Spatial pattern of Bois noir: case study of a delicate balance between disease progression and recovery

Bois noir (BN) is the most important phytoplasma disease of Euro-Mediterranean area and induces severe loss of production and even the death of vines. Understanding the delicate balance between disease progression and recovery of BN infected plants over space and time is crucial to set up management tools. The data collected and analysed allowed to provide insights into dispersal pattern of the disease, caused by’Candidatus Phytoplasma solani’. Point pattern analysis (PPA) was applied to assess the spatial arrangement of symptomatic plants and the spatial correlation of disease levels in four vineyards. For spatio-temporal patterns of BN, a mark-correlation function was applied. Space-time PPA over multiple years (2011–2015) provided graphical visualisation of grapevines more severely affected by BN along the borders of the vineyards, mainly in 2011 when disease incidence was high. PPA across the symptomatic plants in the four vineyards confirmed this visual trend: an overall aggregated pattern at small (<10 m) spatial scales (2013) that were more evident later at all spatial scales (0–15 m). Application of this innovative spatial approach based on point and surface pattern analyses allowed the spread and severity of BN to be monitored, to define the dispersal routes of the pathogen. Such data will contribute to better understand the distribution of symptomatic plants over space and time and to define a model for preventive strategies to reduce future infections.


Materials and methods investigated areas.
This study was carried out in a pilot vineyard cultivated with 'Chardonnay' (MOV), and in three commercial 'Chardonnay' vineyards (MOG, OSI, CAS), with all located within the Marche region. Supplementary Table S1 gives the details of their main features (i.e., location, cultivated area, geographic coordinates, altitude, rootstock, plant spacing, trellis system, year of planting), their surrounding crops/vegetation, their soil and pest management, and the main agronomical practices followed. For OSI and CAS vineyards, IPM practices were adopted, after an accurate monitoring of symptoms. In particular, copper-based formulations were used to control grapevine downy mildew, powdery sulphur for powdery mildew, and Bacillus thuringiensis was used for Lobesia botrana, while weeds were managed through mechanical tools. For the pilot vineyard MOV, data about production (weight, °Brix) from 2009 to 2015, were provided by the owner.
The main climatic data (i.e., mean monthly temperature, cumulative monthly rainfall) of the areas for the relevant years of investigation were provided by the local ASSAM weather stations (Agenzia Servizi Settore Agroalimentare Marche; Marche Region, Ancona, Italy). These data were used to calculate annual and seasonal (i.e., spring, summer, autumn, winter) mean temperatures and rainfall for each vineyard, and to define climatic anomalies respect to regional data collected from 1981 to 2010.

Disease assessment and sanitary status of the vineyards. Visual inspections were carried out for
MOV from 2009 to 2015 in mid-September, and similarly in MOG, CAS, OSI from 2013 to 2015. During the surveys, the positions of symptomatic grapevines were recorded on a two-dimensional map, and the disease severity was evaluated according to an empirical scale 6 . Overall, in the period 2013-2015, one hundred seventy BN-symptomatic, 95 recovered and 115 healthy grapevines were randomly sampled in the surveyed vineyards to assess the presence of 'Ca. P. solani' . The total DNA was extracted according to the protocol suggested by Angelini et al. 33 then the biomolecular assays were based on the use in nested PCR with fStol/rStol, according to Maixner et al. 34 . The phytosanitary status was recorded for each vineyard. The data recorded were used to calculate the following: = aNnbS Annual new symptomatic grapevines, never before ( ) Number of new symptomatic grapevines, which in previous years were always asymptomatic, = Symptom persistence Number of symptomatic grapevines with symptoms for two consecutive years or more, = Recovery rate of symptomatic plants Number of recovered grapevines that were symptomatic for one to six years previously, = tA Total asymptomatic plants per year ( ) Number of asymptomatic grapevines that were never symptomatic in previous years, where c is the value of the severity class, f is the frequency in the severity class, and n is the number of symptomatic grapevines 35 . The mean disease severity was calculated for the always symptomatic grapevines (AS), the annual new symptomatic grapevines (aNS), and the recovered plants that showed again symptoms (SRec), for each year of survey. These data were statistically analysed, applying Turkey's HSD test for multiple comparison of means, at P ≤ 0.05 using R software (ver. 3.1.2, R Development Core Team) equipped with 'car' package. The same approach, method and software was used to statistically analyse the data according to the period of recovery (one to five years) before showing again symptoms, and recovery rate of plants after a different period of symptom persistence (one to six years).
Spatial and temporal analysis of Bois noir in the vineyards. The spatially explicit datasets in the present study were organised and managed in a GIS environment. For each plant, the following attributes were collected: study site (MOV, MOG, CAS, OSI), geographic position (latitude, longitude), phytosanitary status (S, symptomatic; A, asymptomatic; R, recovered) and disease level (0-4, asymptomatic to seriously affected).
For the MOV site only, a natural neighbour spatial interpolation method was adopted to obtain a map of the disease level for the period of 2011-2015. This simple geostatistical method is available in the ArcMap software, and it provided the surface data (spatial resolution, 0.2 m) from a point dataset.
The index of aggregation (nearest neighbour index; NNI) was calculated in the ArcMap software, through the mean nearest neighbour tool (Spatial Statistics), to obtain a unique value for each site and year 36 . The NNI was then used in Pearson's correlation analysis (Pearson's r), against the climatic and main epidemiological data. PPA methods were applied to define the spatial patterns of the symptomatic plants within the four vineyards (CAS, MOG, MOV, OSI), and the spatial correlations of the disease level variable. The pair-correlation univariate function (g(r)), a second-order statistic that provides information at multiple spatial scales 37 , was calculated for each site. The pair-correlation function is non-cumulative and uses only points separated by a certain distance r, allowing specific scales to be identified where significant point-point interactions occur, particularly at small spatial scales 38 . From a preliminary univariate PPA performed on each site separately, a similar overall spatial pattern was found in all sites. For this reason, we conducted the spatial analysis considering the four sites as pseudo-replications 38 , as we were interested in the mean spatial pattern of symptomatic plants within the four vineyards. The results of the four sites were thus combined in one mean graphic pair-correlation function, using the 'combine replicates tool' , as included in the Programita software 37 . The univariate pattern of symptomatic plants was compared with the complete spatial randomness (CSR) null model, to allow for second-order effects 37 . The 95% confidence intervals for the univariate analyses were computed from 1000 Monte Carlo simulations, and the goodness-of-fit (GoF) tests for the null hypothesis (CSR) was performed 37,38 . The analysis was carried out only for the classes with more than 100 points, applying a 1 m lag distance and a maximum distance of 15 m using the grid-based software Programita, with a grid size of 1 m 2 and a ring width of 5 m.
To determine the spatio-temporal patterns of the levels of disease, a normalized mark-correlation function, also known as r-mark-correlation function (K mm (r)), was used on three sites (CAS, MOV, OSI) for the 2013-2015 period 37,39 . This function is used to analyse the spatial relationships among points containing quantitative attributes (marks), and in this study the level of disease (classes 1 to 4) was the quantitative mark, with mark-correlation function used to test for positive or negative correlations between the values of the marks 37 . As a result of this analysis, positive correlation defined points that were closely located and with similar level of disease (mark), and negative correlation defined points that were closely located but with different level of disease (mark). All of the PPAs were computed using the Programita software 37 .

Results
During the period from 2009 to 2015, the mean temperature recorded for these seven years was higher by ~1 °C with respect to the regional data . In particular, in the surveyed locations, the years 2009, 2011, 2012 and 2015 showed constistent anomalies for the mean temperatures during spring (from +0.9 °C, to +3.4 °C), summer (from +0.9 °C, to +3.3 °C), autumn (from +0.5 °C, to +2.3 °C), and winter (from +1 °C, to +3.3 °C). On the other hand, the annual precipitation (mm) that was recorded for the same period for these sites surveyed in the Marche region was generally higher than the mean of regional climatic data (1981-2010).
Sanitary status of the vineyards. Independent of the location, these symptomatic 'Chardonnay' grapevines in the vineyards surveyed showed light BN symptoms at the end of June, which progressively became more severe during the season, to reach their apex in the middle of September. These symptoms were characterised by discoloured yellowing leaves, leaf rolling that often resulted in an angular shape, dried out flowers, bunches on plants that became brown and shrivelled, with numerous small pustules seen along diseased branches ( Figure S1). In 153 out of 170 (90%) of the symptomatic samples analysed in the nested PCR with the primers fStol/rStol, a specific amplicon of ~500 bp of 'Ca. P. solani' was obtained. No amplification was obtained in the recovered and asymptomatic samples collected (data not shown).
The data collected in September during the period 2009 to 2015 for the pilot vineyard MOV, and during the period 2013 to 2015 for the further vineyards MOG, OSI and CAS, were summarised to define how the sanitary status of these vineyards changed over time. The sanitary status was determined from the equilibrium rates of the asymptomatic, symptomatic and recovered grapevines.
In the MOV vineyard, a bell-shaped curve of total symptomatic plants (tS) was recorded over the 7 years, with the highest level reached in 2011, where there were 814 symptomatic plants, of a total of 1629 plants (50%); the lowest number was seen for 2015 (9.2%) ( Table 1). The annual rates of new symptomatic plants (aNS) were around 50% of the total symptomatic plants ( Table 1). Most of aNS were plants that were 'never before symptomatic' (aNnbS) in the previous years (491 in 2009, 320 in 2010, 421 in 2011) (data not shown), while from 2012 there was a large reduction (73 plants) that continued up to 2015, when there were only 10 never before symptomatic plants (data not shown). Over these seven years, there were only 5 plants that were always symptomatic for the seven consecutive years, 9 for six consecutive years, 30 for five consecutive years, 89 for four consecutive years, and 318 for three consecutive years (Table 1). It is worth noting that after seven years of assessment in the pilot vineyard MOV, there were 237 plants (14.5%) that had never shown any BN symptoms (Table 1).
Considering the annual rates for the recovered plants (aRec), calculated respect to the symptomatic plants in the previous year, these generally increased from 19.5% (96/491) in 2010 to 80.8% (328/406) in 2014, then decreased at 70.2% (144/205) in 2015 (Table 1). Starting from 2012, the number of plants that recovered annually was usually higher than the number of plants that maintained the symptoms. In the last two years, there was a sharp increase in the permanently recovered plants (pRec) ( Table 1).
The rate of recovered plants which showed again symptoms (SRec) was not statistically different considering the duration of previous recovered period (from one to five years) ( Table 2). Also the recovery rates of plants was not statistically different according to the duration of persistence of symptoms (one to 6 years) ( Table 3). A high mean of recovery rate (≥60%) was recorded for plants that showed symptoms for only one year and also for longer periods (up to five years) ( Table 3).
The disease severity, calculated for each of seven year of assessments, was generally higher in the always symptomatic plants (AS) compared to the new symptomatic plants (aNS), and to the recovered plants that showed symptoms again (SRec). In all cases though, there were no statistically significant differences seen year by year ( Table S2).   Table 4).     (Fig. 2). This underlying clustered pattern was valid for all of the vineyards studied except for MOG and MOV in 2015, which showed random patterns (Fig. 2,  inset). This situation is given as the mean overall pattern from the use of the 'combine replicates' technique, and it showed a general increase in the scale of aggregation between the symptomatic plants.

Year of survey
The spatio-temporal patterns of the levels of disease were analysed according to univariate normalised mark-correlation functions. These showed significant positive correlations for the 2013-2015 period for CAS   www.nature.com/scientificreports www.nature.com/scientificreports/ only, within the range of the 1-3 m scale (GoF: p ≤ 0.002) ( Table 6). This positive correlation means that neighbouring plants tended to have similar levels of disease.

Discussion
In the pedoclimatic conditions of central-eastern Italy, grapevines infected by 'Ca. P. solani' generally start to show symptoms in mid-June, according to the seasonal trend. Endeshaw et al. 10 described the progressive development of symptoms in 'Chardonnay' grapevines, along with the changes in their physiological parameters (e.g., net photosynthesis, stomatal conductance, transpiration), from before the appearance of symptoms to the harvest (August-September) 10 . In these months, grapevines are considered to be diseased when at least three of the typical parameters are observed, from among the following: (i) shrivelled berries or clusters; (ii) lack of lignification in shoots; (iii) black pustules on the shoots; (iv) partial or total leaf yellowing and/or downward leaf curling; and (v) www.nature.com/scientificreports www.nature.com/scientificreports/ leaf blade fall, with the petiole still attached to the plant, as a late symptom. In this way, the picture of the sanitary status from the visual assessment is more reliable, because the grapevines, showing unambiguous symptoms, have generally high phytoplasma titre, thus there is a robust correlation between symptomatic plants and detection of 'Ca. P. solani' by molecular tools 6,40 . The different locations assessed in the present study (MOV, MOG, Montalto delle Marche; OSI, Osimo; CAS, Castelplanio) showed small differences in the temperatures recorded. However, climatic parameters were not directly correlated with the NNI index, calculated for the symptomatic plants, and epidemiological parameters determined (e.g., rate of newly infected and recovered plants, disease severity), having same pedoclimatic conditions. On the other hand, Panassiti et al. 11 using a Bayesian model demonstrated that the presence of the BN disease increased with the presence of the vectors (H. obsoletus) and it was correlated with environmental conditions (altitudes, mean annual temperature), as well as with grapevine cultivars. This dataset for the BN epidemic presented and analysed here is of particular importance because of the density of the measurements (as individual plants), the tracking of the symptomatic and recovered plants over time and space, and the length of time over which these measurements were taken. Indeed, a similar approach was also followed recently to describe FD epidemiology in a vineyard of cv 'Barbera' in Piedmont 41 .
In the pilot vineyard MOV, where the visual assessments were carried out annually in September from 2009 to 2015, there were two severe outbreaks of BN in 2010 and 2011, with peak incidence levels of about 44% and 50%, respectively. The rate of infection then progressively decreased, to reach about 9% in 2015.
The rates of annual new infections here were high, as generally 50% of the total symptomatic plants. These data suggest the intense activity of the BN vector H. obsoletus, as demonstrated by research carried out in vineyards of Marche region 42 . Although the vectors usually feed on weeds in vineyards, in the absence of these, they are stimulated to move onto the grapevines more frequently. Panassiti et al. 43 demonstrated that the low density of the host plant U. dioica and high vector abundance increase pathogen occurrence. In our situation, the weed management inside the vineyard chosen by the growers saw weed elimination in May, when the grapevines can suffer from their competition. This then directed the vectors towards feeding on the grapevines, which had become the only host plants available for these insects 11,44 .
Indeed, such cultural practices against these host plants, such as weeding and herbicide application, should not be carried out during the flight period of H. obsoletus, to prevent movement of the vectors towards the grapevines 45,46 .
For the persistence of symptoms, it is worth noting that only a few of the grapevines persistently showed symptoms, in agreement with Rott et al. 47 ; hence, the sanitary status of vineyards infected by BN can be interpreted as a dynamic situation. After the two BN epidemic years (2010, 2011), in 2012 there was initially an increment in symptom remission that was related to the temporary recovered grapevines (tRec; with 1-2 years of recovery). In the following years, the trend towards recovered plants was confirmed, which reached a high rate (>75%) for these grapevines (pRec; recovery of>3 years), of around 50%. The total production of the vineyards was also dramatically reduced around the epidemic years, but then increased for the last 3 years, to reach acceptable grape quantity and quality (Table S3).
The persistence of symptoms was not related with the rate of recovery here. The chance for these 'Chardonnay' grapevines to recover was significantly higher independent of the persistence of symptoms. It appears not profitable the replacement of plants affected by BN when the recovery is the most course of the disease 4 . Not only for BN, high rates of natural recovery from FD have been reported for 'Prosecco' grapevines, as well as for 'Chardonnay' grapevines 48 .
The duration of the previous recovery period also did not influence significantly the overall average of the rate of grapevines that showed disease symptoms again after undergoing recovery. However, several studies theorized the physiological basis of recovery. Cytological and biochemical analyses revealed that in the recovered plants, the genes encoding chalcone synthase, phenylalanine ammonia-lyase, and class III chitinase appear to be consistently involved in the recovery phenomenon, with their gene expression not affected by plant phenology 21 . Specifically, as well as in apple and apricot, it has been shown in grapevine that recovery coincides with the accumulation of hydrogen peroxide in the sieve tubes, which often induces increased resistance 26 . In addition, there is abnormal callus accumulation, along with the associated proteins, due to the overexpression of the genes coding for callus www.nature.com/scientificreports www.nature.com/scientificreports/ synthase and protein P, as has been seen for recovered apple plants 49,50 . This defined the hypothesis that the recovered plants can develop resistance mechanisms dependent on Ca 2+ signalling activity. Generally, if we consider the history of the grapevines in the present study, the risk of symptom expression in those showing recovery for 3-5 years was lower than in the always healthy grapevines, but without any statistical significance.
The space-time PPA carried out here in the pilot vineyard MOV over multiple years (2011-2015) allowed statistical description of the BN progression and regression dynamics. Considering the distribution of the grapevines in the vineyard according to the severity of BN symptoms, the maps for MOV obtained through natural neighbour spatial interpolation provide graphical visualisation of the grapevines that were more severely affected by BN. In 2011, these were positioned mainly along the borders of the vineyard, when the BN incidence was higher 51,52 . Mori et al. 53 analysed the factors affecting the spread of "bois noir" disease, and they demonstrated that the incidence of border sides with nettle on vineyard surface was positively correlated to disease incidence in the vineyards with aggregate distribution of symptomatic plants. However, in our situation, we were not able to record nettle plants, but bindweed inside and around, as well as alfalfa plantations on the border. In a previous studies carried out in the vineyard MOV, we analysed most of symptomatic samples collected in 2011 and 2012 and most of them were molecularly characterized as 'Ca. P. solani' tuf-type b1 and only sporadic one were identified as tuf-type a 6,35 .
The general tendency to move from a diffuse mixed-severity pattern to a patchy spatial pattern would appear to be due to recovery of the less severely affected plants at the borders of the symptomatic patches. This geostatistical analysis allowed the hypothesis that the inoculum source was indeed outside the vineyard, and that it corresponded to the border where alfalfa (Medicago sativa) was grown. Moreover, when this approach was applied in the 2011-2015 period, it allowed evaluation of the progress of the disease in the field, to trace the epidemic history. This thus provides an understanding of dispersal routes of the pathogens, as well as the impact and efficiency of some of the control strategies and the role of the vectors, as has been reported for FD 41 . Indeed, the positions of the infected grapevines confirmed a natural source of inoculum and the activity of potential vectors in spreading BN in this vineyard. Recently, Quaglino et al. 54 identified new 'Ca. P. solani' vectors to grapevines, and they hypothized, in Northen Italian pedoclimatic conditions, the spreading of phytoplasma to vineyard boarders through H. obsoletus and its further transmission within the vineyard by the feeding activity of the alternative vectors.
It is worth noting the gradual reduction in the disease levels in the fields from 2011 to 2015. The recovery in this vineyard over these years appears to be related to the plants that showed mild symptoms in the previous year, while most of the grapevines that showed more severe symptoms (i.e., classes 3, 4) remained symptomatic. The analysis of data showed that the mean disease severity recorded for the always symptomatic grapevines (AS) was generally higher, even if not statistically significant, than that of the annual new symptomatic vines (aNS). It was significantly evident in 2014 and 2015 in OSI and CAS vineyards. These data showed that in our pedoclimatic conditions the plants expressing symptoms from long time, showed systemic symptoms on more than 50% of canopy. On the other hand, new symptomatic plants generally showed localized symptoms on one or few shoots.
The PPA for the symptomatic plants in these four studied vineyards confirmed the visual trends through the geostatistical approach. Indeed, we observed an overall aggregated pattern at small (<10 m) spatial scales in 2013 that later became more evident at all spatial scales (0-15 m). The increased aggregation observed for the 2013 to 2015 period was significant for all of the sites studied. The overall spatial patterns of the symptomatic plants suggested that the length scale of the sampling is an important parameter in the complete spatial randomness validation, as the time at which the assessment is carried out. For vineyards OSI and CAS, a clustered distribution was recorded. The spatial arrangements of symptomatic plants in clusters was particularly evident for CAS, where there was positive spatial correlation between disease levels at the small spatial scales (1-3 m; mark-correlation function). This would appear to be due to the patchy distribution of C. arvensis in the area. In particular, the association of U. dioica and C. arvensis with BN symptomatic grapevines and insect vector captures, and their different distribution clustering within the examined vineyards, have been well documented in previous studies carried out in Europe under different pedoclimatic conditions 44,53,55,56 .
In the investigated vineyards, we did not observe an anisotropic pattern along the rows, as reported in a recent study by Maggi et al. 41 . This can be ascribed to the transmission of FD by the vector Scaphoideus titanus from grapevine to grapevine 57 , while in contrast, the behaviour of the vector(s) responsible of the spread of BN is not strongly affected by the linear pattern of the grapevines, but rather by the C. arvensis islands inside the vineyards 14,55,56 .
In conclusion, the data collected for a large number of single grapevine plants throughout a long period of monitoring, elaborated by this innovative approach based on PPA and SPA, allowed to highlight dynamics in BN incidence and severity fluctuation in time and space, to define the sanitary status of a vineyard as a dynamic balance between its symptomatic, recovered and healthy grapevines, and to trace and understand the dispersal routes of the pathogens. With this information supported by innovative 58 and more sensitive molecular tools 59 , as well as molecular typing of 'Ca. P. solani' , it will be very useful for farmers for their decisions whether to replace or maintain a BN-infected grapevine.