Nature - USA (2020-01-23)

Nature | Vol 577 | 23 January 2020 | 579

overlap ZD-K20 and nascent-strand peaks as ‘plus-ZD-ORC1’ (the 20,978
ZD-ORC1 peaks in Fig. 3d), which represents those replication origins
that are regulated by the H2A.Z–SUV420H1–H4K20me2–ORC1 axis,
and the remaining ZD-ORC1 peaks as ‘minus-ZD-ORC1’. We found that
H2A.Z, H4K20me2, ORC1 and nascent-strand signals were all higher
around plus-ZD-ORC1 peaks than around minus-ZD-ORC1 or ZI-ORC1
peaks (Fig. 3e). These results support the essential role of H2A.Z in
activating replication origins with a higher firing efficiency through
the SUV420H1–H4K20me2–ORC1 axis.
Analysis of the genome-wide distribution of plus-ZD-ORC1 showed
that some of these replication origins were located at promoter regions
(Extended Data Fig. 5k). In addition, we analysed the DNA sequence
features in the 20,978 plus-ZD-ORC1 peaks. Consistent with previous
observations^19 ,^20 , we found that G/C-rich and asymmetric A/T-rich
motifs were significantly (P < 0.01) enriched (Extended Data Table 1),
indicating cooperative regulation of origin selection and firing by
genetic and epigenetic elements.

H2A.Z regulates early-replication origins

To investigate whether the origins regulated by H2A.Z show any
preferred timing for replication, we performed BrdU-seq to map

replication timing in HeLa cells. As shown in Extended Data Fig. 6a,

about 80% of the early-replication domains that we identified overlap

with those defined previously by BrdU Repli-seq^21. We also found that

plus-ZD-ORC1 peaks were preferentially enriched at early-replication

domains, whereas both minus-ZD-ORC1 and ZI-ORC1 peaks were more

enriched at late-replication domains (Fig. 4a). Moreover, plus-ZD-ORC1

peaks were more enriched at the centres of early-replication domains

than were the other ORC1 peaks (Extended Data Fig. 6b). Of note, the

10-min BrdU signal (for cells labelled for 10 min immediately after being

released from G1/S arrest) in plus-ZD-ORC1 is higher than in other ORC1

peaks, and it decreased obviously after H2AFZ knockdown (Fig. 4b).

These results suggest that the H2A.Z–SUV420H1–H4K20me2–ORC1

axis preferentially licenses and activates early-replication origins—a

conclusion supported by real-time PCR analyses of H4K20me2, ORC1

and nascent strands (Fig. 4c and Extended Data Fig. 6c). Next we ana-

lysed dynamic changes in replication timing after H2AFZ knockdown,

finding that, although the replication timing of early-replication

domains did not change substantially, the timing of late-replication

regions was advanced (Fig. 4d, e and Extended Data Fig. 6d, e). We

then arrested H2A.Z-depleted cells at the G1/S boundary. After release,

H2A.Z-depleted cells progressed through S phase without any defects

(Extended Data Fig. 7a, b), suggesting that, after H2AFZ knockdown,

21,685

11,522

7,951

4,119

29,390

2,939

4,692

4,769

20,978

34,036

5,917

12,703

d

ZD-K20

(45,277)

Nascent

strands

(41,850)

ZD-ORC1

(73,633)

f

H2A.Z

siNCsiH2A.Z

H4K20me2

siNC siH2A.Z

ORC1

siNC siH2A.Z

Nascent strands

siNCsiH2A.Z

0200 200 0.6 0 0.6 0 0.4 0 0.4 0 0.8 0 0.8

–5K05K–5K05K–5K05K–5K05K–5K05K–5K05K–5K05K–5K05K

c

H4K20me2

35

28

21

14

7

0

ORC1

12

8

4

0

RPKM

15

10

5

0

H2A.Z

4

6

8

2

0

Nascent

strands

22,076 19,774

a

22,852

23,548

24,329

36,320

36,530

9,26110,170

8,776

7,474

19,055

29,326

31,492

0

1

2

3

4

RPKM of nascent strands

OverlapOthers

H2A.Z

(58,642)

H4K20me2

(99,574)

ORC1

(100,917)

Nascent strands

(41,850)

<2.2 × 10 –16

H4K20me2

ORC1

Nascentstrands

H2A.Z

siNC

siH2A.Z

RefSeq

siNC

siH2A.Z

siNC

siH2A.Z

b

100 kb

RFPL3 BPIFC FBXO7

0

1.5

0

0.3

0

0.2

0

0.2

0

0.2

0

1

0

2

0

1.5

0

0.8

0

0.8

0

0.4

0

0.4

0

2.5

0

2.5

FBXO7

Chromosome22:

31,065,220–31,432,359

0.0068

Plus-ZD-ORC1 Minus-ZD-ORC1 ZI-ORC1

RPKM

0

1

2

0

2

5

1

3

4

0

2

4

0

2

6

4

H2A.Z H4K20me2 Nascent strands

e

ORC1

0. 0068

0

2

1

3

4

0

2

0

2

4

SUV420H1

Origin selection

H4K20me2

Helicase Canonical nucleosome H2A.Z nucleosome

SUV420H1 ORC1Replication machine

Origin ring

1

ORC1

1.1 × 10 –128^234

3.1 × 10 –236

9.4 × 10 –27

4.2 × 10 –16

4.3 × 10 –45

5 × 10 –4

<2.2 × 10 –16

<2.2 × 10 –16 <2.2 × 10 –16 <2.2 × 10 –16 <2.2 × 10 –16

Fig. 3 | H2A.Z regulates the recruitment of ORC1 to chromatin in order to
license origins with higher firing efficiency. a, The top Venn diagram shows
the overlap among H2A.Z, H4K20me2 and ORC1 ChIP-seq peaks. The bottom
Venn diagram shows the overlap between the 31,492 ORC1 peaks and nascent-
strand peaks. The box plot shows the nascent-strand signal in the nascent-
strand peaks that overlap the H2A.Z, H4K20me2 and ORC1 peaks (‘overlap’,
n = 19,774) and the other nascent-strand peaks peaks (‘others’, n = 2 2 ,076).
RPKM, reads per kilobase per million mapped reads. b, Genome tracks show
the signals of H4K20me2, ORC1 and nascent strands. RFPL3, BPIFC and FBO7 are
RefSeq gene names. c, Heat maps and violin plots show the signals of H2A.Z,
H4K20me2, ORC1 and nascent strands around the centres of H2A.Z peaks that
overlap both H4K20me2 and ORC1 (n = 19,055). d, Venn diagram showing the
overlap among ZD-K20, ZD-ORC1 and nascent-strand regions. e, Box plots
showing the signals of H2A.Z, H4K20me2, ORC1 and nascent strands in plus-

ZD-ORC1 (n = 20,978), minus-ZD-ORC1 (n = 52,655) and ZI-ORC1 (n = 27, 2 8 4)

regions. f, Working model. Top, origin selection: H2A.Z nucleosomes bind

SUV420H1 directly (step 1) to establish H4K20me2 on chromatin (step 2), which

then recruits ORC1 (step 3) to bind to replication origins (red mark; step 4).

Bottom, origin firing: the H2A.Z–SUV420H1–H4K20me2–ORC1 axis selectively

licenses and activates early replication origins. The experiment in b was

independently repeated two times with similar results. Data in a, c, e were

analysed by two-tailed Wilcoxon test; P-values are shown in the box plots. For

the box plots in a and e, the centre line represents the median, the box limits are

the 25th and 75th percentiles, and the whiskers are the minimum to maximum

values. For the violin plots in c, the box centres represent the median, the box

limits are the 25th and 75th percentiles, and the upper and lower limits make

the 95% confidence interval.