The first transcriptomes from field-collected individual whiteflies (Bemisia tabaci, Hemiptera: Aleyrodidae):  a case study of the endosymbiont composition

Endosymbionts and their role in B. tabaci

Viral-vector interactions within B. tabaci are further influenced by bacterial endosymbionts forming a tripartite interaction. B. tabaci has one of the highest numbers of endosymbiont bacterial infections with eight different vertically transmitted bacteria reported^14–17. They are classified into two categories; primary (P) and secondary (S) endosymbionts, many of which are in specialised cells called bacteriocytes, while a few are also found scattered throughout the whitefly body. A single obligate P-endosymbiont P. aleyrodidarum is systematically found in all B. tabaci individuals. Portiera has a long co-evolutionary history with all members of the Aleyrodinae subfamily¹⁶. In this study, we further explore genes within the P. aleyrodidarum retrieved from individual whitefly transcriptomes, including the transcription termination/antitermination protein NusG. NusG is a highly conserved protein regulator that suppresses RNA polymerase, pausing and increasing the elongation rate^18,19. However, its importance within gene regulation is species specific; in Staphylococcus aureus it is dispensable^20,21.

The S-endosymbionts are not systematically associated with hosts, and their contribution is not essential to the survival and reproduction. Seven facultative S-endosymbionts, Wolbachia, Cardinium, Rickettsia, Arsenophonus, Hamiltonella defensa and Fritschea bemisae and Orientia-like organism have been detected in various B. tabaci populations^14,22–25. The presence of S-endosymbionts can influence key biological parameters of the host. Hamiltonella and Rickettsia facilitate plant virus transmission with increased acquisition and retention by whiteflies²⁵. This is done by protection and safe transit of virions in the haemolymph of insects through chaperonins (GroEL) and protein complexes that aid in protein folding and repair mechanisms²⁰.

Application of next generation sequencing in pest management of B. tabaci

The advent of next generation sequencing (NGS) and specifically transcriptome sequencing has allowed the unmasking of this tripartite relationship of vector-viral-microbiota within insects^24,26–28. Furthermore, NGS provides an opportunity to better understand the co-evolution of B. tabaci and its bacterial endosymbionts²⁶. The endosymbionts have been implicated in influencing species complex formation in B. tabaci through conducting sweeps on the mitochondrial genome²⁷. Applying transcriptome sequencing is essential to reveal the endosymbionts and their effects on the mitogenome of B. tabaci, and predict potential hot spots for changes that are endosymbiont induced.

Several studies have explored the interaction between whitefly and endosymbionts^29,30 and have resulted in the identification of candidate genes that maintain the relationship^31,32. This has been explored as a source of potential RNAi pesticide control targets^32–34. RNAi-based pest control measures also provide opportunities to identify species-specific genes for target gene sequences for knock-down. However, to date all transcriptome sequencing has involved pooled samples, obtained through rearing several generations of isolines of a single species to ensure high quantities of RNA for subsequent sequencing. This remains a major bottle neck in particular within arthropoda, where collected samples are limited due to small morphological sizes³². In addition, the development of isolines is time consuming and often has colonies dying off mainly due to inbreeding depression³⁴.

It is against this background that we sought to develop a method for single whitefly transcriptomes to understand the virus diversity within different whitefly species. We did not detect viral reads, probably an indication that the sampled whitefly was not carrying any viruses, but as proof of concept of the method, we validated the utility of the data generated by retrieving the microbiota P-endosymbionts and S-endosymbionts that have previously been characterised within B. tabaci. In this study we report the endosymbionts present within field-collected individual African whiteflies, as well as characterisation and evolution of the NusG genes present within the P-endosymbionts.

Whitefly sample collection and study design

In this study, we sampled whiteflies in Uganda and Tanzania from cassava (Manihot esculenta) fields. In Uganda, fresh adult whiteflies were collected from cassava fields at the National Crops Resources Research Institute (NaCRRI), Namulonge, Wakiso district, which is located in central Uganda at 32°36’E and 0°31’N, and 1134 meters above sea level. The whiteflies obtained from Tanzania were collected on cassava in a countrywide survey conducted in 2013. The samples: WF2 (Uganda) and WF1, WF2a, and WF2b (Tanzania) used in this study were collected on CBSD-symptomatic cassava plants. In all the cases, the whitefly samples were kept in 70% ethanol in Eppendorf tubes until laboratory analysis. The whitefly samples were used for a two-fold function; firstly, to optimise a single whitefly RNA extraction protocol and secondly, to unmask RNA viruses and endosymbionts within B. tabaci as a proof of concept. In addition, we obtained a NusG sequence from a Brazilian NW2 isolate (De Marchi, unpublished) and other downloaded and published NusG sequences from GeneBank) to ensure phylogenetic representation across whitely species.

Extraction of total RNA from single whitefly

RNA extraction was carried out using the ARCTURUS^® PicoPure^® kit (Arcturus, CA, USA), which is designed for fixed paraffin-embedded (FFPE) tissue samples. Briefly, 30 µl of extraction buffer were added to an RNase-free micro centrifuge tube containing a single whitefly and ground using a sterile plastic pestle. To the cell extract an equal volume of 70% ethanol was added. To bind the RNA onto the column, the RNA purification columns were spun for two minutes at 100 x g and immediately followed by centrifugation at 16,000 x g for 30 seconds. The purification columns were then subjected to two washing steps using wash buffer 1 and 2 (ethyl alcohol). The purification column was transferred to a fresh RNase-free 0.5 ml micro centrifuge tube, with 30 µl of RNAse-free water added to elute the RNA. The column was incubated at room temperature for five minutes, and subsequently spun for one minute at 1,000 x g, followed by 16,000 x g for one minute. The eluted RNA was returned into the column and re-extracted to increase the concentration. Extracted RNA was treated with DNase using the TURBO DNA free kit, as described by the manufacturer (Ambion, Life Technologies, CA, USA). Concentration of RNA was done in a vacuum centrifuge (Eppendorf, Germany) at room temperature for 1 hour, the pellet was suspended in 15 µl of RNase-free water and stored at -80°C awaiting analysis. RNA was quantified, and the quality and integrity assessed using the 2100 Bioanalyzer (Agilent Technologies, CA, USA). Dilutions of up to x10 were made for each sample prior to analysis in the bioanalyzer.

cDNA and Illumina library preparation

Total RNA from each individual whitefly sample was used for cDNA library preparation using the Illumina TruSeq Stranded Total RNA Preparation kit as described by the manufacturer (Illumina, CA, USA). Subsequently, sequencing was carried out using the HiSeq2000 (Illumina) on the rapid run mode generating 2 x 50 bp paired-end reads. Base calling, quality assessment and image analysis were conducted using the HiSeq control software v1.4.8 and Real Time Analysis v1.18.61 at the Australian Genome Research Facility (Perth, Australia).

Analysis of NGS data using the supercomputer

Assembly of RNA transcripts: Raw RNA-Seq reads were trimmed using Trimmomatic. The trimmed reads were used for de novo assembly using Trinity³⁵ with the following parameters: time -p srun --export=all -n 1 -c ${NUM_THREADS} Trinity --seqType fq --max_memory 30G --left 2_1.fastq --right 2_2.fastq --SS_lib_type RF --CPU ${NUM_THREADS} --trimmomatic --cleanup --min_contig_length 1000 -output _trinity min_glue = 1, V = 10, edge-thr = 0.05, min_kmer_cov = 2, path_reinforcement_distance = 150, and group pairs distance = 500.

BLAST analysis of transcripts and annotation: BLAST searches of the transcripts under study were carried out on the NCBI non-redundant nucleotide database using the default cut-off on the Magnus Supercomputer at the Pawsey Supercomputer Centre Western Australia. Transcripts identical to known bacterial endosymbionts were identified and the number of genes from each identified endosymbiont bacteria determined.

Phylogenetic analysis of whitefly mitochondrial cytochrome oxidase I (COI): The phylogenetic relationship of mitochondrial cytochrome oxidase I (mtCOI) of the whitefly samples in this study were inferred using a Bayesian phylogenetic method implemented in MrBayes (version 3.2.2)³⁶. The optimal substitution model was selected using Akaike Information Criteria (AIC) implemented in the Jmodel test³⁷.

Sequence alignment and phylogenetic analysis of NusG gene in P. aleyrodidarum across B. tabaci species: Sequence alignment of the NusG gene from the P-endosymbiont P. aleyrodidarum from the SSA1 B. tabaci in this study was compared with another B. tabaci species, Trialeurodes vaporariorum and Alerodicus dispersus using MAFFT (version 7.017)³⁸. The Jmodel version 2³⁷ was used to search for phylogenetic models with the Akaike information criterion selecting the optimal that was to be implemented in MrBayes 3.2.2. MrBayes run was carried out using the command: “lset nst=6 rates=gamma” for 50 million generations, with trees sampled every 1000 generations. In each of the runs, the first 25% (2,500) trees were discarded as burn in.

Analysis and modelling the structure of the NusG gene

The structures for Portiera aleyrodidarum BT and B. tabaci SSA1 whitefly were predicted using Phyre2³⁹ with 100% confidence and compared to known structures of NusG from other bacterial species. All models were prepared using Pymol (The PyMOL Molecular Graphics System, Version 1.5.0.4).

RNA extraction and NGS optimised for individual B. tabaci samples

In this study, we sampled four individual adult B. tabaci from cassava fields in Uganda (WF2) and Tanzania (WF1, WF2a, WF2b). Total RNA from single whitefly yielded high quality RNA with concentrations ranging from 69 ng to 244 ng that were used for library preparation and subsequent sequencing with Illumina Hiseq 2000 on a rapid run mode. The number of raw reads generated from each single whitefly ranged between 39,343,141 and 42,928,131 (Table 1). After trimming, the reads were assembled using Trinity resulting in 65,550 to 162,487 transcripts across the four SSA1 B. tabaci individuals (Table 1).

Comparison of endosymbionts within the SSA1 B. tabaci samples

Comparison of the diversity of bacterial endosymbionts across individual whitefly transcripts was conducted with BLASTn searches on the non-redundant nucleotide database and by identifying the number of genes from each bacterial endosymbiont (Supplementary Table 1). We identified five main endosymbionts including: P. aleyrodidarum, the primary endosymbionts and four secondary endosymbionts: Arsenophonus, Wolbachia, Rickettsia sp, and Cardinium spp (Table 2). P. aleyrodidarum predominate across all four SSA1 B. tabaci study samples based on number of core gene families identified for WF1 (74.82%), WF2 (72.18%), WF2a (53.17%) and WF2b (71.70%). This was followed by Arsenophonus, Wolbachia, Rickettsia sp, and Cardinium spp, which occurred at an average of 18.0%, 5.9%, 1.6% and <1%, respectively across all four study samples (Table 2). The secondary endosymbiont Hamiltonella had the least number of core genes (n=1) and was detected in only one of the SSA1 B. tabaci (WF2a) samples.

Phylogenetic analysis of single whitefly mitochondrial cytochrome oxidase I (COI)

B. tabaci is recognized as a species complex of 34 species based on the mitochondrion cytochrome oxidase I^1,38,39. We therefore used cytochrome oxidase I (COI) transcripts of the four individual whitefly to ascertain B. tabaci species status and their phylogenetic relation using reference B. tabaci COI GenBank sequences found at http://www.whiteflybase.org. All four COI sequences clustered within Sub-Saharan Africa 1 clade (SSA1) species with greater than 99% identity (www.whiteflybase.org).

Sequence alignment and Bayesian phylogenetic analysis of NusG gene

Nucleotide and amino acid sequence alignments of the NusG in P. aleyrodidarum were conducted for several whitefly species including: B. tabaci (SSA1), Mediterranean (MED) and Middle East Asia Minor 1 (MEAM1), New World 2 (NW2), T. vaporariorum (Greenhouse whitefly) and Aleurodicus dispersus. The alignment identified 11 missing amino acids in the NusG sequences for the SSA1 B. tabaci samples: WF2 and WF2b, T. vaporariorum (Greenhouse whitefly) and Aleurodicus dispersus. However, all 11 amino acids were present in samples WF1 and WF2a, MED, MEAM1 and NW2 (Figure 1). Bayesian phylogenetic relationships of the NusG sequences of P. aleyrodidarum for the different whitefly species clustered all four SSA1 B. tabaci (WF1, WF2, WF2a and WF2b) within a single clade together with ancestral B. tabaci from GenBank (Figure 2). The SSA1 clade was supported by posterior probabilities of 1 with T. vaporariorum and Aleurodicus, which formed clades at the base of the phylogenetic tree (Figure 2).

Figure 1. Sequence alignment of nucleotide sequences of NusG gene in P. aleyrodidarum across whitefly species sequences using MAFFT v 7.017.

Figure 2. Bayesian phylogenetic tree of NusG gene of P. aleyrodidarum across whitefly species using MrBayes -3.2.2.

Structure analysis of Portiera NusG genes

Structures of the NusG protein sequence of the primary endosymbiont P. aleyrodidarum in the four SSA1 B. tabaci samples were predicated using Phyre2 with 100% confidence, and compared to known structures of NusG from other bacterial species including (Shigella flexneri, Thermus thermophilus and Aquifex aeolicus; (PDB entries 2KO6, 1NZ8 and 1M1H, respectively) and Spt4/5 from yeast (Saccharomyces cerevisiae; PDB entry 2EXU)^18,40,41. The 11-residue deletion was found in a loop region that is variable in length and structure across bacterial species, but is absent from archaeal and eukaryotic species (Figure 3 and Figure 4A). The effect of the deletion appears to shorten the loop in NusG from the African whiteflies (WF2 and WF2b). A model of bacterial RNA polymerase (orange surface representation; PDB entry 2O5I) bound to the N-terminal domain of the Thermus thermophilus NusG shows that the loop region is not involved in the interaction between NusG and RNA polymerase (Figure 4B).

Figure 3. Primary endosymbiont Portiera aleyrodidarum whole genome from GenBank CP003868 showing the section of the NusG gene included in the analyses (position 76,922).

Figure 4. Structure analysis of NusG from P. aleyrodidarum in B. tabaci and other endosymbionts.

A. Phyre2 based structure prediction of NusG from Candidatus Portiera aleyrodidarum in B. tabaci SSAI whitefly and comparisons to the structures of NusG from other bacterial species as indicated and of Spt4/5 from yeast. NusG is coloured in grey, the loop region in magenta and the 11-residue deletion is shown in green in the C. Portiera aleyrodidarum structure. B. A model of bacterial RNA polymerase (orange surface representation) bound to the N-terminal domain of the T. thermophiles NusG (grey cartoon representation).

NusG deletion and implications within P. aleyrodidarum in SSA B. tabaci

We report the presence of the primary endosymbionts P. aleyrodidarum and several secondary endosymbionts within SSA1 transcriptome. Furthermore, P. aleyrodidarum in SSA1 B. tabaci was observed to have a deletion of 11 amino acids on the NusG gene that is associated with cellular transcriptional processes within another bacteria species. On the other hand, P. aleyrodidarum from NW2, MED and SSA1 (WF2a, WF1) B. tabaci species did not have this deletion (Figure 1). The deleted 11 amino acids were identified in a loop region of the N-terminal domain of NusG protein, resulting in a shortened loop in the SSA1 WF2b sample. This loop region has high variability in both structure and length across bacterial species, and is absent from archaea and eukaryotic species.

NusG is highly conserved and a major regulator of transcription elongation. It has been shown to directly interact with RNA polymerase to regulate transcriptional pausing and rho-dependent termination^15,20,42,43. Structural modelling of NusG bound to RNA polymerase indicated that the shortened loop region seen in the WF2b sample is unlikely to affect this interaction. Rho-dependant termination has been attributed to the C-terminal (KOW) domain region of NusG, therefore a shortening of the loop region in the N-terminal domain is also unlikely to affect transcription termination. Yet, there has been no function attributed to this loop region of NusG, and thus the effect of variability in this region across species is unknown. However, the deletion could represent the results of evolutionary species divergence. Further sequencing of the gene is required across the B. tabaci species complex to gain further understanding of the diversity.

Why the single whitefly transcriptome approach?

The sequencing of the whitefly transcriptome is crucial in understanding whitefly-microbiota-viral dynamics and thus circumventing the bottlenecks posed in sequencing the whitefly genome. The genome of whitefly is highly heterozygous⁴². Assembling of heterozygous genomes is complex due to the de Bruijn graph structures predominantly used⁴³. To deal with the heterozygosity, previous studies have employed inbred lines, obtained from rearing a high number of whitefly isolines^35,44. However, rearing whitefly isolines is time consuming and often colonies may suffer contaminations, leading to collapse and failure to raise the high numbers required for transcriptome sequencing.

We optimised the ARCTURUS^® PicoPure^® kit (Arcturus, CA, USA) protocol for individual whitefly RNA extraction with the dual aim of determining if we could obtain sufficient quantities of RNA from a single whitefly for transcriptome analysis and secondly, determine whether the optimised method would reveal whitefly microbiota as proof of concept. Using our method, the quantities of RNA obtained from field-collected single whitefly samples were sufficient for library preparation and subsequent transcriptome sequencing. Across all transcriptomes over 30M reads were obtained. The amount of transcripts were comparable to those reported in other arthropoda studies from field collections³². However, we did not observe any difference in assembly qualities³²; probably due to the fact that our field-collected samples had degraded RNA based on RIN, and thus direct comparison with 32 was inappropriate.

Degraded insect specimen have been used successfully in previous studies⁴⁵. This is significant, considering that the majority of insect specimens are usually collected under field conditions and stored in ethanol with different concentrations ranging from 70 to 100%^46–48 rendering the samples liable to degradation. However, to ensure good keeping of insect specimen to be used for mRNA and total RNA isolation in molecular studies, and other downstream applications such as histology and immunocytochemistry, it is advisable to collect the samples in an RNA stabilizing solution such as RNAlater. The solution stabilizes and protects cellular RNA in intact, unfrozen tissue, and cell samples without jeopardizing the quality, or quantity of RNA obtained after subsequent RNA isolation. The success of the method provided an opportunity to unmask vector-microbiota-viral dynamics in individual whiteflies in our study, and will be useful for similar studies on other small organisms.

Endosymbionts diversity across individual SSA1 B. tabaci transcriptomes

In this study, we identified bacterial endosymbionts (Table 2) that were comparable to those previously reported in B. tabaci⁴⁹ and more specifically SSA1 on cassava^23,37. Secondary endosymbionts have been implicated with different roles within B. tabaci. Rickettsia has been adversely reported across putative B. tabaci species, including the Eastern African region^23,50,51. This endosymbiont has been associated with influencing thermo tolerance in B. tabaci species⁴⁹. Rickettsia has also been associated with altering the reproductive system of B. tabaci, and within the females. This has been attributed to increasing fecundity, greater survival, host reproduction manipulation and the production of a higher proportion of daughters all of which increase the impact of virus⁴⁹. In addition, Rickettsia and Hamiltonella play a role in plant virus transmission in whiteflies²⁵ by protecting the safe transit of virions in the haemolymph of insects through chaperonins (GroEL) and protein complexes that aid in protein folding and repair mechanisms²⁰. However, Hamiltonella was reported to be absent in the indigenous whitefly populations studied elsewhere^{15,50,52–54} and in Malawi, Nigeria, Tanzania and Uganda^50,55 as also confirmed in our study. Arsenophonus, Wolbachia, Arsenophonus and Cardinium spp have been detected within MED and MEAM1 Bemisia species^14,50. In addition, 50 and 22 reported Arsenophonus within SSA1 B. tabaci in Eastern Africa that were collected on cassava. These endosymbionts have been associated with several deleterious functions within B. tabaci that include manipulating female-male host ratio through feminizing genetic males, coupled with male killing^56,57.

Within the context of SSA agricultural systems, the role of endosymbionts in influencing B. tabaci viral transmission is important. Losses attributed to B. tabaci transmitted viruses within different crops are estimated to be in billions of US dollars⁴⁶. Bacterial endosymbionts have been associated with influencing viral acquisition, transmission and retention, such as in tomato leaf curl virus^58,59. Thus, better understanding of the diversity of the endosymbionts provides additional evidence on which members of B. tabaci species complex more proficiently transmit viruses, and thus the need for concerted efforts towards the whitefly eradication.