Prevalence and genetic diversity of endosymbiotic bacteria infecting cassava whiteflies in Africa

The whitefly, Bemisia tabaci (Hemiptera: Aleyrodidae) has gained importance as one of the most important agricultural pests owing to its wide geographic spread, large host range of over 500 hosts, and most significantly as a vector of over 100 different plant viruses in the tropical and subtropical regions of the world [1,2]. B. tabaci is a cryptic species complex comprising at least 24 morphologically indistinguishable species [3] with a proposed origin in sub-Saharan Africa (SSA) and with high variability in mitochondrial cytochrome oxidase I (mtCOI) nucleotide sequences amongst major geographical clades [2,4]. Cassava, a key food security crop throughout SSA, suffers devastating yield losses due to B. tabaci-borne cassava mosaic begomoviruses (CMBs) and cassava brown streak viruses (CBSVs). These cause cassava mosaic disease (CMD) and cassava brown streak disease (CBSD), respectively [5,6]. Five genetically distinct groups of B. tabaci, named Sub-Saharan Africa 1 to 5 (SSA 1–5) colonise cassava in SSA. These have been generally referred to as cassava whiteflies in this and other studies. SSA1 occurs throughout the SSA, SSA2 in East and West Africa, SSA3 in Cameroon and Togo, SSA4 only in Cameroon and SSA5 in South Africa [6]. Based on mtCOI sequence divergences, SSA1 was further divided into four subgroups; SSA1- subgroup 1 (SSA1-SG1), SSA1-SG2, SSA1-SG3 and SSA1-SG4 [6].

Superabundant B. tabaci populations, commonly numbering more than 1000 adults per top five leaves, of cassava plants, have been associated with the rapid spread of CMD pandemic in East and Central Africa since the late 1990s. SSA2, the then super abundant population which was also described as the ‘invader or UG2’, was associated with the spread of the CMD pandemic [7,8]. In recent years, a shift in the B. tabaci population has occurred with the relative frequency of SSA1-SG1 increasing from 24.6% to 89.2%, while the frequencies of SSA2 and SSA1-SG2 decreasing significantly from 63.9% to 1.4%, and 11.5% to 1.4%, respectively, between 1997 and 2010 [6]. The reasons for this natural shift in cassava population remain unknown. Similar shift in genetic diversity of populations are reported for B. tabaci species, mainly with respect to the population replacement by the invasive Middle East-Asia Minor 1 (MEAM1, previously B-biotype) and Mediterranean (MED, previously Q-biotype) populations [9-11].

In addition to the primary endosymbiont Portiera aleyrodidarum, the species B. tabaci has been reported to harbour six vertically transmitted secondary endosymbionts, Arsenophonus, Wolbachia, Hamiltonella, Cardinium, Fritschea and Rickettsia [12-14]. Recently, a new bacterium named Candidatus hemipteriphilus asiaticus was also found to infect B. tabaci from China [15]. Several of these endosymbionts can affect the biology and behaviour of B. tabaci. Wolbachia and Cardinium in particular are known to induce cytoplasmic incompatibility (CI), a process in which the host reproduction is manipulated to allow rapid spread of bacteria through insect populations [16]. Whether such phenotypes are induced by these bacteria in B. tabaci remains unknown. Rickettsia, when infecting MEAM1, provided fitness benefits by increased fecundity and survival [17], increased heat stress tolerance [18], defence against pathogens [19] but occasionally also increased the susceptibility to insecticides [20]. Fritschea has reported negative impact with reduced fecundity and narrowing the host range of infected New World species of whiteflies [21]. Endosymbionts can also alter the vector ability of B. tabaci. Hamiltonella in MEAM1 and Arsenophonus in Asia II populations facilitated virus transmission by releasing a bacterial chaperonin GroEL that binds and protects virus particles during their transit through the insect body [22,23]. Hamiltonella in the MED and Rickettsia in MEAM1 populations are also reported to increase acquisition, retention and transmission of Tomato yellow leaf curl virus [24,25]. The study of intracellular bacterial communities in these whiteflies and their impact on the host was essential for understanding the dynamics of insect populations and their vector abilities. In this study, we identified the endosymbionts infecting cassava whiteflies, determined their infection frequencies in different populations and characterised the diverse bacterial species by sequencing. We have also developed a cost effective and reliable restriction fragment length polymorphism (RFLP) diagnostic method for the molecular typing of the cassava whitefly populations.

The main aim of this study was to determine the prevalence and genetic diversity of secondary endosymbionts infecting cassava whiteflies in SSA. Whiteflies harbour multiple bacterial symbionts that play essential roles on insect biology, evolution and virus transmission. Understanding cassava whitefly diversity and the bacterial communities co-existing, within the cassava ecosystem is essential to understand the near extinction of some cassava populations in recent years, or the development of superabundant populations and the resultant epidemics of CMD and CBSD in Eastern and Central African countries in recent years [8,26,27].

At first, the genetic diversity of cassava whiteflies from Uganda, Tanzania, Malawi and Nigeria was studied by mtCOI sequence. This was done to establish the correlation between the prevalence of symbionts in different whitefly populations. Cassava in these countries was colonised by five genetically different whitefly populations. SSA1 and its various sub-groups was predominant in the countries sampled, the only other group present was SSA3 in Nigeria, while SSA2 was not detected. Only SSA1-SG3 was found in coastal Tanzania, while Malawi had high proportions of SSA1-SG3 (89.1%) than Uganda SSA1-SG1 (69.4%) (Figure 2). Based on mtCOI phylogeny, a new population was found in Nigeria, which we referred to as SSA1-SG5 (Figure 3). SSA1-SG5 was predominant (60.3%) in Nigeria, followed by SSA3 (35.3%) and a very few individuals of SSA1-SG1 (4.4%). Overall, these results are concurrent with the previous studies that have also shown high levels of genetic diversity amongst the cassava whitefly populations in SSA [6,27-29].

As seen above and in previous studies, mtCOI is shown to be a reliable marker for separating whitefly species and sub-populations. However, using this as a marker requires sequencing and thus incurs high costs and time. In addition, the threat of the two cassava virus disease pandemics spread by the superabundant B. tabaci populations requires simpler monitoring system for effective disease management. We therefore developed a robust RFLP method for typing cassava whiteflies relatively quickly. Using the two-step method and three restriction enzymes described in this study, we were able to reliably assign whiteflies to phylogenetic groups and subgroups found in this study, and thus saving costs as well as time.

Typing the various bacteria infecting these whiteflies, however, proved far more challenging as some of the methods and primers described in the literature did not work initially on cassava whitefly endosymbionts. This was probably because of the high genetic diversity seen in both cassava whiteflies and the various bacteria that infected them. New primers were therefore developed where necessary and the DNA extraction methods and PCR conditions were optimised. Diagnosis of various bacteria confidently was a pre-requisite to understand the genetic diversity of bacteria infecting cassava whiteflies.

Using the above methods, genetically diverse bacteria were found to infect cassava whiteflies in SSA. Rickettsia, Arsenophonus, Wolbachia and Cardinium were detected in cassava whiteflies, but not Hamiltonella and Fritschea. Hamiltonella was also absent in other native whitefly populations in India and China [30,31], but was reported to be present in SSA1 cassava whiteflies from Tanzania [32]. This is contrasting to our study, and we cannot clearly explain the differences between the two studies at this time. Some of the possible explanations, however, include high site to site variation seen in endosymbiont profiles of cassava whiteflies within a country (Tajebe L., pers comm), and that our samples may have been collected coincidentally from Hamiltonella-free sites. Other reasons include the low titre of the bacteria in our samples which was beyond the limits of PCR detection, or primer mismatch in PCR reactions. We did obtain unspecific amplification of Arsenophonus from Hamiltonella-specific primers in initial studies, which indicated primer mismatch. The Hamiltonella-specific primers, therefore, should be used with care in future studies, while the Hamiltonella quandary between Tajebe et al. [32] and this study remains to be resolved. We used MLST to characterise Wolbachia. All five MLST alleles were amplified from all our populations except only coxA was amplified from SSA1-SG5 and none from SSA3 after exhaustive efforts. Difficulties in amplification of MLST alleles have been reported previously, and could be due to high variability of these genes or low titres of the symbiont [33,34]. The surface protein wsp was therefore used as an alternative marker and this marker also confirmed the high diversity of Wolbachia infecting cassava whiteflies.

Overall, about 77.3% of cassava whiteflies were infected with at least one secondary symbiont, while the remaining 22.7% were completely free of the tested bacteria. These results were similar to the incidences of secondary symbionts seen in other B. tabaci, which ranged from 78% to 100% [14,30,35,36]. A high percentage of the superabundant SSA1-SG1 from the CMD pandemic areas [32] and SSA1-SG5 whiteflies [37] were also reported to be free of secondary symbionts. Further studies comparing the fecundity and life cycle of bacteria-infected and uninfected cassava whiteflies is essential to understand the reasons behind the development of superabundant whiteflies, and the supposed interactions between symbionts and cassava whiteflies.

Single infections of bacteria were more prevalent (59% of total infections) in cassava whiteflies than double (37%) and triple (4%) infections. This was slightly contrary to other studies in which co-infections were more common (>60%) than single infections [14,36,32]. The reasons or the implications of this is unknown but could be due to competition for space and resources among the symbionts [38] or the tolerance of the host to harbour many bacterial communities [35]. Although this is yet to be investigated thoroughly for cassava whiteflies, but specific interactions between bacterial strains and whitefly populations was clearly evident. For example, SSA1-SG1 and SSA1-SG2 were both infected with similar strains of Wolbachia, which were similar to those bacteria infecting butterflies and mosquitoes, whereas SSA1-SG3 was infected with a different Wolbachia. Infection levels of Rickettsia were highest in SSA1-SG3 (54%), which was also similar to the invasive Rickettsia sp. nr Bellii strain that invaded the whitefly population MEAM1 in the USA with fitness benefits to the infected host [17]. However, infection with the same strain of Rickettsia in MEAM1 populations from Israel had no selective advantage to the host [39] and this further indicates specific interaction between symbiont and host genotype or the environment. When and how the Rickettsia invaded cassava whiteflies is unknown, but it remains to be seen if they also provide fitness benefits or not on cassava plants. Another puzzle in the jigsaw of whitefly-bacterial interactions was the detection of three different strains of Arsenophonus in cassava whiteflies. Strain A3 in particular was highly divergent, 7% nucleotide differences, compared to other Arsenophonus infecting B. tabaci across the world. A3 is closely related to the male killing Arsenophonus nasoniae [40], which again might influence the population dynamics and remains the focus of our future investigations. In summary, our findings provide insights to the diverse bacterial species infecting cassava whiteflies in African countries, and that these should be considered in future studies aiming to better understand the changing population dynamics in African cassava fields.

Genetically diverse bacteria infect cassava whiteflies in Africa and their prevalence varied across the different whitefly populations and geographies. Optimising the diagnostic protocols and the characterisation of endosymbionts infecting cassava whiteflies will be highly useful for future investigations on the role of the bacteria on whitefly biology, population development and virus transmission.