Orange jasmine as a trap crop to control Diaphorina citri

Novel, suitable and sustainable alternative control tactics that have the potential to reduce migration of Diaphorina citri into commercial citrus orchards are essential to improve management of huanglongbing (HLB). In this study, the effect of orange jasmine (Murraya paniculata) as a border trap crop on psyllid settlement and dispersal was assessed in citrus orchards. Furthermore, volatile emission profiles and relative attractiveness of both orange jasmine and sweet orange (Citrus × aurantium L., syn. Citrus sinensis (L.) Osbeck) nursery flushes to D. citri were investigated. In newly established citrus orchards, the trap crop reduced the capture of psyllids in yellow sticky traps and the number of psyllids that settled on citrus trees compared to fallow mowed grass fields by 40% and 83%, respectively. Psyllids were attracted and killed by thiamethoxam-treated orange jasmine suggesting that the trap crop could act as a ‘sink’ for D. citri. Additionally, the presence of the trap crop reduced HLB incidence by 43%. Olfactometer experiments showed that orange jasmine plays an attractive role on psyllid behavior and that this attractiveness may be associated with differences in the volatile profiles emitted by orange jasmine in comparison with sweet orange. Results indicated that insecticide-treated M. paniculata may act as a trap crop to attract and kill D. citri before they settled on the edges of citrus orchards, which significantly contributes to the reduction of HLB primary spread.

. Preliminary field studies in São Paulo (SP) State, Brazil, showed that the use of suitable host plants for D. citri as barriers reduced the number of marked psyllids recaptured on yellow stick traps deployed on citrus trees 7 . The lowest recapture rates were recorded when orange jasmine [Murraya paniculata (L.) Jack, syn. Murraya exotica L.] was used as a barrier, envisaging its potential use as a trap crop to manage HLB. Orange jasmine is a preferred host for D. citri in comparison with other rutaceous host plants 17,18 and this effect has been related to its volatiles 19 . Although orange jasmine can be infected with 'Candidatus Liberibacter asiaticus' (CLas), bacterial titer is much lower in this host than in Citrus species and cultivars and decreases over time, probably due to the lower multiplication rates in the former, being CLas infections usually transient in orange jasmine 20,21 . Moreover, recent results showed that D. citri could not acquire CLas upon feeding and developing on CLas-qPCR positive orange jasmine seedlings 22 . These results indicate that orange jasmine is a poor host for CLas (as well as a 'dead-end' host plant 23 ) and suggest that it may be used as a potential trap crop, attracting psyllids which could be controlled in these plants by insecticides, thus limiting spread of HLB. In this study, the effect of orange jasmine as a border trap crop on D. citri settlement and dispersal was assessed in sweet orange orchards. Additionally, the volatile emission profiles and relative attractiveness of both orange jasmine and sweet orange flushes to D. citri were investigated. The results of this study may help citrus growers to control immigrating psyllids that arrive to the edges of citrus orchards from external inoculum sources and thus limit HLB incidence inside citrus orchards.

Effects of orange jasmine as a trap crop on Diaphorina citri natural infestation and HLB incidence.
The presence of orange jasmine as a trap crop on the edge of a new (6-month-old) citrus orchard (Area A, Fig. 1a) reduced incidence of D. citri in the orchard, when compared to the control, over a 45-month survey (d.f. = 1; P = 0.0030). In this area significantly fewer psyllids were captured on yellow sticky traps placed in citrus trees bordered on the east side by the trap crop compared to identical traps placed in control trees bordered on the east side by fallow mowed grass (F = 5.74; d.f. = 1, 38; P = 0.0216) (Fig. 1a). In contrast, when the orange jasmine trap crop was planted on the edge of well established (7-year-old) citrus plots (Area B, Fig. 1b), cumulative trends in numbers of psyllids captured on yellow sticky traps placed in trees within the plots did not differ significantly, regardless of the presence or absence of an orange jasmine trap crop on the orchard edge (d.f. = 1; P = 0.8196), and the total numbers of psyllids recorded on the yellow sticky traps in both the trap crop and control treatments over 45 months did not differ significantly (F = 0.00; d.f. = 1, 38; P = 1.0000) (Fig. 1b).
Regarding the assessments of 'Ca. L. asiaticus' or 'Ca. L. americanus'-positive trees in the new citrus orchard, the HLB incidence in the control plot was ~2.8-fold (control: 1.4%; trap crop: 0.5%) and ~1.8-fold (control: 2.8%; trap crop: 1.6%) higher than on orange jasmine trap crop treatment, for assessments performed on May 2014 and on January 2015, respectively. Assessments on release platforms showed that approximately 90% of the released psyllids had taken off from the platforms at 1 day after release (DAR) and no psyllids were found there by 3 DAR. In general, 83% of all marked psyllids that were found in the experimental area were in the first citrus row regardless of the assessment time (Fig. 3). Overall, 150, 205 and 56 marked psyllids (out of 16800) were found in the experimental area at 1, 3 and 7 DAR, respectively. At all inspection times, most psyllids (94.3, 78.7 and 86.3% at 1, 3 and 7 DAR, respectively) were found on control plots. Psyllids released in front of orange jasmine plots were mostly restricted to the first citrus row. In contrast, individuals released in front of control plots were found mainly in the first four citrus rows, and one psyllid reached the last citrus row (35 m from the release platforms).

Effects of orange jasmine as a trap crop on
Visual assessments on orange jasmine trees at 1 and 3 DAR indicated the presence of 115 and 158 marked D. citri, respectively (Fig. 4). Few psyllids (5 individuals) were found on orange jasmine trees at 7 DAR. In addition, among all psyllids that were found on orange jasmine trees, 19% were individuals released in the control treatments.
olfactometric assays and volatile emission analysis. In 4-arm olfactometer assays, results indicated that D. citri females spent more time on orange jasmine than on the clean air fields (V = 1270.00; d.f. = 1, 60; P = 0.0185) (Fig. 6a). However, no significant effect was observed on the D. citri female foraging activity to the 'Pera' sweet orange vs. clean air (V = 1036.00; d.f. = 1, 64; P = 0.8139) (Fig. 6b). This different D. citri behavior to sweet orange and orange jasmine suggests different volatile profiles between both genotypes. To gain insight into the attractiveness of orange jasmine volatiles, comparative untargeted volatile analysis of sweet orange and orange jasmine flushes was performed. Principal component analysis revealed two separated clustering groups, one for each studied genotype (Fig. 7a). PC1 explained at least 75% of the variance at two independent replicate dates. Area integration of peaks corresponding to relevant loadings at both sampling dates (detailed in Supplementary Table 1) revealed important quantitative and qualitative differences in volatiles emitted from both genotypes (Fig. 7b). For example, 37 compounds (of which 26 are sesquiterpenes) were detected only in the orange jasmine emission profile. On the other hand, the emission of monoterpenes characteristic of sweet orange leaves, such as αand β-phellandrene, d-limonene and linalool, was highly reduced in orange jasmine.

Discussion
In the current study, we evaluated first the use of orange jasmine as a trap crop in two types of orange orchards, recently planted and several years old, that differed in plant height and biomass (new orchard: 1 m-tall; well established orchard: 3 m-tall; Supplementary Fig. 1). The presence of orange jasmine as a trap crop in the border of new citrus orchards caused overall reductions of 40% in the accumulated number of psyllids captured (Fig. 1a) and 83% in the number of psyllids settled on citrus trees (Fig. 2), when compared to the fallow field plots  Regarding D. citri spatial distribution, most psyllids were found in the first rows of citrus (Fig. 3) regardless of the treatment and assessment period, reinforcing psyllid preference for trees located in the border of the orchards 10,25,26 . Psyllids released in front of the trap crop plots did not disperse further than the first row of citrus, whereas individuals released in front of control plots were able to cross the first row and were present in almost all rows. On the other hand, a high number of individuals were observed on orange jasmine trap crop during the first assessments (Fig. 4), including insects that were released on control plots. These results may be explained by a strong attraction effect associated to the orange jasmine trees, which could have reduced psyllid movement into the orchard, and suggest that the psyllid has the ability to detect and move to the preferred host rather than performing a passive random dispersal. This hypothesis is in accordance with previous studies that report the importance of visual and olfactory cues in the host plant finding ability of D. citri [27][28][29] .
The low effect of orange jasmine trap crop on D. citri infestation when tested on well-established citrus orchards also supports the hypothesis of visual and olfactory cues for D. citri behavior. Probably, large differences in plant size between orange jasmine and citrus trees on mature plots may lead to higher presence of visual/olfactory cues from citrus trees than from orange jasmine plants, thus explaining the absence of trap crop effect over the D. citri population in this case. According to Shelton and Badenes-Perez 12 , a successful trap crop relies on the combination of trap crop (e.g. size, phenology, and attractiveness) and pest characteristics.  To evaluate the effectiveness of orange jasmine olfactory cues as D. citri attractants, 4-arm olfactometric assays were performed (Fig. 6). An increase of ~30% in D. citri preference to orange jasmine volatiles in relation to that of 'Pera' sweet orange was observed. High attractiveness of orange jasmine flushes to D. citri has been also observed in Y-olfactometer 19 . Therefore, our olfactometric data reinforces the D. citri preference to volatiles from orange jasmine instead of sweet orange flushes. In this sense, the orange jasmine odors plume could contribute for psyllid mobility towards orange jasmine, which increases the D. citri settling preference to this host, as observed in field experiments. Regarding the volatile analysis, we found that orange jasmine volatile profile is clearly distinguishable to that of sweet orange (Fig. 7). Then, analyses were conducted to identify which compounds were relevant to make this distinction either quantitatively or qualitatively (Supplementary Table 1). Green leaf volatiles, such as hexenyl acetates 30 , were more emitted by orange jasmine than by sweet orange leaves, as previously reported 31 . Monoterpene emission was also different between genotypes, being all them less abundantly emitted by orange jasmine. Lower emission of d-limonene and β-ocimene was also found in orange jasmine in relation to lemon (Citrus × limon (L.) Osbeck), rough lemon (Citrus × taitensis Risso syn. Citrus jambhiri Lushington), sweet orange, grapefruit (Citrus × aurantium L. syn. Citrus paradisi Macfad.) and Citrus × macrophylla Wester in other studies 19,31,32 . Orange jasmine emits less β-pinene and linalool than 'Red Rio' grapefruit, 'Meyer' lemon (Citrus × limon) and 'Valencia' sweet orange 19,31 . Most of the relevant compounds emitted by orange jasmine corresponded to sesquiterpenes, usually absent or poorly emitted by Citrus leaves 19,[31][32][33] . Nearly all of the identified sesquiterpenes have been reported before as emitted by orange jasmine leaves 19,31,32 . These results are in accordance with the highest attractiveness of psyllids to orange jasmine volatiles in the olfactometer device (chemical cues) and field experiments (chemical and visual cues). Delving in this area until determination of which compound/s are responsible for higher attraction of orange jasmine leaves to D. citri would allow the development of more efficient traps to control/monitor insect populations.
Besides reducing D. citri population and its spread, the presence of orange jasmine trap crop at the edge resulted in 43% reduction in HLB incidence at the orange orchard. Similarly, a cereal border crop reduced in 51.5% the incidence of Bean yellow mosaic virus (transmitted by aphids) compared to fallow fields in narrow-leaved lupin (Lupinus angustifolius L.) 34 . Despite orange jasmine has been reported as a CLas host, bacterial titers are much lower in this host than on citrus trees 20,21,[35][36][37] , and consequently, it may be considered as an irrelevant inoculum source epidemiologically. Additionally, it is known that D. citri adults are less efficient than nymphs to acquire CLas [38][39][40] , and the transmission process demands a latent period of at least 7-10 days before psyllids become bacteriliferous 41 . Moreover, at 7 DAR the number of psyllids on orange jasmine plants decreased 97% (Fig. 4), probably due to the treatment with thiamethoxam. This observation was confirmed with data from the experiment in which psyllids confined on thiamethoxam-treated orange jasmine plants presented mortality close to 100% at 7 DAC (Fig. 5). The low number of released psyllids recaptured on citrus trees (0.3%) in comparison to that of insects recaptured when no trap crop was used 7,25 supports that thiamethoxam treated-orange jasmine acted as a 'sink' for D. citri, attracting, killing and consequently, reducing psyllid movement into the citrus orchard. Therefore, it is reasonable to consider that the acquisition of CLas from a psyllid that landed on treated orange jasmine and subsequent inoculation in heathy citrus trees from a commercial orchard would be close to zero.
In summary, this study demonstrated for the first time that M. paniculata treated with insecticides may act as a trap crop to attract and kill D. citri before settling on the border of citrus orchards. However, the use of orange jasmine as a trap crop should be implemented before, at the same time or soon after citrus tree planting. Moreover, our work opens the possibility of performing studies assessing the trap crop integration with other tactics (e.g. kaolin), as a 'push' and 'pull' strategy, which could decrease further infestation rates inside citrus orchards. Finally, in order to avoid the use of insecticides, a genetically modified trap crop, able to interfere with D. citri survival, could be used in the citrus edges to attract and kill D. citri. An analogous approach has been used in Hawaii, where borders of transgenic papaya plants resistant to Papaya ringspot virus reduced the spread of aphid-vectors and then viral incidence on non-transgenic papaya crops 16 .  Supplementary Fig. 1c,d) with spacing of 6.5 × 2.8 m. Both areas were historically subjected to continuous influx of D. citri from neighbor areas with high HLB incidence.

Effects of orange jasmine as a trap crop on Diaphorina citri natural infestation and HLB inci
Each area was divided into two plots of 100 × 120 m. In one plot of each area, the orange jasmine trap crop was planted on east (area A) and north (area B) edges, 20 m from the first citrus tree, in two double-rows (1 m separated) with spacing of 0.4 × 0.4 m per plant. In total, 325 orange jasmine trees (~0.6 m-tall) were planted per row forming a canopy with 2 m in width and 120 m in length. The remaining plots of each area were maintained as fallow mowed grass and used as controls.
Orange jasmine trees were treated with drench applications of thiamethoxam (Actara ® 250 WG, Syngenta Proteção de Cultivos Ltda., Paulínia, SP, Brazil) 10 days before planting (0.25 g tree −1 ) and every 70 days thereafter (0.31 g of active ingredient per meter of tree height). In addition, foliar applications of insecticides with different modes of action (pyrethroid, organophosphate and neonicotinoid) were applied at interval of 14 days to both, trap crop and citrus orchards. The trap crop was fertilized every 60 days with NPK (10-10-10) at 100 g tree −1 .  (Fig. 8). Control plots were maintained as fallow mowed grass. The orange jasmine trees were treated with drench application of thiamethoxam (0.47 g tree −1 ) 10 days before planting. In order to ensure the insecticide efficacy, 10 nursery M. paniculata trees treated with thiamethoxam and 10 untreated were planted on the central west side of the experimental area at 2 m from the last citrus row. In these plants, groups of 10 adult psyllids (10-15 days old) were confined on a young shoot of each orange jasmine tree using sleeve cages, and D. citri mortality was assessed at 1, 3 and 7 days after the beginning of the experiment. Psyllid confinements on untreated orange jasmine trees were used as a control. Adult psyllids, obtained from a colony free of 'Ca. Liberibacter spp. ' , maintained for several generations on orange jasmine seedlings at Fundecitrus (Araraquara, SP State, Brazil), were marked with different colors of fluorescent powder (Day-Glo Color Corp., Cleveland, OH, USA) 43 in order to differentiate psyllids released on orange jasmine and fallow plots. Before release, marked insects were acclimated for 48 h on orange jasmine seedlings. Field release was conducted as described by Tomaseto et al. 7 in 1.5 m-tall platforms located on the east side of the area at 10 m from the first citrus trees of each plot.
The experiment was replicated three times, and psyllid releases were always performed in the afternoon (~15:00), which is the period of the highest D. citri flight activity 27,44,45 , with 800-1000 marked psyllids per plot. The number of settled psyllids was assessed by visual inspection on 24 citrus trees (central trees of each row) from each plot (Fig. 8), at 1, 3 and 7 days after release (DAR). Temperature and rainfall were monitored using a weather station (Vantage Pro2 6152; Davis Instruments, Hayward, CA, USA) 20 m far from the experimental area. For the first replicate, the mean temperature ranged from 22.0 to 25.1 °C with total precipitation of 19.2 mm, while for the second and third replicates, temperature and rainfall values ranged from 23.3 to 27.8 °C with 64.5 mm and from 21.7 to 24.06 °C with 6.1 mm, respectively ( Supplementary Fig. 2). olfactometric assays. The assays were carried out in a climate-controlled room at temperature 25 ± 2 °C, relative humidity 65 ± 10%, and 3000 lux luminosity. The preference of D. citri toward volatiles was investigated using a 4-arm olfactometer (30.0 × 30.0 × 2.5 cm in length, width, and height, respectively), adapted with a yellow acrylic floor arena, essentially as described in Zanardi et al. 46 . Individual constant (0.1 L min −1 ) charcoal-filtered humidified airflow, provided by an oil-free air compressor (Schulz MSV6, Schulz, Joinville, SC, Brazil) converged through 0.635 cm-diameter individual PTFE tubes (Sigma-Aldrich, Bellefonte, PA, USA) to the center of the acrylic arena. A single mated 7-15-day old female was released on the center of the arena. Psyllids that did not perform a choice after 5 min were recorded as "no response" in the analysis. In case of response, 10 min were allowed to observe the time spent in each one of the four odor fields. For each psyllid (replicate), the time spent in each odor source was recorded. Data were collected from 10:00 a.m. to 4:00 p.m. from five different assay days and a total of 61 (orange jasmine × clean air) and 65 ('Pera' sweet orange × clean air) replicates were used. Two of the four possible arms received plant volatiles whereas the two remaining arms received clean air 33 . Odor sources were switched each assay, and the arena was rotated after each responding insect to prevent bias. Orange jasmine and 'Pera' sweet orange nursery trees (~1 year old) to be used as odor sources were pruned 20 days before assays to stimulate the emergence of new shoots in a greenhouse.
Volatile emission analysis. New flushes from 1m-tall M. paniculata and Citrus × aurantium sweet orange trees were detached and kept 30 min with the petioles submerged in water to acclimate. Subsequently they were enclosed in 20 mL screw-cap Pyrex tubes carrying a Teflon septum on the top and containing 1 mL of milli-Q water (for avoiding leaf hydric stress) and maintained 24 h at a controlled temperature of 22 °C. Volatile capture and GC-MS analysis were performed using a 6890 N gas chromatograph (Agilent Technologies Inc., Las Rozas, Spain) coupled to a Thermo DSQ mass spectrometer equipped with a DB-5 ms (Agilent J & W Columns) column (60 m × 0.25 mm i.d. × 1.00 μm film) as described before 33 . In short, SPME fiber (100 μm poly (dimethyl) siloxane/divinylbenzene (Supelco Inc., Bellefonte, PA) was exposed for 30 min at 22 °C and immediately afterwards transferred to GC injector (220 °C) where thermal desorption was prolonged to 4 min. The GC interface and MS source temperatures were 260 °C and 230 °C, respectively. Oven programming conditions were 40 °C for 2 min, 5 °C min −1 ramp until 250 °C, and a final hold at 250 °C for 5 min. Helium was the carrier gas at 1.5 mL min −1 in the splitless mode. Data was recorded in a 5975B mass spectrometer (Agilent Technologies Inc., Las Rozas, Spain) in the 35-250 m/z range at 7 scans, with electronic impact ionization at 70 eV.
Samples from orange jasmine and sweet orange were analyzed by triplicate and at two different sampling dates. At each sampling date, a quality control sample (composed by mixtures of both genotypes) was analyzed by triplicate. Datasets were processed independently by the MetAlign software (www.metalign.nl) for full mass spectral alignment, baseline correction, noise estimation, and ion-wise mass spectral alignment. The MetAlign output scan peak values were corrected by leaf fresh weight and volatile accumulation time and then subjected to PCA analysis with Past3.17 software (folk.uio.no/ohammer/past). Significant scans were manually analyzed on TIC chromatograms to quantify the area and tentatively identify compounds by matching the acquired mass spectra with those stored in reference libraries (NIST, MAINLIB, REPLIB). The area for each compound was corrected by leaf fresh weight and volatile accumulation time and used to generate a heat-map using ClustVis 47 . Data analysis. Data from the commercial and the experimental citrus orchards were analyzed by Poisson generalized linear mixed models (GLMM) using the "glmmADM" 48 (zero-inflated) and the "lme4" 49 packages, respectively. Treatment was considered as fixed effect, while each assessment dates (commercial citrus orchards) or repeated measures on assessed trees (experimental orchard) as random. Time was considered fixed effect on experimental orchard data. The effect of treatment, time and interaction was assessed by likelihood-ratio tests (P < 0.05) between a full and a reduced model. In order to compare the effect of trap crop on cumulative mean number (commercial citrus orchards) or on counts of marked psyllids on citrus trees on each independent assessment time (experimental orchard), data were analyzed by a quasi-Poisson generalized linear model (GLM) 50 . Goodness-of-fit was assessed through half-normal plots with simulation envelopes using the "hnp" package 51 . In case of significant differences, means were separated by computing the 95% confidence intervals for linear predictors using the "lsmeans" package 52 . A 2 × 2 chi-squared contingency table was used to determine the treatment effects on frequency of D. citri detection (percentage of assessments with psyllid detection in relation to total number of assessments) in each commercial orchard plot. In order to analyze the efficacy of systemic insecticide applied on orange jasmine trees, survival rates in the whole assessment time were compared by a log-rank test (P < 0.05) and survival data at each time after release were compared by a quasi-binomial GLM. Infestation maps were generated using the Surfer ® software (Golden Software Inc., Golden, CO, USA) by the inverse of square of the distance interpolation method, considering the sum of insects found in all replicates.
For the olfactometric assays, each pair of a plant (orange jasmine or 'Pera' sweet orange) and control (clean air) in behavioral measurements (time spent in each odor field) was compared by using two-tailed, Wilcoxon matched-pairs signed rank test. All analyses were performed using the statistical software "R", version 3.3.1 53 .

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.