Nature | Vol 578 | 13 February 2020 | 313Fig. 3g). Type-N piRNA-2986 and 304, but not type-E piRNA-1528 and 66,
have downstream ping-pong sites with readily detectable complemen-
tary piRNAs, which can guide PIWI-catalysed slicing of pre-pre-piRNAs.
We examined small RNAs deriving from these loci in Tri-KO cells by
northern blotting, and confirmed the accumulation of correspond-
ing pre-piRNAs (Fig. 2b, mock, red lines). BmZuc depletion decreased
type-E, but not type-N, pre-piRNA levels (Fig. 2b), reinforcing the idea
that type-E pre-piRNAs are generated via BmZuc-mediated cleavage.
Notably, type-N pre-piRNAs showed many shorter heterogeneous
RNA fragments (Fig. 2b). These data suggest that pre-piRNAs gener-
ated by PIWI-catalysed slicing are intrinsically unstable and prone
to non-specific degradation, at least in the absence of Trimmer. This
could explain why the peak lengths of type-N small RNAs were 30 nt
or shorter in Tri-KO cells (Fig. 1c). By contrast, type-E pre-piRNAs were
more stable, especially pre-piRNA-66.
Given our observation that the 2′-O-methylation level was generally
higher for longer small RNAs than shorter ones in Tri-KO cells (Fig. 1b
and Extended Data Fig. 2i, j), we predicted that type-E pre-piRNAs are
efficiently 2′-O-methylated. Indeed, type-E pre-piRNA-1528 and pre-
piRNA-66 were refractory to NaIO 4 treatment (Fig. 2b). By contrast,
type-N pre-piRNA-2986 and pre-piRNA-304, as well as their degra-
dation products, were mostly—if not completely—shortened by one
nucleotide by a NaIO 4 β-elimination reaction (Fig. 2b). Thus, type-E
pre-piRNAs produced via BmZuc-mediated cleavage are more effi-
ciently 2′-O-methylated than type-N pre-piRNAs generated by down-
stream piRNA-guided slicing. Consistently, type-E pre-piRNAs 1528
and 66 became prone to nonspecific degradation upon depletion of
the 2′-O-methyltransferase BmHen1 in Tri-KO cells, whereas type-N
pre-piRNAs 2986 and 304 and their degradation products, which are
intrinsically poorly 2′-O-methylated, were largely unaffected (Fig. 2c).
Mature piRNAs were fully 2′-O-methylated in naive BmN4 cells, regard-
less of how their pre-piRNAs are generated (Fig. 2b, Naive BmN4), sup-
porting the model that 3′-end trimming by Trimmer is tightly coupled
with 2′-O-methylation by BmHen1^24.
Type-N and type-E piRNA loci are heterogeneously distributed even
within a single transposon (Extended Data Fig. 3h), suggesting that
how pre-piRNAs are produced is determined at the level of individual
piRNA loci. We also note that the separation between the type-N and
type-E groups is not absolute; there are many cases in which pre-piRNAs
are produced by both mechanisms, as represented by pre-piRNA-304
(Fig. 2b, asterisks) and pre-piRNA-1249 (Extended Data Fig. 3i).In vitro analysis of BmZuc activity
We next sought to recapitulate pre-piRNA production in vitro. We
previously established a cell-free system to monitor the 3′-end-trim-
ming reaction by Trimmer using mitochondria-containing 1,000g
pellets^24 (Extended Data Fig. 4a, left). Both Trimmer and BmZuc area5 ′ ends of antisense piRNAs in this region1 20355060 100 (nt)5 ′^5 ′5 ′ ends of sense piRNAs in this region120355060 1005 ′5' piRNA-1528
Type-ESiwipiRNA-66BmAgo3
Type-EpiRNA-2986Siwi
Type-NpiRNA-304
BmAgo3Type-N: RNAi
Naive BmN4Tri-KOMockBmHen1c
NaIO 4 /β :–+–+–+BmN4NaiveTri-KOMockBmZuc: RNAi
piRNA-1528
Type-ESiwipiRNA-66BmAgo3
Type-EpiRNA-2986Siwi
Type-NpiRNA-304
BmAgo3Type-N **b051015200510152025Antisense strand Sense strand2030405060708090100 2030405060708090100(^4152) Type-N
RPM per piRNA locus
0
5
10
15
20
0
5
10
15
20
25
Position from 5′-end of piRNA loci (nt)
2030405060708090100 2030405060708090100
(^4152) Type-E
Type-NType-E**P = 3.42 × 10 –3
Fig. 2 | The 3′ ends of type-E, but not type-N, pre-piRNAs are eff iciently 2′-O-
methylated. a, The 5′ ends of piRNAs mapped to 20−100 nt downstream of
piRNA loci were mapped on the antisense (left) or sense (right) genomic strand.
Type-N piRNAs have more antisense piRNAs at ~41−52 nt from the 5′ ends than
type-E piRNAs (two-sided Wilcoxon signed-rank test, n = 12). See also Extended
Data Fig. 3f. b, c, Northern blot analysis of the four representative piRNAs in
naive or Tri-KO BmN4 cells depleted of the indicated protein by RNAi. In b, total
RNAs were treated with or without NaIO 4 β-elimination. Asterisks, BmZuc-
dependent fragments. Red bars, putative pre-piRNAs. See also Extended Data
Fig. 3g.
36 nt –
27 nt –
80 nt –
a b
d
c
eg
h i
–+– +
BufferTri-KO 1,000
g pellet
(BmZuc + BmArmi)
NaIO 4 /β:
- BmZuc
Siwi
Tri-KO1,000g pellet
~36 nt
all U
Trimmer
f
1,000Tri-KOg pellet
: RNAi
BufferNaive 1,000MockBmZuc
g pellet
BmArmiBmGPAT1BmGasz
80 nt —
27 nt —
36 nt —
BmZuc
: RNAi
BmArmi
BmGPAT1
BmGasz
BmPapi
MockBmZucBmArmiBmGPAT1BmGasz
8&$$$$$&8$$&$88888&$$&8888^80
8&$$$$$&8$$&$88888&$$&8888^2840
28–8028–40U
U^1
1 Buffer 2 Naive 3 Tri-KO (BmZuc + BmArmi) 4 Tri-KO 5 Naive
1,000g pellet
Lane 3+
WB: Flag(BmZuc)
BmZuc cleavage
product
BufferNaive BmN4
Tri-KO
MockBmZuc (WT)BmZuc (HN)
1,000g pellet
36 nt — 40 nt —
27 nt —
BufferMockBmZuc (WT)BmZuc (HN)
BmArmi (WT)BmArmi (KA)
1,000Tri-KOg pellet
80 nt —
36 nt —
27 nt —
GFP (BmArmi)
WBFlag (BmZuc)
BmZuc (WT) + BmArmi (WT) BmZuc (WT) + BmArmi
(KA) : RNAi
BmGasz
piRNA-1528Siwi
Type-E
piRNA-66BmAgo3
Type-E
piRNA-2986Siwi
Type-N
piRNA-304
BmAgo3 Type-N
Naive BmN4
Tri-KO
MockBmZucBmArmiBmGPAT1
- BmZuc
Siwi
BmZuc + BmArmioverexpressed
Tri-KO1,000g pellet
Naive1,000g pellet
Trimmer
l
j
37
8&$$$$$&8$$&$88888&*$$^15 &&&&&&&&&^8 $^0
27–50C
:T 1,000ri-KOg
pellet
k
50 nt —
36 nt —
BufferMockBmZuc + BmArmiBmZuc (HN)BufferMockBmZuc + BmArmiBmZuc (HN)BufferMockBmZuc + BmArmiBmZuc (HN)BufferMockBmZuc + BmArmiBmZuc (HN)
allC 37U 37A 37G
0 0.5 1 1.5 2
allC
37U
37A
37G
36 nt signal / all signal (37A = 1)
36 nt –
80 nt –
Fig. 3 | BmZuc requires BmArmi, BmGPAT1 and BmGasz for cleavage of Siwi-
loaded pre-pre-piRNAs in vitro. a, Schematic of in vitro BmZuc cleavage assay.
See also Extended Data Fig. 4a. b, RNA substrates used in c–e, h and i. c–e, Siwi-
loaded 28–40U RNA (top, c) or 28–80U RNA (bottom; d, e) was incubated with
1,000g pellet from naive or Tri-KO cells overexpressing BmZuc(WT) or the
catalytic mutant BmZuc(HN) (c, d) and/or BmArmi(WT) or the ATP-binding
mutant BmArmi(K A) (d), or 1,000g pellet from Tri-KO cells depleted of the
indicated protein by RNAi (e). The expression of Flag-tagged BmZuc and GFP-
tagged BmArmi was confirmed by western blot (c, d, bottom). See also
Extended Data Fig. 4c. f, Western blot analysis of the 1,000g pellet used in e.
See also Extended Data Fig. 4e. g, Northern blot analysis of representative
type-E or type-N pre-piRNAs in Tri-KO cells depleted of the indicated proteins
by RNAi. Asterisks, BmZuc-dependent fragments. Red bars, putative pre-
piRNAs. h, i, Siwi-loaded 28–80U RNA was incubated with 1,000g pellet from Tri-
KO cells overexpressing BmZuc and BmArmi. After incubation, RNAs were
extracted and treated with NaIO 4 followed by β-elimination (h), or naive 1,000g
pellet was added and further incubated (i). j, RNA substrates used in k. k, Siwi-
loaded 2 7–5 0C RNAs bearing U, A or G at position 37 were incubated with 1,000g
pellet from Tri-KO cells overexpressing BmZuc and BmArmi or BmZuc(HN).
l, Quantification of the 36-nt cleavage fragments produced by 1,000g pellet
from Tri-KO cells overexpressing BmZuc and BmArmi in k. Data are mean ± s.d.
from four technically independent experiments. Bonferroni-corrected
P values from two-sided paired t-tests are as follows: P = 0.0163; P = 0.00106;
***P = 0.000485.