Phylogenetic inference enables reconstruction of a long-overlooked outbreak of almond leaf scorch disease (Xylella fastidiosa) in Europe

ALSD is widespread in Majorca

The summer of 2017, following the Xf official detection in October 2016, provided a good scenario for testing whether Xf was implicated in the almond decline. An episode of high water demand (three consecutive days of T_max > 33 °C and R_H < 60%; Fig. 1a) in mid-June 2017 triggered the synchronous onset of ALSD symptoms throughout Majorca. Scorch affected both the leaves attached to single terminal shoots and, more frequently, those on main branches or on the whole crown (Fig. 1b, c). In mid-July, distinctive leaf scorch and ‘golden death’ symptoms, as described elsewhere⁹, were prevalent (Fig. 1d). Such an explosion of ALSD symptoms concurred with a European Commission audit, which motivated changes in the recommendation from eradication to containment measures [DG (SANTE) 2017–6216; EU 2017/2352]. ALSD symptoms emerged two weeks later in the summer of 2018 than in 2017 due to wetter and milder late-spring early-summer weather, stressing the annual variability in ALSD symptom expression and other Xf-related diseases driven by environmental factors^43,44.

ALSD is caused by two Xf subspecies in Majorca

Seventy-six out of 105 almond leaf samples across Majorca brought to the LOSVIB in the summer of 2017 tested positive for Xf in qPCR analyses (using both Harper et al.⁴⁵ and Francis et al.⁴⁶ qPCR protocols). Multilocus sequence typing (MLST) analysis (n = 23) based on DNA extracts from almond leaf samples and some recovered Xf isolates revealed three STs: ST1 belonging to subsp. fastidiosa; a new ST, ST81, belonging to subsp. multiplex; and ST7 belonging to subsp. multiplex, detected only once (https://figshare.com/s/0b571280bcd7738a524d; 10.6084/m9.figshare.12378302). ST81 differs from known Californian isolates within the ‘almond’ clonal complex ST6 and ST7 in only one of the seven alleles used for MLST. ST81 shared an identical MLST profile with the isolate ‘Fillmore’ (Acc. n°: CP052855.1) recovered from an olive tree in California. Both ST1 and ST81 were well distributed across the island, with ST81 also infecting the widespread wild olive trees among other hosts, whereas strain ST1 caused Pierce’s disease as well⁶. No clear spatial pattern in their distribution was observed, suggesting long-term spread and mixed-infection scenarios (Supplementary Fig. 1).

In total, 55 isolates (62%; n = 89 attempts) were obtained from different almond trees recovered from petioles of symptomatic leaves (qPCR positive) in the summers of 2017 and 2018. Colony morphotypes similar to those reported by Chen et al.⁴⁰ in almond orchards of California were distinguished on periwinkle wilt GelRite (PWG) medium, i.e. almond A-genotypes and grape G-genotypes corresponding to the subsp. multiplex and fastidiosa, respectively (Supplementary Fig. 2). Both subspecies were once isolated from the same tree, and in the same orchard; Xf-DNA of both subspecies was detected in the growth rings of two out of three trees examined, strongly indicating frequent coinfections (Supplementary Fig. 2).

The current ALSD incidence and tree mortality preclude a recent introduction

By counting the trees that showed disease symptoms previously attributed to fungal trunk pathogens as an advanced stage of ALSD (see Methods), the incidence for the disease in 2017 was visually determined to be approximately 79.5% ± 2.0 (mean ± SEM) (Supplementary Data 1; Supplementary Software 1). From this percentage, we estimated that ~1,250,308 almond trees, including dead trees, would have been infected by Xf (2017 almond plantation census: 19,417 ha with an average density of 81 trees ha⁻¹; https://www.mapama.gob.es). Similarly, we extrapolated that at least ~552,869 dead trees remained in the fields (Supplementary Data 1). This number is actually an underestimation of the trees that have died since the beginning of the outbreak, given that many dead trees were removed from the field. Disease incidence ranged between 16.6 and 100%, with half of the orchards exhibiting ALSD incidences over 90 and 30% mortality (Supplementary Data 1). ALSD incidence affected trees of different ages equally (P = 0.50); however, younger plantations (trees ≤ 30 years) suffered less mortality than plantations with trees older than 30 years (χ = 5.37; df = 1, P < 0.020).

Such a high ALSD incidence across the island suggest a relatively old entry of the pathogen. To estimate the rate of disease spread, we needed some historical reference to the ALSD incidence. The Google street view panoramic-image repository has been used to assess the distribution and prevalence of other pest and diseases in a territory such the pine processionary moth in France⁴⁷. This approach provided the means to approximately assess the ALSD incidence and mortality in 2012, given that the pictures covered a large part of the territory that year (see Methods). Thus, 249 orchards distributed throughout the island were visually examined, and the average incidence of ALSD was estimated at 53.4% ± 1.6 (mean ± SEM) for 2012 (Supplementary Data 2). This figure already exceeded the inflection point in the disease progression curve projected in a logistic model (y) = ln (y/(1 − y)), where y is the proportion of infected trees in a predominant tree-to-tree transmission, while the rate of new infections decreases in proportion to the number of remaining uninfected almonds.

We predicted a greater accumulation of Xf-infected plants in those areas of previous almond decline. As expected, the orchards closest to the putative main focus area (radius < 20 km) and therefore with a longer exposure time to Xf showed a higher incidence (χ = 25.21, df = 1, P = 0.0001; Supplementary Data 1) and mortality (χ = 26.49, df = 1, P = 0.0001; Supplementary Data 1), supporting the hypothesis that Son Carrió was an expanding focus. Likewise, the incidence and severity of Pierce’s disease had previously been shown to be higher in orchards closer to the main propagation centre in Son Carrió⁶. To further capture the spatial-temporal spread of ALSD, we mapped the estimated disease incidence (%) and mortality (%) distribution among orchards for 2012 and 2017 using QGIS⁴⁸ geographic information system software (Fig. 2). A careful review of the maps revealed a heterogeneous dispersion with several emerging expansion centres, in addition to the main centre in the Son Carrió area (Fig. 2). This distribution pattern could be explained considering two factors: (i) the occasional long-distance dispersal through infected grafts—a common practice in Majorca—at the beginning of the epidemic and (ii) the regional differences in the rate of expression of disease severity and mortality due to environmental and biological factors, such as soil type, agricultural practices, differing susceptibilities of almond varieties and precipitation. In the case of Son Carrió, the combined effect of an early introduction, low rainfall, and shallow soils could have increased the severity of ALSD and therefore resulted in an earlier disease awareness than in any other area of the island.

ALSD precedes symptoms caused by fungal trunk pathogens

We initially doubted whether the trunk pathogenic fungal complex could be the cause of the almond epidemic, and Xf infections were an additional factor that would aggravate the disease. In mid-summer 2017, we noted that trees with ALSD symptoms commonly intermingled in the same orchard with others that exhibited different stages of general decline, shoot and branch diebacks, or a combination of both (Supplemental Data 1 and 3). In our evaluations of disease incidence, we found a significant statistical dependence between ALSD incidence (%) and tree mortality (%) within orchards (2017 disease incidence assessment: Spearman’s rank correlation coefficient: ρ (126) = 0.88; P < 0.0001; 2012 disease incidence assessment ρ (249) = 0.89; P < 0.0001; Supplementary Data 1 and 2). Friedman’s test showed that there were significant differences among repeated measures of tree severity scores in the time series, which corresponded to the sequences of symptom development, from leaf scorch to shoot and branch diebacks to tree death, over time (χ = 42.41, df = 4, P < 0.0001: Kendall’s coefficient of agreement = 0.69; Supplementary Fig. 3; Supplementary Data 3). Furthermore, ALSD symptoms preceded shoot and branch death in 96% of cases, while trees without ALSD symptoms remained healthy until images were no longer available (Supplementary Data 3). Together, these results suggest a continuum in the process from systemic Xf infection to tree death, as illustrated in Fig. 1c. Although these general observations do not prove causality, they show that Xf infections mainly preceded any fungal trunk symptoms, thus corroborating the explanation for Xf-induced water stress as a disturbance factor favouring fungal activation from an endophytic commensal lifestyle to a virulent pathogenic stage⁴.

A combination of dendrochronology and qPCR revealed Xf old infections

Some direct evidence of Xf spread before 2003 emerged after examining the presence of Xf-DNA in the tree rings of the wood sections (see Methods). We plotted for each felled tree (n = 34) the relationship between the quantitative cycle (Cq) values obtained in the qPCR and the dating of the tree rings from which sawdust samples were taken. Consistent with the assumption of annual cyclic colonization by Xf of newly formed xylem tissue in spring (see Methods), the Xf-DNA concentration measured in Cq values significantly increased from the oldest to the youngest rings (linear model: F₁,₁₈₃ = 66.11, P < 0.0001; Supplementary Figs. 4 and 5; Supplementary Data 4A, B). For example, Xf-DNA extracted from growth rings corresponding to 2008 ± 2 was amplified by qPCR in 12 of the 34 trees examined, while this number decreased to nine trees in the 2004 ± 2 rings. In a tree sample (XYL 739/17), Xf-DNA was amplified (Cq = 29) from a growth ring dated 1998 ± 2. In contrast, Xf-DNA was not detected in rings dated prior to 1995 (Supplementary Data 4A, B). Additionally, specific primers and probes targeting genome-specific regions in a duplex qPCR assay⁴⁵ allowed the differentiation between fastidiosa and multiplex subspecies in rings of 25 trees, with nine and 19 trees showing infection by subspecies fastidiosa and multiplex, respectively, and three trees showing a mixed infection. The analysis enabled dating the infection back to 1998 (subsp. fastidiosa ST1) and before 2000 (subsp. multiplex ST81) in Maria de la Salut and Binissalem, respectively (Supplementary Data 4C). After plotting the infection frequencies of dated rings in the 34-almond samples, we observed that the infection frequency sharply increased from the inner older rings to the outer rings, fitting this distribution to a logistic disease progression curve from 1993 to 2017 (Fig. 3a; Supplementary Data 4A). Notably, 1993 was the extrapolated date for epidemic onset (disease incidence < 0.01%) in the logistic model and very closely fits the date of the most plausible introduction event (Fig. 3b).

Survival analysis

Taking advantage of the Xf qPCR data from dated trunk rings for each felled tree, we calculated the time from infection to tree death (decrepitude). When the trees were cut down in 2017 and 2018, most were still alive, so they were right censored in the survival analysis (Supplementary Data 4D). The Kaplan–Meier median (50%) survival estimate was 14 years (95% CI: 11–17), considerably longer than the 3–8-year-period reported by Mircertich et al.⁹ in California (Fig. 4). Our results nevertheless were closer to those reported by Sisterson et al.⁴³ in a 6–7 year monitoring of almond plantations affected by ALSD in California, where they found that 91% of infected trees survived to the end of the study.

Both Xf subspecies are pathogenic and reproduce ALSD symptoms

To explain the high disease incidence and mortality observed, the ‘spreading alien pathogen’” hypothesis requires Xf strains to be broadly pathogenic to more than 87 local and 23 nonlocal almond varieties that are growing throughout the island and identified in the almond germplasm collections in Majorca. The extent of their pathogenesis, however, was only partly revealed in the inoculation assays of 2018. Strain XYL 2055/17 (ST1) from grapevine caused typical almond leaf scorch symptoms in 24 out of the 160 saplings in nine out of 11 varieties 16 weeks after inoculation (Supplementary Table 1). Two out of eight plants of cv. Filau inoculated with strain XYL1752/17 belonging to ST81 in October 2017 developed scorch symptoms in August 2018. In both inoculation experiments, the pathogen could be detected by qPCR on leaves more than 10 cm from the point of inoculation and reisolated from the petioles, thus fulfilling Koch’s postulates. In 2019, we realized that the almond varieties used in the inoculation experiments were among the less-susceptible varieties in the field after monitoring ALSD infection incidence in almond germplasm collections distributed in four fields around Majorca, which may explain the low rate of disease symptoms observed. Seventy-eight of 110 almond varieties naturally exposed to Xf for several years showed ALSD symptoms, and infections by the bacterium were confirmed by qPCR (Supplementary Table 2).

ALSD is transmitted by spittlebugs

Based on ‘olive quick decline syndrome’ epidemiological studies in Italy²¹ and Pierce’s disease in Majorca⁶, P. spumarius was the main candidate vector for ALSD. In the transmission experiment, P. spumarius adults reared from nymphs and caged on ALSD symptomatic trees vectored Xf to healthy almond saplings. Xf was detected by qPCR in seven out of eight almond plants exposed to ‘infected’ P. spumarius and reisolated from symptomatic leaves on six of those plants. Both control plants exposed to uninfected insects (n = 2) and almond plants unexposed to insects (n = 2) were qPCR negative. Xf DNA was detected in all post-mortem P. spumarius adults (n = 35) used in the inoculation treatments, while all insects (n = 10) in two control replicates proved PCR negative. On the other hand, we investigated the prevalence of Xf subspecies in infected P. spumarius adults in the field by analysing the MLST of Xf-DNA within a subsample of 55 adults that were Xf positive for qPCR collected in or near three almond orchards. Among the positive spittlebugs analysed by MLST (25 samples), for 19 of them, some MLST loci could be amplified; two were infected by subsp. multiplex with alleles belonging to ST81, and 17 insect samples had alleles of subsp. fastidiosa ST1 (https://figshare.com/s/0b571280bcd7738a524d; 10.6084/m9.figshare.12378302).

Fungal trunk pathogens are nonspecifically related to almond decline

In previous works^2,3,49, Koch’s postulates were fulfilled for most of the pathogenic trunk fungi found associated with almond decline; however, none of them could be ascribed as the aetiological agent of the decline. Therefore, the term ‘complex disease’ was used in those studies. We collected data from the work of Olmo et al.⁴⁹ to determine whether the regional diversity of fungal species associated with almond decline could be related to Xf infection. We observed that fungal diversity increased in those areas where ALSD severity was the highest and therefore with the longest exposure time to the pathogen (Fig. 2; Supplementary Fig. 6). Both of these observations agree with the explanation that the endophytic fungi would be activated to their pathogenic phase as a result of the stress induced by the Xf infection in the tree (i.e. the disturbance factor). Related to this lack of specificity of wood fungal pathogens, we found that two of the most pathogenic and frequently isolated fungal species on almonds in Majorca, Neofusicoccum parvum and Pleurostomorpha richardsiae (Supplementary Table 3), were also initially implicated in ‘olive quick decline syndrome’ caused by Xf in Apulia, Italy⁵⁰. In contrast, in the nearby province of Foggia, Apulia (300 km from Gallipoli), still free of Xf, the same fungi provoked olive decline without causing death⁵⁰.

Biologically induced water stress versus climatic drought

Among farmers, the almond decline was generally attributed to the increase in drought periods due to climate change. Drought has also been partly associated with the appearance of the almond wood fungal complex⁴⁹. If there was a relationship between drought and almond decline, an increase in drought episodes would be expected beginning at the time the disease was first noticed around 2003 (see Methods; Supplementary Data 5). To test this, we compared the crop average cumulative water deficit index (CWDi) during the periods 1988–2002 (before decline) and 2003–2017 (after decline). We found that the mean CWDi before 2003 was lower (CWDi = −263.4), indicating greater water stress, than that after 2003 (CWDi = −235.7), but this difference was not statistically significant (t = 1.25, df = 28, P = 0.22). Because it could be argued that drought periods prior to 2003 could have exceeded an irreversible threshold that activates pathogenic fungi to the present day, we looked for similar drought periods before 1988. We found that other severe episodes of water deficit anomalies occurred during 1963–1968 and 1981–1984⁵, without triggering any almond decline in the following decade. Surprisingly, the wettest period (i.e. highest CWDi) of the general data series between 1988 and 2017 occurred between 2004 and 2010 (see Supplementary Data 5), matching the period of the exponential growth of the ALSD outbreak (Fig. 3a). Regardless of the effect of weather as the main driver of ALSD epidemics, the most plausible explanation for the emergence of fungal trunk pathogens is the disturbance (i.e. induced drought) caused by infection and colonization of xylem vessels by Xf.

From the 1990s to 2003: the cryptic period

Taken together, the data presented thus far strongly support the “spreading alien pathogen” hypothesis (i.e. Xf introduction) as the original cause of almond decline in Majorca (beginning in 2003¹) over previous explanations of ‘endemic’ fungal trunk pathogens being activated by abnormal environmental changes. This almond disease showed a well-defined main focus in Son Carrió, several spreading centres with wave-pattern fronts, and a clonal genetic population structure, all of which are distinctive traits for an “introduced spreading pathogen”. We conclude that ALSD was overlooked and confounded with the advanced stages of Xf infection when the action of a complex of fungi aggravated the disease. As occurred with the initial diagnosis of ‘olive quick decline syndrome’ in Apulia, Italy⁵⁰, the almond decline was also associated with a complex of fungal trunk pathogens^1,2, comprising a similar assortment of fungal species in both diseases (Supplementary Table 3). This is a key piece in the reconstruction of the ALSD epidemic and the main chronological link between the introduction event and the awareness of the presence of Xf in Majorca in October 2016.

Similar to other introduced plant pathogens, Xf underwent a lag period of inoculum build-up at the early epidemic stage before damage became apparent⁵¹. Given our median survival estimate of 14 years for ALSD-infected trees, the disease emergence in Son Carrió around 2003, the lack of detection of Xf-DNA in growth rings prior to 1995 and the disease logistic model projecting 1993 as the time with a 0.01 disease incidence (Fig. 3b), we proposed the beginning of the 1990s as the most likely time for Xf introduction. In 2017, we performed an epidemiological investigation to find connections between California and Majorca related to ALSD. A relevant publication was found in which it described a visit to the Central Valley of California in August 1993 of a group of main stakeholders of the Balearic almond sector (agricultural extensionists, almond cooperative members, etc.) to learn about crop management in California⁵². Although there could have been other older non-documented visits and pathways of introduction, the most plausible explanation of how two coetaneous ALSD-related Xf strains only known at that time in California reached Majorca is that infected scions were brought from California into Majorca and grafted onto local rootstocks. To address this hypothesis, we tested whether infected almond buds could transmit Xf and develop ALSD. In this transmission experiment, only two of the 13 (~15%) plants grafted with buds from Xf-positive trees in June showed scorch symptoms on both the rootstock and the scion three months after grafting, and Xf could be detected by qPCR and isolated. In our grafting transmission experiment, buds were collected from infected trees at the beginning of June, when trees are mostly asymptomatic and the bacterial load in the vascular system is lower and thus may be irregularly distributed along the wood tissue, which could be one of the reasons for the number of successful infections being lower than expected. Nonetheless, a 15% transmission rate extrapolated to a common plantation of one hectare considerably increases the likelihood of Xf establishment in new areas. Moreover, in our further efforts to link epidemiological events between California and Majorca, we identified two orchards, one in the municipality of Consell (centre of the islands) and the other in Son Carrió, where Californian almond varieties had been previously grafted onto local rootstocks. According to a farmer in Son Carrió, the plantation began to decline several years after grafting, which led him to graft again with a local almond variety onto the Californian variety. We counted the growth rings in two wood sections of a single tree and determined 1995 ± 1 as the date of grafting. We expected to detect Xf-DNA in growth rings around 1995, but in one single tree analysed, the oldest ring with Xf-DNA was from 2006 ± 2 (Supplementary Data 4).

The ALSD outbreak in Majorca began with the introduction of Xf strains from California

To delimit the divergence between the Xf isolates of California and Majorca, we separately analysed the phylogenies of the fastidiosa and multiplex subspecies. Details of the de novo assembled genomes and those retrieved from GenBank that are included in the phylogenetic analyses are provided in Supplementary Data 6.

Xf subsp. fastidiosa

In total, 2239 single nucleotide polymorphisms (SNPs) were identified from the set of 1872 monocopy core genes shared by 12 USA isolates and 15 from Majorca. Maximum likelihood (ML) phylogenetic analysis confirmed that all Xf isolates from subspecies fastidiosa from Majorca form a monophyletic cluster (1000 bootstrap scores) within the Californian clade (Supplementary Fig. 7). Similar topologies in phylogenetic trees were inferred using Bayesian analysis at the divergence time applying different combinations on the model parameters (Supplementary Table 4). Among these models, the uncorrelated relaxed exponential clock model combined with the Bayesian Skyline coalescent model produced the best-fit phylogeny. Root-to-tip dating provided little temporal signal but was significant enough to calibrate the molecular clock without tree priors (Supplementary Fig. 8a). Thus, we calculated the time of the most recent common ancestor for the Majorcan clade to be around 2004 (95% highest posterior density, HPD: 1982–2015). The average substitution rate calculated was 7.71 × 10⁻⁷ substitutions nucleotide-site⁻¹ year⁻¹ (95% HPD: 1.20 × 10⁻⁷–1.69 × 10⁻⁶), almost identical to that reported previously^12,53, and corresponding to ∼1.3 SNPs core-genome⁻¹ year⁻¹.

Although these and other phylogenetic trees built without restrictions at clade nodes and tree roots identified clades from the eastern United States, California and Majorca, they all showed wide 95% HPD bars at their root heights and internal nodes. Such uncertainties were mainly due to the variation in the substitution rates in the branches due to the small number of genomes that covered the entire timespan of the divergence analysis¹². To overcome this drawback, we introduced strong informative tree priors to our model based on the data presented herein and in other published works^54,55 (Table 1). Thus, we legitimately placed 1998 as the minimum age limit of the Majorcan monophyletic clade (our equivalent of a fossil record in an unequivocally well-dated stratum) and 1993 as the expected node age with a normal logarithmic distribution (the expected introduction event). In addition, we limited the minimum age for the tree root to 1892, the year Pierce’s disease was first reported in Southern California⁵⁵. Once we entered these tree priors into the model, we estimated the time of the most recent common ancestor for the Majorcan clade to be 1994 (95% HPD: 1990–1997), close to the expected introduction event in 1993. The closest relative to the Majorcan clade was the strain CFPB 8071, isolated from an almond tree in California, which diverged around 1983 (95% HPD, 1976–1986; posterior probability, pp = 1). Our time-calibrated phylogenetic tree correctly assigned the dates for the interior nodes of each clade related to the chronology of the appearance of ALSD in California⁹ and the estimated date of 1882 (95% HPD: 1878–1886; pp = 1) for the introduction of Xf ST1 in the USA⁵⁵ (Fig. 5).

Despite the lower number of sequenced isolates (n = 6), the California clade showed a greater genetic polymorphism (1074 SNPs; nucleotide diversity, π = 2.6 × 10⁻⁴) and a deeper phylogenetic structure (π = 1.8 × 10⁻⁶) than the 16 SNPs found within the Majorcan clade (n = 15) and its shallow comb-shaped clade topology (Fig. 5). This tree pattern is expected for the introduction of a small subset of genotypes from a larger source population into a new area (founder effect) after undergoing genetic drift⁵⁶. Such differences in the number of SNPs between both transcontinental populations also reflect a longer timespan from the introduction event⁵⁵ (~1892) and the largest geographic area in California (423,970 km²) with respect to the recent introduction in Majorca (~1993) and its limited geographical range (3640 km²). On the other hand, the presence of two subclades within the Majorcan clade suggests some incipient evolutionary divergence from the introductory event. The genetic variability between and within these subclades was compared by calculating the fixation index (Fst), which measures population differences due to genetic structure. A higher Fst (0.21) than expected for a founder introduction was obtained. However, contrary to expectations, both subpopulations were apparently not genetically structured either by their hosts or by their geographical distribution, since the isolates of vines and almonds throughout the island were interspersed within both subclades (Fig. 5). Furthermore, when the Majorcan isolates were grouped according to the host and analysed for their genetic diversity within and between each group (vines and almond trees), we found an Fst = ~0. These data strongly suggest the existence of cross infections between almond trees and vines by ST1 strains of subspecies fastidiosa, a limited gene flow between both subclades and a high clonality within them, since no recombination was detected using ClonalFrameML⁵⁷ (Supplementary Fig. 9). Before drawing conclusions about the genetic structure of both subclades, more samples would need to be included in the genetic analysis. Nonetheless, these preliminary data agree with our hypothesis of the introduction of a single or few Xf ST1 genotypes through infected buds or scions and their subsequent long-distance spread through grafting across the island.

Xf subsp. multiplex

Differences among ST81 isolates were limited to 11 SNPs of the core genome (1,664,716 base pairs), a range similar to that found in the ST1 population in Majorca (independence test; χ = 0.42, P = 0.52). Furthermore, both subspecies showed a similar nucleotide diversity (subsp. multiplex, π = 2.5 × 10⁻⁶), which supports the idea that they have undergone parallel demographic trajectories since the time of their introductions. As also occurred with ST1, the same haplotypes of ST81 were detected in two hosts, in this case on almonds and wild olive trees. Cross infections between both hosts need to be tested in inoculation experiments to confirm whether the widespread distribution of wild olive trees has contributed to the spread of ALSD across the island (see Supplementary Fig. 1), as well as the possible spread to 12 other hosts infected with ST81, including the fig and the ash of narrow leaves. Identical haplotypes have also been found on the islands of Majorca and Menorca, which are interspersed in the same clade in the phylogenetic analysis, indicating that the subpopulation of the island of Menorca very possibly derives from an introductory event from the main island. Hereafter, we denote this clade formed by ST81 isolates as (Majorca (Menorca)). Possible routes of entry include the transfer of infected plant material for grafting or the introduction of infected insects into ships, vehicles or containers that move daily between Majorca and Menorca.

The concatenated sequence alignment of 25 core genomes comprised 1,615,082 nucleotides. A total of 2511 SNPs (544 singletons, i.e. mutations appearing only once among the sequences) were identified, of which 1341 SNPs (π = 4.1 × 10⁻⁴) belonged to the Californian clade sensu stricto. ML analysis confirmed that the isolates of the multiplex subspecies from (Majorca (Menorca)), Alicante and Corsica, together with isolates belonging to ST6 and ST7 from California, form a robust monophyletic group previously described by Landa et al.³¹ (Supplementary Fig. 7b). Because all European isolates were collected between 2015 and 2018 (n = 18), only a weak temporal signal was obtained that incurs pseudo-replication to calibrate the molecular clock for the panel of 25 genomes (Supplementary Fig. 8b). To solve this problem, we divided the panel into four monophyletic clades, i.e. California, (Majorca (Menorca)), Corsica and Alicante, based on the ML analysis (Supplementary Fig. 7b) and ran the best-fit model (Supplementary Table 5). For each monophyletic clade, we calibrated the age of the nodes based on the tree priors and transferred their uncertainties to the probability distribution (Table 1). For example, we used an intermediate introduction scenario (1993) for the Corsican clade due to the lack of references on the Xf strains for each of the scenarios in the mathematical model, as well as the few genomes available to infer the time of introduction⁵⁴. For the clade (Majorca (Menorca)), DNA detection of the multiplex subspecies in a ring dated 2000 ± 2 in one almond tree in Majorca (Supplementary Data 4) provided a minimum limit for the age of this node. On this basis, it was estimated that the time of the most recent common ancestor of ST81 in the Balearic Islands would be 1995 (95% HPD: 1991–1998) and the introduction of Alicante ~2005 (95% HPD: 1996–2017) and the isolates CFPB8417 and CFPB8418 of Corsica ~2000 (95% HPD: 1995–2002) (Fig. 6). These results are consistent with our proposed introduction event in 1993 and are also in line with the current epidemiological stage of ALSD in Alicante, while supporting one of the introduction scenarios proposed for Corsica in 2001 by Soubeyrand et al.⁵⁴, and the age of the nodes agrees with the chronology of the appearance of the ALSD in California⁹.

Source: Ecology - nature.com