312 | Nature | Vol 578 | 13 February 2020
Article
Trimmer protein and the in vitro trimming activity (Extended Data
Fig. 2d−f ). Tri-KO cells lacked mature 27−28 nt piRNAs and accumulated
longer RNAs of about 30–40 nt (Extended Data Fig. 2g, red line) that
co-immunoprecipitated with Siwi or BmAgo3 (Fig. 1a). Overexpres-
sion of wild-type (WT) but not catalytically inactive Trimmer E30A
(EA) recovered mature-length piRNAs (Extended Data Fig. 2h). These
results suggest that silkworm pre-piRNAs are about 30–40 nt in length
and are trimmed by Trimmer for maturation, irrespective of which
PIWI protein they bind.
To characterize the pre-piRNAs in Tri-KO cells, we sequenced 20−50-nt
small RNAs from Tri-KO cells with or without NaIO 4 treatment, which
enables specific detection of 2′-O-methylated species. Small RNAs
mapping to well-defined 3,236 piRNA loci^3 showed a sharp distribution
at 27–28 nt in naive BmN4 cells, but had a broad length distribution
around 30–40 nt in Tri-KO cells (Fig. 1b). The small RNAs in Tri-KO cells
were largely protected from NaIO 4 treatment (Fig. 1b, Extended Data
Fig. 2i), suggesting that they are 2′-O-methylated, with longer species
more efficiently methylated than shorter ones (Extended Data Fig. 2j).
We determined the most frequent small RNA length (peak length) for
each piRNA locus in the NaIO 4 -treated library and plotted the peak
lengths for all 3,236 piRNA loci without considering the small RNA
abundance from each locus (Fig. 1c, Extended Data Fig. 2k). In con-
trast to mature piRNAs in naive cells (Extended Data Fig. 2k), the peak
lengths of Tri-KO small RNAs showed a clear bimodal distribution: one
peak at 27–28 nt and a broader peak around 35 nt (Fig. 1c, bottom). For
simplicity, we refer to the piRNA loci with peak length up to 30 nt as
type-N (non-extended) and those with peak length greater than 30 nt
as type-E (extended).
To test the requirement for Bombyx mori Zucchini (BmZuc) dur-
ing processing of pre-pre-piRNAs into pre-piRNAs in silkworms, we
knocked down BmZuc in Tri-KO cells. Notably, depletion of BmZuc did
not affect the length distribution of type-N small RNAs, but strongly
decreased peaks of length greater than 30 nt in the type-E small RNAs
(Fig. 1d, Extended Data Fig. 2l and Supplementary Note 1), suggest-
ing that BmZuc is required to produce pre-piRNAs from type-E loci.
Supporting this idea, the genomic nucleotide immediately following
the 3′ end of type-E small RNAs in Tri-KO cells tended to be U (Fig. 1e),
a proposed hallmark of Zucchini-mediated cleavage called the ‘+1U
bias’^2 ,^4 –^6 ,^12 ,^13. Moreover, type-E small RNAs in Tri-KO cells were fre-
quently accompanied by immediately downstream piRNAs on the same
genomic strand (Fig. 1e), a pattern typically observed in trailing piRNAs
or pre-piRNAs^2 ,^4 ,^12 ,^13. By contrast, these signatures were nearly absent
in Tri-KO type-N small RNAs (Fig. 1e). Taken together, we conclude
that BmZuc mediates the production of type-E pre-piRNAs, whereas
type-N pre-piRNAs are generated via a BmZuc-independent pathway.
We next investigated whether BmZuc generates pre-piRNAs for
both Siwi and BmAgo3, the two PIWI proteins in silkworms. To this
end, we first defined Siwi- and BmAgo3-dominant piRNA loci (Extended
Data Fig. 3a, see Methods). We then plotted the peak length of Tri-KO
small RNAs separately for Siwi- or BmAgo3-dominant piRNA loci, and
observed similar bimodal distributions corresponding to type-N and
type-E (Extended Data Fig. 3b). BmZuc depletion reduced the peak-
length populations of type-E small RNAs for both Siwi- and BmAgo3-
dominant piRNA loci (Extended Data Fig. 3c), suggesting that BmZuc
mediates type-E pre-piRNA production regardless of which PIWI protein
is bound to the pre-pre-piRNA. However, compared with Siwi-dominant
type-E pre-piRNAs, BmAgo3-dominant type-E pre-piRNAs showed a
weaker +1U bias (Extended Data Fig. 3d) and had a lower frequency
of immediately downstream piRNAs (Extended Data Fig. 3e). Thus,
even though BmZuc mediates pre-pre-piRNA cleavage for both Siwi
and BmAgo3, the production of downstream trailing piRNAs is largely
restricted to Siwi.
We next investigated how pre-piRNAs in the type-N group are gener-
ated. In flies, most Ago3-bound piRNAs are processed from pre-piRNAs
generated by piRNA-guided slicing at a downstream position^23 (Supple-
mentary Discussion). To determine whether the 3′ end of silkworm pre-
piRNAs can be generated by downstream slicing, we analysed sense and
antisense piRNAs mapped to the downstream region of type-N or type-E
piRNA loci. The abundance of sense piRNAs in the downstream region
was similar for type-N and type-E piRNA loci (Fig. 2a, sense strand). By
contrast, antisense piRNAs at approximately 41−52 nt from their 5′ ends
were observed more frequently in the downstream region of type-N
loci (Fig. 2a, antisense strand), for both Siwi-dominant and BmAgo3-
dominant loci (Extended Data Fig. 3f ). Antisense piRNAs in this region
can, in theory, guide slicing of pre-pre-piRNAs and generate the 3′ end
of 31−42 nt pre-piRNAs. Therefore, unlike in flies^12 ,^13 ,^23 , downstream
slicing of pre-pre-piRNAs in silkworms is probably determined by the
context of the cleavage site and not by the identity of PIWI proteins.
However, the peak lengths of Tri-KO small RNAs in the type-N group
(less than 30 nt) were shorter than the expected pre-piRNA lengths
based on the downstream slicing sites (31−42 nt) (Fig. 1c), implying
that they are somehow fragmented into shorter species in Tri-KO cells.
To further investigate pre-piRNA generation, we examined four
piRNA loci representing a Siwi-dominant (piRNA-1528) or BmAgo3-
dominant (piRNA-66) type-E locus, as well as a Siwi-dominant (piRNA-
2986) or BmAgo3-dominant (piRNA-304) type-N locus (Extended DataWB : Siwi30 nt —
20 nt —40 nt —50 nt —abcdeTri-KONaive20 2530 35 40 45RPM (×103 )0100200300400Length (nt)Select 1 piRNA locuspiRNA-1543Per cent in piRNA-1543 locus0(^105)
1520
25
20 25 30 35
34
40 45 (nt)
Per cent of piRNA loci^0
5
10
15
20
Tri-KO NaIO 4 (+)
20 25 30 35 40
Peak length (nt)
Type-N (≤ 30 nt) Type-E (≥ 31 nt)
20 25 30 35 40 45 (nt)
piRNA loci (3,236 species)
Type-E
Type-N
100%
0%
50%
————
————
—————
—————
Naive Naive Tri-KO
IP :IgGISiwi gGBmAgo3
Cell line :
Bound RNAs (5′-radiolabelling)
BmAgo3
Naive Naive Tri-KO
30 nt —
20 nt —
40 nt —
50 nt —
Tri-KOType-N
Tri-KOType-E
RPM (×10
3 )
202530354045
Peak = 28 nt
0
5
10
15
(^20) Mock RNAi
Z 28 = 3.9
BmZucZ RNAi
28 = 4
202530354045
Peak = 29 nt
0
5
10
(^15) Mock RNAi
Z 29 = 3.8
BmZucZ RNAi
29 = 3.7
Length (nt)
RPM (×10
3 )
202530354045
Peak = 35 nt
0
10
20
30
(^40) Mock RNAi
Z 35 = 4.5
BmZucZ RNAi
35 = 1.9
202530354045
Peak = 36 nt
0
10
20
30
(^40) Mock RNAi
Z 36 = 4.3
BmZucZ RNAi
36 = 2.1
A
GC
U
(3,180)Naive
Tri-KO
Type-N (≤ 30 nt)(1,456) Type-E (≥ 31 nt)(1,708)
Average frequency ofRelative position (nt, 3′ end of preceding small RNA = 0)
piRNA 5'-end (%)
+1
(^0) –100+10
2
4
6
8
(^10) +1
(^0) –100+10
2
4
6
8
(^10) +1
(^0) –100+10
2
4
6
8
10
Z+1 = –0.9 UGAACZ+1 = –0.2 UGACZ+1 = 4.9
Fig. 1 | Two types of small RNAs accumulate in Tri-KO cells.
a, Immunoprecipitated (IP) Siwi or BmAgo3 from naive or Tri-KO BmN4 cells
was analysed by western blotting (WB) (top) and bound RNAs were detected by
5′ radiolabelling (bottom). IgG, immunoprecipitation with non-immunized
rabbit IgG. b, Length distribution of small RNAs mapped to 3,236 piRNA loci in
the total small RNA library from naive or Tri-KO BmN4 cells. See also Extended
Data Fig. 2i, j. c, The most abundant small RNA length among the reads sharing
the same 5′ end was defined as the peak length for each piRNA locus (for
example, peak length = 34 nt for piRNA-1543, top). The 3,236 piRNA loci were
aligned in the order of their peak lengths in the Tri-KO-NaIO 4 librar y (middle).
piRNA loci with peak length of ≤30 nt were defined as type-N and those with
peak length of ≥31 nt were defined as type-E (bottom). See also Extended Data
Fig. 2k. d, Changes in the length distribution of NaIO 4 -treated Tri-KO small
RNAs bearing peak lengths of 28 or 29 nt (type-N), or 35 or 36 nt (type-E) caused
by depletion of BmZuc. Mock indicates knockdown for Renilla luciferase. See
also Extended Data Fig. 2l. Zn denotes the z score at position n. RPM, reads per
million. e, Mean occurrence of piRNA 5′ ends relative to the peak position of
each piRNA locus. Pie charts show the nucleotide composition immediately
after the peak position of each piRNA locus. The numbers of analysed piRNA
loci are shown in the parentheses. The per cent nucleotide composition in the
silkworm genome corresponding to positions 11−45 of piRNA loci is
26:23:22:29 (T:G:C:A).