Molecular epidemiology of the citrus bacterial pathogen Xanthomonas citri pv. citri from the Arabian Peninsula reveals a complex structure of specialist and generalist strains

Abstract Molecular epidemiology studies are essential to refine our understanding of migrations of phytopathogenic bacteria, the major determining factor in their emergence, and to understand the factors that shape their population structure. Microsatellite and minisatellite typing are useful techniques for deciphering the population structure of Xanthomonas citri pv. citri, the causal agent of Asiatic citrus canker. This paper presents a molecular epidemiology study, which has improved our understanding of the history of the pathogen's introductions into the Arabian Peninsula, since it was first reported in the 1980s. An unexpectedly high genetic diversity of the pathogen was revealed. The four distinct genetic lineages within X. citri pv. citri, which have been reported throughout the world, were identified in the Arabian Peninsula, most likely as the result of multiple introductions. No copper‐resistant X. citri pv. citri strains were identified. The pathogen's population structure on Mexican lime (their shared host species) was closely examined in two countries, Saudi Arabia and Yemen. We highlighted the marked prevalence of specialist pathotype A* strains in both countries, which suggests that specialist strains of X. citri pv. citri may perform better than generalist strains when they occur concomitantly in this environment. Subclade 4.2 was the prevailing lineage identified. Several analyses (genetic structure deciphered by discriminant analysis of principal components, RST‐based genetic differentiation, geographic structure) congruently suggested the role of human activities in the pathogen's spread. We discuss the implications of these results on the management of Asiatic citrus canker in the region.


| INTRODUC TI ON
Agricultural industries are threatened by plant disease emergences, a consequence of many factors, including the globalization of plant product trade, a marked increase of human international travel, crop intensification linked to the increase in human population size and climate change (Anderson et al., 2004;Paini et al., 2016;Savary et al., 2019). Refining our understanding of the emergence, spread, and evolution of plant pathogenic populations requires a thorough comprehension of the determinants of their fitness and their eco-evolutionary dynamics at the within-host up to agricultural landscape scales in response to environmental selective pressures (McDonald & Stukenbrock, 2016;Penczykowski et al., 2016;Plantegenest et al., 2007). Bacterial pathogens exhibit a highly variable degree of host specialization and can be broadly divided into specialists versus generalists on the basis of their host range width (Baumler & Fang, 2013;Shaw et al., 2020). The degree of host specialization of pathogens has major implications for disease progress and management. Heterogeneous agricultural landscapes have a negative impact on the natural dispersal of highly specialized pathogens (McDonald & Stukenbrock, 2016;Plantegenest et al., 2007).
Moreover, the degree of host specialization of a pathogen markedly influences surveillance strategies (Gandon et al., 2013;Morris et al., 2022;Parnell et al., 2017). In the specific case of some perennial crops affected by highly specialized pathogens, scion replacement using top grafting strategies can be efficient management options (Mudge et al., 2009). Among factors involved in the emergence of pathogens, the impact of host specialization on disease dynamics and epidemiology is a major issue. Population biology, which has been very successful in furthering our knowledge on the dynamics and evolution of single-host pathogens, has the ability to refine the understanding on the key question of the ecological and evolutionary advantages and disadvantages of host specialization (Woolhouse et al., 2001). There is a strong need for more experimental data to (i) evaluate the performance of specialist vs. generalist pathogens in situations where they occur concomitantly and (ii) compare their fitness (Barrett et al., 2009).
Citrus is an important crop worldwide (FAS, 2019). For several decades, it has been threatened by various diseases. One of the most severe, Citrus bacterial canker (CBC), has damaged the citrus industries and is now endemic in many hot-humid citrus growing regions of the world . CBC has a wide host range among citrus species and relatives (Gottwald et al., 1993). Canker symptoms occur on all aerial organs and severe CBC causes fruit blemishing, defoliation, and early fruit drop (Gottwald et al., 2002). Two pathovars of Xanthomonas citri (i.e., infra-specific groups of bacteria with similar pathological characteristics such as host range and disease syndrome) cause visually indistinguishable canker-like symptoms on some Citrus species and sometimes on other rutaceous genera.
Xanthomonas citri pv. citri, the causal agent of Asiatic citrus canker (ACC), infects a single plant family, Rutaceae. In agricultural terms, this is by far the most significant form of CBC . CBC is an agriculturally major citrus disease worldwide.
It causes direct (premature leaf and fruit drop, decrease in fruit quality and yield) and indirect losses (restricted access to export markets, increase in production costs due to implemented surveillance and management options, undesirable side effects of pesticide application), thereby resulting in a significant socioeconomic impact on impacted citrus industries Stall & Seymour, 1983). The annual cost of CBC in Florida only (approximately 0.3 million hectares of commercial citrus at the time of assessment) was estimated as ca. 350 million US$ (Gottwald et al., 2002).
Strains of X. citri pv. citri differ in host range. Three pathotypes (A, A* and A w ), i.e., groups of strains causing canker on a distinct range of citrus species and/or distinct defense phenotypes on nonhost citrus, were described (Rybak et al., 2009;Sun et al., 2004;Vernière et al., 1998) (Derso et al., 2009;Escalon et al., 2013;Vernière et al., 1998). Pathotype A w has the unique feature of inducing a hypersensitive reaction (HR) when inoculated at high titer into some non-host citrus lines, such as grapefruit (C x paradisi) and sweet orange (C. x sinensis) (Rybak et al., 2009;Sun et al., 2004). This phenotype was found to be associated with the specific presence in A w strains of the type III effector avrGf1 (Rybak et al., 2009;Webster et al., 2020). This suggests that the genetic basis of the restricted host range in A* and A w is likely to be different. Although pathotype A w strains were first reported in Florida, it is now thought that they originated in India (Bui Thi Ngoc, Vernière, Jarne, et al., 2009;Schubert et al., 2001). While A* and A w strains have a globally lower economic impact due to their narrow host range, they are still economically important in countries with a well-developed lime industry. For example, extensive cankers and dieback caused by A* strains were observed on Mexican lime in Thailand (Bui Thi Ngoc et al., 2007).
Given the high agricultural significance of X. citri pv. citri, numerous genotyping techniques for subtyping outbreak strains have been developed over the years (Bui Thi Ngoc, Vernière, Jarne, et al., 2009, Bui Thi Ngoc, Vernière, Vital, et al., 2009Cubero & Graham, 2002;Jeong et al., 2019;Pruvost et al., 2014). Only some of the techniques are powerful. Minisatellite and CRISPR genotyping are currently the most suitable techniques for locating outbreak strains in relation to the global diversity of X. citri pv. citri (Jeong et al., 2019;Pruvost et al., 2014). Online databases make it possible to compare new outbreak strains with reference strains (http://www.biopr ed.net/ mlva/). These techniques may be valuable alternatives to whole genome sequencing (WGS) for surveillance of not highly critical cases.
They can also be helpful in countries where bioinformatic analysis of WGS data is not straightforward. Microsatellite genotyping has also proved useful in molecular epidemiology studies in order to (i) assess the population structure of X. citri pv. citri at micro geographical scales, (ii) characterize new subclades, (iii) identify pathways associated with pathogen spread, and (iv) decipher invasion routes (Leduc et al., 2015;Pruvost et al., 2015Pruvost et al., , 2019Vernière et al., 2014).
To date, all these studies have analyzed outbreaks associated with a single pathotype. The molecular methods outlined above have the potential to analyze more complex epidemiological situations, where several pathotypes infect Mexican lime and/or its very close relative, alemow (C. x macrophylla). species of X. citri pv. aurantifolii and X. citri pv. citri (Rossetti, 1977).
It was hypothesized that the putative pathogen was introduced into Yemen via contaminated citrus trees imported from India in 1981 (Dimitman, 1984). CBC was first observed in Saudi Arabia on Mexican lime at the end of 1983, in Sabya (Jizan region). It was thought to have originated from Yemen (Fadlallah, 1986). Additional observations of the disease were recorded from 1985 on the same species in several citrus nurseries of the Najran region and three years later in orchards (Fadlallah, 1986). The pathogen isolated from these regions at that time was identified as X. citri pv. citri pathotype A* (Vernière et al., 1998). A more recent study suggested that both pathotypes A and A* occur in Saudi Arabia (Al Saleh et al., 2014).
Using a detached leaf assay, a recent study confirmed that there was some variation in pathogenicity on grapefruit among strains from Saudi Arabia (Ibrahim et al., 2019). As both pathotype A and a subclade of pathotype A* were found to produce lesions on grapefruit after inoculation (Escalon et al., 2013), Ibrahim et al. (2019) concluded that precise genotyping data is needed to determine whether or not pathotype A occurs in Saudi Arabia. CBC was reported in 1984 in the United Arab Emirates in the Dhaid area (Bové, 1995).
Strains obtained from this outbreak produced a virulence pattern similar to that of pathotype A strains (El Goorani, 1989). Lastly, CBC was first observed in Oman in 1986 on Mexican lime and grapefruit in two remote areas located near Ibri and Salalah, respectively (Bové, 1995). Bacterial strains isolated at that time were highly diverse and were assigned to all pathotypes reported to date (A, A*, and A w ). Besides India, Oman is the sole country that hosts all three pathotypes of X. citri pv. citri (Gordon et al., 2015). Attempts to eradicate the pathogens were recommended in the four countries at that time (Bové, 1995). However, the action plans were ineffective and the pathogen has continued to spread geographically at least in Saudi Arabia (Ibrahim et al., 2017).
In the context of (i) the high regional genetic diversity of X. citri pv. citri (Gordon et al., 2015;Pruvost et al., 2015) and (ii) uncertainties with regard to the lineage(s) behind the emergence of CBC on the Arabian Peninsula, the present study set out to characterize the CBC-causing pathogen(s) in Saudi Arabia and Yemen, with special emphasis on Mexican lime (i.e., a widely distributed citrus species that can host all X. citri lineages reported to date and on which CBC is highly prevalent). We conducted a precise genetic and pathological assessment of the diversity of Saudi and Yemeni strains. Drawing on a recent comprehensive strain collection, supplemented with strains collected during the first outbreaks in the 1980s in Saudi Arabia, Yemen, and Oman, we compared them to an international collection of reference strains. Here, we used genotyping data to refine the understanding of the histories of CBC establishments in Saudi Arabia and Yemen and confirmed the presence of pathotypes A and A* in both countries (Dimitman, 1984;Ibrahim et al., 2019). This complex epidemiological situation prompted an extensive sampling on Mexican lime on which we deciphered the distribution and genetic structure of several populations differing in host range, and highlighted a marked prevalence of specialist strains in this environment.

| Surveyed areas
Surveys on CBC were conducted in Saudi Arabia (SA) and Yemen (YE) in citrus-growing areas. We surveyed: thirty-two citrus blocks on commercial farms and four citrus nurseries in Al-Baha, Al-Madina, Aseer, and Jizan in Saudi Arabia; and twenty-one blocks on commercial farms in Al-Hudaydah, Lahj, and Taiz in Yemen ( Figure 1a and Table S1). When available, ten citrus leaves with putative CBC symptoms were randomly collected from each tree at each site (30 trees/site). Samples were sealed in plastic bags and taken to the laboratory at the King Saud University (College of Food and Agriculture Sciences, Plant Protection Department) for the culture of Xanthomonas spp.

| Isolation of bacterial strains
Infected leaves were washed in running water and surface-sterilized by wiping with 75% ethanol. Bacteria were isolated from diseased leaf tissues. This involved cutting leaves with canker lesions into small pieces, crushing them in sterile distilled H 2 O and leaving them for 10 min at room temperature. One loopful of the bacterial suspension was streaked onto KC semi-selective medium containing yeast extract 7 g L −1 , peptone 7 g L −1 , glucose 7 g L −1 , agar 18 g L −1 , cephalexin 40 mg L −1 , kasugamycin 20 mg L −1 , propiconazole 20 mg L −1 ; pH 7.2 (Pruvost et al., 2005). The plates were incubated at 28°C for 3 days. Single bacterial colonies displaying characteristics of Xanthomonas spp. were selected and purified by re-streaking on YPGA plates (yeast extract 7 g L −1 , peptone 7 g L −1 , glucose 7 g L −1 , agar 18 g L −1 ; pH 7.2) for 24 h at 28°C. The bacterial strains were stored in 50% glycerol at −80°C until use. Although strains originated from several Citrus species, Mexican lime was sampled more extensively (Table S1). Whenever possible, local populations (i.e., collection including ≥14 strains originating from distinct plants sampled from a single grove or nursery; n = 23) were assembled. Twentyfour and three strains isolated in the 1980s from Saudi Arabia and Yemen, respectively, were included as reference.

| Phenotypic characterization
Pure cultures of putative Xanthomonas spp. were assayed using biochemical and physiological tests, including gelatine and casein hydrolysis and growth in the presence of 3% NaCl, as described by Vernière et al. (1991). A subset of 36 strains (Table S1) were assayed for (i) copper resistance, as described previously (Richard et al., 2017) and pathogenicity on Mexican lime and mandarin cv.
Kinnow, using the attached leaf assay. Briefly, adult leaves from the youngest vegetative flush were infiltrated with bacterial suspensions containing ca. 1.10 5 cfu mL −1 . Disease development was scored weekly. Technical details are available in previous studies (Escalon et al., 2013;Vernière et al., 1998).

| Polymerase chain reaction (PCR) assay
The bacterial strains used in this study are listed in Table S1. Bacterial cultures were grown in YP broth tubes (yeast extract 7 g L −1 ; peptone 7 g L −1 ; pH 7.2) and incubated for 16-18 h at 28°C in an orbital shaker. F I G U R E 1 (a) Map of the Arabian Peninsula showing the three regions from Saudi Arabia and the three governorates from Yemen where citrus canker samples were collected (including the number of authenticated Xanthomonas citri pv. citri strains and their genetic assignation). Note that the exact isolation place is unknown for Omanese strains from a previous study (Vernière et al., 1998). (b) Close-up map. Blue solid lines link pairs of subclade 4.2 local populations (blue dots) for which no significant genetic differentiation (p > 0.05 based on R ST ) was found. DNA was purified using the CTAB method (Ausubel et al., 1987).
DNA was precipitated with isopropanol at −20°C overnight and then washed with 70% ethanol. After drying, the pellet was re-suspended for the PCR assay in 25 μl of DNase-free water.
The identity of citrus canker bacteria was verified using primers 2 and 3 (Hartung et al., 1993). PCR reactions were conducted at 25 μl reaction volume and PCR programs were conducted according to Cubero and Graham (2002).

| Genotyping
Genotyping was performed on bacterial suspensions prepared for the whole strain collection (Table S1), plus reference strains, using a set of 14 microsatellites (MLVA-14), mostly as reported previously . The MLVA-14 scheme was supplemented with two additional PCR primer pairs, which respectively targeted the copL and copD genes, specific to each of the two copper resistance systems identified so far in X. citri pv. citri (Richard et al., 2017). For this, the primer pairs copL-F (5′ CCGTGTCAGCCTCCTCACTTCTAC 3′)/copL-R (5′ FAM-CAGCGGCATGACATCCAGGCC 3′) and copD-F (5′ CGACACGGATCACCCACGTCA 3′)/copD-R (5′ NED-TCTCCATCCGTCTCGCGCTCT 3′) were included in the multiplex PCR pools 1 and 2, respectively (Behlau et al., 2013;Richard et al., 2017). The controls in each plate were the strains LN002-1 (IAPAR 306 in which the pLH201.1 plasmid hosting the copLAB system was introduced by conjugation) and LN173-1 (IAPAR 306 in which the pLM199.1 plasmid hosting the copABCD system was introduced by conjugation). The number of tandem repeats (TR) at each locus was computed from fragment length for all assayed strains, using GeneMapper 4.0 (Applied Biosystems, Courtaboeuf, France) and was used as input data. Genetic diversity indices for MLVA-14 data were calculated using the poppr 2.8.3 package (Kamvar et al., 2014) in R version 3.6.1 (https://www.R-proje ct.org/). Allelic richness (A) was computed using the rarefaction procedure for unequal sample sizes and the hierfstat 0.04-22 R package (Goudet, 2005). Private alleles were identified using poppr 2.9.3 (Kamvar et al., 2014). The algorithm recommended for MLVA data was used to produce minimum spanning networks, combining global optimal eBURST and Euclidean distances in PHYLOViZ v1.2 (Francisco et al., 2012). The population structure of X. citri pv. citri was then subject to the discriminant analysis of principal components (DAPC) using the adegenet 2.1.1. R package (Jombart, 2008;Jombart et al., 2010). This method is free of any assumption linked to a population genetic model (e.g., Hardy-Weinberg equilibrium or linkage equilibrium), which makes it suitable for analyzing datasets produced from predominantly clonal bacteria. Genetic differentiation (based on R ST ) was computed using the genepop 1.1.3 package in R (Rousset, 2008). The occurrence of spatial structure was estimated using the global.rtest and related functions of the adegenet package.
A subset of strains selected on the basis of microsatellite data was further genotyped using 31 minisatellites (MLVA-31). The aim was to compare them to a worldwide X. citri pv. citri strain collection of (http://www.biopr ed.net/MLVA/) and assign them to genetic lineages, which were delineated for this bacterium as described previously .
Drawing on minisatellite data, strains were assigned to genetic lineages and sub-clusters, as previously reported (Leduc et al., 2015).

| Surveyed areas and strain characterization
CBC was widely detected in most sampled regions in Saudi Arabia and Yemen (Figure 1a and Table S1) with the single exception of Al-Madina region where citrus is grown commercially but no canker symptoms were found. All strains with a cultural morphology typical of Xanthomonas produced amplicons when assayed by PCR with the 2/3 primers (Hartung et al., 1993). This suggests that no strains of X. citri pv. aurantifolii were present in our Saudi and Yemeni strain collection. This finding was further confirmed by phenotypic assays (i.e., strains hydrolyzed gelatin and casein and were able to grow in the presence of 3% NaCl). Moreover, when microsatellite genotyping was performed on the whole strain collection (MLVA-14), amplicons were produced for all strains at all targeted loci. This is typical of X. citri pv. citri, but unlike X. citri pv. aurantifolii (Bui Thi Ngoc, Vernière, Jarne, et al., 2009;Bui Thi Ngoc, Vernière, Vital, et al., 2009). Apart from the positive controls, not a single strain yielded PCR amplification with primers that target copper resistance genes (which are involved in the two known resistance systems in X. citri pv. citri).
This suggests that they are susceptible to copper (a pesticide widely used for controlling CBC). Consistent with these results, none of 36 strains (selected on the basis of their genetic diversity- Table S1) grew on YPGA medium supplemented with copper. All assayed strains yielded a typical growth on the basal medium, thus phenotypically confirming them as copper-susceptible. to situate strains from the present study in the known global diversity of X. citri pv. citri . The DAPC analysis assigned these strains to genetic lineages 1 (Yemen), 2 (Saudi Arabia), and 4 (both countries). All strains assigned to either lineage 1 or 2 (i.e., pathotype A strains) originated from several Citrus Conversely, only lineage 1 and 2 strains produced typical cankerlike lesions on mandarin, whereas lineage 4 strains produced small blister-like lesions, surrounded by a water-soaked area, typical of pathotype A and A* strains, respectively (Vernière et al., 1998). We found no obvious difference in the phenotype observed in response to inoculations between Saudi or Yemeni pathotype A* strains, which were assigned to distinct subclades using minisatellite analysis.

|
In summary, our results suggest that both countries host strains that are genetically related to pathotypes A and A*, but not A w .
However, the latter was previously identified in neighboring Oman.
Strains genetically related to pathotype A were detected in both countries, but assigned to distinct genetic lineages. All strains from Mexican lime were assigned to lineage 4, using minisatellite genotyping. Subclade 4.2 was detected in both countries and was overall prevalent in the analyzed strain collection. Each country hosted a distinct additional subclade (4.1 and 4.4 in Saudi Arabia and Yemen, respectively).  and Yemen (n = 161) was found genetically related to pathotype A strains ( Figure 2). This was clearly observed in the case of a Mexican lime block (JP2, n = 18- Table S1) that was the direct neighbor of a lemon block (JP1) where only lineage 2 strains (pathotype A) were detected.

| Genetic structure based on microsatellite data
We further focused our microsatellite analyses on the prevailing subclade 4.2. No global spatial structure (p = 0.094) was found.
In contrast, a significant local structure (p = 0.001) was revealed, indicating strong genetic differences between strains sampled over short distances (i.e., a same site). All Yemenese strains (sampled from groves in the Al Hudaydah region) grouped in a single DAPC cluster (cluster #2) that contained no Saudi strain, a result that was consistent with the produced minimum spanning tree (Figure 3). Moreover,  (Vernière et al., 1998) are shown as red dots.  (Figure 3). Strains sampled from a same site were often assigned to several DAPC clusters, consistent with the detection of a significant local spatial structure. A significant genetic differentiation (p < 0.05) was observed among 95% (187/197) of remote (separated by ≥3 km) local population pairs. Analysis on local populations showed that 5% (5/98) of pairs of populations located in distinct regions were not differentiated (p > 0.05). The corresponding citrus groves were located in Jizan and Al Bahah regions at distances >300 km (Figure 1b). Consistently, some strains from these two regions grouped in the same DAPC clusters.

| Genotyping data informs on the history and epidemiology of X. citri pv. citri in the Arabian Peninsula
We genotyped strains associated with the pioneer outbreaks of CBC in the Arabian Peninsula. The first outbreak reported in Yemen was based solely on visual observations. Therefore, the pathovar status (pv. citri vs. pv. aurantifolii) of this outbreak was at that time unresolved (Dimitman, 1984). Three strains isolated from Yemen in the 1980s, including one sampled during the very first outbreak reported in the country, were assigned to genetic lineage 1. They were further identified as pathotype A strains, based on pathogenicity assays. Minisatellite genotyping revealed that these strains were closely related. Their nearest relatives were strains from the Southwest Indian Ocean region and India (data not shown), which is consistent with the putative Indian origin of Yemeni strains presumed in the original report (Dimitman, 1984).
All strains from our collection recently isolated from Mexican lime in Yemen were related to lineage 4. More specifically, lineage 4 subclade 4.2 (the prevalent subclade in Saudi Arabia) was identified in the Al Hudaydah region. Lineage 4 subclade 4.4 was identified in all sampled regions and is, therefore, the prevailing subclade in Yemen. Subclade 4.4 was previously reported in the Sistan-Baluchistan region in Iran and sporadically in Mauritius (Pruvost et al., , 2021. However, it should be pointed out that the main areas for sweet orange and mandarin cultivation in Yemen were not accessible for security reasons at the time of sampling. However, it is likely that pathotype A strains still occur in these areas of citrus cultivation, since no drastic eradication measures have been adopted (Dimitman, 1984). Indeed, there is no evidence that the pathogen will disappear given the prevailing disease management practices and there is no similar situation having occurred anywhere else.
TA B L E 1 Genetic diversity parameters for Xanthomonas citri pv. citri strains originating from Saudi Arabia (n = 563) and Yemen (n = 164)  (Nei, 1978). b Allelic richness computed by the rarefaction method.
In Saudi Arabia, strains from the first documented outbreak sampled in the Jizan area were assigned to genetic lineage 4 subclade 4.1 and further identified as pathotype A*, based on pathogenicity assays. The current hypothesis that X. citri pv. citri was introduced in Saudi Arabia from Yemen (Fadlallah, 1986) is clearly not supported by the genetic data. Interestingly, subclade 4.1 was only detected during the recent extensive survey from a single grove in exactly the same area, but not in other locations. This suggests that this subclade is unlikely to be widespread in citrus nurseries and not widely transmitted through the sale of plants for planting, a highly efficient pathway for the long-distance spread of X. citri pv. citri (Leduc et al., 2015;Vernière et al., 2014); and supports that the prevailing environmental conditions in the region are unlikely to lead to long-distance natural spread of the pathogen (Irey et al., 2006). Within groves, the pathogen is primarily spread through wind-driven rain, overhead irrigation, and grove maintenance operations . In the mid-1980s, another outbreak in Saudi Arabia was reported in the Najran region (Fadlallah, 1986). pv. citri has never been documented over such a distance even in association with extreme weather events (Irey et al., 2006). This suggests that the pathogen has likely been disseminated by contaminated citrus propagative material, a pathway that was associated with long-distance spread of the pathogen in several countries (Leduc et al., 2015;Vernière et al., 2014).
Altogether, the presence of both pathotypes A and A* were ascertained in Saudi Arabia and Yemen. In the experimental design used herein, two citrus species, Mexican lime and mandarin, were used for pathogenicity tests. Therefore, it remains unknown whether some strain from the Arabian Peninsula displays a distinctive pathogenicity on other citrus species, a feature that was previously reported from Japan (Shiotani et al., 2007). In the present study, pathotype A strains were found to be clearly different, with lineage 1 and 2 identified from Yemen and Saudi Arabia, respectively, thus representing distinct introductions. We did not identify any pathotype A w in our large strain collection, a group that has been reported from neighboring Oman (Gordon et al., 2015). However, sampling was conducted in two main areas and it cannot be en-

| Lineage 4 strains are highly prevalent on Mexican lime
Many crop pathogenic bacteria display a high host specialization, similar to bacteria that are pathogenic to vertebrates Gilbert et al., 2012;Shaw et al., 2020). Indeed, bacterial pathogens spread naturally in agroecosystems, where its host abundance is high and its diversity is very low (if not null;McDonald & Stukenbrock, 2016;Thrall et al., 2007). Although the causes and consequences of host specialization have been theoretically studied, there is a need from experimental data comparing the performance and epidemiology of pathogenic strains varying in host specialization in natura (Barrett et al., 2009).
Xanthomonas citri pv. citri overall is clearly on the specialist side of the specialist-generalist continuum that describes the pathogens' host range (Barrett et al., 2009;Graham et al., 2004). However, at the infrapathovar level, this pathogen still displays significant differences in host specialization (i.e., the X. citri pv. citri pathotypes). The present study represents the first extensive population analysis of X. citri pv. citri in two countries where both pathotypes A and A* occur. We focused on bacterial strains isolated from Mexican lime, a highly susceptible host species for all known pathotypes causing CBC (Gottwald et al., 2002). One of the main findings in the present study was that specialist strains of a crop bacterial pathogen suggested an ability to outcompete generalist strains in natura. Consistent with theoretical studies (Gandon, 2004;Leggett et al., 2013) and studies conducted on other organisms (Baumler & Fang, 2013; Morris & Moury, 2019), we suggest a fitness cost linked to wide host range pathotype A strains. Generalist bacterial haplotypes with a lower fitness are expected to be counter-selected and may be ultimately removed from the population (Sheppard et al., 2018). The analysis of biological interactions in several ecosystems suggests that hostparasite specialization tends to be promoted by host abundance, which is clearly the case for agroecosystems (Vazquez et al., 2005).
However, several other factors, such as cooperation or competition between bacterial actors through several mechanisms (e.g., bacteriocins, toxins, effectors exported by type 4 and 6 secretion systems…), involving other partners (e.g., bacteriophages, predating amoeba, the microbiota…) or other parameters (e.g., human activities and prevailing environmental conditions) can also modulate how specialist and generalist pathogens can successfully maintain in scenarios of co-infection (Baldeweg et al., 2019;Bayer-Santos et al., 2018;Gottwald et al., 2002;Leggett et al., 2013;Riley & Wertz, 2002;Sieiro et al., 2020;Souza et al., 2015;Vannier et al., 2019;Zhu et al., 2020). Therefore, further experiments under controlled conditions are required to assess virulence and/or transmission ability of specialized and generalist bacterial strains on Mexican lime and to determine the underlying mechanisms associated with potential differences. Indeed, the virulence of some pathogens that are able to infect multiple hosts can be host-dependent (Leggett et al., 2013).
Consistent with the general concept that a precise characterization and surveillance of outbreak populations have clear fallout on disease control, the present study has implications for ACC management in Saudi Arabia (Bull & Koike, 2015;Woolhouse, 2002).
First, the wide genetic diversity revealed in this paper is likely to complexify the intended use of bacteriophages for ACC control in Saudi Arabia (Ibrahim et al., 2017) and should orient its future development. Second, it suggests that resistance to copper compounds, widely used to control ACC worldwide, has not yet developed in the sampled regions. Indeed, our results suggest that copper sprays are still efficient but regular surveillance is required in future.
Copper resistance in X. citri pv. citri was first reported in pathotype A strains from Argentina in the mid-1990s  and more recently in remote territories (Huang et al., 2021;Richard et al., 2016Richard et al., , 2017. Moreover, a recent genomic study revealed its presence in pathotype A* strains from Southeast Asia (Webster et al., 2020). delay the emergence of more invasive strains (Zhan et al., 2015).

ACK N OWLED G M ENTS
This research was funded by the National Plan for Science, Technology, and Innovation (MAARIFAH), King Abdul-Aziz City for Science and Technology, Kingdom of Saudi Arabia, Grant Number (13-BIO 1222-02). We thank Dr. C. Vernière for internal review. This manuscript is dedicated to his memory.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The microsatellite data that support the findings of this study are openly available in the CIRAD dataverse at https://datav erse.cirad.