Stepwise articial evolution of a plant disease

resistance gene
C. Jake Harrisa, Erik J. Slootwegb, Aska Goverseb, and David C. Baulcombea,1
a
Department of Plant Sciences, University of Cambridge, Cambridge CB2 3EA, United Kingdom; and bLaboratory of Nematology, Department of Plant
Sciences, Wageningen University, 6708 PB, Wageningen, The Netherlands

Genes encoding plant nucleotide-binding leucine-rich repeat (NBLRR) proteins confer dominant resistance to diverse pathogens.
The wild-type potato NB-LRR protein Rx confers resistance against
a single strain of potato virus X (PVX), whereas LRR mutants
protect against both a second PVX strain and the distantly related
poplar mosaic virus (PopMV). In one of the Rx mutants there was
a cost to the broad-spectrum resistance because the response to
PopMV was transformed from a mild disease on plants carrying
wild-type Rx to a trailing necrosis that killed the plant. To explore
the use of secondary mutagenesis to eliminate this cost of broadspectrum resistance, we performed random mutagenesis of the Nterminal domains of this broad-recognition version of Rx and
isolated four mutants with a stronger response against the PopMV
coat protein due to enhanced activation sensitivity. These mutations are located close to the nucleotide-binding pocket, a highly
conserved structure that likely controls the switch between active and inactive NB-LRR conformations. Stable transgenic plants
expressing one of these versions of Rx are resistant to the strains
of PVX and the PopMV that previously caused trailing necrosis. We
conclude from this work that articial evolution of NB-LRR disease
resistance genes in crops can be enhanced by modication of both
activation and recognition phases, to both accentuate the positive
and eliminate the negative aspects of disease resistance.
plant immunity

| genetically modied | arms race | NLR | plant defense

lant resistance (R) genes confer dominant resistance against

diverse pests and pathogens including viruses, fungi, and
invertebrates (1). The vast majority of R genes encode NB-LRR
(nucleotide-binding leucine-rich repeat) proteins (2) that form
a subgroup within the STAND ATPases (signal transduction
ATPases with numerous domains) (3). NB-LRRs are thought to
function as a molecular switch, existing in ATP-bound active
or ADP-bound inactive states (47). Activated NB-LRRs initiate a downstream signaling cascade triggering defense responses,
which often culminate in a form of programmed cell death know
as the hypersensitive response (HR) (8).
Activation of NB-LRRs occurs following molecular recognition in which a pathogen-derived elicitor interacts, either directly
or indirectly, with the LRR and/or other domains of the protein
(915). In different R proteins the molecular mechanisms may
vary but, in Rx from potato (16), which confers resistance to
potato virus X (PVX), the domains required to initiate downstream signaling (17, 18) are likely exposed by conformational
changes (19, 20) following recognition.
NB-LRR genes represent a useful target for generating diseaseresistant crops, and have long been selected unknowingly by crop
breeders. More recently, transgenic approaches have been used
to transfer R genes between plant species (2123). However, the
usefulness of NB-LRR genes in both conventional breeding and
transgenic approaches is limited by the availability of R protein
genes with useful recognition specicities. In addition, there may
be a cost to carrying NB-LRR R genes. The disease-resistant
plants may have reduced tness in competition with susceptible
plants (24), there may be a tradeoff with plant growth (25),
the plants may exhibit hybrid incompatibility (26), or they may
www.pnas.org/cgi/doi/10.1073/pnas.1311134110

exhibit a partial resistance against some strains of pathogen so

that there is a spreading necrosis that kills the plant (2730).
Articial evolution through random mutagenesis has been
explored previously as a strategy to expand the number of useful
variants of Rx (27). The evolved products were forms of Rx with
mutant LRR domains that recognized more strains of PVX than
the wild-type protein. The elicitor of Rx-mediated resistance is
the viral coat protein (CP), and the wild-type protein responds to
strains in which there is a T and a K at positions 121 and 127,
respectively (CPTK), but not those with a K and an R at these
positions (CPKR) (31). The articial evolution focused on the
LRR domain (27), and the selected mutants were elicited by
both CPKR and the original CPTK. However, this broad recognition had a cost in that one of the mutants, RxM1 (N846D),
displayed systemic necrosis when the plants were challenged with
poplar mosaic virus (PopMV). This virus is a distant relative of
PVX, and on plants with wild-type Rx, or without Rx, it only
induces a mild mosaic. To explain the lethal symptom associated
with PopMV on RxM1 plants we proposed that the mutation
mediated weak recognition of a PopMV structure that was invisible to the wild-type Rx. The systemic necrosis would have
arisen because the weak recognition by Rx triggers a delayed HR
response that is too late to prevent spread of the virus from the
site of initial infection.
This side effect of the M1 mutation illustrates the balance
between the costs and benets of disease resistance, and it created an opportunity to nd out whether articial evolution could
be used to reduce the costs associated with disease resistance of
NB-LRR proteins. Our approach was random mutagenesis of
RxM1 combined with selection for response to the PopMV coat
protein rather than CPKR. In addition, rather than the LRR, we
mutagenized the amino-terminal coiled coil (CC), NB, and
Signicance
Plant nucleotide-binding leucine-rich repeat (NB-LRR) proteins
are responsible for detecting potential pathogens and triggering a defense response. However, disease resistance can
also involve a cost. Here we show that the trailing necrosis cost
associated with a broad-recognition version of the NB-LRR
protein, Rx, can be overcome by selecting for mutations that
increase the sensitivity of the NB-LRR protein. Using homology
modeling, we predict that these mutations are colocalized to
the ATP/ADP nucleotide-binding pocket, which is thought to
control the switch between active and inactive states.
These results suggest that a stepwise approach to coordinating
recognition and response could be used to design improved
NB-LRRs for use in agriculture.
Author contributions: C.J.H., A.G., and D.C.B. designed research; C.J.H. performed research; E.J.S. contributed new reagents/analytic tools; C.J.H. and E.J.S. analyzed data;
and C.J.H. and D.C.B. wrote the paper.
The authors declare no conict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
1

To whom correspondence should be addressed. E-mail: dcb40@cam.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.

1073/pnas.1311134110/-/DCSupplemental.

PNAS | December 24, 2013 | vol. 110 | no. 52 | 2118921194

PLANT BIOLOGY

Edited by Brian J. Staskawicz, University of California, Berkeley, CA, and approved November 13, 2013 (received for review June 12, 2013)

ARC (shared by Apaf-1, resistance genes, and CED-4) domains.

By selecting mutants with enhanced recognition of PopMV, we
predicted to isolate mutants that give a strong and rapid response
to this virus and eliminate the spreading HR. By focusing on
domains other than the LRR, we reasoned that we would avoid
compromising the benecial aspects of the RxM1 phenotype.
Here we describe ve amino acid changes that gave the selected response to the PopMV-CP. These mutants increase the
activation sensitivity rather than the recognition phase of the Rx
resistance mechanism and, from protein modeling, we conclude
that they are localized around the conserved ATPase nucleotidebinding pocket. Transgenic plants containing these versions of
Rx retained broad-spectrum resistance against PVX strains and
PopMV without systemic necrosis: The cost of RxM1 resistance
had been eliminated. The results show that Rx can be enhanced
by a stepwise process that targets both recognition and activation
mechanisms. Based on these results, we suggest that articial
evolution could be a useful enhancement of all disease resistance
that is to be transferred by genetic manipulation of NB-LRRs.
Results
Articial Evolution Screen Across RxM1 N-Terminal Domains. R genes
can be coexpressed with cognate pathogen proteins in leaf segments of Nicotiana tabacum or N. benthamiana using an Agrobacterium-mediated transient expression assay (32). If the R
gene recognizes the pathogen protein, a hypersensitive response
is induced that manifests as necrosis in the inltrated leaf segment. RxM1 does not induce HR when transiently coexpressed
with the coat protein of PopMV (PopMV-CP), but RxM1
transgenic plants display trailing necrosis when infected with
PopMV, and electrolyte leakage assays suggest that RxM1 mounts
a delayed response to PopMV (27).
To nd versions of RxM1 with an accelerated PopMV response, we generated a library of 1,500 RxM1 variants by randomly
mutagenizing the N-terminal CC-NB-ARC1-ARC2 domains (Fig.
1A) at an error rate of 5.1 bp changes per 1.5-kb molecule, ensuring an average coverage of 5 bp changes per position. The
selection was by transient coexpression of these mutants with
PopMV-CP. The weak recognition of PopMV-CP by RxM1 does
not lead to visible necrosis but, of the 1,500 RxM1 variants, we
identied 22 RxM1 mutants that produce a necrotic response
(Table S1). Of these, there were 18 candidates that also produce
a necrotic response when coexpressed with GFP, indicating that
these mutations cause constitutive activation of the protein. The
remaining four clones, denoted RxS1*M1, RxS2*M1, RxS3*M1,
and RxS4M1, induced necrosis when coexpressed with PopMVCP but no necrosis when coexpressed with GFP (Fig. 1 B and C
and Fig. S1A) and were selected for further analysis. The basis for
this nomenclature is explained in later sections of this text: S is
for sensitized, and the asterisk indicates that the mutants contain
more than one amino acid change.

RxM1

CC
Chlorophyll content (g/mm2)

N846D
NB-ARC

We used a chlorophyll content assay to quantify the level of

necrosis induced by different Rx variant/elicitor combinations
after transient expression in N. tabacum (33). The four candidates, RxS1*M1, RxS2*M1, RxS3*M1, and RxS4M1, retained
broad recognition of PVX-CP, producing a strong necrotic response when coexpressed with both versions of the PVX coat
protein, PVX-CPTK and PVX-CPKR (Fig. 1C). The RxM123
construct, which contains three previously identied broad recognition-conferring mutations combined in the LRR domain
(N846D, N796D, and L607P, respectively) was used as a positive
control for recognition of PopMV-CP (27). Western blots of
inltrated leaf tissue indicated that the mutations in RxS1*M1,
RxS2*M1, RxS3*M1, and RxS4M1 do not increase Rx protein
stability (Fig. S1B). This control observation rules out that the
phenotypes are due to enhanced accumulation of the mutant
NB-LRR proteins.
Sensitized Mutations Are Close to the ATP/ADP Binding Site. The
RxS1*M1, RxS2*M1, and RxS3*M1 proteins each contain more
than one amino acid change in addition to M1 (N846D). RxS1*M1
contains four mutations (L179M, R291C, M293L, K482N);
RxS2*M1 contains three (E55K, T178A, P187Q); RxS3*M1 contains four (A88T, N147D, V337A, V362I); and RxS4M1 contains
only one (G340R) (Fig. 2A, Fig. S2A, and Table S1). To determine
which amino acid changes are responsible for the PopMV-CP
response in the RxS1*M1, RxS2*M1, and RxS3*M1 clones, we
used site-directed mutagenesis to generate single-amino acid
change constructs corresponding to each mutation.
All single-amino acid change constructs produced functional
proteins that respond to PVX coat proteins in transient expression assays (Fig. S2B). The M293L (referred to as S1) and T178A
(referred to as S2) single mutations were necessary and sufcient
to recapitulate the phenotypes of RxS1*M1 and RxS2*M1, respectively, whereas for RxS3*M1, both N147D and V337A (together referred to as S3) were required (Figs. S2C and S3). As
there was only one newly introduced amino acid change in RxS4M1
(G340R), this mutation is referred to as S4.
The ve amino acid changes in these four mutants (S14) are
located in the NB and ARC1 domains of Rx and are dispersed on
the primary structure (Fig. 2A and Fig. S4). However, when these
residues are mapped onto a 3D homology model of the Rx NBARC1-ARC2 domains, these positions colocalize around the nucleotide-binding pocket (Fig. 2B) (20). The N147D and V337A
mutations (derived from RxS3*M1) are positioned on opposite
sides of the nucleotide, in motifs that appear together in the
ATPase family phylogeny (34). N147D is located in the characteristic STAND ATPase hhGRExE motif of the NB domain (34)
and V337A is located in the GxP motif, which provides the main
contribution of the ARC1 domain to the nucleotide-binding
pocket. The G340R mutation (S4) in RxS4M1 is also located in
the GxP motif. A direct contact of this residue with the nucleotide is not likely, but the change from a small glycine to a large

PopMV-CP +

LRR
GFP
PopMV-CP
PVX-CP(TK)
PVX-CP(KR)

0.2
0.18
0.16
0.14
0.12

RxM1

RxS3*
M1

RxS1*
M1

RxS4
M1

0.1
0.08

RxS2*
M1

0.06
0.04
0.02
0

Rx

RxM1 RxS1* RxS2* RxS3* RxS4 RxM123

M1 M1 M1 M1

21190 | www.pnas.org/cgi/doi/10.1073/pnas.1311134110

Fig. 1. N-terminal random mutagenesis of RxM1

and identication of candidates. (A) Schematic of
RxM1. Red, N-terminal domains; green, C-terminal
LRR domain. Arrows depict error-prone PCR target
region. (B) Transient expression in leaf segments of
N. tabacum. PopMV-CP is coexpressed with RxM1
and the RxM1 variants identied in the screen,
RxS1*M1, RxS2*M1, RxS3*M1, and RxS4M1. Photos
were taken 5 d postinoculation (dpi). (Scale bar,
1 cm.) (C) Quantication of necrosis after transient
coexpression of Rx variants with viral CP elicitors in
N. tabacum (5 dpi) by chlorophyll content assay. A
low level of chlorophyll indicates high levels of necrosis. GFP and RxM123 are used as negative and
positive controls, respectively. Each bar represents an
average of ve biological replicates. Error bars represent 95% condence intervals.

Harris et al.

CC

NB

ACR1 ARC2

LRR

N846D

RxM1
RxS1*M1
RxS2*M1
RxS3*M1

L179M R291C

E55K T178A

M293L

K482N

N846D

P187Q

A88T N147D V337A

N846D
V362I

N846D

G340R

N846D

RxS4M1

N147

P332

M293

ARC1

NB

G340
V337

ARC2

K176

H459

T178

Fig. 2. Amino acid changes cluster around the nucleotide-binding pocket.

(A) Schematic of RxM1 variants identied from the screen, with errorprone PCRinduced amino acid changes. Responsible amino acid changes
are encapsulated in red boxes. (B) Homology model of NB-ARC1-ARC2
domains of Rx (20). Red, positions of the responsible amino acid changes
identied; blue, previously reported residues involved in nucleotide
binding. K176 in the P loop (NB); P332 in the GxP motif (ARC1); H549 in
the MHD motif (ARC2). Yellow, ADP nucleotide. The image was generated
by PyMOL software.

positively charged arginine may affect the structure of ARC1, the

interaction of the GxP motif with the nucleotide, and/or the interaction with ARC2. The M293L mutation (S1) in RxS1*M1
lies just outside the nucleotide-binding pocket, in a linker between the fth -strand of NB and the rst -helix of ARC1, as
part of the RNBS-C motif (35). The T178A mutation identied
in RxS2*M1 (S2) is located in the P-loop motif of the NB domain, immediately C-terminal to the GKT sequence. The P loop
is highly conserved and plays an important role in nucleotide
binding (6, 34, 36). Although most mutations in the P loop cause
loss of function, a mutation in the T178A homologous position in the tomato NB-LRR, Mi-1.2 (T557S), causes autoactivation (37).

Sensitization Requires an Intact P Loop. The nucleotide-binding P

loop, also known as the Walker A motif, is essential for the
function of most (6, 34, 41) but not all NB-LRRs (42, 43). To nd
out whether the S14 mutants are P loop-dependent, we introduced
three inactivating mutations (G175A, K176R, T177A) (6, 42) into
the invariant nucleotide-binding GKT motif of the P loop in the
autoactive RxS1234 background.
All three P-loop mutations suppressed autoactivity, but only
the K176R mutation completely abolished RxS1234-induced
necrosis (Fig. 4C). From Western blots, we rule out that this loss
of function is due to protein stability (Fig. 4D). The results are
consistent with nucleotide binding as an essential step in RxS1234
autoactivity. Because the S1, S2, S3, and S4 mutations are in close

0.3

Rx-LRR +
PopMV-CP
M1-LRR +
PopMV-CP

0.2

Rx
0.1

Rx +
PVX-CP(TK)
0

S1

Harris et al.

S2

S3

S4

Rx

Chlorophyllcontent(g/mm2)

Chlorophyllcontent(g/mm2)

Mutations Increase Activation Sensitivity. The increased response to

PopMV-CP in the RxM1 mutants could be due either to improved
recognition of the elicitor or to increased activation sensitivity. To
test these hypotheses, we exploited the weak HR induced when
either Rx or RxM1 is overexpressed in the absence of CP elicitor
(27, 38). If the mutations affect activation, then overexpression in
the absence of elicitor should produce more robust HR than
RxM1 when overexpressed. Alternatively, if the mutations affect
recognition of the elicitor, then overexpression-induced HR should
remain unchanged: the phenotype would be dependent on the
presence of elicitor. In each instance, the overexpression of
RxS1M1, RxS2M1, RxS3M1, and RxS4M1 from the strong CaMV
35S promoter produced a more rapid and robust HR in N.
tabacum leaf segments than 35S::RxM1 (Fig. S5A).
In an additional test of elicitor-independent HR, we coexpressed the original Rx promoter constructs with RanGAP2, an Rx
cytoplasmic retention factor causing overaccumulation of Rx in the
cytoplasm and mild HR with Rx (39, 40). As in the overexpression
assay, each of the S14 mutants produced a stronger HR than
RxM1 (Fig. S5B). These ndings suggest that the S14 mutations
increase the sensitivity for activation of the NB-LRR protein
rather than via improved specic recognition of PopMV-CP.
In a third test, we replaced the M1 LRR mutation (N846D)
with the wild-type LRR sequence (N846) in S14 mutants and
coexpressed them with PopMV-CP. If these mutations had enhanced the recognition of PopMV-CP, there would have been an
HR in the absence of the M1 mutation. However, none of these
constructs without M1 produced a robust HR on coexpression
with PopMV-CP (Fig. 3A). However, with the coat protein from
the resistance-breaking PVX isolate PVX-CPKR, HR was enhanced in the absence of the M1 mutation (Fig. 3B). Presumably,
the wild-type Rx has weak recognition of CPKR that is sensitized
by the S14 mutants.
In a fourth test of sensitization, we assayed the phenotype of
RxS1234M1 and RxS1234 mutants that carry all ve S mutations
(Fig. 4A). We predicted that the combined effects of these
mutants on sensitization would result in a necrotic response in
the absence of an elicitor, whereas recognition-mediated phenotypes would be elicitor-dependent. The results were in line
with the sensitization model (Fig. 4B). Thus, all four tests reinforce our conclusion that the M1 and S14 mutations affect
different stages of the Rx resistance mechanism: M1 mutation
affects recognition, and S14 sensitize the responsiveness.

0.2

0.1

Rx

Rx
+ TK

Rx
RxS1
+ KR + KR

RxS2 RxS3 RxS4

+ KR + KR + KR

Fig. 3. S mutations sensitize the response. (A) Responsible amino acid change constructs S1, S2, S3,
and S4, with or without the M1 mutation in the LRR,
coexpressed with PopMV-CP or (B) S1, S2, S3, and S4
without the M1 mutation coexpressed with PVX-CPKR
and quantied by chlorophyll content assay 5 dpi in
N. tabacum. Rx (light gray bar) and Rx coexpressed
with PVX-CPTK (dark gray bar) are used as negative
and positive controls for necrosis, respectively. Each
bar represents an average of ve biological replicates. Error bars represent 95% condence intervals.

PNAS | December 24, 2013 | vol. 110 | no. 52 | 21191

PLANT BIOLOGY

proximity to the nucleotide-binding pocket (Fig. 2B), they could

increase activation sensitivity by affecting the afnity or catalysis
of the ATP/ADP nucleotide (4).
Transgenic Plants Are Resistant to PopMV. To determine whether
the sensitized versions of RxM1 provide broad-spectrum resistance to PVX and PopMV strains without the systemic necrosis, we generated transgenic lines of N. benthamiana expressing
RxS1*M1, RxS2*M1, RxS3*M1, and RxS4M1 and compared
them with Rx and RxM1. When challenged with PopMV, the
wild-type and Rx plants developed mosaic symptoms and mild
stunting whereas the RxM1 plants exhibited trailing necrosis, as
previously reported (27).
In contrast, the RxS1*M1, RxS2*M1, RxS3*M1, and RxS4M1
transgenic lines showed resistance against PopMV infection. The
resistance in the RxS2*M1, RxS3*M1, and RxS4M1 plants was
variable, and these plants exhibited complete resistance or trailing
necrosis in the T1 generation (progeny of primary transformants)
when challenged with PopMV. RxS1*M1, however, displayed full
resistance to PopMV (Fig. 5A). RT-PCR of systemic leaf tissue
from RxS1*M1 plants inoculated with PopMV revealed a perfect
anticorrelation between expression of the transgene and accumulation of PopMV, demonstrating cosegregation of transgene
and resistance to PopMV (Fig. 5B). The RxS1*M1-expressing
plants were also fully resistant to both strains of PVX (CPTK and
CPKR), like RxM1 (Fig. S6).

Discussion
An initial interpretation of Rx and other NB-LRR proteins was that
the LRR domain would carry out the elicitor recognition function
in disease resistance and that the amino-terminal domains, including NB, would mediate the response phase leading to intracellular signaling and disease resistance. In this and a previous
study (27), we provide strong evidence for these two separate
phases of Rx-mediated resistance. The S14 mutations identied
here, for example, could activate an HR in our transient assay in
the absence of elicitor (Fig. 4 and Fig. S5), and they have the
predicted properties of response-phase mutants. Conversely, the
modied HR responses of the previously identied M13 mutations were absolutely dependent on the presence of elicitor and
have the properties of recognition-phase mutants.
With Rx and other NB-LRRs, there is evidence that the LRR
plays some role in recognition, as predicted by the original
model, because sequence variants in that domain inuence the
specicity of the disease resistance (911, 27, 44). Similarly, it is
likely that the amino-terminal domains, including NB, have some

RxS1234M1
T178A
N147D

role in the response phase, as predicted by the original model

(17, 45, 46). Supporting this idea, we describe ve mutations
affecting the structure of the conserved nucleotide-binding pocket
(formed by the NB-ARC domains; Fig. 2B) that increased activation sensitivity of Rx-mediated responses. However, other analyses have blurred the relationship between structure and function
of NB-LRR proteins (12, 47, 48), and it no longer seems likely that
there is a clean separation of the activation and recognition functions between the amino- and carboxyl-terminal domains. Our data
do not rule out that recognition and response functions are dispersed throughout the structure of Rx.
The Effect of the N-Terminal Mutations on Activation of Rx in the
Response Phase. That individual mutants S1, S2, S3, and S4 in-

crease activation sensitivity of Rx was only evident when they

were overexpressed or expressed together with RanGAP2 (Fig.
S5). However, when the ve mutations were combined in RxS1234,
the Rx HR was activated even when the protein was expressed
from the native Rx promoter and without RanGAP2 coexpression
(Fig. 4 A and B). This effect could be explained in terms of an
ATPase that mediates the switch between ATP-bound on and
ADP-bound off NB-LRR activation states (4). These two forms
of Rx would be in dynamic equilibrium and, in the absence of coat
protein elicitor with wild-type Rx, the pool would be predominantly in the inactive state. The S1, S2, S3, and S4 mutations could
each shift the balance toward the activated form so that, in
RxS1234, it is likely that the full activation was achieved without
any further stimulus from overexpression of Rx or RanGAP2. The
same mechanism is indicated by the increased preference for
binding ATP (49) in an autoactive version of the Flax M NBLRR. Similarly, an autoactive version of the tomato NB-LRR
protein, I-2, is impaired in nucleotide hydrolysis (50). Ultimately,
it will be necessary to perform biochemical tests on these versions of Rx to test these hypotheses.
Several other observations are consistent with this mechanistic
explanation of the S14 mutants. For example, the requirement
for an intact P loop (Fig. 4 C and D) indicates an essential role of
nucleotide binding for the S14 mutations to have their effect. An
intact P loop has also been shown to be required for other
autoactive NB-LRR variants, including Rx and I-2 (38, 50).
Similarly, the proximity of the S14 mutations to the nucleotidebinding pocket (Fig. 2B) is also consistent with an effect on ATP
turnover. However, we cannot exclude the possibility that the
mutations affect the interaction with the LRR domain or
downstream signaling components (18, 19, 40).

M293L

V337A
G340R

N846D

RxS1234M1

RxM1

RxS1234

Rx

RxS1234
M293L

V337A
G340R

0.2

150kDa

0.15
0.1

100kDa

0.05
0

RxS1

234
RxS1
2
G17 34
5A
RxS1
2
K17 34
6R
RxS1
2
T17 34
7A

50kDa

21192 | www.pnas.org/cgi/doi/10.1073/pnas.1311134110

RxS1
234
RxS1
2
G17 34
5A
RxS1
2
K17 34
6
RxS1R
2
T17 34
7A
GFP

Lad
der

0.25

Rx

0.3

Rx

Chlorophyll content ug/mm 2

T178A
N147D

Fig. 4. Combining the S mutations causes autoactivity and requires the P loop. (A) Schematic showing
the combined amino acid change constructs (S14) in
the M1 (Upper) and wild-type (Lower) LRR backgrounds. (B) Transient expression of constructs
encoding S1234 induces elicitor-independent necrosis in
the M1 and wild-type LRR backgrounds. Corresponding
constructs (RxM1 and Rx) with no S mutations in the
N-terminal region are not autoactive. (Scale bars,
0.5 cm.) Pictures were taken 5 dpi. (C) Chlorophyll
content assay to quantify necrosis after transient
expression of the autoactive RxS1234 constructs containing mutations in the GKT motif 5 dpi in N. tabacum.
Rx and RxS1234 are used as negative and positive
controls for autoactivity, respectively. Each bar represents an average of ve biological replicates. Error bars
represent 95% condence intervals. (D) Western blot
for accumulation of the RxS1234 GKT mutants after
transient expression in N. tabacum. RxS1234 does not
show a clear band, as the tissue was extensively necrotic
when collected at 3 dpi. RuBisCo is used as a loading control.

Harris et al.

wt

Rx

RxM1

RxS1*
M1

RxM123

RxS1*M1 + PopMV
PopMV
Rx
GAPDH

Fig. 5. Transgenic plants carrying RxS1*M1 are resistant to PopMV. (A)

Representative plants 6 wk after inoculation with PopMV. Wild-type and Rx
plants show mild stunting and mosaic symptoms; RxM1 plants display the
lethal trailing necrosis phenotype; and RxS1*M1 plants do not present viral
symptoms. The RxM123 transgenic line is used as a positive control for
PopMV resistance (27). (B) RT-PCR of cDNA from noninoculated leaf tissue
harvested 3 wk postinoculation from eight different RxS1*M1 plants. For
each plant, the same cDNA was used in three RT-PCR reactions to amplify
PopMV coat protein, RxS1*M1 transgene, or the housekeeping gene
GAPDH. Because RxS1*M1 plants are in the T1 generation, the transgene is
segregating, providing an additional blinding control in the experiment
(uninfected RxS1*M1 T1 plants are phenotypically identical). These results
were repeated on at least three separate occasions.

Articial and Natural R Gene Evolution in Crop Protection. In this

paper, we have used the RxM1 mutant to illustrate the tradeoff
of costs and benets in disease resistance. We show that a cost
spreading necrosis following PopMV infectioncan be mitigated by secondary mutation and, if PVX were a problem in
crops, it may well be that one or more of the RxS1-4M1 genes
would be more useful than the natural alleles of Rx: they would
confer resistance against both the CPTK and CPKR strains of
PVX. However, the crops with evolved RxM1 alleles would need
to be tested against a range of pathogens to rule out that new
costs had been introduced.
The idea that the LRR and N-terminal domains must coevolve
has been indicated by recombination experiments showing incompatible interactions between domains of closely related NBLRRs (20, 47), and it may well be that other NB-LRR genes
could be enhanced by stepwise articial evolution as with RxM1
(51). The NB-LRRs in crops would have originated in wild species where the tradeoff of costs and benets selected in the
natural environment would have inuenced the resistance phenotype. However, the optimal tradeoff in a natural environment
may not be the same as in a managed agricultural setting, and it
could be that the range of useful R genes for crop protection
could be expanded by stepwise articial evolution. In the rst
stage, a gene could be evolved in one domain to derive mutants
in which the degree or specicity of resistance is modied. The

Harris et al.

mutants could then be further rened by mutation of a second

domain to ensure that costs have been minimized. This approach
might be particularly useful if resistance-breaking strains of
a pathogen have emerged or if the resistance of the natural R
gene is too weak to contain the pathogen. Alternatively, broad
recognition could be paired with a reduced level of activation
sensitivity to minimize the resistance tradeoff with plant growth
(25) and ensuring that only very strongly recognized pathogen
signals initiate a defense response. As we have shown that these
S mutations can act independently of the broad recognitionconferring mutation in the LRR (Figs. 3A and 4B), we can begin
to ne-tune the balance between NB-LRR recognition and
response functions.
How does our sequential in vitro selection approach compare
with evolution in natural populations? According to Ossowski
et al. (52), the spontaneous mutation rate of Arabidopsis is estimated at 7 109 nt changes per site per generation, 1 105
changes per 1.5-kb region per generation. Here we surveyed
around 7,500 changes (1,500 clones at an average mutational
frequency of 5 changes per molecule, 5 1,500 = 7,500) across a
1.5-kb region. Therefore, this study surveys the mutational spectrum equivalent to 750 million Arabidopsis plants (7,500/1
105) in each round of mutation: The combined effect of a stepwise strategy is therefore equivalent to much larger populations.
An additional consideration is that the selection in our articial evolution approach is much more stringent and targeted to
the NB-LRR of interest than selection on individual NB-LRRs
in a natural plant population (53). Stepwise articial evolution
therefore represents a powerful approach to modulate NB-LRR
characteristics that is more efcient in time and space than
evolution in the eld. This study demonstrates that a heuristic
approach, combining in vitro directed evolution and structural
modeling, can be used to design improved NB-LRRs for use in
crop protection.
Materials and Methods
Plasmids. The constructs used are described in SI Materials and Methods.
Homology Modeling. Homology modeling has been described previously (20).
Images were generated using PyMOL software, http://www.pymol.org/.
Transient Expression. Agrobacterium tumefaciens GV3101 or C58C1 was inltrated in 3- to 4-wk-old N. tabacum or N. benthamiana plants using similar
methods described previously (32).
Chlorophyll Content Assay. Based on ref. 33, the protocol is described in detail
in SI Materials and Methods.
Western Blots. Total protein was extracted from inltrated N. tabacum tissue
using methods described by ref. 54. HA-tagged Rx was detected using an
anti-HA rat monoclonal antibody (clone 3F10; Roche 1867423) and goat antirat (LI-COR; IRDye 800CW, 926-32219) secondary antibody and visualized on
a LI-COR Odyssey CLx.
Transgenic Plants. Previously described N. benthamiana transgenic lines used
in this study were wild type, Rx (RxH3), RxM1 (2/227 2B10), and RxM123 (GR3
F11) (27). RxS14*M1 transgenic N. benthamiana were generated by leaf
disc transformation (55).
RT-PCRs. Total RNA was extracted from noninoculated leaf tissue using TRIzol
reagent, and cDNA was generated using random hexamer-primed RT
synthesis (Invitrogen; 18080-051). RT-PCR was performed using gene-specic
primers (Table S2).
ACKNOWLEDGMENTS. We thank our colleagues Attila Molnar and
Natasha Elina for helpful discussions and their critical review of the manuscript, and Wladimir Tameling for facilitating a fruitful collaboration.
This work was supported by the Biotechnology and Biological Sciences
Research Council (C.J.H.), the Gatsby Charitable Foundation, and the
Royal Society Edward Penley Abraham Research Professorship (to D.C.B.).
E.J.S. and A.G. are supported by the European Commission Integrated
BIOEXPLOIT project.

PNAS | December 24, 2013 | vol. 110 | no. 52 | 21193

PLANT BIOLOGY

A PopMV inoculated

1. Jones JDG, Dangl JL (2006) The plant immune system. Nature 444(7117):323329.
2. Martin GB, Bogdanove AJ, Sessa G (2003) Understanding the functions of plant disease resistance proteins. Annu Rev Plant Biol 54:2361.
3. Lukasik E, Takken FLW (2009) STANDing strong, resistance proteins instigators of
plant defence. Curr Opin Plant Biol 12(4):427436.
4. Collier SM, Moffett P (2009) NB-LRRs work a bait and switch on pathogens. Trends
Plant Sci 14(10):521529.
5. Takken FL, Albrecht M, Tameling WI (2006) Resistance proteins: Molecular switches of
plant defence. Curr Opin Plant Biol 9(4):383390.
6. Tameling WIL, et al. (2002) The tomato R gene products I-2 and MI-1 are functional
ATP binding proteins with ATPase activity. Plant Cell 14(11):29292939.
7. Ade J, DeYoung BJ, Golstein C, Innes RW (2007) Indirect activation of a plant nucleotide binding site-leucine-rich repeat protein by a bacterial protease. Proc Natl Acad
Sci USA 104(7):25312536.
8. Heath MC (2000) Hypersensitive response-related death. Plant Mol Biol 44(3):321334.
9. Ellis JG, Lawrence GJ, Dodds PN (2007) Further analysis of gene-for-gene disease resistance specicity in ax. Mol Plant Pathol 8(1):103109.
10. Shen QH, et al. (2003) Recognition specicity and RAR1/SGT1 dependence in barley
Mla disease resistance genes to the powdery mildew fungus. Plant Cell 15(3):732744.
11. Rairdan GJ, Moffett P (2006) Distinct domains in the ARC region of the potato resistance protein Rx mediate LRR binding and inhibition of activation. Plant Cell 18(8):
20822093.
12. Chen Y, Liu Z, Halterman DA (2012) Molecular determinants of resistance activation
and suppression by Phytophthora infestans effector IPI-O. PLoS Pathog 8(3):e1002595.
13. Kanzaki H, et al. (2012) Arms race co-evolution of Magnaporthe oryzae AVR-Pik and
rice Pik genes driven by their physical interactions. Plant J 72(6):894907.
14. Ueda H, Yamaguchi Y, Sano H (2006) Direct interaction between the tobacco mosaic
virus helicase domain and the ATP-bound resistance protein, N factor during the
hypersensitive response in tobacco plants. Plant Mol Biol 61(1-2):3145.
15. van der Hoorn RA, Kamoun S (2008) From guard to decoy: A new model for perception of plant pathogen effectors. Plant Cell 20(8):20092017.
16. Bendahmane A, Kanyuka K, Baulcombe DC (1999) The Rx gene from potato controls
separate virus resistance and cell death responses. Plant Cell 11(5):781792.
17. Rairdan GJ, et al. (2008) The coiled-coil and nucleotide binding domains of the potato
Rx disease resistance protein function in pathogen recognition and signaling. Plant
Cell 20(3):739751.
18. Tameling WIL, Baulcombe DC (2007) Physical association of the NB-LRR resistance
protein Rx with a Ran GTPase-activating protein is required for extreme resistance to
potato virus X. Plant Cell 19(5):16821694.
19. Moffett P, Farnham G, Peart J, Baulcombe DC (2002) Interaction between domains of
a plant NBS-LRR protein in disease resistance-related cell death. EMBO J 21(17):
45114519.
20. Slootweg EJ, et al. (2013) Structural determinants at the interface of the ARC2 and
leucine-rich repeat domains control the activation of the plant immune receptors Rx1
and Gpa2. Plant Physiol 162(3):15101528.
21. Narusaka M, et al. (2013) Interfamily transfer of dual NB-LRR genes confers resistance
to multiple pathogens. PLoS ONE 8(2):e55954.
22. Horvath DM, et al. (2012) Transgenic resistance confers effective eld level control of
bacterial spot disease in tomato. PLoS ONE 7(8):e42036.
23. Wulff BBH, Horvath DM, Ward ER (2011) Improving immunity in crops: New tactics in
an old game. Curr Opin Plant Biol 14(4):468476.
24. Tian D, Traw MB, Chen JQ, Kreitman M, Bergelson J (2003) Fitness costs of R-genemediated resistance in Arabidopsis thaliana. Nature 423(6935):7477.
25. Todesco M, et al. (2010) Natural allelic variation underlying a major tness trade-off
in Arabidopsis thaliana. Nature 465(7298):632636.
26. Bomblies K, Weigel D (2007) Hybrid necrosis: Autoimmunity as a potential gene-ow
barrier in plant species. Nat Rev Genet 8(5):382393.
27. Farnham G, Baulcombe DC (2006) Articial evolution extends the spectrum of viruses
that are targeted by a disease-resistance gene from potato. Proc Natl Acad Sci USA
103(49):1882818833.
28. Nam M, et al. (2011) Arabidopsis TTR1 causes LRR-dependent lethal systemic necrosis,
rather than systemic acquired resistance, to tobacco ringspot virus. Mol Cells 32(5):
421429.
29. Komatsu K, et al. (2010) Viral-induced systemic necrosis in plants involves both programmed cell death and the inhibition of viral multiplication, which are regulated by
independent pathways. Mol Plant Microbe Interact 23(3):283293.
30. Vallejos CE, et al. (2006) Genetic and molecular characterization of the I locus of
Phaseolus vulgaris. Genetics 172(2):12291242.

21194 | www.pnas.org/cgi/doi/10.1073/pnas.1311134110

31. Goulden MG, Khm BA, Santa Cruz S, Kavanagh TA, Baulcombe DC (1993) A feature
of the coat protein of potato virus X affects both induced virus resistance in potato
and viral tness. Virology 197(1):293302.
32. Peart JR, Cook G, Feys BJ, Parker JE, Baulcombe DC (2002) An EDS1 orthologue is
required for N-mediated resistance against tobacco mosaic virus. Plant J 29(5):
569579.
33. Pruzinsk A, et al. (2005) Chlorophyll breakdown in senescent Arabidopsis leaves.
Characterization of chlorophyll catabolites and of chlorophyll catabolic enzymes involved in the degreening reaction. Plant Physiol 139(1):5263.
34. Leipe DD, Koonin EV, Aravind L (2004) STAND, a class of P-loop NTPases including
animal and plant regulators of programmed cell death: Multiple, complex domain
architectures, unusual phyletic patterns, and evolution by horizontal gene transfer.
J Mol Biol 343(1):128.
35. Meyers BC, et al. (1999) Plant disease resistance genes encode members of an ancient
and diverse protein family within the nucleotide-binding superfamily. Plant J 20(3):
317332.
36. Hanson PI, Whiteheart SW (2005) AAA+ proteins: Have engine, will work. Nat Rev
Mol Cell Biol 6(7):519529.
37. Gabrils SHEJ, et al. (2007) An NB-LRR protein required for HR signalling mediated by
both extra- and intracellular resistance proteins. Plant J 50(1):1428.
38. Bendahmane A, Farnham G, Moffett P, Baulcombe DC (2002) Constitutive gain-offunction mutants in a nucleotide binding site-leucine rich repeat protein encoded at
the Rx locus of potato. Plant J 32(2):195204.
39. Tameling WIL, et al. (2010) RanGAP2 mediates nucleocytoplasmic partitioning of the
NB-LRR immune receptor Rx in the Solanaceae, thereby dictating Rx function. Plant
Cell 22(12):41764194.
40. Sacco MA, Mansoor S, Moffett P (2007) A RanGAP protein physically interacts with the
NB-LRR protein Rx, and is required for Rx-mediated viral resistance. Plant J 52(1):
8293.
41. Aravind L, Dixit VM, Koonin EV (1999) The domains of death: Evolution of the apoptosis machinery. Trends Biochem Sci 24(2):4753.
42. Bonardi V, et al. (2011) Expanded functions for a family of plant intracellular immune
receptors beyond specic recognition of pathogen effectors. Proc Natl Acad Sci USA
108(39):1646316468.
43. Collier SM, Hamel L-P, Moffett P (2011) Cell death mediated by the N-terminal domains of a unique and highly conserved class of NB-LRR protein. Mol Plant Microbe
Interact 24(8):918931.
44. Krasileva KV, Dahlbeck D, Staskawicz BJ (2010) Activation of an Arabidopsis resistance
protein is specied by the in planta association of its leucine-rich repeat domain with
the cognate oomycete effector. Plant Cell 22(7):24442458.
45. Bernoux M, et al. (2011) Structural and functional analysis of a plant resistance protein TIR domain reveals interfaces for self-association, signaling, and autoregulation.
Cell Host Microbe 9(3):200211.
46. Maekawa T, et al. (2011) Coiled-coil domain-dependent homodimerization of intracellular barley immune receptors denes a minimal functional module for triggering cell death. Cell Host Microbe 9(3):187199.
47. Qi D, DeYoung BJ, Innes RW (2012) Structure-function analysis of the coiled-coil and
leucine-rich repeat domains of the RPS5 disease resistance protein. Plant Physiol
158(4):18191832.
48. Burch-Smith TM, et al. (2007) A novel role for the TIR domain in association with
pathogen-derived elicitors. PLoS Biol 5(3):e68.
49. Williams SJ, et al. (2011) An autoactive mutant of the M ax rust resistance protein
has a preference for binding ATP, whereas wild-type M protein binds ADP. Mol Plant
Microbe Interact 24(8):897906.
50. Tameling WIL, et al. (2006) Mutations in the NB-ARC domain of I-2 that impair ATP
hydrolysis cause autoactivation. Plant Physiol 140(4):12331245.
51. Brunner S, et al. (2010) Intragenic allele pyramiding combines different specicities of
wheat Pm3 resistance alleles. Plant J 64(3):433445.
52. Ossowski S, et al. (2010) The rate and molecular spectrum of spontaneous mutations
in Arabidopsis thaliana. Science 327(5961):9294.
53. Bergelson J, Kreitman M, Stahl EA, Tian D (2001) Evolutionary dynamics of plant Rgenes. Science 292(5525):22812285.
54. Slootweg E, et al. (2010) Nucleocytoplasmic distribution is required for activation of
resistance by the potato NB-LRR receptor Rx1 and is balanced by its functional domains. Plant Cell 22(12):41954215.
55. Horsch RB, et al. (1985) A simple and general method for transferring genes into
plants. Science 227(4691):12291231.

Harris et al.