Distribution and phylogenetics of whiteflies and their endosymbiont relationships after the Mediterranean species invasion in Brazil

The Bemisia tabaci is a polyphagous insect and a successful vector of plant viruses. B. tabaci is a species complex and in Brazil native species from the New World (NW) group, as well as the invasive species, Middle East-Asia Minor 1 (MEAM1) and Mediterranean (MED) were reported. For better understanding the distribution of the different species four years after the Mediterranean species invasion in Brazil, whiteflies were collected from 237 locations throughout the country between the years of 2013 and 2017, species were identified and the facultative endosymbionts detected. The survey revealed that MEAM1 was the prevalent species found on major crops across Brazil. It is the only species present in North, Northwestern and Central Brazil and was associated with virus-infected plants. MED was found in five States from Southeast to South regions, infesting mainly ornamental plants and was not associated with virus-infected plants. The prevalent endosymbionts identified in MEAM1 were Hamiltonella and Rickettsia; and the mtCOI analysis revealed low genetic diversity for MEAM1. In contrast, several different endosymbionts were identified in MED including Hamiltonella, Rickettsia, Wolbachia and Arsenophonus; and two distinct genetic groups were found based on the mtCOI analysis. Monitoring the distribution of the whiteflies species in Brazil is essential for proper management of this pest.


Results
MED species was found in Minas Gerais, São Paulo, Paraná, Santa Catarina and Rio Grande do Sul States ( Fig. 1) colonizing ornamental plants as was recently observed 31 and on tomatoes and sweet pepper plants cultivated in greenhouses, in which ornamentals were previously grown. It was also found on cucurbits and broccoli in open fields near greenhouses where MED was previously detected. Interestingly, MED was found heavily infesting a sweet pepper (greenhouse) in the county of Cerqueira Cesar (São Paulo State) with no apparent connection and kilometers apart from any ornamental crop. The sweet pepper plants were all symptomless and no viruses were detected in ornamental plants associated to MED. Scientific Table 1).
The distribution of endosymbionts in MED was heterogeneous showing an increase in the percentage of Hamiltonella, Rickettsia and Wolbachia from 2015 to 2017. In contrast, the endosymbiont distribution remained homogeneous for MEAM1 species and the percentages of all the endosymbionts identified remained similar between 2015 and 2017 (Fig. 3).
In addition, it was noticed the presence of whiteflies populations that were free of facultative endosymbionts. This was reported in some MEAM1 individuals from sampling sites 21 and 35; some MED individuals from  (Fig. 6). The phylogenetic tree revealed a genetically homogeneous population of MEAM1 across the country, regardless of the host or the production system (greenhouse or open field) (Fig. 4). The MEAM2 clade is not present in the phylogenetic tree due to a recent study suggesting that MEAM2 is a pseudogene artifact and so not a real species 38 . In contrast, MED populations in Brazil seem to be more variable genetically as two distinct genetic groups were observed, Western Mediterranean (Q1) and Middle Eastern (Q2) (Fig. 5). It seems to have no correlation between specific endosymbiont constitution and populations separated in different sub-clades observed for MED.
Populations of T. vaporariorum analyzed presented low diversity among them. However, samples from the South Region of the country (Santa Catarina and Rio Grande do Sul) were grouped in a different sub-clade from samples collected in the Midwest Region (São Paulo and Minas Gerais States) (Fig. 6).
The Hamiltonella phylogenetic tree (Fig. 7A) as well as the Rickettsia (Fig. 7B) grouped different endosymbiont populations from MED and MEAM1 in the same clades, suggesting low genetic diversity of both Hamiltonella and Rickettsia among the whitefly populations analyzed.
The phylogenetic analysis of Wolbachia (Fig. 7C) separated the endosymbionts populations from MED collected in the same State (São Paulo) in two different main clades. In addition, Wolbachia populations from NW and B. tuberculata were grouped in a sub-clade.

Discussion
Herein, we report the current scenario of geographical distribution, phylogenies, hosts, endosymbionts and main viruses associated to whiteflies in Brazil four years after MED introduction 29 and 27 years after MEAM1 introduction 28 . The invasive MEAM1 is still the prevalent species across the country with an established set of endosymbiont, low genetic diversity and is the main species associated to plant viruses. On the other hand, MED was not found in the Northern States and is mainly found in ornamental plants. In addition, MED presents two distinct genetic groups with a highly variable endosymbionts constitution and is not associated to plant viruses yet.
The first report of MED in Brazil has posed many questions regarding the possible impacts on the local agriculture 29 . MED is known to be able to out-compete MEAM1 under conditions of high insecticides use 36,39,40 . The displacement of MEAM1 by MED in major crops such as soybeans and cotton would change the entire pest management system and tactics of control of this pest in Brazil. MED invasion into row crops could increase the cost of production and threaten the whole agricultural system. In this study, our survey revealed this displacement has not occurred yet and MEAM1 is still likely to be the prevalent species in major crops in important agricultural regions of Brazil in the Midwest and Northeast. However, the presence of MED in Minas Gerais, São Paulo, Paraná, Santa Catarina and Rio Grande do Sul States is still threatening Brazilian agriculture and a preventive management must be carried out to avoid the spreading of MED to other regions. Although MED is still strongly associated to ornamental plants or vegetables cultivated in greenhouses, we also found this species colonizing plants in open field close to greenhouses with ornamental plants, which is a different scenario than the one verified two years before 31 . Recent studies in Florida revealed that after twelve years in which MED, was associated with greenhouse-grown ornamental horticulture, it started also to be detected in residential landscapes and open field agricultural production in 2016 41 . In Florida, different factors may explain the detection of B. tabaci MED in the landscape, such as changes in the use of insecticides that are less effective to MED and environmental conditions that may have been favorable for the buildup of B. tabaci in this scenery. MED was also recently detected in open-field bean crop in Argentina, together with MEAM1 and NW 42 . B. tabaci MED is not usually a problem in the landscape, but the tropical weather and presence of cultivated hosts and weeds during the year can contribute for the adaptation of MED to the landscape and dynamic of the insect in these regions needs to be carefully surveyed.
The emergence of viruses in several regions may also be related to presence of a specific species of whitefly. In Brazil, this was the case for begomoviruses in tomatoes after MEAM1 invasion 43   displacement of both MEAM1 and the native species in this country 46,47 . A similar situation has been reported in Mediterranean countries 48 . TYLCV, one of the most devastating viruses in the world, has not yet been reported in Brazil according to previous surveys 49 . However, it has been reported in the neighboring country, Venezuela 50 , being a menace to the Brazilian agriculture. In this study, virus infected plants were mainly associated to MEAM1 and the viruses identified were CpMMV for soybean; ToSRV and ToCV for tomatoes. However, it is known that whiteflies in Brazil are also transmitting the begomovirus (BGMV), causing several economic losses to the bean crop. BGMV was not identified in the sampling sites where this survey was carried out.
Other studies suggest the higher ability of MED to colonize pepper compared to MEAM1 51,52 as well as the predominance of MED over MEAM1 due to the intensive use of insecticides 36 . It's interesting to mention that we found sweet pepper plants heavily infested with MED species, under high pressure of insecticide use, in a greenhouse with no connection with ornamental plants, and it was impossible to presume the origin of MED is this region. In this study, virus symptoms were not observed on sweet pepper plants infested with MED species suggesting that these whiteflies may not be as successful in transmitting viruses as MEAM1.
The Native American B. tabaci species (NW) was found on weeds (Sida spp., Euphorbia heterophylla and Ageratum spp.) and was the prevalent species in soybean in Rio Grande do Sul, the southernmost State of Brazil. It was thought that the indigenous species from the Americas would have been completely displaced by MEAM1 in Brazil. However, they were reported since 2012 colonizing E. heterophylla, Xanthium cavanillesii Schouw (Asteraceae), soybean 12 and Ipomoea sp. (Convolvulaceae) 33 . The NW species was demonstrated to be a good vector of the begomovirus, BGMV and the carlavirus, CpMMV, to beans 53 . In this study, no viruses were associated to plants where NW species were collected. In Argentina, NW was also reported in soybeans and beans 54 , which may indicate a suitable condition for this species to become a concern to the these crops in this region. In Brazil, B. tuberculata has been previously reported infesting cassava 37 . In this study, all whiteflies collected on cassava were identified as B. tuberculata, except for sample ID 36 in which B. tabaci MED adults were found. Even though the report of a B. tabaci species on cassava in Brazil is unusual, it's not possible to prove whether MED was actually colonizing and completing the life cycle on this cassava plant since only adults were analyzed and they could be only feeding on this cassava plant. A different situation is observed in African countries where different In this study, T. vaporariorum was detected mainly in greenhouses. T. vaporariorum is a vector of the crinivirus Tomato chlorosis virus, ToCV 56 and Tomato infectious chlorosis virus, TiCV 57 . The later has not yet been reported in Brazil. ToCV is also transmitted by B. tabaci and is becoming a serious threat for tomato production in Central Brazil 58 . In this survey, ToCV infected tomato plants were found in a greenhouse in Rio Grande do Sul State highly infested by T. vaporariorum, reinforcing the importance of this species as a vector.
The phylogenetic analysis of the mtCOI gene suggested very low genetic variability among MEAM1 populations. MEAM1 individuals collected in Rio Grande do Sul State (Southernmost region) and Pará State (Northernmost Region) were in the same branch in the Phylogenetic Tree (Fig. 3), indicating that the population of MEAM1 in Brazil is very homogeneous across the country and all the individuals are very close genetically. This could also be verified by testing the endosymbiont composition which was very homogeneous for MEAM1 populations as 89,5% of the specimens analyzed presented a co-infection of Hamiltonella and Rickettsia (Fig. 2). In contrast, mtCOI analysis of MED revealed a highly variable population, as individuals collected in the same State (São Paulo State) were placed in distinct genetic groups in the Phylogenetic Tree (Fig. 5). The facultative endosymbionts found in MED also revealed high variability (Fig. 2). Many different facultative endosymbionts were found infecting MED populations. However, most of the specimens analyzed (31%) were infected only by Hamiltonella (Fig. 2). It was possible to verify a co-relation between sampling sites and endosymbionts for MED (Supplementary Table 1  The presence of whitefly individuals free of facultative endosymbionts occurred mainly in MED populations in this survey. The absence of facultative endosymbionts in populations must be reported and may have biological implications on the whitefly vector such as better virus acquisition and retention, and a faster development time 59 . Endosymbionts analyses of MED over the years revealed an increasing percentage of Hamiltonella, Rickettsia and Wolbachia from 2015 to 2017 (Fig. 3). This potentially indicates different introductions of MED species in Brazil. It is already known that secondary endosymbiont composition differs among different MED groups 60,61 . No correlation between facultative endosymbionts and different genetic groups of MED (Western Mediterranean and Middle Eastern) was observed in this study.
The high incidences of Hamiltonella in Brazilian populations of B. tabaci may have serious implications to virus transmission in agriculture. Hamiltonella has been identified as a key factor in the transmission of the begomovirus, Tomato yellow leaf curl virus (TYLCV) encoding a GroEL chaperonin homologue protein that safeguards begomoviruses in the haemolymph 62 . The effects of Hamiltonella in the transmission of plant viruses found in Brazil are still unknown.
The phylogenetic analyses of the endosymbionts suggest that Hamiltonella and Rickettsia from MED and MEAM1 are genetically close, which might be a result of horizontal transmission of these endosymbionts between different whitefly species. Samples from different States were placed in the same clades in the phylogenetic tree (Fig. 7A,B).
The analysis of Wolbachia has placed the endosymbionts present in MED species in different clades (Fig. 7C). Horizontal transmission of endosymbionts between whitefly populations can be influential to the agriculture system. The acquisition of an endosymbiont that was not present in the population may rapidly change the biology of the insect. Previous studies reported that Rickettsia-infected whiteflies produced more offspring, had higher survival to adulthood, developed faster, and produced a higher proportion of daughters compared with uninfected whiteflies 63 .
In summary, this is the most comprehensive study on surveying and identifying whiteflies species in Brazil, which showed the distribution, the phylogenetic relationships and the coinfection of endosymbionts from the main species found in the country. This kind of survey is essential for the determination of the suitable management strategies of these pests. This study opens new doors and raises questions that must be further clarified over MED behavior in Brazil. Biological aspects must be verified, such as the insecticide resistance under the Brazilian conditions, the competition with the MEAM1 species in different plant hosts, and its ability to transmit local begomoviruses and criniviruses.  Table 1). B. tabaci adults were homogenized in 60 µl of 5% Chelex solution in a 1.5 ml tube. The tube was vortexed for few seconds, and then incubated at 56 °C for 15 min and at 99 °C for 8 min. After centrifugation at 13000 rpm for 5 min, the supernatant was then collected and used as a template for the PCR amplification. Primers sequence and annealing temperatures of PCR reactions used for whiteflies identification are available on Supplementary  Table 3.

Methods
All DNA samples were first subjected to PCR analysis to differentiate MEAM1 from MED using the primers pair Bem23F and Bem23R, which amplifies a microsatellite locus of about 200 bp and 400 bp for MEAM1 and MED, respectively 34,65,66 . Later, samples that did not amplify were screened with the generic insect primers C1-J-2195 and TL2-N-3014 that amplify a fragment of the mtCOI 67 followed by Restriction fragment length polymorphism (RFLP) technique 12,68 to identify NW. RFLP analysis of the amplicons consisted of 5 µl of each PCR (880 bp) digested with one unit of TaqI at 65 °C for 2 hours in a final volume of 15 µl. The restricted DNA was visualized by electrophoresis in 2% agarose gel stained with ethidium bromide. Samples that still did not amplify or had unexpected RFLP pattern were analyzed with specific primers for T. vaporariorum, TvapF and Wfrev 69 .
A representative number (121 out of 237) of PCR products amplified from mtCOI of the whiteflies were purified (QIAquick Gel Extraction Kit Qiagen) and sequenced (Macrogen, South Korea) in both directions using C1 -J-2195/TL2-N-3014 for B. tabaci and B. tuberculata or TvapF/Wfrev for T. vaporariorum. The nucleotides sequences were analyzed and compared with those present in GenBank database using BLAST tools (http://blast. ncbi.nlm.nih.gov/Blast). Nucleotide sequences from mtCOI were deposited in GenBank and accession numbers are available in Supplementary Table 1 75 and Fritschea 76 that were reported from whiteflies, using genus-specific primers targeting the 16S or 23S rDNA genes. PCR cycling was performed as described by 37 . The endosymbiont presence confirmation was performed by sequencing the amplified sequences from representative individuals. Primers sequence and annealing temperatures of PCR reactions used for endosymbiont screening are available on Supplementary Table 3. PCR products of the facultative endosymbionts were purified (QIAquick Gel Extraction Kit Qiagen) and sequenced (Macrogen, South Korea). The nucleotides sequences were analyzed and compared with those present in GenBank database using BLAST tools (http://blast.ncbi.nlm.nih.gov/Blast). Nucleotide sequences were deposited on GenBank and accession numbers are available in Fig. 7.
Phylogenetic analysis. Phylogenetic analyses were carried out using the 121 B. tabaci mtCOI sequences obtained in this study added to the new global B. tabaci mtCOI dataset 13 downloaded from GenBank on November 1, 2017 and includes 1040 sequences. The mtCOI, sequences ranged from 562 to 626 bp in length. In addition, phylogenetic analyses were carried out for the facultative endosymbionts Hamiltonella, Rickettsia, Wolbachia and Arsenophonus. Partial 16S rDNA gene was analyzed for Hamiltonella, Rickettsia and Wolbachia. For Arsenophonus, partial 23S rDNA gene was analyzed. The alignment consisted of sequences of 628 bp, 743 bp, 569 bp and 606 bp in length for Hamiltonella, Rickettsia, Wolbachia and Arsenophonus respectively. Multiple sequence alignment was prepared using MAFFT 77 within the Geneious 9.1.5 software.
Subsequently, Bayesian analyses were conducted using Mr Bayes v. 3.2.2 78 and were run in parallel across 384 nodes on the Magnus supercomputer (located at the Pawsey Centre, Western Australia). Analyses were run for 30 million generations with sampling every 1000 generations. Each analysis consisted of four independent runs, each utilizing four coupled Markov chains. The run convergence was monitored by finding the plateau in the likelihood scores (standard deviation of split frequencies <0.0015). The first 25% of each run was discarded as burn-in for the estimation of a majority rule consensus topology and posterior probability for each node. Trees were visualized, edited and rooted using FigTree v1.4.2.

Whitefly-transmitted virus detection. Field plants colonized by whiteflies showing typical viral disease
symptoms such as mosaic leaf pattern, crinkled leaves, yellowed leaves and plant stunting were collected for further molecular analysis. Nucleic acid extractions were carried out using the Viral RNA/DNA Mini Kit by Invitrogen. Subsequently, different PCR's were carried out for the detection of the main viral diseases for each crop in Brazil. For soybean, the detection of Cowpea mild mottle virus (CpMMV) 53 was carried out, tomato plants were analyzed for the presence of Tomato severe rugose virus (ToSRV) 79 and Tomato chlorosis virus (ToCV) 80 . Ornamental plants were analyzed for the presence of Torradovirus 81 . Primers details and PCR conditions are available in Supplementary Table 3.