Cauliflower mosaic virus protein P6‐TAV plays a major role in alteration of aphid vector feeding behaviour but not performance on infected Arabidopsis

Abstract Emerging evidence suggests that viral infection modifies host plant traits that in turn alter behaviour and performance of vectors colonizing the plants in a way conducive for transmission of both nonpersistent and persistent viruses. Similar evidence for semipersistent viruses like cauliflower mosaic virus (CaMV) is scarce. Here we compared the effects of Arabidopsis infection with mild (CM) and severe (JI) CaMV isolates on the feeding behaviour (recorded by the electrical penetration graph technique) and fecundity of the aphid vector Myzus persicae. Compared to mock‐inoculated plants, feeding behaviour was altered similarly on CM‐ and JI‐infected plants, but only aphids on JI‐infected plants had reduced fecundity. To evaluate the role of the multifunctional CaMV protein P6‐TAV, aphid feeding behaviour and fecundity were tested on transgenic Arabidopsis plants expressing wild‐type (wt) and mutant versions of P6‐TAV. In contrast to viral infection, aphid fecundity was unchanged on all transgenic lines, suggesting that other viral factors compromise fecundity. Aphid feeding behaviour was modified on wt P6‐CM‐, but not on wt P6‐JI‐expressing plants. Analysis of plants expressing P6 mutants identified N‐terminal P6 domains contributing to modification of feeding behaviour. Taken together, we show that CaMV infection can modify both aphid fecundity and feeding behaviour and that P6 is only involved in the latter.

vectors and alter some of their characteristics to optimize transmission. The interactions may be direct, that is, changes in vector behaviour or fitness following virus acquisition by and retention in the vector, or indirect, that is, plant traits like odour, colour, and nutritive value modified by viral infection impact vector behaviour and fitness (for review see Dáder et al., 2017;Fereres & Moreno, 2009;Mauck et al., 2018). The ways viruses modify or even manipulate plant hosts and vectors depend on the transmission mode and tissue tropism of the virus (Mauck et al., 2012). Transmission modes are classified by two criteria: retention time in the vector (persistence) and vector interaction, with circulative viruses cycling through the vector body before being inoculated as a saliva component into a new host plant and noncirculative viruses interacting only with the vector mouthparts. Circulative and persistent viruses are most often phloemrestricted and characterized by long acquisition and retention times in the vector, resulting in prolonged, often lifelong transmissibility.
General predictions assume that these viruses would benefit from fast and prolonged access of vectors to the phloem, which would facilitate virus acquisition. The circulative and persistent viruses also tend to increase food quality of the host, resulting in an improvement in vector fitness and an increase in vector population (Dáder et al., 2017;Fereres & Moreno, 2009;Mauck et al., 2018). In contrast, noncirculative and nonpersistent viruses are often tissue generalists, with fast acquisition and short retention times. Plants infected with noncirculative and nonpersistent viruses may attract vectors for virus acquisition (Fereres & Moreno, 2009) and subsequently encourage them to leave the plants rapidly for fast dispersal of the virus, as shown for cucumber mosaic virus (Mauck et al., 2010). Vector departure is often related to the poor taste and low nutritive value of infected plants (Mauck et al., 2014). While many reports support these models, there are also a number of examples that do not follow the expectations. Outcome of virus-host-vector interactions may vary depending on the specific virus-host-vector combination and probably other factors (Mauck & Chesnais, 2020;Mauck et al., 2018).

Cauliflower mosaic virus (CaMV) (genus Caulimovirus, family
Caulimoviridae) has features of nonpersistent and persistent viruses and is often classified as a semipersistent virus. Being a tissue generalist, CaMV can infect all cell types and is acquired from epidermis and mesophyll cells like nonpersistent viruses, but also from the phloem sap like persistent viruses (Palacios et al., 2002). With regard to vector interaction, CaMV is a noncirculative virus that binds to stylin receptors in the stylet tips of its aphid vectors (Uzest et al., 2010;Webster et al., 2018). Virus particles require the virusencoded transmission helper protein P2 for vector interaction that forms a complex with the virions (Leh et al., 1999;Moreno et al., 2005). Formation of this P2-virus complex, mandatory for transmission, can occur in infected plant cells or in the stylets, allowing simultaneous or sequential binding of P2 and virus to the vector . Simultaneous binding is associated with fast CaMV acquisition from epidermis and mesophyll, whereas sequential binding favours uptake of CaMV from the phloem sieve tubes that contain virions but are devoid of P2 (Palacios et al., 2002).
Whereas the molecular and cellular mechanisms of CaMV acquisition are well studied, information on the effect of CaMV infection on transmission-relevant changes in aphid behaviour and performance is scarce. Chesnais et al. (2019) reported that the green peach aphid Myzus persicae and the cabbage aphid Brevicoryne brassicae did not show any preference for CaMV-infected Camelina sativa plants.
Both aphid species created more intracellular punctures, ingested less phloem sap, and displayed reduced fecundity on infected plants compared to healthy plants, as expected for a nonpersistent virus.
In this study, we tested how different isolates of CaMV affect M.
persicae feeding behaviour and fitness on the model plant Arabidopsis thaliana. We then addressed the specific role of the multifunctional viral protein P6-TAV in modification of aphid traits. P6 is a key player in CaMV infection (for review see Pooggin & Ryabova, 2018;Schoelz et al., 2016). It is required for translation of the polycistronic CaMV 35S RNA and interacts, among others, with TOR kinase. TOR is involved in translation initiation, but plays also a pivotal role in controlling cell homeostasis of catabolic and anabolic processes. P6 binding triggers TOR phosphorylation, which then promotes protein translation and simultaneously suppresses potential antiviral autophagy and innate immunity (Hafrén et al., 2017;Zvereva et al., 2016).
P6 is also a suppressor of RNA silencing and sufficient to induce typical symptoms of infection by itself (Yu et al., 2003). Finally, P6 is the matrix protein of the viral factories, where replication occurs and virus particles are stored (Schoelz & Leisner, 2017). The P6 functions in dampening plant defences and in inducing symptom expression could impact aphid-plant interactions. P6 is, therefore, a perfect candidate to modify aphid behaviour and performance.

| Effect of viral infection on M. persicae
We first compared the impact of CaMV infection on M. persicae feeding behaviour and performance. For this, we chose the well-characterized CaMV isolates Cabb B-JI (referred to as JI) and Cm1841 that accumulate in infected Brassicaceae at high and low levels (Lung & Pirone, 1973) and cause severe and mild symptoms in Arabidopsis, respectively (Cecchini et al., 1998). Unlike JI, Cm1841 is a nontransmissible CaMV isolate, due to a mutation of amino acid 94 in the transmission helper protein P2 (Woolston et al., 1987). For better comparison, we reverted this mutation to wild type (wt) and The duration of the pathway phase was significantly reduced on infected Arabidopsis for both isolates (GLM, df = 2, χ 2 = 11.14, p = .004). The duration of phloem ingestion was significantly longer on JI-infected Arabidopsis compared to CM-infected and mockinoculated plants (GLM, df = 2, χ 2 = 6.27, p = .04; Figure 2b).

| Effect of P6 on aphids
We used transgenic Arabidopsis expressing wt and mutant P6 proteins from the constitutive 35S promoter to determine whether P6 is involved in plant-aphid interactions.

| D ISCUSS I ON
Many studies report changes in vector behaviour and performance after infection of plants with viruses, but only a few identify viral determinants involved in vector manipulation. Also, most work centres shows the same blot revealed with a short (left) or a longer (right) exposure time to visualize weak bands. Extracts prepared from untransformed Col-0 and from CM-infected leaves were loaded as negative and positive controls, respectively. (j) Detection of P6-JI-Eki and P6-CM-Δd23-HA by western blot. The two proteins that were not detected in the blot shown in (i) could be revealed in a different experiment using more concentrated extracts. Each lane presents extract from a different transgenic plant. Signals from wild-type P6 loaded on the same blots are shown to the left of the panels and either the blots were exposed much shorter (P6-JI-Eki) or extracts were diluted (P6-CM-Δd23-HA). Note that the mutant P6 concentration varied considerably. Ponceau red staining of the large RuBisCO subunit is shown as a loading control on persistent and nonpersistent viruses; studies on semipersistent viruses are scarce. To fill this gap, we studied the feeding behaviour and fecundity of aphids on Arabidopsis infected with a severe and a mild isolate of CaMV, a virus with transmission properties of both nonpersistent and persistent viruses. Then, we studied the potential role of the multifunctional viral protein P6 in virus-induced alterations of host-plant traits and vector responses.

| Effect of CaMV infection on aphid fecundity
The severe and high-accumulating CaMV isolate JI reduced aphid fecundity on infected Arabidopsis significantly (c.2.5 fewer nymphs produced per adult after 5 days compared to mock-inoculated plants), while infection with the mild and low-accumulating isolate CM had only a moderate impact on aphid fecundity, below the significance threshold (c.1 nymph less produced per adult after 5 days compared to mock-inoculated plants; Figure 2a). Thus, both increased symptom severity and higher viral accumulation correlated with reduced aphid fecundity. Whether symptom severity is related to CaMV accumulation is still unclear and seems to depend on the host-virus association (Doumayrou et al., 2013). Lower fecundity and other negative effects of plant infection on vector performance such as reduced vector survival or delayed development are predicted for nonpersistent viruses. These modifications may promote rapid dispersion of vectors and subsequently virus transmission (Mauck et al., 2018). Indeed, mathematical models show that increasing departure rates from hosts lead to increased vector movements between host plants, which is beneficial for virus dissemination (Carr et al., 2020;Shaw et al., 2017). Whether this applies for CaMV on Arabidopsis remains to be determined because although reduced fecundity is a proxy for aphid fitness, it does not indicate that aphids will really This suggests that the effect on fecundity is not host plant-specific, but rather virus isolate-specific.

| Effect of CaMV infection on aphid feeding behaviour
M. persicae on CaMV-JI-infected plants spent significantly less time in the pathway phase in the leaf tissue and more time ingesting phloem sap than aphids on healthy plants (Figure 2b). This is expected to be counterproductive for transmission of nonpersistent viruses (Carr et al., 2020;Eigenbrode et al., 2018;Mauck et al., 2012).
In fact, typical nonpersistent viruses like potyviruses or cucumber F I G U R E 4 Myzus persicae fecundity 5 days after deposit on untransformed or transgenic 5-week-old Arabidopsis plants expressing P6 proteins. Different letters show significant differences between plants as tested by GLM followed by pairwise comparisons using "emmeans" (p < .05; method: Tukey, n = 21-24). No statistically significant differences were recorded F I G U R E 5 Aphid feeding behaviour parameters recorded by EPG on untransformed Col-0 and transgenic P6-expressing 5-week-old Arabidopsis. Different letters indicate significant differences between plants as tested by GLM followed by pairwise comparisons using "emmeans" (p < .05; method: Tukey, n = 25-31) mosaic virus (single-stranded RNA viruses of families Potyviridae and Bromoviridae, respectively) are acquired during short intracellular stylet punctures in plant epidermis and mesophyll (Martin et al., 1997). They are lost from the stylets when aphids stay on plants for longer times and when the stylets reach the sieve tubes that do not contain transmission-competent virus forms (Kloth & Kormelink, 2020;Wang & Ghabrial, 2002). The case is different for CaMV that is acquired efficiently from both cells and phloem sap and after both short and long acquisition periods (Bouchery et al., 1990;Markham et al., 1987;Palacios et al., 2002). Therefore, both improved palatability (characterized by few probes, rapid access to the phloem, and long phloem ingestion) and reduced palatability (characterized by many probes, impairment in phloem access, and reduced phloem generally transmitted by less than a dozen. This means that these noncirculative viruses need to interact with multiple vectors, and it is unlikely that they can engage in manipulations that are specific for each vector species. Rather, they must target a common and accessible vector feature, for example a conserved receptor in the aphid stylets (Webster et al., 2018), to which they can bind easily and already after short vector-plant contact. This enables most aphids, even noncolonizers, to engage in behaviours conducive with nonpersistent virus acquisition and inoculation (e.g., epidermis and mesophyll probing). As a consequence, there may be little (or no) advantage and no strong selection pressure for CaMV and other noncirculative viruses to manipulate vector feeding behaviours (Mauck & Chesnais, 2020). This would mean that other viral determinants (proteins or RNAs), either independently or in concert with P6, induce changes in plants that reduce aphid fecundity. Indeed, it has been suggested that the CaMV proteins P1-5 are mainly responsible for leaf malformation, while P6 causes chlorosis and dwarfism (Yu et al., 2003). The second explanation is that P6 levels in the transgenic plants were too low to reduce aphid fecundity. Indeed, we observed that P6 levels were considerably lower and variable in transgenic plants compared to infected plants (Figure 3i,j). However, in CaMV-infected plants a major fraction of P6 protein is located in inclusion bodies/replication factories (Schoelz & Leisner, 2017) and only a small fraction of P6 may be available in a soluble form to fulfil its other functions (reviewed by Pooggin & Ryabova, 2018). Finally, it is also possible that phloem sap composition (e.g., amino acid and/or sugar concentration and composition) was changed in infected but not in transgenic plants.

| A role for P6 in aphid manipulation
Further experiments are required to solve this issue.
Some aphid feeding behaviour parameters were significantly al-

| P6 domains involved in aphid manipulation
The comparison of the aphid feeding behaviour on transgenic Arabidopsis expressing various P6 versions yielded evidence for the involvement of its N-terminus in modification of plant-aphid interactions ( Figure 5). P6 from the D4 isolate, although accumulating to much higher levels in transgenic plants than P6-JI or P6-CM ( Figure 3i), had no effect on aphid behaviour, and plants displayed a wt phenotype (Figure 3c), as previously reported by Yu et al. (2003).
Because P6-D4's RNA silencing suppression domains are functional in Arabidopsis (Shivaprasad et al., 2008;Zvereva et al., 2016), we conclude that the silencing suppression domain (probably located in the C-terminal portion of P6; see Figure 6 for an overview of P6's functional domains) has no impact on aphid behaviour or on symptom expression. P6-D4's Vir/Avr domain, responsible for symptom expression, and the mini-TAV domain, required for translation transactivation and TOR-mediated immunity, are nonfunctional in Arabidopsis (Yu et al., 2003;Zvereva et al., 2016), precluding any de- CaMV isolate-specific. In fact, the Vir/Avr domain is one of the most variable domains of P6 among CaMV isolates. Notably, despite previous controversial evidence (Laird et al., 2013), the P6-CM Vir/Avr domain was found to be crucial for virus infectivity and virulence but not suppression of RNA silencing in Arabidopsis (Zvereva et al., 2016). Moreover, this domain contributes to suppression of salicylic acid-dependent autophagy and effector-triggered innate immunity in Arabidopsis, although it was not absolutely essential for this TORdependent function of P6-CM (Zvereva et al., 2016). We speculate P6-mediated dampening of pattern-and/or effector-triggered innate immunity may explain the observed changes in aphid feeding behaviour. Indeed, aphids are known to induce innate immunity responses in Arabidopsis and deliver effector proteins to suppress these responses (Mugford et al., 2016;Prince et al., 2014). A role for the P6 mini-TAV region with TOR-binding domain in modification of aphid behaviour remains disputed. Despite its high accumulation levels in transgenic plants, comparable to those in virus-infected plants, P6-JI-ΔdsR (no TOR interaction) only slightly affected aphid behaviour. This indicates that the TOR-binding domain is not required but other elements of the mini-TAV region preserved in the mutant might still be important for modification of aphid behaviour.
Finally, P6-JI-Eki did not change aphid behaviour significantly, suggesting that the extreme N-terminus of P6 may not be involved in aphid interactions. However, in this case the mutant protein accumulated in transgenic plants at extremely low levels, precluding any definitive conclusion.
Taken together, this report shows that aphid responses to CaMV infection depend on the virus isolate, and that part of the effects, but not the entire response, relies on P6.

| Cloning
Cm1841-Rev, where the mutation of amino acid 94 in P2 in the Cm1841 genome is reverted to wt, was obtained by site-directed mutagenesis using the QuikChange Lightning Kit (Agilent) following the manufacturer's instructions and using the oligonucleotides QuikReve-F (5′-GTCAGTTTTTAATACTGCAAAAAACATTTTTAAA AGTGGGGGGGTTGATTACTCG-′) and QuikReve-R (5′-CGAGTAAT CAACCCCCCCACTTTTAAAAATGTTTTTTGCAGTATTAAAAACTG AC-3′). The two oligonucleotides contain a silent G → A mutation at nucleotide 277 of the P2 sequence to create a DraI restriction site for easy identification of recombinants and an A → G mutation at nucleotide 280 of P2 to revert amino acid R 94 to wt G 94 . The pCa122 plasmid containing 1.2 copies of the Cm1841 genome (Tsuge et al., 1994) was used as a matrix. The cycling programme was 2 min at 95 °C; followed by 18 cycles of 20 s at 95 °C, 10 s at 60 °C, and 7 min F I G U R E 6 Functional domains of P6 and P6 mutants used in this study. Only relevant domains and interacting proteins are shown. The N-terminal half of P6 contains three major regions: The N-terminus (amino acids 4-31) contains one of several P6/P6 self-interaction domains required for P6's function as the matrix protein of virus inclusions. This region is followed by the Vir/Avr domain delimited by the d23 region, which is involved in chlorosis induction and dwarfism. The mini-TAV domain comprises amongst others an N-terminal region binding to ribosomal proteins eL13 and eL18, the central double-stranded RNA-binding (dsR) region that interacts with TOR kinase, and a C-terminal region that interacts with reinitiation supporting protein (RISP) and contains also a nuclear localization signal (NLS1). The mini-TAV region is important for antiviral autophagy, innate immunity, and polycistronic translation. One of the functions of the C-terminal half of P6 is probably suppression of RNA silencing, in which also the N-terminal nuclear export signal (NES) and the two nuclear localization signals (NLS1 in the mini-TAV domain and NLS2 in the C-terminal region) play a role. The dsR region is deleted in P6-JI-ΔdsR. In P6-JI-Eki, the conserved Eki motif just preceding the NES is substituted by three alanines. CM-HA contains an N-terminal HA-tag as modification and CM-Δd23-HA in addition a deletion of the d23 region. RBa, RBb, RNA-binding regions. Figure adapted from Pooggin and Ryabova (2018) at 69 °C; and a final extension step of 5 min at 68 °C. The plasmid pCa122rev was verified by sequencing the mutated region.

| Virus inoculation
Initial inoculation was performed with the pCa122rev plasmid or the pGreen-BJI plasmid (Khelifa et al., 2010)

| Aphid rearing
The M. persicae green peach aphid clone was originally isolated in the Netherlands. Aphids were reared on Chinese cabbage (Brassica rapa pekinensis) in a growth chamber at 20 ± 1 °C, under a 16 hr photoperiod. Only wingless forms were used in assays. For synchronization, we placed adults on detached Chinese cabbage leaves that were spread on 1% agarose in a Petri dish. The adults were removed 24 hr later and the newborn larvae used in experiments after another 7 days when they had reached the adult stage.

| Aphid feeding behaviour
We used the electrical penetration graph DC-system as described by Tjallingii (1988) to investigate the effects of plant CaMV infection on the feeding behaviour of M. persicae. Eight aphids were connected to the Giga-8 DC-EPG amplifier and each one was placed on the leaf of an individual A. thaliana plant. The recordings were performed continuously for 8 hr during the photophase. Each aphid-plant system was placed inside a Faraday cage at 21 ± 1 °C.
Acquisition and analysis of the EPG waveforms were carried out with PROBE 3.5 software (EPG Systems). Relevant aphid behaviour EPG parameters were calculated with EPG-Calc 6.1 software (Giordanengo, 2014) and were based on different EPG waveforms described by Tjallingii and Hogen Esch (1993). The following parameters were selected because they are relevant and important for the acquisition of CaMV by aphids: the total duration of "probing time", "pathway phase", and "phloem sap ingestion phase" and the time needed by the aphid to reach the phloem. For each condition (healthy or infected and wt or mutant, respectively), EPGs of 25-30 individual aphids were analysed. Aphids that produced signals (i.e., total duration of probing time) for fewer than 5 hr were excluded from the analysis.

| Aphid fecundity
Synchronized wingless adults (8 ± 1 days old) were randomly selected from the aphid pools and transferred onto Arabidopsis plants (one aphid per plant) to study adult fecundity. The number of nymphs produced were recorded after 5 days. Adult aphids that died before day 5 were excluded from the analysis. Data on CaMV-infected or mock-inoculated Arabidopsis were collected in three repetitions, and data on Arabidopsis mutants were collected in four repetitions, comprising altogether 21-33 aphids per condition (infected or mockinoculated and wt or mutant, respectively).

| Statistical analysis
We used GLMs with a likelihood ratio and the chi-square (χ 2 ) test to assess whether there was an effect of plant infection or Arabidopsis mutants on M. persicae feeding behaviour. When a significant effect was detected, a pairwise comparison using estimated marginal means (R package "emmeans"; p value adjustment with Tukey method) at the .05 significance level was used to test for differences between treatments. Data on aphid feeding behaviour (probing, pathway, and phloem sap ingestion phases) were not normally distributed; accordingly we carried out a GLM using a gamma (link = "inverse") distribution. Data on aphid feeding behaviour (t1 < E2) were modelled using the Cox proportional hazards model and we treated cases where the given event did not occur as censored. The assumption of validity of proportional hazards was checked using the functions "coxph" and "cox.zph", respectively (R package "survival"). When a significant effect of one of the main factors was detected or when an interaction between factors was significant, a pairwise comparison using estimated marginal means (R package "emmeans") (p value adjustment with Tukey method) at the .05 significance level was used to test for differences between treatments. The fit of all GLMs was controlled by inspecting residuals and QQ plots. All statistical analyses were performed using R software v. 3.3.2 (www.r-proje ct.org/).

| Western blot
Total leaf extracts were prepared as follows: Leaves were frozen in a mortar with liquid N 2 and ground with a pistil to a fine powder. The powder was transferred to a 1.5 ml reaction tube and 2× Laemmli buffer (Laemmli, 1970) was added at a ratio of 1:1 (wt/vol). The samples were heated for 5 min at 80 °C and centrifuged for 10 min at 16,000 × g, and aliquots of the supernatants were charged on 6/12.5% discontinuous sodium dodecyl sulphate (SDS)-polyacrylamide gels.
Proteins were transferred after SDS polacrylamide gel electrophoresis onto nitrocellulose membranes using the wet blotting technique (Towbin et al., 1979). Efficiency of transfer was controlled by Ponceau red staining. Membranes were blocked for 30 min with 5% low-fat milk powder in Tris-buffered saline (TBS) and incubated overnight at 4 °C with primary antibodies. After three washes with TBS, membranes were incubated for 3-4 hr at room temperature with secondary antibodies. After another three washes with TBS, protein bands were visualized by enhanced chemiluminescence using a G-Box. The following 1:2,000 dilutions of primary antibodies were used: anti-P2 (Blanc et al., 1993), anti-P4 (Champagne et al., 2004), and anti-P6 (Khelifa et al., 2010). Secondary antibodies were horseradish peroxidase conjugates, which were used at a 1:25,000 dilution.

ACK N OWLED G EM ENTS
The authors are grateful to Claire Villeroy for aphid care. This work was funded by Human Frontier Science Program grant RGP0013/2015 (to M.D.) and Agence Nationale de la Recherche grant ANR-18-CE20-0017-01 (to M.D. and M.P.). The funding sources had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.