Cucumber mosaic virus populations in Tunisian pepper crops are mainly composed of virus reassortants with resistance-breaking properties

Authors


Abstract

Cucumber mosaic virus is one of the most prevalent viruses in Tunisian pepper crops, where it has been detected in 68% of plants developing mosaic symptoms, making it essential to characterize the molecular and biological properties of local CMV populations. Two hundred and seventy-eight isolates collected in the late 1990s, 2006 and 2008–2010 were characterized genetically. Isolates belonging to the three phylogenetic subgroups of CMV (IA, IB and II) were detected, but surprisingly, 90% of the isolates were reassortants between subgroups IA and IB, with two predominant haplotypes, IB-IA-IA and IB-IA-IB (nomenclature according to the subgrouping of the three genomic RNAs). The IB-IA-IA haplotype was present in all regions surveyed, while IB-IA-IB was observed only in northern Tunisia. This situation was unexpected, because CMV reassortants were previously thought to be counterselected in nature, and this raises the questions of the origin of IB strains in Tunisia and of the widespread distribution of these two reassortant types. Phylogenetic studies revealed low diversity within haplotypes, whatever the locality or the year of sampling. However, analysis of haplotype frequencies revealed a high genetic differentiation between CMV populations, which was better explained by the localities of sampling than by years. Geographic distances affected the differentiation of CMV populations, mainly between north and central Tunisia. When tested against a polygenic resistance to CMV movement in pepper, 55 of 57 isolates tested were able to break the resistance, indicating that this resistance would not be useful for controlling CMV in Tunisian pepper fields.

Introduction

Pepper (Capsicum annuum) is an important vegetable crop in Tunisia, grown mostly for local consumption. It is affected by different viral diseases, among which those induced by Cucumber mosaic virus (CMV) and Potato virus Y (PVY) seem to be the most damaging (Fakhfakh et al., 1999; Ben Khalifa et al., 2009). The two viruses often occur in mixed infection, which results in increased severity of the disease through synergy (see review by Syller, 2012). The molecular and biological properties of Tunisian PVY populations have been studied in detail (Ben Khalifa et al., 2009, 2012), but very little is known about CMV populations in Tunisia.

Cucumber mosaic virus, the type member of the genus Cucumovirus, has a tripartite genome composed of single-stranded positive-sense RNAs (RNAs 1, 2 and 3), encoding five proteins, numbered according to that of the genomic segments (Palukaitis & García-Arenal, 2003; Jacquemond, 2012). Proteins 1a and 2a are components of the replication complex. ORF 2b, which overlaps ORF 2a, encodes a viral RNA silencing suppressor, which is also involved in virus movement and overall virulence. Proteins 3a and 3b correspond to the movement protein (MP) and the capsid protein (CP), respectively. Based on phylogeny, CMV strains fall into three main subgroups, IA, IB and II, with inter-subgroup nucleotide identities ranging from 72 to 92% (subgroup II versus subgroups IA/IB, subgroup IA versus IB, respectively). While subgroups IA and II are distributed worldwide, subgroup IB is mainly found in Asia, and has only been reported elsewhere since the 1980s (Palukaitis & García-Arenal, 2003; Jacquemond, 2012).

The molecular epidemiology of CMV populations has been studied in depth in only two countries, Spain (Fraile et al., 1997; Bonnet et al., 2005) and the USA (California; Lin et al., 2003, 2004). Only subgroups IA and IB isolates were present in California, while all three subgroups were identified in Spain. Reassortants resulting from exchange of genomic RNAs from different subgroups represented a low proportion of the populations in both cases, and were generally found in only one sample in the above studies. Moreover, few of the possible reassortant types were recovered from naturally infected plants (Fraile et al., 1997; Lin et al., 2004; Bonnet et al., 2005). This is in contrast with experimental data showing that reassortants are viable, independently of the subgrouping of the donor strains, which suggests that reassortants are at a selective disadvantage in nature (Fraile et al., 1997). However, this assumption is not compatible with the hypothesis that the main advantage conferred by a multipartite genome is to facilitate adaptation to varying environmental conditions through, precisely, reassortment.

There are very few resistance traits against CMV in pepper, and most of them confer a quantitative resistance to the virus. Tunisian cultivars are local and mainly provided by cooperatives or propagated by the growers themselves. Very little is known about their genetic characteristics, but they all seem highly susceptible to the virus.

Characterization of the properties of CMV populations is essential in order to propose control methods for reducing the impact of outbreaks in a crop that has an important social and economic impact in Tunisia. A study of the genetic diversity of CMV infecting pepper in important open-field producing areas was initiated. The study also considered the ability of the virus to infect pepper harbouring a polygenic resistance to CMV movement, in order to evaluate the potential usefulness of such resistances for controlling CMV in Tunisian pepper.

Materials and methods

Field isolates and reference strains

Fruits of pepper plants showing viral disease symptoms were collected from 2008 to 2010 in northern and central Tunisia, the main growing areas of peppers (Fig. 1; Table 1). Samples were first analysed by DAS-ELISA (Clark & Adams, 1977) for the presence of CMV and PVY. More than 100 CMV-positive samples in 2008 and 2009 were used to inoculate a tobacco (cv. Xanthi) and a melon (cv. Charentais), the latter in order to eliminate PVY if present; tissues from both hosts were kept desiccated on calcium chloride. The collection was completed by samplings done in 2006 in northern Tunisia for which total RNA from the collected fruits were already available (Ben Khalifa et al., 2009), and by a few isolates collected at the end of the 1990s, under conditions not fully known, and kept since then as desiccated infected tissues. Finally, surveys were also done in 2010 and 2011 in species other than pepper in central and southern Tunisia.

Table 1. CMV field isolates
Year1990s2006200820092010a
Characteristics for samplesbGenotypedGenotypedCollectedELISA > 0GenotypedCollectedELISA > 0GenotypedCollectedELISA > 0Genotyped
RegionLocation (abbreviation)           
  1. a

    Surveys were done late in 2010 when most of the crops in North Tunisia were already pulled up.

  2. b

    Collected: number of samples collected in one location; ELISA > 0: number of samples that tested positive by ELISA; Genotyped: number of samples for which the haplotypic composition was determined. Figures in column ELISA > 0 correspond to the number of collected samples and the percentage of infected samples is in brackets.

  3. c

    Region (either Korba or Kairouan) where these three isolates were collected was unknown.

BizerteGhar el Melh (GM) 52018 (90)10156 (40)6   
Sounine (S) 11212 (100)463 (50)3   
Ras el Jebel (RJ)  2422 (92)597 (78)754 (80)4
Metline (M)  169 (56)2      
Bejou (B)  4227 (64)998 (89)8   
Douar Hmouda (DH)  1310 (77)6      
Oued Saadane (OSA) 21912 (63)198 (89)8   
El Garia (EG)  72 (29)1101 (10)1   
Aousja (AOU) 14821 (44)71913 (68)13   
Thouar (T)  3016 (53)61710 (59)6   
Nouvelle Utique (NU) 3217 (33)31010 (100)10   
Zaounine (Z)  4820 (42)71612 (75)12   
Kalaat el Andalous (KAA)  174 (24)260    
Cap BonDhhari (DHH)  4949 (100)183022 (73)21   
Bou Charray (BC) 85252 (100)193126 (84)26   
Korba (KO)4          
KairouanKairouan (KA)4       2923 (79)22
UnknowncUnknown (UK)3          
Total 1120418281100187126121342726
Figure 1.

Sampling sites in the regions of Bizerte (northern Tunisia), Cap Bon (northeastern Tunisia), Kairouan (central Tunisia) and Gabes (southern Tunisia).

In addition, three CMV strains for which infectious cDNA clones are available in the laboratory were used as reference strains: I17F was isolated from tomato in France in 1975 (accession numbers HE793683, HE793684, Y18137); R was isolated from Ranunculus asiaticus in France in 1969 (accession numbers HE793685, HE7936686, Y18138); Vir was isolated from pepper in northern Italy in 1988 (accession numbers HE962478, HE962479, HE962480). Tobacco plants were infected with infectious cDNA clones by bombardment, from which viral RNAs were purified and used as controls in all RT-PCR experiments.

Genotyping CMV isolates

Total RNA was extracted using TriReagent® (Molecular Research Center, Inc.), according to the manufacturer's instructions, directly from material collected in the open field (2008, 2010 and 2011), or from hosts infected experimentally (2008 and 2009), c. 2 weeks after inoculation. In the latter case, total RNA was extracted from a systemically infected tobacco leaf if CMV was the sole virus detected by ELISA in a sample, or a melon leaf if PVY was also initially present.

cDNAs corresponding to parts of the three genomic segments were amplified by reverse transcription-PCR (RT-PCR). For RNA1, a region in the 5′ half of the genomic segment was sequenced, because from known sequences, no restriction sites could be identified for differentiating the different subgroups. A region encompassing the end of ORF 2a, all of ORF 2b and the beginning of the 3′-UTR, which corresponds to the most variable region of the viral genome flanked by more conserved sequences, was chosen for RNA2. Two regions were considered for bicistronic RNA3, covering either the MP ORF or the CP ORF and the beginning of the 3′-UTR. For the MP region, the RT-PCR products were sequenced because again, from known sequences, no restriction enzymes could be used to distinguish the three subgroups. Primers were designed following alignment of a large set of sequences available in the gene banks, and are listed in Table 2. Primers polyvalent for all subgroups could be designed for RNAs 1 and 3. In addition, primers specific for Tunisian subgroup IA/IB isolates (pair G1.2 in Table 2) were designed in the third part of ORF1a for RFLP mapping. Two primer pairs, corresponding to either subgroups IA and IB or subgroup II, were designed for RNA2. It should be noted that primers designed for subgroups IA/IB strains gave faint amplification from subgroup II isolates, while the converse was not true. Specificity of amplification from RNA2 with the two primer pairs was thus indicative of the subgrouping of this RNA. RT-PCR was carried out in a single tube reaction, using AMV reverse transcriptase and Taq DNA polymerase (Promega), according to the manufacturer's instructions. PCR was performed with an annealing temperature corresponding to the lower Tm of the primer pair and elongation duration of 1 min per kb, for 35 cycles.

Table 2. Primers used for RT-PCR amplification of CMV regions
NamePolarityPrimer sequence (5′ to 3′)Target RNABinding site
  1. a

    Nucleotide positions in I17F CMV.

  2. b

    Nucleotide positions in R CMV.

G1.1-F+GTTTTATTTACAAGAGCGRNA11–18a
G1.1-RGATGATATCACGTCCCATTCRNA11605–1621a
G1.2-F+GCTCAGACACGTTCCCCRNA11821–1840a
G1.2-RACAGTCGGACATTCATTAAGRNA12406–2425a
G2-I-F+GGCTGCTTTAATGTTAGGCGRNA22257–2276a
G2-I-RGGATGGACAACCCGTTCACCRNA22844–2863a
G2-II-F+CCGCCCAAGTTTTGGGCRNA22257–2273b
G2-II-RCCTTTGGGGCATGACTGRNA22863–2879b
G3-MP-F+CAGTCCTCAGCGGCTACGTCRNA3157–176a
G3-MP-RAAGACCGTTAACCACCTGCGRNA3938–957a
G3-CP-F+CGTCCAACTATTAACCACCCRNA31405–1424a
G3-CP-RCTGGATGGACAACCCGTTCRNA32011–2030a

Genotyping the RT-PCR products by restriction fragment length polymorphism (RFLP) was done as presented in Table 3. For RT-PCR products derived from RNA2, an RsaI site was expected at different positions in subgroups IA and IB isolates. For subgroup IB, this site was identified in most, but not all, sequences retrieved from the databanks, and was not present in the region amplified from the Tunisian isolates. Identification of IB isolates was thus essentially based on no digestion with RsaI, while subgroup IA isolates were efficiently cut by RsaI (other position), but also StuI and HpaII. For products derived from RNA3, only one site, located near the 3′ end of the RT-PCR products, was expected in IA isolates and needed delicate 4% polyacrylamide gels to be implemented. A SalI site was present in most but not all IB sequences and was found in 80% of Tunisian IB sequences. As Tunisian isolates proved to be close to the reference strains, an AccIII site cutting IB, but not sIA, sequences was also considered for a confident genotyping by RFLP. Finally, as expected, a NruI site was specific for RT-PCR products derived from subgroup II isolates. All digestions were performed using 1/25–3/25 of the amplified RT-PCR products, with one unit of restriction enzymes from Promega and following the manufacturer's instructions. Digestion products were analysed on 1·5% agarose gels or 4% polyacrylamide gels. Sequencing was performed by Beckman Coulter Genomics or GATC Biotech.

Table 3. Restriction patterns of RT-PCR fragments derived from CMV RNAs of the different subgroups
Target RNAPrimer pairbRestriction enzymeLengtha of the RT-PCR products after digestion (bp)
Subgroup IA (I17F)Subgroup IB (Vir)Subgroup II (R)
  1. a

    Referring to reference strains I17F, Vir and R.

  2. b

    Primers are listed in Table 

  3. c

    No amplification (RNA1) or only a faint amplification (RNA2) occurred, because the primer pairs were designed for subgroups IA/IB strains.

  4. d

    The RsaI site is located at different positions in sequences deriving from subgroups IA and IB RNA2 and is present in most, but not all, IB RNA2.

  5. e

    The SalI site is present in most, but not all, IB RNA3 sequences.

RNA1G1.2-F–G1.2-RNcoI605405, 200c
XhoI454, 152605c
RNA2G2.I-F–G2.I-RRsaId113, 494612 (or ≅ 358, 234)dc
StuI406, 201612c
HpaII464, 143612c
RNA3G3-CP-F–G3-CP-RSalI626461, 165 (or ≅ 630)e461, 180
AclI596, 30626641
NruI626626559, 82
AccIII626326, 300326, 315

Data analysis

Sequences were aligned using ClustalW (Thompson et al., 1994). Phylogenetic trees were constructed using the neighbour-joining (NJ) method implemented in mega 4 software (Tamura et al., 2007) and the robustness of nodes was evaluated with 1000 bootstrap replicates.

Ten haplotypes could be defined on the basis of RFLP patterns and/or sequence of part of each genomic segment (from haplotype A to haplotype J). The Ewens–Watterson neutrality test (Watterson, 1978) was applied, using arlequin v. 3.5 (Excoffier & Lischer, 2010), to compare the observed haplotype distribution and that expected in a neutral evolution model.

Pairwise FST values (Weir & Cockerham, 1984) between pairs of populations were computed using arlequin v. 3.5, and their significance was estimated by 1000 random permutations of individuals among populations. The pattern of isolation by distance (IBD) was tested through the correlation between genetic distance (measured as FST/(1−FST)) and geographic distance expressed as the natural logarithm of the distance (Rousset, 1997) for each pair of CMV populations. Significance of the correlation was assessed by a Mantel test (10 000 permutations) using the software xlstat 2012 (Addinsoft SARL). The hypothesis of IBD was explored for the whole data set and for the region of Bizerte only.

Hierarchical structure was assessed by two amova models (analysis of molecular variance; Excoffier et al., 1992) using arlequin v. 3.5. Three hierarchical levels were considered for each model: within population, among populations of the same region (or of the same year) and among regions (or years).

Phenotyping CMV isolates

The resistant pepper cultivar Milord was obtained by recurrent breeding and presents a partial polygenic resistance to long-distance movement (Caranta et al., 2002). The susceptible cultivar Yolo Wonder was used as control. The resistance test was carried out as described by Caranta et al. (2002). Briefly, plants grown until they developed 6 and 7 leaves were cut under the fifth leaf and inoculated with the virus 4–5 days later on the third leaf. Inoculums consisted of crude extracts of infected tobacco plants. Development of symptoms in the third and fourth axillary shoots was surveyed for 7 weeks following inoculation. Presence (or absence) of the virus in the inoculated leaves was confirmed by ELISA on day 7, and in the developed axillary shoots 3 and 6 or 7 weeks after inoculation.

Results

Presence of CMV in Tunisian pepper crops

Four hundred and eighteen and 187 fruit samples were collected from plants with symptoms in 2008 and 2009, respectively (Table 1). In 2008, 281 samples tested positive for CMV by ELISA, of which 108 also tested positive for PVY, while 35 tested positive for PVY alone. In 2009, 126 samples tested positive for CMV and 44 for both viruses, while 16 tested positive only for PVY alone. Attempts to detect other viruses were not done. CMV was recovered following experimental inoculation of tobacco and melon plants, from 100 and 121 samples in 2008 and 2009, respectively. An additional survey was done in 2010, mainly in the central region, resulting in detection of CMV in 27 out of the 34 samples collected (Table 1).

Experimental approach

Ideally, unambiguous RFLP genotyping of viral RNAs should be based on restriction enzymes that specifically cut the sequence derived from each subgroup, or that cut each subgroup with different patterns. In addition, it is essential to take into account two possibly confounding features: intragroup sequence variation occurring at restriction sites and incomplete digestion of the RT-PCR products, which could be misinterpreted as indicating mixed infections. This study encountered two cases where a restriction site that was expected for a certain RT-PCR product was absent in a significant proportion of the Tunisian sequences. Sequence data obtained for most of the isolates collected in 2009 made it possible to circumvent these difficulties, by identifying additional enzymes differentiating the Tunisian subgroup sequences, or designing new primers specific for the Tunisian sequences. Confident genotyping by RFLP was achieved by using three and four restrictions enzymes for fragments deriving from RNA2 and the CP region, respectively. Mixed infections were clearly identified when analyses for all restriction enzymes were congruent.

Most of the RT-PCR products generated from the 5′ half of RNA1 were sequenced, except samples collected in 2010. When RT-PCR or sequencing failed, and also for the 2010 samples, a second primer pair was used that allowed genotyping of Tunisian subgroups IA/IB sequences by RFLP, using two subgroup-specific restriction enzymes. For RNAs 2 and 3, all samples were analysed by RT-PCR-RFLP. In addition, most of the products amplified from RNA2 and the CP region of the samples collected in 2009 were sequenced. RT-PCR products from the RNAs of non-reassortant isolates (see below), RNAs 2 and 3 assigned to subgroup II, and the RT-PCR products derived from RNA2 of a few randomly chosen 2008 samples, were also sequenced. Finally, the RT-PCR products for which sequencing was unsuccessful were cloned. Clones were characterized by RFLP and one clone per restriction profile was sequenced. The whole data set is presented in Table S1. Sequences were deposited at the EMBL nucleotide bank database and received the following accession numbers: RNA1: HE971277 to HE971473; RNA2: HE971474 to HE971599; RNA3: HE971600 to HE971708.

For each isolate, a haplotype was defined by the subgroup to which each of its genomic RNAs belonged. For example, the haplotype IA-IA-IA corresponds to an isolate whose three genomic RNAs belong to subgroup IA.

For all samples collected in 2008 and 2009 (but not in 2006, 2010 and 2011) a multiplication host was used in order to keep the isolates alive for biological studies, but this required validation. Total RNA was extracted from 100 fruits collected in 2008 from which the virus could be recovered on a multiplication host. A fragment corresponding to the CP region of RNA3 was successfully produced for 91 samples. These samples were further analysed by RFLP for their RNAs 1 and 2 (64 and 53 successful amplifications, respectively). Genotyping of pepper fruit CMVs showed that genotyping with RNA preparations from a multiplication host (tobacco or melon) did not modify the genetic composition of the original CMV isolates.

Genetic composition of Tunisian isolates

This study was initiated by the characterization of CMV isolates collected in 2008. Most isolates were reassortants, composed of RNA1 of subgroup IB, RNA2 of subgroup IA, and RNA3 of either IA or IB (haplotypes D and E in Table 4). This subgrouping was further confirmed using other primer pairs for each RNA and sequencing, on a set of 10 random isolates (not shown). Moreover, this unexpected result was confirmed for the samples collected in 2009 (Table 4). Total RNAs previously purified from fruits of naturally infected peppers collected in 2006 for the study of Tunisian PVY populations (Ben Khalifa et al., 2009) were then analysed. The situation was similar for samples collected in 2006. Finally, older isolates that were collected at the end of the 1990s were recovered from desiccated infected tissue. Seven of these 11 isolates also proved to be reassortants corresponding to haplotype D (Table 4).

Table 4. Genetic characterization of CMV isolates
Infection statusSingle infectionsMixed infections
SubgroupIA/IBIIIA/IB/IIIA/IBIA/IB/II
HaplotypeaRegionABDECFIB+DB+ED+EC+DC+EC+JE+FG+HNDb
  1. a

    Each haplotype was defined by the subgroup to which each of its genomic RNAs belonged. A: IA-IA-IA; B: IB-IB-IB; C: II-II-II; D: IB-IA-IA; E: IB-IA-IB; F: IB-IA-II; G: IB-II-IA; H: IB-II-II; I: II-IA-IB; J: II-II-IB.

  2. b

    ND: haplotypes not determined (samples in which five genomic segments were characterized).

  3. c

    Region (either Korba or Kairouan) where these three isolates were collected was unknown.

1996–1998Cap Bon     4          
Kairouan  4             
Unknownc  3             
2006Bizerte  62       2   2
Cap Bon  7       1     
2008Bizerte1236201    3      
Cap Bon  3321        1  
2009Bizerte 22231   139  1 14
Cap Bon  433 1          
2010Bizerte    3 1         
Kairouan  22             
Total 14176585511312121116

In general, haplotype D (IB-IA-IA) was prevalent, representing 62–100% of the subgroup IA/IB reassortant population, depending on the year and the region, and was present in all regions. On the other hand, haplotype E (IB-IA-IB) was abundant in the region of Bizerte, only occasionally found in the region of Cap Bon, and not detected in Central Tunisia, although sampling was more limited here than in the other regions.

Non-reassortant isolates were rare, as only four subgroup IB isolates and one of subgroup IA were identified in singly infected plants. Mixed infections within subgroups IA and IB were observed, resulting in the presence of the two major haplotypes D and E in 12 samples (four genomic segments identified), or one reassortant (either haplotype D or haplotype E) and a subgroup IB isolate in four samples (Table 4). Globally, subgroup IA/IB reassortants were present in 250 samples, representing 90% of the total.

Subgroup II isolates were detected only in northern Tunisia (regions of Bizerte and Cap Bon; Fig. 1), in nine of the 16 locations surveyed, and in 23 samples (8·3% of the population), for all years studied (Table 4). In most cases (18 samples), they occurred in mixed infection with subgroup IA/IB components. Six isolates were reassortants between subgroups IA/IB and II, because three genomic segments were identified (haplotypes F and I in Table 4). The presence of two distinct haplotypes was proposed as the simplest explanation for three samples in which six genomic segments were identified. For three other samples, sequencing the products derived from RNAs 1 and 2 similarly suggested the presence of two haplotypes, either a subgroup II isolate and a new subgroup I/II reassortant (haplotype J), two I/II reassortants (haplotypes G and H) or a mixture of reassortants already characterized in single infections. Finally, the situation appeared more complex for six samples for which five genomic segments were identified and for which the haplotype composition could not be assigned simply (Table 4).

In 2008 and 2009, the proportion of PVY-infected plants differed according to the localities. Particularly, the proportion of plants co-infected with PVY and CMV in pepper crops in the region of Cap Bon was significantly higher than in the Bizerte region in 2009, or globally in 2008. There was no significant correlation between the frequency of co-infection with PVY and particular CMV haplotypes (not shown); however, sufficient numbers of samples for these analyses were available only for haplotypes D and E in the region of Bizerte in 2008 and 2009.

Phylogeny of Tunisian CMV isolates

For RNA1, 197 sequences of a 906 nt fragment in the 5′ part of the molecule were used to construct a phylogenetic tree (Fig. 2a). Of these, two subgroup II sequences showed a three-nucleotide deletion, resulting in the loss of one amino acid (position 254 in R-CMV). Apart for the presence of the three subgroups, the neighbour-joining (NJ) tree revealed a general low diversity (average diversity value of 0·009 within cluster IB; Fig. 2a). Subclusters of two or a few sequences were observed within subgroup IB, but the bootstrap values were often lower than 65 and the branches remained very short.

Figure 2.

Neighbour-joining phylogenetic trees of the nucleotide sequence of the RT-PCR products amplified from RNA1 (a), RNA2 (b), the CP region of RNA3 (c), and of the sequences resulting from concatenation of the three previous ones for each isolate (d). Robustness of nodes was evaluated with 1000 bootstrap replicates. Bootstrap values lower than 65% are not indicated. Scale bars indicate branch lengths in substitutions per nucleotide. Reference strains are in bold and italics. Non-reassortant isolates are underlined in the concatenated tree. Name and composition of the different haplotypes are indicated in the right of part (d). The two recombinant isolates in the CP region are indicated by arrows.

For RNA2, sequences of 539, 544 and 517 nt, corresponding to the amplified region of RNA2 of subgroups IA (112 sequences), IB (three sequences) and II (11 sequences), respectively, were similarly exploited (Fig. 2b). Within subgroup IA, some length variations were observed. For two isolates, a deletion of three nucleotides in the coding regions (positions 2626–2628 in I17F-CMV) resulted in the deletion of amino acid 848 of protein 2a and amino acid 70 of protein 2b (numbering as I17F-CMV). Deletions of 1–3 nucleotides at three sites in the 3′-UTR, alone or in combination, were observed in 31 isolates. Despite these differences, the NJ tree constructed using the entire data set showed, as for RNA1, a low genetic diversity (average diversity value of 0·011 within cluster IA; Fig. 2b).

The sequences analysed for the CP gene and the beginning of the 3′-UTR of RNA3 were 530, 531 and 541 nt long, for subgroups IA, IB and II, respectively, and no gaps were observed. The NJ tree revealed a high homogeneity within subgroup IA, but a higher diversity within subgroup IB (Fig. 2c). Particularly, three isolates that clustered separately (BC3.3/09, DHH1.1/09 and DHH1.2/09) accounted for most of the genetic diversity within this subgroup (average diversity of 0·004 when they were excluded vs 0·013 when included). In addition, two other sequences also clustered separately, either closer to subgroup IA (T1.1/09-1) or within subgroup IB (T1.1/09-2). These two sequences were present in isolate T1.1/09, and were sequenced following cloning of the RT-PCR product. They proved to be exact converse recombinants between a IA and a IB sequence. The crossover was in a stretch of identity in the middle of the CP gene (positions 1555–1628 in I17F-RNA3 and 1558–1631 in Vir-RNA3). Recombination probably did not occur during in vitro amplification, because several clones of the RT-PCR product were analysed and all were recombinant. In addition, this crossover site corresponds to the dominant hotspot for recombination in RNA3 that was observed between a subgroup IA and a subgroup II isolate under experimental conditions (Turturo et al., 2008).

The sequence of a c. 750 nt-long region in the MP ORF was also established for 16 randomly chosen isolates collected in 2009 that belonged to either haplotype D or E. In all cases, the sequence of the MP region had the same subgrouping as the CP region (data not shown); thus no MP/CP recombination events were identified.

The sequences of the three genomic RNAs used above were concatenated for the 98 isolates for which all three sequences had been obtained, for further visualization of the reassortment and recombination events (Fig. 2d). A first clustering differentiated subgroups IA/IB and II isolates, with an intermediate clade corresponding to haplotype F (reassortant IB-IA-II). A second clustering differentiated subgroup IA isolates (one isolate and the I17F reference strain) from all other isolates, whatever their genetic characteristics. The 58 haplotype D isolates (IB-IA-IA) clustered apart with a high bootstrap value. All isolates closer to subgroup IB (haplotypes IB-IB-IB and IB-IA-IB) showed the same clustering as previously observed for their RNA3 alone, indicating that this RNA accounted for most of the diversity.

Genetic structure of CMV populations in pepper fields

Haplotypic richness was established for the 16 localities (considered as 16 populations) where sampling was done between 2006 and 2010. When isolates were grouped by year of sampling, an amova analysis showed that the year did not contribute to the total variation (data not shown), making it valid to pool together isolates collected in different years for genetic structure analysis of the CMV populations. The number of distinct haplotypes per CMV population ranges from 1 to 6, and the haplotypic richness from 0·045 to 0·5 (or 1·0 when considering populations containing only two isolates; Table 5). Comparison of the observed and expected haplotypic distributions with the Ewens–Watterson test did not reveal any significant departure from neutrality for each population because all probabilities were higher than 0·05 (Table 5). Further analysis of the structure of these populations could thus be done, because no important selective or demographic biases were expected. However, populations represented by only two samples in Table 5 (EG, KAA and M) were discarded for measures of genetic differentiation between populations.

Table 5. Haplotypic richness, observed and expected haplotypic distributions under the infinite-allele model and probability of the Ewens–Watterson test of neutrality
PopulationsNumber of samplesNumber of haplotypesHaplotypic richnessaObs.bExp.b P c
  1. a

    Haplotypic richness: number of haplotypes/number of samples.

  2. b

    Obs. and Exp.: observed and expected (under the neutral model) haplotype distributions.

  3. c

    P: the probability associated with the Ewens–Watterson test. > 0·05 indicates that the observed and expected haplotype distributions are not significantly different and reveal no departure from neutrality.

  4. d

    nc: not calculable because all haplotypes were different.

  5. e

    nc: not calculable because there was a single haplotype in the population.

AOU2740·1480·5010·4700·651
B2020·1000·5200·7330·124
BC5540·0730·7670·5330·878
DH630·5000·3890·4300·594
DHH3920·0510·8580·7690·638
EG221·000ncdncdncd
GM2130·1430·8230·5571·000
KA2210·045ncencence
KAA221·000ncdncdncd
M210·500ncencence
NU1830·1670·6300·5510·682
OSA1220·1670·5000·6860·056
RJ1750·2940·2530·3370·130
S810·125ncencence
T1460·4290·2650·2560·711
Z2050·2500·3800·3570·680

Pairwise genetic differentiation between the 13 populations was estimated by the Wright's diversification index (FST). The matrix of pairwise FST values showed that three populations, BC, DHH (from Cap Bon) and KA (from Kairouan), were highly differentiated from the populations from the region of Bizerte, except for GM and S (Fig. 3). However, among the Bizerte populations, 20 FST values were significant (< 0·05) among the 45 possible pairwise comparisons ((n2n)/2). Therefore, the pattern of isolation by distance was tested both for the whole data set (13 populations) and for isolates from Bizerte only (10 populations). Distance matrices of genetic differentiation (measured as FST/(1−FST)) and of geographic distance (expressed as the natural logarithm of the distance in km) were compared by a Mantel test. A significant correlation was found either with all populations (= 0·398 and = 0·0003; Fig. 4) or with populations in the region of Bizerte separately (= 0·378 and = 0·0104, not shown). The high genetic differentiation and the pattern of isolation by distance suggest low dissemination of the virus between the different localities.

Figure 3.

Matrix of pairwise FST between the 13 CMV populations using haplotype distributions. Significant FST are indicated by a white star (< 0·05).

Figure 4.

Isolation by distance pattern between genetic differentiation (using haplotype distributions) and geographic distance for pairwise CMV populations. Pearson's coefficient of correlation (r) and Mantel test P-value (P) are indicated on the graph.

amovas were carried out to complete these results and determine what factor(s) mainly condition the observed genetic differentiation. Results for all years were considered and three groups (Bizerte, Cap Bon and Kairouan) were defined. Computing conventional F-statistics from haplotype frequencies revealed that the region contributed significantly to the total variation (20·0%; Table 6). This probably reflects the heterogeneity of distribution of haplotype E (IB-IA-IB), which was mainly found in the region of Bizerte. The different localities within each region contributed 17·4% of the variation, but the largest part of the variance was within-population (62·6%; Table 6). Because sampling was done in only six and two localities in 2006 and 2010, respectively, the 2008 and 2009 samples were also analysed separately (data not shown). Similar results were obtained, suggesting that temporal variation was unlikely to have occurred during this short period.

Table 6. Analysis of molecular variance when CMV populations are grouped by region
Source of variationd.f.Sum of squaresVariance components% variationF-statistics
Between region212·10·0620·0FCT = 0·20
Between populations in each region1011·40·0517·4FSC = 0·22
Within populations26648·40·1862·6FST = 0·37
Total27872·00·29  

CMV in crops other than pepper

In 2010, other crops were sampled in the vicinity of the pepper crop sampled at Ras el Jebel. These included eight mature snake melon and six marrow plants, all of which expressed strong mosaic symptoms, and 33 tomato plants. Watermelon mosaic virus (WMV, genus Potyvirus) and Cucurbit aphid-borne yellows virus (CABYV, genus Polerovirus) were detected in the cucurbits, but not CMV. Only one tomato plant among the 33 that were sampled was infected with CMV, and the isolate proved to be of the D haplotype. In 2011, a larger sampling was done in southern Tunisia (Gabes, location 18 in Figure 1) in snake melon crops and in volunteer cucurbits growing in either tomato or pepper crops. CMV was detected in 13 samples, and all genomic segments were successfully PCR amplified from 10 of them. Here also, all isolates corresponded to haplotype D (data not shown).

Ability of Tunisian CMV isolates to infect resistant peppers

Twenty-eight and 19 isolates collected in 2008 or 2009, respectively, and 10 isolates collected at the end of the 1990s were tested for their ability to infect peppers resistant to CMV movement. Isolates were chosen according to the genetic diversity within the two main reassortant populations. Non-reassortant IA, IB and II isolates were included when they had been saved alive. Isolates from each period were tested in three independent assays, which also included the subgroup IA strain I17F (which cannot infect the resistant cultivar) and the subgroup IB strain Vir (which overcomes the resistance). Presence of the virus in the inoculated leaves was determined by ELISA, and confirmed a successful infection in all cases (data not shown). All isolates except two successfully infected the resistant line: a subgroup II isolate and the unique isolate of haplotype A, that of I17F-CMV (Table 7). Isolates possessing virulence properties towards a polygenic resistance to CMV movement are thus already prevalent in Tunisian crops.

Table 7. Properties of Tunisian CMV isolates in susceptible and resistant pepper cultivars
YearaRegionHaplotypeNumber of isolates% infected plantsb
Yolo WonderMilord
  1. a

    Isolates collected at the same period were tested simultaneously.

  2. b

    Eight to 15 plants of the susceptible Yolo Wonder and resistant Milord cultivars were inoculated with each isolate. Results presented correspond to the mean value determined for all isolates possessing the same haplotype.

  3. c

    Region (either Korba or Kairouan) where these two isolates were collected was unknown.

  4. d

    The reference I17F (haplotype IA-IA-IA) and Vir (haplotype IB-IB-IB) strains were included in all experiments.

1996–1998KairouanIB-IA-IA481·298·4
KorbaIB-IA-II487·495·2
UnknowncIB-IA-IA275·0100·0
Reference strainsdIA-IA-IA170·00·0
IB-IB-IB190·0100·0
2008BizerteIA-IA-IA1100·010·0
IB-IB-IB2100·0100·0
II-II-II1100·00·0
IB-IA-IA1196·493·2
IB-IA-IB8100·096·9
Cap BonIB-IA-IA4100·098·8
IB-IA-IB1100·0100·0
Reference strainsdIA-IA-IA1100·00·0
IB-IB-IB1100·0100·0
2009BizerteIB-IA-IA380·0100·0
IB-IA-IB977·8100·0
Cap BonIB-IA-IA670·0100·0
IB-IA-IB1100·0100·0
Reference strainsdIA-IA-IA1100·00·0
IB-IB-IB1100·0100·0

Discussion

In this work, the genetic characterization of natural CMV isolates from peppers was primarily based on the digestion pattern of RT-PCR products produced from the three viral genomic RNAs. RT-PCR-RFLP is less expensive than sequencing, but can lead to misinterpretations. This study showed that it can give reliable results, provided that several enzymes are used for a given PCR product and that some sequence data are available to validate its use. Most of the Tunisian pepper CMV isolates were reassortants, composed of genomic RNAs belonging to different subgroups. It is believed that this is not due to an artefact for several reasons: (i) when applied to a mixture of the reference strains, the protocols allowed amplification of all genomic RNAs, and if similar RNAs 1 and 2 were found in most of the isolates, they differed for their third component; (ii) non-reassortant isolates were identified; (iii) sequencing nearly half of the RT-PCR products was fully concordant with the RFLP results, whatever the genomic RNA; and (iv) the extremely high prevalence of the reassortant haplotypes is not compatible with this being due to an artefact of amplification.

Based on sequence analysis of each part of the genome, the CMV population in Tunisia showed a low intra-subgroup genetic diversity. This is consistent with previous studies in Spain, California and Brazil (Eiras et al., 2004; Lin et al., 2004; Bonnet et al., 2005). Despite this low molecular diversity, a significant genetic differentiation was observed between CMV populations in terms of haplotype distribution. At least part of the differentiation could be attributed to the geographic distance separating populations, as shown by the significant IBD at both spatial scales tested here. Accordingly, amovas showed that the genetic variation (haplotype frequencies) was better explained by regions than by years. This observation is probably due to the abundance of haplotype E which seemed to vary along a negative north–south geographic gradient. However, because sampling was not done the same year in all regions, a more general survey would be required to confirm this tendency. Also, it should be noted that while samplings in the regions of Bizerte and Cap Bon were similar in 2008 and 2009 (total sampling and proportion of samples in the two regions), the relative proportions of the two major haplotypes D and E indicated a decrease of haplotype D and an increase of haplotype E. The difference was at the limit of significance (Fisher exact test, = 0·048), but this suggests that a replacement of populations could be occurring. Additional surveys in the coming years would be required to evaluate this tendency.

The lack of long-distance dissemination of CMV, and the resulting genetic differentiation between geographically close populations, may be related to the characteristics of its dissemination in the field. For CMV, which is transmitted by aphids in a non-persistent manner, it has been proposed that the average distance of dissemination is less than a few kilometres (Sacristán et al., 2004). Moreover, genetic drift due to strong bottlenecks during both CMV movement within the plant (Ali & Roossinck, 2010) and aphid transmission (Ali et al., 2006; Betancourt et al., 2008) could also favour genetic differentiation between CMV populations. Similar conclusions were drawn for PVY populations in peppers in Tunisia (Ben Khalifa et al., 2009), a virus submitted to similar strong bottlenecks (Moury et al., 2007; Fabre et al., 2012).

Mixed infections are the prerequisite for reassortment or true recombination in a genomic RNA. Mixed infections represented 10% of the samples, less than the 16% previously observed in CMV populations in several crop species over several years in Spain (Fraile et al., 1997). Recombination in coding regions has only rarely been observed in natural CMV isolates, and is thought to have a fitness cost (Bonnet et al., 2005); it is thus not surprising that recombination within an ORF was observed in only one sample. On the other hand, recombination between IA and IB isolates by the exchange of ORFs in the bicistronic RNA3 is more frequent in Spain, where both IA-IB and IB-IA recombinants were observed, and the frequency of the IB-IA type increased considerably from 1996 to 2002, when it reached 81% of the recombinants and nearly 14% of the entire population (Bonnet et al., 2005). It is noteworthy that this recombinant was found mainly in melon and in weeds, where it represented 45% of the CMV population (Sacristán et al., 2004). In the present study, in the samples which were sequenced within both the MP and CP ORFs, IA-IB or IB-IA recombinants were not observed; considering the number of sequences obtained, if IA-IB or IB-IA recombinants were present, they would represent <6% of the population.

In Spain, reassortants between subgroups IA and IB represented 4% of the population (Fraile et al., 1997). Only four of the six possible reassortants were observed, suggesting that some combinations were not favoured. The two major reassortants reported in Spain (62·5% of the total reassortants) correspond to the major haplotypes in Tunisian pepper described here, D and E. However, these reassortants did not establish in Spain, because in a subsequent study their prevalence was low (Bonnet et al., 2005). A single subgroup IA/IB reassortant was also described in California among 81 isolates collected on cucurbits (Lin et al., 2004). The uniqueness of the Tunisian CMV population is that the two predominant subgroup IA/IB reassortants were not only well established in pepper crops, but they were also the dominant haplotypes, and were much more abundant than non-reassortant IB and IA isolates, which represented 2·88 and 0·36% of the total, respectively. Although surveys were not carried out in all pepper growing areas (and particularly not in northwestern Tunisia), as well as in crops other than pepper, a broad dissemination of at least reassortant IB-IA-IA is suggested by its prevalence in cucurbit plants in southern Tunisia.

Worldwide, reassortment between subgroups IA/IB and II is less frequent, which may be due to subgroups IA/IB and II only occurring together in regions with a temperate climate, because subgroup II strains are sensitive to high temperatures. Three different reassortants between IA/IB and II isolates have been observed in Spain, Japan and China (Sacristán et al., 2004; Chen et al., 2007; Maoka et al., 2010). In Tunisia, these reassortants between subgroups IA/IB and subgroup II were not found, but five other ones were found: most were present in just one location, whereas haplotype IB-IA-II was found in six populations in 1996, 2008 and 2009. Haplotype IB-IA-II could correspond to a secondary RNA3 reassortment between one of the two major haplotypes (IB-IA-IA/IB) and a subgroup II isolate. The results thus confirm the overall rarity of reassortants between subgroups IA/IB and subgroup II, which however does not preclude their emergence, at least locally.

IB strains are mainly present in Asia, and it is generally thought that they were introduced recently into other continents in contaminated plant materials. Their introduction in southern Europe was associated with severe CMV outbreaks in tomato crops in the mid-1980s. In Italy, IB strains represented c. 50% of the CMV population at the end of the 1990s (Gallitelli, 2000). In Spain, IB strains represented 39% of the population in the middle of the 1990s (Fraile et al., 1997). They were present essentially in tomato crops in eastern Spain, were genetically very close to the Italian isolates characterized during this period, and then apparently disappeared (Bonnet et al., 2005). However, similar isolates reappeared in tomato crops in northeastern Spain at the beginning of the 2000s (Aramburu et al., 2007). Three European IB strains have been fully sequenced: Vir from northern Italy, Tfn from southern Italy, and PI.1 from Spain. Phylogenetic studies showed that Vir was closer to isolate IA from Indonesia, while Tfn and PI.1 were closer to This suggests a common origin for IB strains in southern Italy and Spain, but at least two introductions from Asia. If it is considered that the Tunisian isolates are similar to Vir, and that IB components were already present in CMV isolates collected in 1996 in Tunisia, it is tempting to propose that IB strains were introduced into Tunisia at about the same time as into northern Italy, or shortly thereafter. This is consistent with the import of tomato and pepper plants into Tunisia from Italy during this period, the local production in nurseries being insufficient.

A further point concerns the origin of the two major subgroup IA/IB reassortants in Tunisia. In a first scenario, the reassortants would have been directly introduced into the country via imported plant materials. In a second scenario, the reassortants would have emerged in Tunisian crops, in plants infected by subgroups IA and IB isolates, following introduction of IB isolates. The available data do not allow discrimination between these possibilities.

The most remarkable feature of the results is the predominance of two reassortant haplotypes, a situation that had not been described previously for CMV. This suggests that either the local pepper genotypes or some ecological feature exerts selection pressure in favour of these reassortants. In this regard, the observation that all the CMV isolates of IB-X-X haplotypes overcame a polygenic resistance to CMV movement in pepper that is effective against non-reassortant IA and II isolates may be significant. This resistance has a complex determinism, as seven regions in the pepper genome are involved (Caranta et al., 2002), and it is thus distinct from the dominant Cmr1 locus studied by Kang et al. (2012). Although the presence of any or all of these determinants in Tunisian pepper cultivars is not known, introgression of resistance factors against CMV cannot be ruled out, and if present may have contributed to selection of resistance-breaking isolates. In any case, the results show that deployment of this resistance in Tunisian cultivars would not be useful. Current research aims to provide further information on the potential selection pressure exerted by pepper genotypes on complex CMV populations.

Finally, it is noteworthy that the Tunisian reassortants, as well as the Spanish recombinant, are examples of successful emergence of novel CMV haplotypes. Taking into account the selective pressures and strong bottlenecks within and between plants exerted on new viral genotypes, as written by Elena et al. (2011), ‘the emergent viruses represent just the few lucky cases that have been able to surmount all these limitations’.

Acknowledgements

This study was supported by a Franco-Tunisian Utique program (PHC-Utique 10G 0905) and the Plant Health Department of INRA. The authors acknowledge the Tunisian Ministry of Agriculture for help with surveys.

Ancillary