Nature | Vol 578 | 13 February 2020 | 315(Fig. 4b, e). The small RNAs derived from the reciprocal plasmid library
that immunoprecipitated with BmAgo3 also exhibited preferences for
–10A and +4C. However, the nucleotide bias around the BmZuc cleavage
site (–1U, 0U and +1U) was markedly lower for small RNAs immunopre-
cipitating with BmAgo3 than for those immunoprecipitating with Siwi
(Fig. 4e). Thus, BmZuc has similar but distinct nucleotide preferences
for Siwi-loaded and BmAgo3-loaded pre-pre-piRNAs.
To determine whether this difference is also observed in the endog-
enous piRNA loci, we analysed the nucleotide frequency around the
3′ end of Siwi- or BmAgo3-dominant type-E pre-piRNAs in the NaIO 4 -
treated Tri-KO small RNAs (Extended Data Fig. 3b). The –10A, –2A and
+4C sequences were fairly well conserved in the type-E pre-piRNAs for
both Siwi and BmAgo3 (Extended Data Fig. 5f ). Notably, as observed
in the randomized libraries, Siwi-dominant and BmAgo3-dominant
Type-E pre-piRNAs showed differences around the BmZuc cleavage site
(–1U, 0U and +1U), with the +1U bias nearly lost in BmAgo3-dominant
type-E pre-piRNAs (Extended Data Fig. 3d, 5f ). These data suggest that
the identity of PIWI proteins can influence the substrate specificity
of BmZuc (Fig. 4e, Extended Data Fig. 5f ) as well as the production of
downstream trailing piRNAs (Extended Data Fig. 3e). In other words,
although +1U is tightly linked with trailing piRNA production, it is not
a prerequisite for BmZuc-mediated cleavage per se. The strong +1U
bias observed for fly Piwi-bound piRNAs^12 ,^13 and mouse MILI- and MIWI-
bound pachytene pre-piRNAs^2 ,^4 –^6 might reflect the efficient production
of trailing piRNAs for these PIWI proteins.
Having uncovered BmZuc motifs that help to define the 3′-end of
endogenous type-E pre-piRNAs, we determined whether we could
use these motifs to predict where BmZuc cleaves in pre-pre-piRNAs.
We defined the similarity score to the weighted BmZuc motif (Fig. 4e,
Extended Data Fig. 5g, see Methods) and plotted this against the posi-
tions of the predicted BmZuc cleavage sites. The position that gave
the maximum similarity score agreed well with the actual peak length
(that is, the actual BmZuc cleavage site) of endogenous type-E pre-
piRNAs (Fig. 4f). Moreover, the in vitro cleavage patterns of Siwi-loaded
84497 or 111750 RNAs matched well with their calculated similarity
scores (Fig. 4g, Extended Data Fig. 5h), highlighting the critical role of
the BmZuc motif in determining the cleavage site. In fact, the simple
presence or absence of BmZuc motifs in pre-pre-piRNAs can dictate by50 nt —34 nt —a
–10 –2–1 0 +1 +4Putative BmZuc cleavage site
Relative position :
N N A N N N N N N N A U U U N N CbcdefBufferBmZuc + BmArmiBmZuc (HN) BufferBmZuc + BmArmi BmZuc (HN)BufferBmZuc + BmArmi BmZuc (HN) BufferBmZuc + BmArmi BmZuc (HN) BufferBmZuc + BmArmi BmZuc (HN) BufferBmZuc + BmArmiBmZuc (HN)Original All mut.+1U mut. –1U/0U mut.
exc. for Mut.
+1U –10A/–2A/+4Cmut.Siwi-loaded 84497^84497OriginalAll mut.+1U mut.Mut. exc. for +1U–1U/0U mut.–10A/–12A/+4C mut.
4
3(^21)
0
BmZuc score
4038 for Siwi
32
Nucleotide position
34
36
84497 original
8$&&8&8888&$&8&8$8&&8$&$*8*$$888 3488 &8&&$$&$&
Expected BmZuc cleavage site by deep seq
–10 –2 0 +1+4
84497 –1U/0U mut.84497 –10A/–2A/+4C mut. $$&8&888&
84497 +1U mut.84497 mut. exc. for +1U $&&8$88&&
84497 all mut. &&&
h
g
pre-pre-piRNA
Modest +1U
Siwi
Siwi
Me
Trimmer
U Trailing piRNA production
BmZuc motif
Mature piRNA
Me: 2′-O-methylation
pre-piRNA
pre-piRNA
U
BmZuc
BmAgo3
Trailing piRNA production
pre-pre-piRNA
BmZuc motif No +1U
~31–42 ntBmZuc
~31–42 ntPIWI
ACA Me
~31–42 nt
~31–42 ntPIWI
AAUUC Me
N
BmAgo3-dominant
piRNA loci (1,259 species)
Type-N
01
23
piRNA loci (1,946 species) 3132333435363738394041
(log
)Normalized 2
nucleotide frequency
Siwi-IP
Relative position (3′end = 0)
–10 –5–2 01 45
3 ′-end
–1.2
–0.8
–0.4
0
0.4
0.8
1.2
UG
CA
–10 –5–2 01 4 5(nt)
BmAgo3-IP
3 ′-end
–1.2
–0.8
–0.4
0
0.4
0.8
1.2
Small RNA library derived from plasmid transcripts
Deep sequencing
Restore original sequences
randomizedLibrary of
region
Siwi BmZuc
(^16) NNNNNNNNNNNNNNN
Identify cleavage sites
NNNN....NNNNN
BmAgo3-dominant piRNA target site Randomized region(N35) BmAgo3-dominant piRNA target site
Plasmid-based random seq library
MockBmZuc + BmArmi BmZuc (HN)
Tri-KO cells
NNNNN....NNNNN NNNNNNNNNNNNNNN
5 ′ 3 ′
~2.2 × 10^5 variations
score forBmZuc
Siwi
0
2
4
Siwi-dominant
3132333435363738394041
Type-N
Type-E Type-E
Fig. 4 | Identif ication of BmZuc consensus motifs. a, Schematic of the screen
to identify BmZuc motifs. b, Top 6 nucleotides with highest frequencies
(coloured) in small RNAs derived from the randomized library, aligned to the
BmZuc cleavage site. c, RNA substrates used in d. The top 6 nucleotides in the
BmZuc motif are shown in colour and their mutations (mut.) are shown in black
(exc., except). d, BmZuc-mediated cleavage assay using Siwi-loaded
84497-derived RNAs. Each gel image was adjusted to equalize the loading
signal. See also Extended Data Fig. 5b, c. e, Nucleotide frequency around the
BmZuc cleavage sites for Siwi-bound (left) or BmAgo3-bound (right) small
RNAs with peak lengths of 31−44 nt derived from the randomized sequence
libraries. The frequency was normalized to the nucleotide composition in the
library of the randomized region. The 6 nucleotides in the BmZuc motif are
highlighted. See also Extended Data Fig. 5d−f. f, Similarity scores with the
weighted BmZuc motif (BmZuc score) for Siwi or BmAgo3 were calculated for
extracted genomic sequences from Siwi- or BmAgo3-dominant piRNA loci in
sliding windows and plotted in the same order as in Fig. 1c. Red lines indicate
the actual 3′ ends of Tri-KO type-E small RNAs in Fig. 1c. See also Extended Data
Fig. 5g, i. g, BmZuc scores for Siwi were calculated for 84497-derived RNAs in
sliding windows and plotted as in d. See also Extended Data Fig. 5g. h, A model
for pre-piRNA production and 3′-end maturation in silkworms. Silkworm pre-
piRNAs are generated via two parallel endonucleolytic mechanisms, BmZuc-
mediated cleavage and PIWI-catalysed slicing. BmZuc shows similar but
distinct nucleotide preferences between Siwi- and BmAgo3-loaded pre-pre-
piRNAs. The ‘+1U’ bias, the previously recognized hallmark of Zucchini-
mediated cleavage, is modest for Siwi but absent for BmAgo3. The 3′ ends of
pre-piRNAs are matured by Trimmer and Hen1, regardless of how they are
produced.