Aphid lineages, and bacterial strains and plasmids used in this study are listed in Supplementary Tables S1 and S2, respectively. All aphid lineages were reared in a growth chamber at 18°C with a 16 h day/8 h night photoperiod on the broad bean, Vicia faba (Castel), at low density to avoid the production of winged individuals. Escherichia coli and A. tumefaciens strains were grown on Luria-Bertani medium at 37°C and 30°C, respectively. For solid media, agar was added at a final concentration of 1.5% (w/v). Antibiotics were used at the following concentrations: for all bacteria, 50 μg/ml kanamycin; for A. tumefaciens, 50 μg/ml rifampicin; for E. coli, 10 μg/ml gentamycin.

Pisum sativum (Supplementary Table S3) and Medicago sativa plants were grown in a growth chamber at 18°C with a 16 h day/8 h night photoperiod for 2 and 3 weeks, respectively.

Life traits of five aphid lineages from pea (Ar_Po_28, Ar_Po_58), alfalfa (L9Ms14) and clover (YR2, T8005) biotypes (Supplementary Table S1) were measured on P. sativum cultivars ZP1130 and ZP1109 (Supplementary Table S3). Adult aphids were installed on both pea cultivars and removed 24 h later, giving them enough time to produce 10 larvae that were left on the plants (day 1). Survival rate of the 10 larvae was measured at day 9 (when they reach adulthood), three surviving adult aphids were then reinstalled on the plants and biomass (the cumulated weight of the three adults and their offspring) of the aphid population was weighted at day 17. The biomass is a good proxy of the number of nymphs produced by adult aphids and reflects well their overall fitness (Peccoud et al., 2009). Five replicates for each aphid lineage on the two tested plants were performed.

All primers used in this study are listed in Supplementary Table S4. The genes encoding eGFP and the β-glucuronidase with a plant derived intron (GUSi; Vancanneyt et al., 1990) were amplified using GFP-Fw/GFP-Rv primers and GUS-Fw/GUS-Rv primers, respectively, and were added complete attB1 and attB2 sequences by the second PCR with attB1 and attB2 primers. In order to clone aphid salivary genes, cDNAs produced from aphid head total RNA were used to enrich transcripts encoding salivary genes. Adult aphids feeding on V. faba were flash frozen in liquid nitrogen, and decapitated with a scalpel between the first and second pairs of legs. Head RNA was extracted from 10 to 20 individuals using the RNeasy plant mini kit (Qiagen). cDNA synthesis was performed with poly-T primers using the AMV reverse transcriptase system (Promega) according to the manufacturers’ instructions. ACYPI009919 (Ap25) and ACYPI008617 (ApC002) open reading frames encoding mature proteins were amplified from the cDNA of the Ar_Po_58 line (pea biotype) with Phusion DNA polymerase (ThermoFisher Scientific) using AP25-Fw/AP25-Rv and APC002-Fw/APC002-Rv primers (Supplementary Table S4), respectively. attB1 and attB2 sites were added with a second PCR using attB1 and attB2 primers. All amplicons, eGFP, GUSi and two salivary genes, were recombined by BP reaction into pDONR207 (Invitrogen) using BP clonase II (Invitrogen) and produced entry vectors (Table S2). Entry vectors were recombined by LR reaction using LR clonase II (Invitrogen) into pEAQ-HT-DEST1 expression vector (Supplementary Table S2; Sainsbury et al., 2009). Expression vectors were transformed in electro-competent A. tumefaciens cells (Supplementary Table S2).

Agrobacterium tumefaciens-mediated transient expression was performed as described (Rivas et al., 2002). Freshly cultured cells were resuspended in induction buffer [10 mM MgCl₂, 10 mM Mes (2-(N-morpholino) ethanesulfonic acid), pH 5.6, and 150 μM acetosyringone] to an O.D.₆₀₀ (optical density at 600 nm) of 0.5. Cells were syringe infiltrated into leaves of 2 week-old P. sativum (Supplementary Table S1) and 3 week-old M. sativa plants.

Plant leaves infiltrated with Agrobacterium were detached 3 days post infiltration (dpi) and GUS activity was visualized as described (Jefferson, 1987). Briefly, leaves were vacuum infiltrated with GUS staining solution (61 mM Na₂HPO₄, 39 mM NaHPO₄, 0.1% triton X-100, 10 mM EDTA, 0.3% H₂O₂ and 1.5 mM 5-bromo, 4-chloro, 3-indolyl glucuronide (X-Glc, Biosynth), pH 7.0) and incubated overnight at 37°C. Then chlorophyll discoloration was performed with successive washes with ethanol at 37°C.

Three leaf disks per leaf were sampled using a cork borer (area = 0.79 cm²) at 0, 7, and 10 days post-infiltration for GFP protein detection. Leaf disks were flash frozen in liquid nitrogen and stored at -80°C. Proteins were extracted in 120 μl extraction buffer (50 mM tris pH 7.5, 1 μM Dithiothreitol, glycerol 10%, 1 mM PMSF (Phenylmethylsulfonyl fluoride), 0.05% triton X-100). Extracts from pea plants were prepared as described (Canonne et al., 2011) and supernatants were resuspended in 5X loading buffer (0.5 M Tris pH 6.8, SDS 10%, glycerol 50 and 0.001% bromophenol blue). Fifteen microliters of samples were separated by SDS-PAGE (12% polyacrylamide) and transferred on PVDF (Polyvinylidene fluoride) membranes, (Merck Millipore) as described (Witte et al., 2004) with following modifications: PVDF membranes were soaked in methanol before and after transfer, and then washed in water. Methanol in transfer buffer was replaced by ethanol. The rabbit anti-GFP antibody (Biorad) and secondary antibodies (polyclonal goat anti-rabbit antibody peroxidase conjugated; Sigma–Aldrich) were both used at 1:10000. Detection was performed by chemiluminescence using Clarity Western ECL Substrate (Biorad) and CL-XPosure^TM Film (Lifetechnologies) according to manufacturer’s instructions. Coomassie stains were performed with 0.2% Coomassie Brilliant Blue R250 (Sigma) in 50:40:10 water, methanol, acetic acid.

One young leaf of the P. sativum ZP1109 cultivar was syringe-infiltrated with A. tumefaciens AGL-1 strain harboring expression vectors. Three days later (at 3 dpi), 6 new-born aphids (1 day-old) born on V. faba were installed on P. sativum agroinfiltrated leaves in custom-built clip cages (area = 2.54 cm²). When aphids were 8 days-old (10 dpi), clip cages were opened and the number of surviving aphids was recorded to estimate the survival rate. From the survivors, one average sized aphid was selected and transferred to a new P. sativum leaf that was infiltrated with the same construct of Agrobacterium 3 days before the transfer. Clip cages were opened when aphids were 12 and 15 days-old to assess the fecundity by counting the number of nymphs produced by each aphid. The nymphs were removed after each counting to avoid overcrowding of the cages. In one experiment, 10 replicates per gene were performed and the same experiment was repeated twice, producing 20 replicates. All the experiments were conducted at 20°C, 16 h day/8 h night photoperiod.

All statistical analyses were conducted in R version 3.1.2 (R Core Team, 2014). Data were checked for approximate normal distribution by graphical visualizing of residuals. The effects of the different factors (pea cultivar, aphid lineage, expressed gene) were tested and the simplest model explaining the data was used. Analyses of survival rates (Figures 2A and 3A) and fecundity counts (Figure 3B) were performed by classical linear regressions using generalized linear models (GLM) with binomial and Poisson distributions, respectively. Both tests were followed by multiple comparisons of means by the Tukey contrast method implemented in the package “multcomp” (Hothorn et al., 2008). The influence of pea cultivars and aphid lineage on aphid biomass (Figure 2B) was analyzed by a two-way ANOVA. Tukey’s post hoc multiple comparisons of means from the R package “agricolae” (De Mendiburu, 2014) were used to reveal differences between groups.

Combinations of A. tumefaciens and various M. sativa and P. sativum cultivars were tested using the β-glucuronidase containing a plant derived intron (GUSi) as a reporter gene (Vancanneyt et al., 1990). Green fluorescence protein (GFP) could not be used as a reporter due to strong auto fluorescence induced in the leaf surface by the infiltration. Initially, we tested two plant expression vectors pGWB402Ω (Nakagawa et al., 2007) and pEAQ-HT-DEST1 (Sainsbury et al., 2009) in some pea cultivars, but the difference in expression levels between the two vectors was not very clear or slightly better when pEAQ-HT-DEST1 was used. Therefore, we used pEAQ-HT-DEST1 for the rest of screening. Also, our initial test showed that a bacterial suspension with an O.D.₆₀₀ less than 0.3 resulted in a weak transgene expression and more than 0.7 triggered leaf chlorosis a few days after infiltration. Therefore, for the rest of the screening, agroinfiltrations were performed using syringe infiltration method and a bacterial suspension with an O.D.₆₀₀ = 0.5. Seventeen P. sativum (Supplementary Table S3) and five M. sativa cultivars were selected based on geographic origin and phylogenetic groups in order to screen a large genetic diversity. Each cultivar was infiltrated with three Agrobacterium strains [C58C1, GV3101 and AGL-1 (Supplementary Table S2)] each harboring pEAQ-HT-DEST1-GUSi to identify the combination of plant and bacterium genotypes that produce high amount of GUS proteins. Leaves were analyzed histochemically for GUS activity at 3 dpi. At least three independent experiments were performed for each combination and results are summarized in Table 1. None of the M. sativa cultivars was suitable for Agrobacterium-mediated transient expression in leaves as no GUS staining could be observed in these plants under the tested conditions. High differences between pea cultivars were observed. Most of the pea cultivars had no or weak intensities of GUS staining. Of the three Agrobacterium strains used in this study, AGL-1 induced the highest expression of GUS, and C58C1 was the lowest inducer. Two pea genotypes, ZP1130 and ZP1109, inoculated with AGL-1 showed most intense coloration during GUS staining (Figure 1A). GUS staining could be observed at 3 dpi for both cultivars, ZP1130 and ZP1109. To confirm protein expression in these two cultivars, transient expression of eGFP and detection by western-blot was performed (we could not visualize GFP fluorescence due to autofluorescence induced by wounding). eGFP protein was detected at 7 and 10 dpi for both ZP1109 and ZP1130 (Figure 1B). During this study, yellowing of the leaves starting at 9–10 dpi for ZP1130 and at 12–13 dpi for ZP1109 was observed. This leaf yellowing was probably due to AGL-1 infection as the yellowing was observed in the leaves infiltrated with Agrobacterium with empty vector control, and no yellowing was observed in buffer infiltrated leaves (data not shown). Taken together, we identified two pea cultivars, ZP1130 and ZP1109, and the A. tumefaciens strain AGL-1 as the combinations that are suitable for transient gene expression, and we presumed that 3–8 dpi for ZP1130 and 3–10 dpi for ZP1109 are the timing to examine the effect of transgene expression in the plant or plant–aphid interactions.

Survival rate and biomass of the five A. pisum lineages belonging to three biotypes (pea, alfalfa and clover; Supplementary Table S1) were assessed on the ZP1130 and ZP1109 pea cultivars we identified as suitable for agroinfiltration (Figure 2). Analysis revealed that the two plant cultivars did not influence the survival rate and produced aphid biomass [χ²= 0.14, P = 0.243; F_(5,44) = 129.7, P = 0.261; for survival and biomass, respectively], but pea aphid lineages differed significantly in their survival rates (χ²= 19.04, P < 0.001) and biomass production [F_(4,45)= 128.9, P < 0.001]. The pea adapted lineages Ar_Po_28 and Ar_Po_58 showed a higher survival rate on the pea cultivars at day 9 compared to L9Ms14 (alfalfa biotype), YR2 and T8005 (clover biotype). The difference in survival was very pronounced between pea and alfalfa specialized lineages, and intermediate for lineages of the clover biotype (Figure 2A). On both ZP1130 and ZP1109 cultivars, only the lineages of the pea biotype (Ar_Po_28 and Ar_Po_58) produced a substantial biomass. Although Ar_Po_28 had a significantly higher biomass than Ar_Po_58, both lineages performed well on the tested cultivars that they seem to recognize as favorable hosts. By contrast, alfalfa and clover adapted lineages hardly reproduced on the pea cultivars that seem to be non-host plants in these interactions (Figure 2B). Thus, the ZP1130 and ZP1109 cultivars are selective hosts for A. pisum biotypes, allowing to assess host and non-host interactions using agroinfiltration experiments.

Next, we expressed two salivary genes in ZP1109 by agroinfiltration using strain AGL-1 and examined their effects on A. pisum feeding on the infiltration site. We chose Ar_Po_58 as a test aphid line as it belongs to the pea biotype and harbors no secondary symbiont, which may interfere with plant–aphid interactions. Mature proteins encoding ACYPI008617 (ApC002) and ACYPI009919, which we named Ap25, were transiently expressed using pEAQ-HT-DEST1 vector. The genes were expressed by CaMV 35S promoter, which is known to be ubiquitously and constitutively activated in various plant tissues including epidermal, mesophyll and phloem tissues (Stockhaus et al., 1989). In the process of establishing phloem feeding, A. pisum punctures various tissues and salivates (Schwarzkopf et al., 2013). When the aphid attempts to feed on non-host legume plant, it punctures epidermal and mesophyll cells but cannot establish phloem feeding: therefore, the factors that determine the compatibility between A. pisum and host plants are present in those tissues (Schwarzkopf et al., 2013). Based on these informations, we thought it is important to express salivary proteins ubiquitously to fully assess their functions in plants and used 35S promoter for transient expression. 35S promoter has been successfully used in other studies on aphid salivary proteins (Bos et al., 2010; Naessens et al., 2015). ApC002 was chosen because it is one of the most studied salivary proteins and is shown to be essential for A. pisum to feed on the universal host plant (V. faba; Ferrari et al., 2008; Peccoud et al., 2009). Ap25 was selected because the gene presents the same features as that of ApC002: the gene was identified in salivary glands by transcriptomic analyses (Carolan et al., 2011), is specifically expressed in salivary glands (Akiko Sugio et al., unpublished data), and encodes a signal peptide and a small (13.9 kDa) mature protein with no predicted function. Although many genes are duplicated in A. pisum genome, Ap25, like ApC002, is single copy in A. pisum and its orthologues exist only in the Aphididae family (Hélène Boulain et al., unpublished data).

In this study, transient protein expression was observed from 3 (detected by GUS activity) to 10 days (detected by western blot) at 20°C after infiltration of Agrobacterium. A. pisum starts to reproduce around 9th day after birth, reaches its peak of reproduction around 5 days later, and slows down but continues to reproduce until its death at an age of approximately 30 days (Tsuchida et al., 2004). By supplying newly infiltrated leaves, we extended the duration of the experiment to characterize the effect of transgene expression on aphid fecundity. Leaves of ZP1109 were infiltrated with AGL-1 harboring expression plasmids of eGFP, ApC002 or Ap25. Three days after the infiltration, six new-born aphids of the pea adapted clone Ar_Po_58 were clip caged on the infiltrated leaves. When the aphids were 8 days-old (at 10 dpi) the cages were opened to count the number of survivors. One aphid was transferred to a new 3-day-post-infiltrated leaf. Production of nymphs of the caged adult was measured when the aphid was 12 and 15-day old corresponding to the peak of reproduction of adults. Survival rate and total number of nymphs of the aphids are shown in Figure 3. There was no difference in the survival rate of the aphids that were fed on the leaves expressing the three tested genes (χ²= 0.01, P = 0.96). Production of nymphs of Ar_Po_58 feeding on ApC002 expressing leaves was same as that of aphids feeding on eGFP expressing leaves, while the aphids produced approximately 12% more offspring on Ap25 expressing leaves than on eGFP expressing leaves (20 biological replicates, χ²= 18.75, P < 0.001). The results indicate that Ap25 plays a role in promoting A. pisum feeding on P. sativum.