sensitive to cycloheximide and MG132, in-
hibitors of protein synthesis and the protea-
some (fig. S7). We conclude that NIN appears
to be proteolytically processed at a site internal
to the protein to produce a c40kDa N-terminal
product and a c55kDa C-terminal product, which
are only apparent 10 days afterSinorhizobium
melilotiinfection (Fig. 1C and fig. S8).DNF1
is a nodulation-induced component of a SPC
(fig. S9) that is expressed in the zone of the
nodule where bacterial infection occurs ( 12 ).
The DNF1 complex has been implicated in
processing the signal peptide from nodule-
specific cysteine-rich (NCR) peptides ( 12 , 13 ),
which are necessary for bacteroid develop-
ment ( 14 , 15 ).dnf1mutants produce small,
white nodules ( 12 ) similar to two weaknin
alleles (fig. S5, G to I) ( 16 , 17 ), and all show
sparse intracellular bacteria that lack differ-
entiation (Fig. 2B). These internalized bacteria
do not transition into the nitrogen-fixing state
[nitrogenase iron protein (NifH) is lacking]
(Fig. 2A), and the nodules lack expression of
genes such as leghemoglobins and many NCR
and glycine-rich peptides (GRPs) (Fig. 3, A to C).
The similarity in the mutant phenotypes of
dnf1-2andnin-16ledustoassesswhetherthe
DNF1 complex may be associated with the
proteolytic processing of NIN. To this end, we
assessed NIN accumulation indnf1-2and found
neither the c40kDa nor the c55kDa products
(Fig. 2C), which cannot be explained by the ef-
fects ofdnf1mutation on transcription ofNIN
(Fig. 3B and fig. S10). The catalytic subunit of
SPC is conserved in eukaryotes (fig. S11A), and,
on the basis of prior knowledge ( 18 ), we pre-
dicted a NIN cleavage site at A400 (fig. S11, B
and C, and table S5). Mutation of A400 blocked
processing (fig. S11D) and recapitulated the
dnf1-2andnin-16phenotypes (Fig. 2D). This
alanine residue is conserved in NIN-like pro-
teins (NLPs), but the neighboring structure
differs, which may affect the ability of SPC to
process NLPs (fig. S11, E and F). Unexpectedly,
we did not observe reciprocal accumulation of
full-length NIN in any of the mutants that affect
processing; only a small accumulation of full-
length NIN was observed after treatment with
MG132 indnf1-2(fig. S12). We propose that
full-length NIN is unstable, and degradation
offull-lengthNINmaybefurtherfacilitatedby
other induced proteases indnf1-2(fig. S13).
The SPC is located on the endoplasmic retic-
ulum membrane ( 19 ), and we found that NIN
c40kDa and c55kDa were associated with mem-
brane and nuclear fractions, with nuclear pref-
erence (Fig. 2C and fig. S14). SPC primarily
removes signal peptides of proteins targeted
for secretion ( 20 ), however, other unusual tar-
gets, such as NIN, have been identified, and
the human SPC processes viral proteins lack-
ing signal peptides ( 21 , 22 ).
Overexpression ofNINin the absence of
S. melilotileads to the accumulation of low
levels of NIN c55kDa and higher levels of the
predicted full-length product (fig. S15). We
suggest that processing of overexpressed
NIN results from the constitutively expressed
SPC (fig. S9B). Overexpression ofNINis suf-
ficient to activate nodulation (fig. S16) and the
expression of many late genes associated with
nitrogen fixation that were not expressed in
nin-16anddnf1-2(Fig. 3, A to C). Transactiva-
tion assays reveal that the processed C-terminal
form of NIN can promote the expression of late
nodule-associated genes but not early nodule-
associated genes (Fig. 3D and fig. S17) from
which we conclude that NIN is sufficient to
activate early genes involved in nodule organ-
ogenesis and late genes associated with ni-
trogen fixation, but the latter step requires
processing by the DNF1 SPC.
The DNF1 SPC was proposed to process
NCR peptides ( 12 , 13 ),butherewesuggestan
additional, or alternative, target. WhereasNIN
is essential in many legumes, includingLotus
japonicus, NCR peptides are limited to legume
species that show bacteroid differentiation( 14 , 23 ), which excludesL. japonicus. To assess
the relative importance of the potential targets
of DNF1 processing, we generated adnf1mu-
tant inL. japonicus(Ljdnf1) (fig. S18). This mu-
tant recapitulated the phenotypes observed in
M. truncatula dnf1: small, white nodules with
few bacteria and reduced accumulation of NifH
(Fig. 4, A and B). In addition, while symbio-
somes in wild-type nodules often contained
multiple bacteria, symbiosomes ofLjdnf1con-
tained only single bacteria, whose membranes
were deformed, providing evidence of prema-
ture degradation (Fig. 4C). Using antibodies
generated againstL. japonicusNIN (a-LjNIN),
we found evidence for NIN processing in
L. japonicus, comparable to that observed in
M. truncatula(Fig. 4, D and E, and fig. S19),
with c60kDa and c38kDa fragments as primary
products (Fig. 4, D and E). As inM. truncatula,
accumulation of the processed forms occurred
10 days after inoculation withMesorhizobium
lotiand were depleted inLjdnf1(Fig. 4E). The
combination of comparable phenotypes fordnf1
mutants, along with the comparable processing630 29 OCTOBER 2021¥VOL 374 ISSUE 6567 science.orgSCIENCE
Fig. 2. DNF1 complexÐmediated processing of NIN during the transition to nitrogen fixation.(A) Bacterial
NifH in the nodules of wild-type (A17 and R108),dnf1-2, andnin-16.(B) Transmission electron microscopy
(TEM) images showing infection threads (it) and bacteroids (b) in wild-type (A17 and R108),dnf1-2, andnin-16
nodules. Scale bar, 2mm. (C) The c40kDa and c55kDa processed NIN products are present in membrane
(M) and nuclear fractions (N) of wild-type (A17 and R108) but absent innin-13,nin-16, anddnf1-2. Red asterisks
indicate NIN-specific products.a-H+ATPase (hydrogen adenosine triphosphatase) is loading control for M, and
a-histone H3 is the control for N. (D)Wild-type(A17)andnin-1transformed withNINandNINA400 mutants,
28 days afterS. melilotiinoculation. Number of transformed plants with nodules relative to total number
of transformed plants is shown for each image.RESEARCH | REPORTS