits, such as LexA-bicoid, LexA-Ku86 and the parent vector, pEG202. The diploid progeny were
streaked onto a glucose/-ura/-his/-trp master plate, and then replica plated onto a galactose/-ura/-his/-trp/-leu plate. Interaction results in
activation of the Lex A op-Leu2 reporter and growth in the absence of leucine. Human FEN-1 causes strong growth in the presence of the human PCNA bait, and
weak growth in the presence of the Ku86 bait, bicoid bait, and pEG202 bait
(Fig.
1
A).
To test whether the strong genetic indication of an interaction could be
documented physically, we bound human PCNA to protein G beads via an anti-human PCNA monoclonal antibody. As a control, we used anti-human c-myc antibodies. Soluble FEN-1 was then incubated with the beads and found to
associate with the PCNA beads to a greater extent than to the anti-myc control beads (Fig.
1
B). The interaction was stable at 300 mM NaCl and required 600 mM to elute.
Hence, the physical interaction is detectable even at salt concentrations more
than twice physiologic.
FEN-1 is also a ds DNA exonuclease that is most active at nicks. Its activity
and binding at gaps decreases with increasing gap size. We were interested in
whether PCNA stimulates FEN-1 exonucleolytic activity at a DNA nick. We find that it does. However,
the magnitude of the stimulation is somewhat smaller than for the
endonucleolytic activity (Fig.
3
). The products are predominantly mononucleotides but dinucleotides
exonucleolytic products are also generated.
Yeast PCNA stimulates yeast FEN-1. FEN-1 is 61% identical and 78% similar between yeast and human (
3
). PCNA is 30% identical between yeast and human. Hence, we were interested in
the extent to which these were interchangeable. We find that human PCNA can
stimulate yeast FEN-1 to an extent that is equivalent to human FEN-1 (Fig.
4
A). However, yeast PCNA does not stimulate human FEN-1 (Fig.
4
B).
FEN-1 acts on substrates with 5' flaps as an endonuclease and at nicks or recessed 5' DNA ends as a 5' -> 3' exonuclease. We were interested in
whether substrates that lack a 5' flap or a recessed 5' end could function as FEN-1 substrates when PCNA is provided. We tested heterologous
DNA loops (Fig.
5
). At the site where the loop departs from double-stranded DNA conformation, these substrates share some features with 5' flap structures and pseudo-Y branched DNA structures. Specifically, if the heterologous
loop were nicked at its upstream attachment point to the double-stranded DNA, then it would become the 5' flap structure optimal for FEN-1 cutting. The only difference then is that there is no free
5' flap terminus. When tested, FEN-1 is able to distinguish the heterologous loop from the 5' flap structure, and it does not cut the loop. We reasoned
that in the cell, FEN-1 is acting in the presence of PCNA. We wondered if this failure to cut
such a similar structure could be overcome by PCNA stimulation. We find that
PCNA does not help FEN-1 to cleave heterologous loops. Hence, a free 5' terminus is necessary for FEN-1 cleavage.
PCNA loads onto linear DNA substrates by diffusion onto the double-stranded DNA termini (
19
). For FEN-1 loading, we wondered if the 5' flap had to be single-stranded over its entire length, only near the site of
cleavage, or only at its most 5' terminal end.
Escherichia coli
polymerase I has a 5' -> 3' nuclease domain that has very similar endonucleolytic and
exonucleolytic properties to FEN-1. In fact, FEN-1 is the counterpart of this domain for eukaryotic polymerases (
1
,
7
,
20
,
21
). There is significant amino acid homology between the FEN-1 and
E.coli
pol I 5' nuclease domain (24% overall and 52% in a selected 63 aa region) (
4
). The 5' nuclease of
E.coli
pol I appears to slide down the single-stranded flap until it reaches the branch point where it cleaves. Double-stranded character along this 5' flap is known to prevent
E.coli
pol I from acting (
22
). One might assume that FEN-1 would function in a similar fashion. However,
S.cerevisiae
RAD2 is a member of the highly conserved FEN-1 family of nucleases (
3
), and it does not require a free 5' terminus to cut at the corresponding position as FEN-1 (
23
).
We wondered which mechanism of loading FEN-1 uses and whether the manner of loading is modified by PCNA. In order to
examine this issue, we generated 5' flap structures as we have described previously, and we tested the
ability of oligonucleotides annealed to various locations along this flap to
prevent FEN-1 action (Fig.
6
). In these cases, the flap is 30 nt long. The oligonucleotide being annealed to
the flap is 15 nt. When we anneal the 15 nt oligonucleotide to the most distal
portion of the flap, cleavage is entirely blocked. When we anneal in the middle
of the flap 7 nt from the 5' terminus and leaving 8 nt of single-stranded character before the branch point, cleavage is also
entirely blocked. We have done the same experiments with a 65 nt flap and a 15
nt oligonucleotide annealed in the middle of its length. This leaves 25 nt of
single-stranded DNA on each side, with the duplex portion located in the middle
of the flap. We find that cleavage is still blocked (not shown). Hence, FEN-1, like the 5' nuclease domain of
E.coli
pol I (
22
) and unlike
S.cerevisiae
Rad2 (
23
), requires fully single-stranded character between the 5' terminus and the branch point.
In a two-hybrid search using human PCNA, we found that human FEN-1 is detected as the predominant interactor. However, does this
binding reflect a functional interaction in the nucleus? Based on the enzymatic
stimulation of FEN-1 by PCNA, we infer that it does. PCNA stimulates the endonucleolytic
activity at 5' DNA flap structures. It also stimulates the exonucleolytic activity of
FEN-1 at DNA nicks.
Human FEN-1 in the absence of PCNA is markedly inhibited by increasing
concentrations of monovalent salt. The binding between FEN-1 and PCNA results in substantially more endonucleolytic and
exonucleolytic activity than would otherwise occur at these salt concentrations
(
1
). Therefore, the functional binding with PCNA explains how FEN-1 can be active at higher salt concentrations.
In DNA replication, PCNA is localized to the 3'-OH of the primer DNA strand by RF-C (
19
). At the replication fork, there is no free ds DNA end onto which PCNA can
diffuse. Therefore, RF-C binds to the 3'-OH and then catalyzes PCNA assembly from monomer to trimer
around the axis of the ds DNA. It is reasonable that PCNA would then bind FEN-1 because this nuclease is required for Okazaki fragment processing.
Specifically, RNase H degrades the RNA primer of each lagging strand down to a
point where the last ribonucleotide remains (
20
,
21
). Pol [delta] extends the upstream strand to a nick. At that point, either of two
pathways achieve the same result. In one pathway, the polymerase extends
further to displace the downstream strand. This generates a 5' flap structure, which FEN-1 can then cleave endonucleolytically. In the second pathway, FEN-1 functions exonucleolytically, cleaving the last ribo- and deoxyribonucleotide off exonucleolytically as a
dinucleotide. The localization of FEN-1 to the replication fork by PCNA and the stimulation of FEN-1 activity by PCNA are consistent with their functions in
replication.
We start to see endonucleolytic stimulation of FEN-1 by PCNA at a molar ratio of 12 PCNA trimer molecules per each FEN-1 molecule. The stimulation continues for as far as we carried out
the titration, which was a ratio of 800 PCNA trimers per FEN-1. In the absence of RF-C, PCNA loads onto DNA as a toroid diffuses onto the end of a rod.
This is a diffusion-limited process that has obvious steric requirements. This type of PCNA
loading is much less efficient than that catalyzed by RF-C. The requirement for stoichiometric excess of PCNA under conditions for
diffusional loading has been extensively documented (
19
). Hence, it is not surprising that FEN-1 stimulation by PCNA would require a large stoichiometric excess and that
it would increase progressively with increasing PCNA.
It is noteworthy that this PCNA-FEN-1 interaction extends from yeast (
15
) to humans. The features of the stimulation are similar overall. However, there
is one interesting difference that we have noted. We did not detect stimulation
of yeast FEN-1 exonucleolytic activity at linear ds nick sites by yeast PCNA even in
the presence of yeast RF-C. We could only detect stimulation of yeast FEN-1 by PCNA and RF-C at nick sites on circular M13 molecules. We reasoned that
this was because the PCNA molecules diffused off of the linear DNA molecules
too quickly. For human FEN-1 and PCNA, we do detect some stimulation of FEN-1 exonucleolytic activity at simple nicked ds linear DNA molecules
(this expands the nick to a gap). Hence, the residence time of the human PCNA
may be longer, allowing for stimulation of FEN-1. This could be due to a difference between yeast and human PCNA or a
difference between yeast and human FEN-1. The human PCNA may diffuse across the nicked linear DNA slower or the
human FEN-1 may bind more tightly to the nick site than the yeast FEN-1.
In addition to replication, the interaction between FEN-1 and PCNA may have broader implications in DNA metabolism. FEN-1 mutants are adversely affected in mismatch DNA repair (
9
). The specific enzymatic steps in eukaryotic mismatch repair are not yet
sufficiently well-defined to permit specification of the precise step for FEN-1 activity or the involvement of PCNA. However, the result is marked
instability of dinucleotide repeats, just as is the case for the other mismatch
repair components. In another DNA repair process, nucleotide excision repair,
PCNA is known to be important
(
24
,
25
). Although FEN-1 itself has not been shown to be involved in this reaction, RAD2 (XP-G in higher eukaryotes) is absolutely required (
26
). Based on the involvement of PCNA in nucleotide excision repair and the
presence of homology between FEN-1 and RAD2, it is possible that PCNA may also interact with RAD2 to
facilitate its loading and thereby excision of damaged nucleotides. Though RAD2
has not specifically been shown to be the target of PCNA stimulation in
nucleotide excision repair, the work of Nichols and Sancar (1992) makes it
clear that PCNA is stimulating some form of nucleolytic activity in excision repair. Thus, there may be a common theme in
various aspects of DNA metabolism, in addition to DNA replication, in which a
processivity factor stimulates a structure-specific nuclease in processing nicked and branched DNA intermediates. We
are currently investigating the role of FEN-1 in DNA end joining (double-strand break repair). In this case, PCNA could diffuse onto the free
ds DNA end and stimulate FEN-1 action at a nick, gap or flap. The fact that PCNA stimulation of FEN-1 exonucleolytic activity can occur on linear substrates in the
absence of RF-C in human cells is potentially important in DNA end joining. We did not
find such stimulation in yeast for this particular substrate configuration. It
is interesting that in yeast the predominant mode of DNA end joining is
different from mammalian cells and procedes by homologous recombination
involving the homologous chromosome as a template.
PCNA trimer loads diffusionally onto the end of double-stranded DNA (
19
). Monomeric PCNA is unable to stimulate FEN-1 (
15
). Therefore, the assembled trimer is important for FEN-1 stimulation. In the absence of a DNA terminus, RF-C catalyzes the assembly of PCNA monomers into trimers at a 3'-OH (
19
).
Based on the failure of FEN-1 to cleave at DNA bubbles and heterologous loops, we inferred that a free
5' terminus is important for FEN-1 substrate recognition. We confirmed this by annealing
oligonucleotides at various positions along the single-stranded DNA flap. Whether annealed to the most 5' portion of the flap or along the middle of the flap, FEN-1 endonucleolytic activity was eliminated. Therefore, it
appears that FEN-1 tracks along the single-stranded DNA from the 5' terminus to the cleavage point.
Given these observations, it appears that FEN-1 and PCNA may load from different points of a branched DNA substrate
(Fig.
9
A), with PCNA loading from one end and FEN-1 loading along the single-stranded 5' flap. The simplest model is that PCNA stabilizes FEN-1 at the branch point, increasing its residence time
there, this being the basis for PCNA stimulation of FEN-1 activity.
Are the endonucleolytic and exonucleolytic activities of FEN-1 related or distinct activities? One way of answering this question is by
considering the similarities and differences in binding and cleavage of exo- versus endonucleolytic substrates. Both activities show little sensitivity to the base at the 5' most position at the flap or nick. Both FEN-1 endo- and exonucleolytic substrate binding and cutting are
stimulated by an upstream oligonucleotide (flap adjacent strand or primer) (
27
). This is also the case for
E.coli
pol I (
28
).
Although a 5' OH terminus is a good substrate for FEN-1 loading onto a 5' flap substrate, it serves as a poor substrate when part of
a double-stranded DNA nick. The electrostatic repulsion by the terminal phosphate
may allow increased breathing of the substrate into a pseudo-flap configuration, providing the active form of the substrate for FEN-1 (Fig.
9
B). Such an explanation would indicate a single active site and a single
mechanism of loading of FEN-1 onto the 5' ss-DNA terminus of the flap or pseudo-flap configuration of the nick. Consistent with this
model are our observations that optimal activity at a nick requires very low Mg
2+
and monovalent salt, which destabilize base-pairing, whereas flap cleavage is optimal at moderate Mg
2+
and monovalent salt concentrations. Furthermore, we have previously shown that
one nucleotide flaps are efficient substrates (
1
).
This work was supported by NIH grants CA51105 and GM43236 to M.R.L. and GM32431
to P.M.J.B. M.R.L. is a Leukemia Society of America Scholar.
Plasmid and strains.
The plasmids, pJG4-5 and pEG202, have been described previously (
16
). The human PCNA acidic activation constructs, contain the entire human PCNA
structural gene and is the result of ligation of an
Nde
I blunt fragment from p3038 2xT (
17
) into pEG202 digested with
Eco
RI and blunted. Lex A-Ku86 contains the carboxy-terminal 210 amino acids of human Ku86 protein, which was cloned as
an
Eco
RI-
Xho
I fragment into pEG202 after polymerase chain reaction amplification. Lex A-bicoid was a gift of Roger Brent. The junctions of all constructs were
sequenced. All strains were derived from the parent
S.cerevisiae
strain EGY48 (
MATa trp1 ura3 his3 leu2:: pLexAop6-leu2
). YPH499 (
MAT ura3-52 trp 1-63 his3-200 leu2-1 lys-801 ade 2-101
).
Modified two-hybrid screening.
We developed a yeast liquid mating strategy to screen libraries with the yeast
interaction trap of Brent (
16
). The prey strain, YPH499, was transformed with a HeLa cDNA expression library.
The transformants, selected on tryptophan dropout plates, are harvested,
aliquoted and then stored at -80oC in 32.5% glycerol, 12.5 mM Tris-HCl pH 8.0, 50 mM MgSO
4
. The resulting library strain consists of 1.8 * 10
6
independent YPH499 transformants. The bait strain, EGY48, was transformed with
a lacZ
reportor plasmid pSH 18-34 and lexA-hPCNA (human PCNA) fusion plasmid pEG202-hPCNA. Screening is performed by mating of library and bait
strains followed by selection of leucine prototrophy. In brief, 5 * 10
7
colony forming units of the library strain are combined with 5 * 10
7
cells from the bait strain. This mixture is pelleted, resuspended in 2 ml YPAD
media, divided into 20 aliquots, and then incubated for 12-16 h at 30oC. Yeast were thoroughly washed with sterile water and resuspended
in CM/-ura/-his/-trp/2%galactose/1% raffinose. Incubation in the latter medium
for 3-4 h at 30oC permits induction of the galactose inducible promotor of the
library plasmids. To screen the library for interaction, the mixture containing
complete media/-ura/-his/-trp/-leu/2% galactose/1% raffinose. After 3-4 days of incubation at 30oC, the largest colonies were picked and
further analyzed as described (
16
,
18
).
PCNA
. Recombinant human PCNA was purified as described previously (
17
).
FEN-1
. Recombinant human FEN-1 was purified as described for recombinant murine FEN-1 (
3
).
REFERENCES
Return