Adaptation of flea beetles to Brassicaceae: host plant associations and geographic distribution of Psylliodes Latreille and Phyllotreta Chevrolat (Coleoptera, Chrysomelidae)

Abstract The cosmopolitan flea beetle genera Phyllotreta and Psylliodes (Galerucinae, Alticini) are mainly associated with host plants in the family Brassicaceae and include economically important pests of crucifer crops. In this review, the host plant associations and geographical distributions of known species in these genera are summarised from the literature, and their proposed phylogenetic relationships to other Alticini analysed from published molecular phylogenetic studies of Galerucinae. Almost all Phyllotreta species are specialised on Brassicaceae and related plant families in the order Brassicales, whereas Psylliodes species are associated with host plants in approximately 24 different plant families, and 50% are specialised to feed on Brassicaceae. The current knowledge on how Phyllotreta and Psylliodes are adapted to the characteristic chemical defence in Brassicaceae is reviewed. Based on our findings we postulate that Phyllotreta and Psylliodes colonised Brassicaceae independently from each other.


Introduction
Plant-feeding insects are often classified as specialists or generalists according to their food plant range. While generalist insect herbivores are able to feed on plants that belong to distantly related plant families, specialist insect herbivores feed selectively on one or a few closely related plant species (Schoonhoven et al. 2005). Many phytophagous insects, including numerous leaf beetle species, have a narrow food plant range (Jolivet and Hawkeswood 1995;Forister et al. 2015), which is at least partially determined by toxic and deterrent plant secondary metabolites. Plants produce more than 200,000 different secondary metabolites, and many of them are involved in defence against herbivores (Mithöfer and Boland 2012). The distribution of secondary metabolites in related plant species often correlates with the food plant range of specialised insect herbivores, which evolved strategies to avoid, tolerate, or detoxify these defence compounds (Heckel 2014). Such adaptations presumably played an important role in the species diversification of plant-feeding insects (Ehrlich and Raven 1964;Futuyma and Agrawal 2009), but the specific molecular mechanisms underlying host plant adaptation, and their role in insect ecology and speciation, are largely unknown.
Several genera in the family Chrysomelidae include species that are specialised to feed on plants in the family Brassicaceae (Table 1). In the subfamily Chrysomelinae, the genera Colaphellus, Entomoscelis, and Microtheca feed primarily on Brassicaceae (Jolivet and Petitpierre 1976b;Nielsen 1988), whereas the genus Phaedon is associated with several different plant families, e.g. Asteraceae, Brassicaceae, Scrophulariaceae, and Ranunculaceae (Table 1). In the subfamily Galerucinae, the flea beetle genera Phyllotreta, Psylliodes, Leptophysa, Caeporis, and Hemiglyptus utilise Brassicaceae as host plants (Furth 1979;Nielsen 1988;Jolivet 1991;Nadein 2010). In addition, many other polyphagous chrysomelid genera feed occasionally on this plant family. However, within Chrysomelidae, the genera Psylliodes and Phyllotreta comprise the highest number of crucifer specialists.
Glucosinolates are the characteristic secondary metabolites of Brassicaceae and other families in the order Brassicales (Agerbirk and Olsen 2012). Upon herbivory, glucosinolates are hydrolysed by β-thioglucosidase enzymes (myrosinases) to unstable aglucones, which can generate various hydrolysis products such as isothiocyanates, thiocyanates, and nitriles (Wittstock et al. 2016). Isothiocyanates, the most toxic glucosinolate hydrolysis products, are primarily reactive towards thiol-(-SH) and amino-(-NH 2 ) groups in peptides and proteins (Brown and Hampton 2011). Previous studies revealed that insects developed different strategies to overcome this plant defence (reviewed in Winde and Wittstock (2011) and Jeschke et al. (2016)). For example, Plutella xylostella larvae (Lepidoptera, Plutellidae) prevent glucosinolate breakdown by rapidly converting ingested glucosinolates to stable desulfo-glucosinolates (Ratzka et al. 2002), while Pieris rapae larvae (Lepidoptera, Pieridae) express a nitrile specifier protein (NSP) in their gut, which promotes the formation of less toxic nitriles instead of isothiocyanates (Wittstock et al. 2004). The evolution of NSP activity in Pierinae butterflies is regarded as an evolutionary key innovation that enabled a host shift from Fabales plants to the glucosinolate-containing Brassicales. As predicted by the coevolutionary 'escape and radiate' hypothesis, speciation rates were higher in the clade that colonised Brassicales plants compared to their sister taxon (Wheat et al. 2007;Edger et al. 2015). In contrast, the host shift of Ceutorhynchini weevils from the plant family Lamiaceae to Brassicaeae was not associated with a speciation rate shift (Letsch et al. 2018).
Glucosinolates and their hydrolysis products are well known to affect the behavior of crucifer-feeding Chrysomelidae (reviewed in Mitchell (1988Mitchell ( , 1994, and Nielsen (1988)). Volatile isothiocyanates, for example, attracted high numbers of Phyllotreta spp. and Psylliodes chrysocephala in field trapping experiments, indicating that isothiocyanates might play a role in host plant localisation (Görnitz 1956;Bartlet et al. 1992;Pivnick et al. 1992;Tóth et al. 2007). Glucosinolates, on the other hand, stimulated feeding of Phyllotreta spp., Ps. chrysocephala, Phaedon cochleariae, and Entomoscelis americana in laboratory experiments (Hicks 1974;Mitchell 1978;Nielsen 1978;Bartlet et al. 1994;Reifenrath and Müller 2008). Although these specialists are adapted to the glucosinolate-  , and B. rapa lines selected for high myrosinase activity displayed significantly less feeding damage (ca. 10%) than those with low enzyme activity . In contrast, studies with Ps. chrysocephala did not reveal a correlation between glucosinolate levels and feeding damage (Bartlet et al. 1996;Bartlet et al. 1999).
Here, we provide an overview on the host plants, diet breadth, and geographic distribution of known Phyllotreta and Psylliodes species, as well as their proposed relationships to other genera of Alticini. Diet breadth was classified according to Biondi (1996). Species feeding on one or two closely related botanical genera are considered as monophagous, species feeding on more plant genera of one or two closely related families are defined as oligophagous, and species feeding on many distantly related plant species are considered as polyphagous. For species with limited information on food plants, we did not specify the diet breadth. Data on the geographical distribution of the Palearctic Psylliodes and Phyllotreta species was primarily obtained from Döberl (2010) and is described according to Löbl and Smetana (2010). The zoogeographical regions are abbreviated as follows: Afrotropical Region (AFR), Australian Region (AUR), Nearctic Region (NAR), Neotropical Region (NTR), Oriental Region (ORR), Palearctic Region (PAR). In the second part of this review, we summarise the knowledge on the adaptations of Phyllotreta and Psylliodes spp. to the glucosinolatemyrosinase defence system and other defences in their host plants.
Of all Psylliodes species with known host plants, 50% are specialised on Brassicaceae, followed by 13% feeding on Poaceae, 10% on Solanaceae and 10% on Fagaceae (Fig. 1A). Previous surveys of host plant associations of Psylliodes spp. focused on specific countries or regions and thus included a much smaller total number of Psylliodes species (Furth 1983;Cox 1998;Döberl 2010;Baviera and Biondi 2015). Interestingly, host plant use often correlates with the proposed Psylliodes s. str. species groups, which indicates that presumably closely related Psylliodes species feed on closely related host plants (Suppl. material 2). For example, Psylliodes species in the chrysocephala and pyritosa groups are specialised to feed on Brassicaceae, while species in the luteola group are mainly associated with Fagaceae.
The genus Phyllotreta Chevrolat, 1836 comprises about 242 species and host plant information is available for 117 species (Suppl. material 3). Most Phyllotreta species are specialised on glucosinolate-containing plants in the order Brassicales (Fig. 1B). An analysis of the diet breadth of Phyllotreta species revealed that 31% are monophagous, 64% are oligophagous, and 5% are polyphagous. In Phyllotreta, 63% are specialised on Brassicaceae, whereas 18% feed on plants in more than one family in the order Brassicales ( Fig. 1B).Very few Phyllotreta species feed on plant families, which do not contain glucosinolates, for instance, Phyllotreta cruralis is specialised on Amaranthaceae.
Several Psylliodes and Phyllotreta species are of economic importance. The cabbage stem flea beetle, Ps. chrysocephala is a serious pest of winter oilseed rape in Northern Europe (Zimmer et al. 2014), whereas Phyllotreta striolata and Ph. cruciferae are oilseed rape pests in Canada where their damage causes losses of tens of millions of US dollars annually (Lamb 1989;Hill 2008;Knodel 2017). On the other hand, the Palearctic species Psylliodes chalcomera (feeding on Asteraceae) was introduced to North America in 1997 as a control agent for the invasive weed Carduus nutans (musk thistle), but it likely did not establish in the Nearctic region (Antonini et al. 2008).

Geographic distribution of Psylliodes and Phyllotreta flea beetles
The genus Psylliodes has a worldwide distribution (Biondi and D'Alessandro 2018). The highest number of species occurs in the Palearctic region (160 species, 145 endemic species), followed by the Oriental region (27 species, 19 endemic species), the Nearctic region (13 species, 4 endemic species), the Afrotropical region (13 species, 9 endemic species), the Neotropical region (8 species, 4 endemic species), and the Australian region (8 species, 7 endemic species; Suppl. material 1). A graphical overview of the species distribution is shown in Figure 2A; the host plant associations of all species and endemic species in each zoogeographical region are shown in Figure 2B. Some species are wide-spread in more than one zoogeographical region such as Ps. brettinghami (feeding on Solanaceae), which is found in the Australian, Oriental, and Palearctic regions, while others are strictly endemic to very limited areas, e.g. Ps. tarsata, which is only found on Madeira (Portugal). Psylliodes species that are endemic to the Palearctic region account for 83% of those associated with Brassicaceae. All other Brassicaceae-feeding species are found in other zoogeographical regions except for Australia (Fig. 2B).
The geographic distribution of the genus Phyllotreta shows the highest number of species in the Palearctic region (137 species, 118 endemic species) followed by the Afrotropical region (49 species, 39 endemic species), the Nearctic region (49 species, 40 endemic species), the Oriental region (25 species, 18 endemic species), the Neotropical Region (5 species, 3 endemic species), and the Australian Region (4 species, 3 endemic species; Suppl. material 3). The species distribution is shown in Figure 3A, and the host plant associations of all species and endemic species in each zoogeographical region are shown in Figure 3B. In general, a high percentage of endemic Phyllotreta species is found in all geographical regions (≥ 60%) with highest values in the Palearctic, Afrotropical, and Nearctic regions (≥ 80%). In some areas, especially in the Nearctic region, several species of Phyllotreta are not native and have been introduced from other regions (Milliron 1953;Smith 1985). Most species feeding on Brassicaceae are found in the Palearctic and Nearctic regions. The host plants of a large proportion of the species endemic to the Afrotropical, Australian, and Neotropical regions are unknown ( Fig. 3B; Suppl. material 3).

Phylogenetic relationships of Psylliodes and Phyllotreta to other Alticini
The most comprehensive phylogenetic analyses of the subfamily Galerucinae sensu lato are those of Ge et al. (2011Ge et al. ( , 2012 and Nie et al. (2018), which included about 80 and 70 genera of Alticini (including problematic genera), respectively. Ge et al. (2011Ge et al. ( , 2012 used two mitochondrial (16S rRNA and cytochrome oxidase (cox) 1) and two nuclear genes (18S and 28S rRNA) to infer phylogenetic relationships, while Nie et al. (2018) used the mitochondrial genome and nuclear rRNA genes. In these analyses, Psylliodes and Phyllotreta were never retrieved as sister genera, but instead clustered in distinct clades with other Alticini as summarised in Table 2. All three studies suggest a close phylogenetic relationship of Psylliodes to Chaetocnema and Crepidodera (see Table 2 for Bayesian posterior probability values and/or Maximum Likelihood bootstrap support values). Surprisingly, two different Crepidodera species included in  the analysis of Nie et al. (2018) were not monophyletic. Crepidodera pluta clustered in the Chaetocnema group with Psylliodes, while the second Crepidodera sp. clustered together with two Phyllotreta species in a distant clade. However, the proposed relationships of Phyllotreta to other Alticini differ among the studies, and are usually less supported than those suggested for Psylliodes. None of the genera with proposed close phylogenetic relationships to Phyllotreta and Psylliodes are associated with Brassicaceae plants ( Table 2).

Adaptations of crucifer-feeding flea beetles to chemical plant defences
An unexpected observation revealed that Ph. striolata adults emit low amounts of toxic isothiocyanates, which are derived from glucosinolates that are stored at high concentrations of up to 50 µmol/g fresh weight (ca. 2% of the body weight) in adults (Beran 2011;Beran et al. 2014). When adults were transferred to different crucifer species, they selectively accumulated mainly aliphatic glucosinolates from their food plants, e.g. allyl glucosinolate from Brassica juncea, and 4-methylsulfinylbutyl (4MSOB) glucosinolate from Arabidopsis thaliana. In contrast, adults sequestered only low amounts of the benzenic 4-hydroxybenzyl glucosinolate from Sinapis alba. The glucosinolate accumulation pattern depended both on glucosinolate structure and on the host plant background, suggesting that the plants' glucosinolate composition affects sequestration in Ph. striolata. The ability to accumulate high glucosinolate amounts demonstrates that Ph. striolata can at least partially prevent activation of ingested glucosinolates.

Study
Psylliodes Phyllotreta Ge et al. (2011) Sister genus: Chaetocnema ( However, quantitative feeding studies, for instance with radiolabeled glucosinolates, are needed to determine to which degree ingested glucosinolates are sequestered intact. To activate sequestered glucosinolates, Ph. striolata possesses an insect myrosinase with high activity towards aliphatic glucosinolates, which evolved from insect β-Oglucosidases (Figure 4; Beran et al. 2014). To investigate how Ph. striolata activate sequestered glucosinolates and prevent autointoxication, dissected tissues from adults were analysed for the presence of glucosinolates and myrosinase activity, respectively. Interestingly, both glucosinolates and myrosinase were mainly localised in the hemolymph and elytra (Beran and Ahn, unpublished), but whether both components are stored separately in hemoplasma and hemocytes as previously reported for cyanogenic glycosides and the cyanogenic β-glucosidase in Zygaena filipendulae larvae (Lepidoptera, Zygaenidae; Pentzold et al. 2017), is not yet known.
In the genus Psylliodes, the cabbage stem flea beetle, Ps. chrysocephala, selectively sequesters glucosinolates as well, but compared to Ph. striolata, glucosinolate concentrations are much lower (ca. 4 µmol/g fresh weight; Beran et al. 2018). Although glucosinolates are present in all life stages of Ps. chrysocephala, a defensive function is unlikely, as neither larvae nor adults possess endogenous myrosinase activity (Beran et al. 2018). An analysis of the metabolic fate of ingested 4MSOB glucosinolate in Ps. chrysocephala adults revealed that adults utilise at least three strategies to prevent isothiocyanate formation and toxicity. Ps. chrysocephala sequester intact glucosinolates, detoxify glucosinolates by desulfation, and detoxify dietary isothiocyanates by conjugation to glutathione. The isothiocyanate-glutathione conjugate is metabolised via the conserved mercapturic acid pathway to three different cyclic cysteine conjugates, which are excreted. These three strategies accounted for the metabolic fate of 18.5%, 8%, and 17% of the total ingested glucosinolates, respectively. The amounts of other glucosinolate breakdown products (4MSOBisothiocyanate, -cyanide, -amine, and -acetamide) corresponded to 17.5% of the total ingested glucosinolate (Figure 4; for details, refer to Beran et al. 2018). However, the metabolic fate of about 39% of the total ingested glucosinolate remained unknown in this study.
The detoxification of isothiocyanates in Ps. chrysocephala comes at the expense of the amino acid cysteine. Therefore, interference with protein digestion, for instance by plant proteinase inhibitors or other digestibility reducers, might affect the detoxification capacity for isothiocyanates by limiting the availability of cysteine for glutathione biosynthesis. Interestingly, there is evidence that Ps. chrysocephala can compensate for the ingestion of plant proteinase inhibitors. Ps. chrysocephala larvae reared on a transgenic Brassica napus line that overexpressed the cysteine proteinase inhibitor oryzacystatin I showed doubled proteolytic activity and were heavier than those reared on the corresponding B. napus wild type (Girard et al. 1998). This unexpected result shows that Ps. chrysocephala is not only adapted to glucosinolates but also to plant proteinase inhibitors.
Specialist chrysomelids are well known for discriminating between crucifer species (Feeny et al. 1970;Nielsen 1977;Bartlet and Williams 1991;), but the factors that determine host suitability and preference are often not understood. Although leaf beetles recognise and differentially respond to individual glucosinolates when offered in isolation, there is little evidence that host plant preference relies on specific glucosinolate profiles (Nielsen 1988). Instead, the presence of other toxic secondary metabolites such as cucurbitacins and cardenolides was shown to affect host suitability for Phyllotreta spp. and Phaedon cochleariae (abbreviated Phaedon) (Nielsen 1978). Toxic cucurbitacins B, E, and I present in Iberis spp. deterred feeding of Phyllotreta nemorum but not of Phaedon, an effect that correlated with their feeding behavior towards Iberis plants. On the other hand, Phaedon, Phyllotreta undulata, and Phyllotreta tetrastigma did not feed on cardenolidecontaining Cheiranthus and Erysimum spp., which are accepted as food plants by Ph. nemorum (Nielsen 1978). Upon herbivory, glucosinolates are usually hydrolysed by the plant enzyme myrosinase to an unstable aglucone, which spontaneously rearranges to a toxic isothiocyanate. In the presence of plant specifier proteins, other hydrolysis products such as thiocyanates and nitriles are formed. Both flea beetle species sequester glucosinolates in their bodies, suggesting that not all glucosinolates are hydrolysed in feeding-damaged plant tissue. Sequestered glucosinolates may be activated for defensive purposes by an insect myrosinase in Ph. striolata, but not in Ps. chrysocephala. In addition, Ps. chrysocephala partially detoxifies glucosinolates by desulfation, whereas no glucosinolate sulfatase activity was found in Ph. striolata. According to a quantitative feeding study performed with Ps. chrysocephala, most ingested glucosinolates are activated, and isothiocyanates are detoxified by conjugation to glutathione. The isothiocyanate-glutathione conjugate is metabolized via the mercapturic acid pathway to several cyclic metabolites in Ps. chrysocephala adults (Beran et al. 2018). Examples of three structurally different glucosinolate side-chains are shown in the box. Beetle photos: Anna Schroll.
The oligophagous species Ph. nemorum is used as a model to study the genetic basis of host plant adaptation. The common wild crucifer, Barbarea vulgaris ssp. arcuata (abbreviated B. vulgaris), is an atypical host plant for Ph. nemorum. However, the discovery of two different flea beetle populations using B. vulgaris as natural host plant suggests that Ph. nemorum is extending its host plant range to include B. vulgaris in Denmark (Nielsen 1996;. There are two distinct types of B. vulgaris. The so-called P-type with pubescent leaves is susceptible to all Ph. nemorum genotypes, whereas the G-type with glabrous leaves is resistant to most Ph. nemorum genotypes (Nielsen 1997b). The flea beetle-resistant G-type represents the common B. vulgaris genotype in Western Europe, while the P-type is rare (Hauser et al. 2012;Christensen et al. 2014).
The two B. vulgaris types differ not only morphologically but also regarding their chemical defences, i.e. glucosinolates and saponins. Feeding assays showed that susceptible Ph. nemorum larvae started to mine into the leaves of the G-type, but then either left and refused to feed or died in the mine, showing that the G-type is toxic for them (Nielsen 1997a(Nielsen , 1997b. Resistance of the G-type to Ph. nemorum is linked to the presence of the triterpenoid saponins hederagenin cellobioside, oleanolic acid cellobioside, gypsogenin cellobioside, and 4-epihederagenin cellobioside, and not to distinct glucosinolate profiles (Agerbirk et al. 2001;Kuzina et al. 2009;Nielsen et al. 2010). The toxicity of saponins is at least partially due to their interactions with cell membranes, which can cause cell death (Augustin et al. 2011). The activity of isolated hederagenin cellobioside and oleanolic acid cellobioside was tested separately in nochoice feeding assays with Ph. nemorum adults from five different near-isogenic lines (Nielsen et al. 2010). In these experiments, hederagenin cellobioside had a much stronger negative effect on adult feeding than oleanolic acid cellobioside, whereas the corresponding aglycones of both saponins were not active. An even stronger negative effect on some Ph. nemorum lines was observed for α-hederin, a saponin which is not present in B. vulgaris, and only differs from hederagenin cellobioside in its glycosylation pattern (Nielsen et al. 2010). These results show that aglycone structure as well as glycosylation pattern affect the biological activity of saponins towards Ph. nemorum.
Although the saponin-based defence of B. vulgaris is a dead-end for most Ph. nemorum genotypes, resistant individuals that performed well on the G-type were found at varying frequencies in all sampled populations (Nielsen and de Jong 2005;Nielsen 2012;Vermeer et al. 2012). The ability to use the G-type as a host plant clearly shows that resistant individuals can tolerate or detoxify saponins by an unknown mechanism. In genetic analyses, Nielsen and de Jong identified the presence of dominant resistance-conferring genes (R-genes) in all resistant individuals, but divergent modes of inheritance of these R-genes (autosomal and sex-linked) between populations (Nielsen 1997a;de Jong and Nielsen 2002;Nielsen 2012). For example, in the resistant population from Ejby (Denmark), two major R-genes were linked to the sex chromosomes with additional autosomal R-genes. In a resistant population from Kvaerkeby (Denmark), most individuals were homozygous for a single autosomal R-gene (Nielsen 1997a;. In crossing experiments with resistant males from a Swiss population, an autosomal R-gene was inherited only to female offspring due to non-random segregation. The most likely explanation for this nonrandom segregation of the autosomal R-gene together with the X chromosome is the fusion of an autosome carrying the susceptible allele to the Y-chromosome in Swiss males (Nielsen 2012). When this R-gene was introduced into the genetic background of the susceptible Ph. nemorum line, it showed a normal Mendelian inheritance pattern (Nielsen 2012). These results strongly suggest that the genetic architecture of Ph. nemorum males differs among flea beetle populations, and that this polymorphism affects the inheritance of R-genes that enable the offspring to use the otherwise toxic B. vulgaris G-type as a host plant. Interestingly, attempts to generate Ph. nemorum lines that are homozygous for an autosomal R-gene resulted in very low survival rated of the homozygous larvae Breuker et al. 2007). This observation was surprising as the homozygous resistant genotype was common at least in the B. vulgaris-feeding population from Kvaerkeby, which suggests that co-adapted genes present in the field population counteract the fitness cost of R-genes de Jong and Nielsen 2002). The genetic diversity and population structure of Ph. nemorum makes this species an ideal model to study the genetic basis of host range expansion in an oligophagous herbivore.

Conclusions and future directions
The flea beetle genera Psylliodes and Phyllotreta are closely associated with glucosinolatecontaining plants mainly in the family Brassicaceae. Nevertheless, they differ remarkably in their overall host plant use and their adaptations to glucosinolates, the characteristic defence metabolites in Brassicaceae. While Ph. striolata can utilise sequestered glucosinolates for its defence against predators, Ps. chrysocephala apparently does not possess endogenous myrosinase activity and accumulates much lower amounts of glucosinolates compared to Ph. striolata. In addition, both species differ regarding their ability to detoxify glucosinolates by desulfation (Beran et al. 2014(Beran et al. , 2018.
Despite this progress, our knowledge on the adaptations of Phyllotreta and Psylliodes to the glucosinolate-myrosinase defence is far from complete. It is unknown, for instance, whether Phyllotreta rapidly sequester glucosinolates to prevent their breakdown to toxic isothiocyanates, and whether Phyllotreta gain protection from natural enemies by activating sequestered glucosinolates using their own myrosinase. In Ps. chrysocephala, the importance of the various detoxification strategies and their evolution needs to be investigated. To this end, a robust phylogenetic tree of the genus and comparative studies on how other Psylliodes species are processing dietary glucosinolates are necessary.
A future goal is to place adaptations of Phyllotreta and Psylliodes to their glucosinolate-containing host plants into a broader evolutionary context. While recent phylogenetic studies support the hypothesis that both genera adapted independently to Brassicaceae, their relationships to other genera of Alticini remain largely unresolved (Ge et al. 2011;Ge et al. 2012;Nie et al. 2018). At this background, a comprehensive and well-resolved phylogenetic tree of the tribe Alticini will enable studies on interactions with plants in general and adaptations to plant chemical defences, and how they contributed to the evolutionary success of this megadiverse lineage. Bartlet E, Williams IH (1991)