314 | Nature | Vol 578 | 13 February 2020
Article
mitochondrial outer-membrane proteins, so we anticipated that
the same strategy could be applied to detect BmZuc activity in the
1,000g pellet from Tri-KO cell homogenate (Fig. 3a, Extended Data
Fig. 4a, right). Since Zucchini is thought to cleave 5′ of U, we first used
a series of single-stranded (ss)RNAs bearing a poly(U) sequence as
model substrates (Fig. 3b). Incubation of Siwi-loaded 40-nt poly(U)-
containing RNA with Tri-KO 1,000g pellet produced an RNA fragment
of about 36 nt, which is much longer than the mature trimming prod-
uct observed with naive 1,000g pellet. (Fig. 3c). The 36-nt fragment
was also observed when we used an 80-nt poly(U)-containing RNA
(Extended Data Fig. 4b, ATP+). For both 40- and 80-nt RNAs, overex-
pression of catalytically inactive BmZuc H141N (HN) decreased the
~36-nt signal (Fig. 3c, Extended Data Fig. 4b), suggesting that active
BmZuc is required to generate this fragment. Thus, BmZuc catalyses
the production of the 36-nt RNA fragment in vitro, regardless of the
initial length of Siwi-loaded poly(U)-containing RNAs. This is consist-
ent with the idea that the PIWI proteins themselves position Zucchini
on pre-pre-piRNAs^2.
Depletion of ATP from the in vitro reaction abolished the 36-nt cleav-
age product (Extended Data Fig. 4b, ATP−), suggesting that BmZuc-
mediated cleavage requires ATP. Purified Zucchini cleaves ssRNAs
in an ATP-independent manner^10 ,^11 ,^21 , whereas Armitage (MOV10L1 in
mice)—a factor required for the biogenesis of trailing piRNAs—is an ATP-
dependent RNA helicase^25 ,^26. To examine whether BmArmi is required
for BmZuc-mediated cleavage in vitro, we overexpressed wild-type
BmArmi or its ATP binding mutant, K692A (KA), with or without BmZuc
in Tri-KO cells. Overexpression of wild-type BmArmi alone strongly pro-
moted in vitro cleavage of the Siwi-loaded 80-nt RNA, suggesting that
BmArmi is a rate-limiting factor for BmZuc-mediated cleavage (Fig. 3d).
By contrast, overexpression of BmArmi(KA) inhibited the cleavage
reaction, indicating the importance of the ATPase activity (Fig. 3d).
Knockdown of BmZuc or BmArmi in Tri-KO cells abolished the produc-
tion of the 36-nt RNA fragment in vitro and biogenesis of endogenous
type-E, but not type-N, pre-piRNAs, confirming their requirement for
the cleavage reaction (Fig. 3e–g, Extended Data Fig. 4c). In addition to
BmArmi, BmZuc-mediated cleavage required two other proteins local-
ized on the mitochondrial surface, BmGPAT1 and BmGasz (Fig. 3e–g,
Extended Data Fig. 4c–e, Supplementary Discussion), homologues of
which have been genetically implicated in Zucchini-mediated piRNA
production in flies and mice^12 ,^27 –^30.
We found that the BmZuc in vitro cleavage product of about 36 nt
was at least partly resistant to NaIO 4 treatment, suggesting that it is
protected by 2′-O-methylation (Fig. 3h). Thus, our in vitro system
properly recapitulates BmZuc-mediated cleavage of pre-pre-piRNAs
and the production of 2′-O-methylated type-E pre-piRNAs. Finally, we
examined whether Trimmer can trim the cleavage product generated
by BmZuc to produce mature piRNAs, recapitulating processing in vivo.
The 36-nt BmZuc cleavage product was efficiently converted into 27−28-
nt mature piRNAs by naive 1,000g pellet, which contains endogenous
Trimmer (Fig. 3i). This result validates the stepwise 3′-end maturation
mechanism of type-E piRNAs: BmZuc cleaves pre-pre-piRNAs to gener-
ate pre-piRNAs, which are trimmed to the mature length by Trimmer.
Previous genetic and deep-sequencing analyses have suggested
that Zucchini preferentially cleaves immediately 5′ to U in vivo^2 ,^4 –^6 ,^12 ,^13.
We also observed that the 3′ ends of type-E pre-piRNAs in Tri-KO cells
have a modest +1U bias, especially for type-E pre-piRNAs bound to
Siwi (Fig. 1e, Extended Data Fig. 3d). However, previous biochemical
analyses using purified Zucchini proteins and naked RNAs have failed to
detect this U preference^10 ,^11 ,^21. We applied our new in vitro system using
mitochondria-containing pellets and Siwi-loaded pre-pre-piRNAs to
revisit this inconsistency. We performed the in vitro BmZuc cleavage
assay with a 50-nt RNA bearing a poly(C) sequence as well as variants
that substituted the C at position 37 with U, A or G (Fig. 3j). Compared
with the 37A, 37G and 37C RNAs, the 37U RNA substrate yielded mod-
erately but significantly increased levels of the 36-nt cleavage product,
in a manner dependent on the catalytic activity of BmZuc (Fig. 3k, l).
Thus, our system recapitulates the U preference of BmZuc, consistent
with our bioinformatics analysis of type-E pre-piRNAs (Fig. 1e, Extended
Data Fig. 3d).BmZuc motif dictates piRNA biogenesis
The moderate U preference of BmZuc was apparent for poly(C)-based
sequences in vitro (Fig. 3j−l). However, BmZuc does not always cleave
immediately 5′ to U in our in vitro system (Extended Data Fig. 4f ). More-
over, natural type-E pre-piRNAs showed only a modest +1U bias (Fig. 1e,
Extended Data Fig. 3d). Thus, the proposed U preference does not fully
explain how the cleavage site is chosen by BmZuc. To investigate the
substrate specificity of BmZuc in a comprehensive and unbiased man-
ner, we performed a screen in Tri-KO cells. In brief, we constructed a
plasmid-based library that expresses 35 nt of random sequence flanked
by target sites for an abundant BmAgo3-dominant piRNA^31 (Fig. 4a,
see Methods and Supplementary Note 2). The transcripts are expected
to be sliced by the BmAgo3-dominant piRNA, loaded into Siwi via the
ping-pong pathway as new pre-pre-piRNAs, and cleaved by BmZuc
within the downstream randomized region (Fig. 4a), producing various
type-E pre-piRNAs. We first sequenced the library-derived small RNAs
and examined their peak length distribution (Extended Data Fig. 5a).
We observed library-derived small RNAs around 35 nt, recapitulating
the size range of endogenous type-E pre-piRNAs. These approximately
35-nt RNAs were enhanced by overexpression of BmZuc and BmArmi,
and strongly inhibited by BmZuc(HN), indicating that they are gener-
ated by BmZuc-mediated cleavage. We then aligned them at the 3′ ends
of their peak length (that is, putative BmZuc cleavage sites, defined
as position 0) and analysed the nucleotide frequencies at each posi-
tion. Focusing on the six nucleotides with the highest frequencies, we
identified a sequence motif (−10A, −2A, −1U, 0U, +1U, +4C) (Fig. 4b),
which was also consistently observed in the BmZuc + BmArmi over-
expression condition.
To further investigate the ‘BmZuc motif ’ in Siwi-bound pre-pre-piRNAs,
we analysed two representative sequences from the library (84497
and 111750) that contain all six consensus nucleotides. We generated a
series of mutants that alter the consensus sequence and performed the
BmZuc cleavage assay (Fig. 4c, d, Extended Data Fig. 5b, c). Wild-type
sequences showed site-specific cleavage at 34 or 35 nt, as expected
from the in-cell screen. By contrast, ‘All mut’ sequences, in which all the
six consensus nucleotides were mutated, lacked site-specific cleavage
(Fig. 4d, Extended Data Fig. 5c), providing further supporting evidence
that this motif determines the BmZuc cleavage site. Unexpectedly,
mutating only the +1U, the proposed hallmark of Zucchini-mediated
cleavage, did not inhibit BmZuc-mediated site-specific cleavage of
these sequences (Fig. 4d, Extended Data Fig. 5c, +1U mut.). However,
mutating –1U and 0U together strongly inhibited the cleavage at the
correct position, whereas mutating –10A, –2A and +4C together had
a minor effect. In sum, our findings reveal a previously unrecognized
consensus motif that is important for BmZuc to precisely determine
the cleavage site.
BmZuc generates the 3′ ends for both Siwi- and BmAgo3-loaded pre-
piRNAs (Extended Data Fig. 3b, c). To investigate whether BmZuc has a
different nucleotide preference for pre-pre-piRNAs bound to BmAgo3,
we constructed a reciprocal plasmid library whose transcripts were
loaded into BmAgo3 as pre-pre-piRNAs with a randomized sequence.
We then performed co-immunoprecipitation with Siwi from the Tri-KO
cells transfected with the original library and co-immunoprecipitation
with BmAgo3 from the cells transfected with the reciprocal library,
and analysed the bound small RNAs with peak lengths of 31−44 nt after
NaIO 4 treatment (Extended Data Fig. 5d, e). As expected, the small
RNAs immunoprecipitating with Siwi showed very similar nucleotide
preferences around their 3′ ends as the BmZuc motif identified by the
non-immunoprecipitation experiment using the same plasmid library