The role of the whitefly, Bemisia tabaci (Gennadius), and farmer practices in the spread of cassava brown streak ipomoviruses

Abstract Cassava brown streak disease (CBSD) is arguably the most dangerous current threat to cassava, which is Africa's most important food security crop. CBSD is caused by two RNA viruses: Cassava brown streak virus (CBSV) and Ugandan cassava brown streak virus (UCBSV). The roles of the whitefly Bemisia tabaci (Gennadius) and farmer practices in the spread of CBSD were investigated in a set of field and laboratory experiments. The virus was acquired and transmitted by B. tabaci within a short time (5–10 min each for virus acquisition and inoculation), and was retained for up to 48 hr. Highest virus transmission (60%) was achieved using 20–25 suspected viruliferous whiteflies per plant that were given acquisition and inoculation periods of 24 and 48 hr, respectively. Experiments mimicking the agronomic practices of cassava leaf picking or the use of contaminated tools for making cassava stem cuttings did not show the transmission of CBSV or UCBSV. Screenhouse and field experiments in Tanzania showed that the spread of CBSD next to spreader rows was high, and that the rate of spread decreased with increasing distance from the source of inoculum. The disease spread in the field up to a maximum of 17 m in a cropping season. These results collectively confirm that CBSV and UCBSV are transmitted by B. tabaci semipersistently, but for only short distances in the field. This implies that spread over longer distances is due to movements of infected stem cuttings used for planting material. These findings have important implications for developing appropriate management strategies for CBSD.

was reported to cause reductions of up to 70% in tuberous root yield of susceptible cultivars (Hillocks, Raya, Mtunda, & Kiozia, 2001). In addition to having direct deleterious effects on the growth of cassava plants, the disease causes necrosis of affected roots, making them unfit for consumption or marketing, and thus affecting food security (Legg et al., 2014). The continental significance of CBSD increased greatly from 2004, when the first reports were made of epidemics in mid-altitude areas of Uganda (Alicai et al., 2007). In subsequent years, further outbreaks were reported from other countries in the Great Lakes region of East and Central Africa, including western Kenya, north-western Tanzania, Rwanda, Burundi and Democratic Republic of Congo (Bigirimana, Barumbanze, Ndayihanzamaso, Shirima, & Legg, 2011;Legg et al., 2011;Mahungu, Bidiaka, Tata, Lukombo, & N'luta, 2003;Mulimbi et al., 2012). The disease has potential to spread from the mid-altitude regions of East and Central Africa to the neighbouring cassava-growing areas in southern and West Africa, and eventually to much of SSA with devastating consequences (Legg et al., 2014. Cassava brown streak disease is caused by two distinct species of single-stranded RNA (ssRNA) viruses: Cassava brown streak virus (CBSV) and Ugandan cassava brown streak virus (UCBSV), (genus Ipomovirus, family Potyviridae) (Mbanzibwa, Tian, Mukasa, & Volkonen, 2009;Mbanzibwa et al., 2011;Monger et al., 2010;Winter et al., 2010), which are together referred to as cassava brown streak ipomoviruses (CBSIs). Earlier work on the transmission of CBSIs showed that they can be graft-transmitted from cassava to cassava (Ogbe, Dixon, Huges, Alabi, & Okechukwu, 2006) and mechanically transmitted from cassava to a number of herbaceous hosts (Lister, 1959;Mohammed, Abarshi, Muli, Hillocks, & Maruthi, 2012). In addition, it was suggested that CBSIs spread naturally in the field through the transmission activity of insects, in particular two whitefly species; Bemisia tabaci (Gennadius) (Bock, 1994;Storey, 1939) and Bemisia afer (Priesner & Hosny) (Hemiptera: Aleyrodidae), which were abundant in the CBSD endemic areas (Bock, 1994;Munthali, 1992). Subsequent transmission studies with both species of whitefly and with some species of aphid, however, were unsuccessful (Bock, 1994;Lennon, Aiton, & Harrison, 1986).
The first evidence of CBSV transmission by an insect vector, the whitefly B. tabaci, was obtained in our earlier laboratory studies (Maruthi et al., 2005), which was later confirmed (Mware et al., 2009).
However, virus transmission patterns were inconsistent in both of these studies, and the low rate of transmission observed could not explain the high rate of spread in the field. The lack of correlation between laboratory studies and field observations has led to speculation that CBSIs may also be spread by other means, such as through contact between diseased and healthy plants, through tools contaminated during the process of cassava harvesting, and/or in the process of harvesting cassava leaves (leaf picking) for use as a vegetable.
The aim of this study was therefore to determine whether CBSIs can be transmitted by contaminated tools or during the process of leaf picking, as well as to understand the transmission characteristics of CBSIs by the B. tabaci. The findings from these studies will provide guidance for the development and implementation of control strategies to address what is currently one of Africa's biggest crop production threats.

| Cassava varieties, virus isolates and whitefly colonies used in the study
Two CBSD-susceptible cassava varieties (var.)-Albert and TMS 60444-were grown from stem cuttings and confirmed to be free from CBSIs by reverse transcription polymerase chain reaction (RT-PCR; Abarshi et al., 2010Abarshi et al., , 2012Otti et al., 2016). These were used as target plants for virus inoculations in the UK. Two virus isolates-UCBSV from Kabanyoro, Uganda and CBSV from Naliendele, Tanzania-described previously were used in virus transmission experiments where indicated . Virus-free plants of two cassava vars.-Kiroba and Kaleso-were also used to test the efficiency of virus transmission by whiteflies. Both Kiroba and Kaleso inhibit the multiplication of CBSV upon inoculation and were described as tolerant and resistant to CBSD, respectively (Maruthi, Bouvaine, Tufan, Mohammed, & Hillocks, 2014). Another cassava var. Ebwanateraka infected with either CBSV or UCBSV provided the source of viruses.
The colony of B. tabaci used in this study was collected on cassava originally from Uganda and maintained subsequently on cassava in the quarantine insectary facilities of NRI in the UK (Maruthi, Colvin, & Seal, 2001). This colony was confirmed to belong to the species sub-Saharan Africa 1-subgroup 1 (SSA1-SG1) based on mitochondrial cytochrome oxidase I gene sequences.
Virus-indexed tissue culture plantlets of var. Kiroba, shown to be free of CBSIs using RT-PCR, were hardened off in a screenhouse with insect-proof netting in Kibaha, Pwani Region, Tanzania. These plants were subsequently used to establish the CBSD spread trials in the field and screenhouse in the year 2012, as described below. Field-grown CBSD-affected plants of the same cassava variety were obtained from field experiments at Kibaha for use as the spreader blocks in each of these trials, and B. tabaci adults used in this experiment were similarly obtained from field-grown cassava plants.

| Transmission of CBSV by B. tabaci
Initial CBSV transmission experiments by B. tabaci involved a combination of using long periods of virus acquisition access (AAP) and inoculation access (IAP) of up to 5 days and using high whitefly numbers to increase the probability of virus transmission. Whiteflies were collected from the colony and allowed to feed for 4 days on CBSD-affected cassava plants of var. Ebwanateraka. The suspected viruliferous whiteflies were then collected and immediately released in two groups of either 20-25 or 50-100 on each healthy target plant for 5 days to inoculate the virus. In another experiment, between 25 and 100 whiteflies born on diseased plants were used for transmitting CBSV to each healthy plant (Table 1). Between 10 and 26 plants were inoculated for each category of whiteflies in three replications. All inoculated plants were enclosed individually in insect-proof bread bags to prevent cross-contamination.
Plants were kept in an insectary (28 ± 5°C) and observed for symptom development. Unless otherwise specified, all plants used in controlled experiments in the UK were tested for infection with CBSV and UCBSV by RT-PCR (Abarshi et al., 2010 three months after exposure to adult whiteflies from CBSD-infected plants. Data on the number of plants infected with the viruses were subjected to Chi-squared test using the software package sigmaplot for Windows version 11.0 (Systat Software inc., San Jose, CA, USA).

| Determining the mode of transmission of CBSV by B. tabaci
Transmission experiments were initiated to investigate potential nonpersistent, semipersistent and persistent modes of CBSV transmission by whiteflies. To verify the non-persistent mode of transmission, whiteflies were given three relatively short AAP of 5-10 min, 30 min and 1 hr on a CBSV-infected cassava plant of var. Ebwanateraka.
About 20-25 adult viruliferous whiteflies were immediately introduced to each target plant for a 48 hr IAP. Chi-squared analyses of data from these experiments were conducted in all possible combinations to identify significant differences between the treatments ( Table 2). All non-persistent treatments were compared to all persistent and semipersistent treatments (both 24/48 and 48/48 hr AAP/IAP combinations). Finally, all semipersistent treatments were compared to all persistent treatments using sigmaplot 11.0.

| Determining virus acquisition, inoculation and retention times in B. tabaci
For testing AAP, whiteflies were allowed to feed on CBSV-infected cassava var. Ebwanateraka for 5-10 min, 24 hr values were compared.

| Transmission of CBSV and UCBSV to different cassava varieties
Three cassava var-Albert, Kiroba and Kaleso-were inoculated with CBSV or UCBSV by whiteflies to validate the whitefly transmission method for varieties with contrasting levels of resistance to CBSD.
Albert is susceptible to CBSD, Kiroba is tolerant with delayed expression of root symptoms, and Kaleso is resistant with no root symptoms but with mild leaf symptoms. Negligible amounts of virus accumulate in Kaleso and Kiroba, while high amounts of virus accumulate in Albert (Maruthi et al., 2014). Thirty plants of each variety were each inoculated with 20-25 suspected viruliferous whiteflies that were given an AAP and IAP of 24 hr each. The experiment was conducted in three replicates for each virus-variety combinations.  (Abarshi et al., 2010. Data from the above three experiments were compared using the ANOVA procedure in sigmaplot 11.0.  and was continued for an additional 5 weeks. The ANOVA procedure of sigmaplot 11.0 was used to analyse the pattern of distribution between plots of both CBSD incidence and whitefly abundance, while

| Screenhouse simulation of CBSD spread
Pearson's correlation and linear regression analyses were employed to examine the relationship between whitefly abundance and CBSD.  B. tabaci. Severity was assessed using the standard 1-5 scoring system in which "1" corresponds to symptom-free, "2" to the mildest symptoms and "5" the most severe symptoms (Hillocks & Jennings, 2003;Hillocks et al., 2001). Whitefly abundance was assessed by counting the number of adult B. tabaci on the top five leaves of each plant.

| Field transmission of CBSIs
Unless otherwise indicated, data were recorded at weekly intervals up to 6 MAP. Kruskal-Wallis one-way analysis of variance on ranks was used to test for the significance of gradients in CBSD incidence and whitefly abundance from the nearest to the farthest plot from the spreader. For this test, data for each of the time points (from 4 WAP to 22 WAP) were considered as replicates.

| Verifying the transmission of CBSV by B. tabaci
Highest virus transmission was recorded (53.0%) when 50-100 whiteflies that had up to 5 days each AAP and IAPs were used in the experiments (

| Mode of CBSV transmission by B. tabaci
Whiteflies that had an AAP of 5-10 min were able to acquire and  (Table 2).

| AAP, IAP and retention of CBSV in B. tabaci
This experiment reconfirmed that CBSV can be acquired within 5-10 min of whitefly feeding on CBSD-affected plants (Table 3).
Highest rate of transmission (45.0%) was achieved at 24 hr AAP, although this was not significantly different from those that had AAPs of 1 hr, 4 hr and 48 hr. Whiteflies were also able to transmit CBSV within 5-10 min (IAP) of feeding on a healthy plant ( Comparison of data by Chi-squared tests showed significant differences in transmission efficiencies between whiteflies with 5-10 min AAP and those with 1 hr plus AAP (χ 2 = 4.23, p = .04, df = 1).

| Effect of leaf age, virus species and cassava variety on virus transmission
Whiteflies that fed on younger leaves with no or early symptoms of CBSD achieved a slightly higher rate of transmission (36.3%) compared to those fed on older but fully symptomatic leaves (28.5%). and Kaleso (0.0 ± 0.00, 0%).

| Whitefly abundance
Whiteflies were first recorded from test plots 1 week after their introduction, but over the course of the first 4 weeks of records (4-7 WAP) spread to reach block 4, which was most distant from the spreader (Figure 1a). This means that in the absence of wind in the protective environment of a screenhouse, whiteflies took 7 weeks to move from spreader rows to the farthest block.

| CBSD incidence
The first symptoms of CBSD in test plants were recorded in block 2 at 8 WAP (Figure 1b). CBSD was restricted to blocks 1 and 2 (maximum distance 4 m) up to 13 WAP. Incidences increased greatly in all blocks following the ratooning of the spreader-from 18 WAP onwards. There were strong gradients in the incidence of CBSD from the nearest (highest incidence) to the furthest (lowest incidence) blocks away from the spreader from 18 to 20 WAP, after which the disease became more generally distributed (Figure 1a). Statistically significant gradients were seen in CBSD incidences for both the 18 WAP and 22 WAP data sets (Table 4).
It was evident both from the graphical representation of the data (Figures 1 and 2) and the statistical analyses (Table 4) that gradients in whitefly abundance corresponded with those for CBSD incidences.
To examine this further, Pearson's correlation analyses were run to relate mean whitefly abundances to CBSD incidences for corresponding  2 1.5 8.6 49.7 79.2 119.1 160.9 88.1 48.1 43.5 51.2 82.1 62.8  plots, using both the 18 WAP and 22 WAP data sets (Table 5). The strongest correlation was obtained with whiteflies at 18 WAP and CBSD at 22 WAP. In addition, there was a strongly significant linear regression relationship between whitefly abundance at 18 WAP and CBSD incidence 4 weeks later (CBSD = 0.28 + 0.018 WF; F = 24.0, p < .001, r 2 = .63).  The Kruskal-Wallis ANOVA on ranks test comparing whitefly abundance in each of the experimental plots provided no evidence for differences between plots (H = 6.8, df = 4, p = .150). It was therefore concluded that whiteflies were randomly distributed between plots.

| CBSD incidence
The first symptoms of CBSD on test plants were recorded at 4 WAP, 2 m and 7 m from the spreader plot ( Figure 2a). Incidences of CBSD appeared at 17 m from the spreader plot starting from 7 WAP. The first symptoms at 12 m from the spreader plot were observed at 8 WAP.
There was a strong gradient of declining CBSD incidence from the test plot nearest to the spreader plot (2 m

| DISCUSSION
Research into CBSD and its causal viruses (CBSV and UCBSV) has increased greatly as the spread of the disease was reported into previously unaffected parts of East Africa (Alicai et al., 2007).
However, the mechanisms of transmission of these viruses remain poorly characterized. Our results respond to several of the key questions on transmission and epidemiology. Initial experiments confirmed that CBSV can be transmitted by B. tabaci adults under laboratory conditions. The rate of transmission, however, was moderate (highest 53%) even when using high whitefly numbers (50-100 per plant) and with prolonged acquisition and inoculation access periods of up to 5 days, or when using whiteflies that had emerged from CBSDaffected plants. These results were, however, similar to previous findings (Maruthi et al., 2005;Mware et al., 2009) (Webb et al., 2012). Transmission of CVYV was also moderately efficient. Virus acquisition and inoculation occurred within 10-20 min, but required 30-35 whiteflies to reach a highest transmission rate of 80%. Persistence in the vector was also short, with a dramatic decrease in transmission from 81% to 14% after 2 hr (Harpaz & Cohen, 1965). Similar results were obtained using another isolate of CVYV in the 1990s (Mansour & Al-Musa, 1993 Kaleso, although at differing efficiencies. UCBSV was only transmissible to Albert and Kiroba, but not to Kaleso. This could be due to the relatively mild nature of the virus and low virus quantities in infected plants Winter et al., 2010). CBSV in comparison was transmitted to all three varieties with different efficiencies, including the resistant var. Kaleso, confirming that whiteflies play a significant role in virus spread in the field irrespective of the variety that is grown. Experiments confirmed that neither leaf picking nor the use of contaminated tools for cutting stems resulted in CBSV transmission. It is therefore concluded that neither of these widespread practices contribute to the epidemiology of CBSD in the field, as had been suspected by some researchers. Circumstantial evidence further confirms this finding, as leaf picking is practiced in some regions of East Africa and not in others, and there is no apparent association between the incidence of CBSD and the prevalence of leaf picking.
Similarly, if stem cutting resulted in transmission, significant increases in incidence might be anticipated even in areas where whiteflies are infrequent, which does not match with field data (Jeremiah et al., 2015;Legg et al., 2011). In both experiments, there was a clear association between the abundance of B. tabaci whiteflies and new CBSD infections, both in space and through time. Over the 8 months that data were recorded in the field experiment, the furthest distance that CBSD infections were recorded from the spreader plot was 17 m. Both experiments emphasize the relatively short distances over which CBSIs are spread-a result which is strongly congruent with the semipersistent transmission mechanism described from the laboratory experiments. Whiteflies migrated into the field experiment randomly from the surrounding vegetation. However, the strong gradient of CBSD between the spreader and the test plots, in which no CBSD at all was recorded from the test plot furthest away from the spreader, provides clear evidence that the spreader plot was the only significant source of CBSD. The corollary of this is that neither the natural vegetation immediately surrounding the field experiment, nor the distant (>300 m) cassava fields that had significant incidences of CBSD, had any significant effect on CBSD spread in the test plots of the field experiment.
The results of our experiments present a consistent picture for the pattern of transmission of CBSIs by the whitefly vector-Bemisia tabaci. As well as helping to explain how CBSD is spreading, knowledge of the semipersistent transmission mechanism also allows us to design appropriate and effective control strategies. The relatively poor retention of CBSIs by B. tabaci, and associated short gradients of spread, means that isolation is likely to be more effective in preventing infection from neighbouring virus sources. Using this as a basis, a novel cassava phytosanitation programme has been implemented in Tanzania to remove all CBSD-affected cassava from rural communities to establish CBSD-free zones. Farmers are then given disease-free cassava planting material for cultivation, which is expected to remain diseasefree because of the poor transmission of CBSIs by the whiteflies. If implemented together with the development and dissemination of disease-resistant cassava varieties, this can be a successful strategy for CBSD control in affected countries. Our results also indicate that by far the greatest threat of long-distance spread of CBSIs comes from the inadvertent carriage by people of infected cassava stems. This will require implementing stricter quarantine regulations to prevent the movement of infected cassava material as well as applying rigorous phytosanitary standards when multiplying and disseminating cassava germplasm obtained from regions affected by CBSD.