Functional analysis of Plasmodium genes by classical reverse genetics is currently limited to mutants that are viable during erythrocytic schizogony, the pathogenic phase of the malaria parasite where transfection is performed. Here, we describe a conceptually simple experimental approach to study the function of genes essential to the asexual blood stages in a subsequent life cycle stage by a promoter-swap approach. As a proof of concept we targeted the unconventional class XIV myosin MyoA, which is known to be required for Toxoplasma gondii tachyzoite locomotion and host cell invasion. By placing the corresponding Plasmodium berghei gene, PbMyoA, under the control of the apical membrane antigen 1 (AMA1) promoter, expression in blood stages is maintained but switched off during transmission to the insect vector, i.e. ookinetes. In those mutant ookinetes gliding motility is entirely abolished resulting in a complete block of life cycle progression in Anopheles mosquitoes. Similar approaches should permit the analysis of gene function in the mosquito forms that are shared with the erythrocytic stages of the malaria parasite.
Reverse genetics is a powerful tool to study gene function in apicomplexan parasites, medically important single cell eukaryotic pathogens. Malaria, for example, continues to be the most important vector-borne infectious disease and is caused by apicomplexan parasites of the genus Plasmodium. Their life cycle is a sequence of alternating invasive and replicative stages within vertebrates and mosquitoes, each stage being equipped with a repertoire of shared and tailor-made gene products. The recent availability of near-complete genome sequences of the human pathogen Plasmodium falciparum and of various rodent malaria model species now permits the systematic analysis of gene function to prioritize drug and vaccine targets (Kooij et al., 2006).
Plasmodium transfection and selection for recombinant parasites is performed with cultured asexual blood stages (Janse et al., 2006), which cause all malaria-related mortality and morbidity. Due to the haploid genome of this parasite stage, vital genes cannot be targeted by classical reverse genetics, precluding a direct genetic analysis of some of the most promising targets. Thus, alternative strategies for conditional Plasmodium gene depletion have been proposed (Carvalho et al., 2004; Meissner et al., 2005; Armstrong and Goldberg, 2007; Combe et al., 2009). One approach is based on tetracycline analogue-regulated gene expression and was shown to work for regulated expression of the green fluorescent protein (GFP) and chloramphenicol acetyl transferase (CAT) in P. falciparum (Meissner et al., 2005). In the related parasite Toxoplasma gondii this strategy was used to characterize endogenous gene function in vitro and in vivo (Meissner et al., 2002; Plattner et al., 2008). A different strategy was developed for the rodent malaria model parasite Plasmodium berghei and is based on the Flp/FRT site-specific recombination system originating from yeast(Carvalho et al., 2004; Combe et al., 2009). This approach achieves precise gene excision at the sporozoite stage (Carvalho et al., 2004). Recently this approach was successfully employed to excise the promoter of the merozoite surface protein 1 (MSP1) at the sporozoite stage, which resulted in a severe defect in merozoite release upon mosquito-to-mouse transmission (Combe et al., 2009). A novel approach takes advantage of an in-frame fusion of a ligand-regulatable protein destabilization domain for use in P. falciparum in vitro cultures (Armstrong and Goldberg, 2007). This strategy permitted the first comprehensive characterization of an essential P. falciparum gene in vitro (Dvorin et al., 2010). Whether this method can be adapted to an in vivo model, such as P. berghei in mice, remains to be shown. However, despite these recent advancements, a strategy that facilitates detailed phenotypic in vivo analysis awaits further improvements, including complete gene excision to obtain homogenous loss-of-function parasites.
We therefore employed a simple promoter-exchange approach to characterize the function of a candidate gene that is expressed in blood stages as well as in the mosquito infecting ookinete. The goal was to generate a conditional mutant that would allow us to study a gene of interest in the insect forms but retain the ability of parasite propagation in the mammalian host, a prerequisite for its generation and maintenance. As a target gene we selected myosin A, the founding member of unconventional class XIV myosins (Heintzelman and Schwartzman, 1997; 1999). These myosins are remarkably small, essentially composed of the actin-interacting ATPase domain, followed by a very short tail domain. Class XIV myosins are exclusively found in the Alveolata branch of eukaryotic organisms, including the Apicomplexa (Heintzelman and Schwartzman, 2001; Foth et al., 2006). MyoA was previously utilized to establish anhydrotetracycline-inducible gene expression in T. gondii parasites (Meissner et al., 2002). A conditional mutant was obtained by exchanging the promoter of the gene encoding TgMyoA with a synthetic promoter that can be negatively regulated by addition of tetracycline analogues. The recombinant parasites expressed only residual amounts of TgMyoA in the presence of the drug abolishing all forms of motility. Consequently, host cell invasion and parasite egress were reduced by 84% and 82% respectively (Meissner et al., 2002). Residual expression of MyoA or the contribution of other myosins might cause the incomplete blockage of life cycle progression in presence of the drug. By analogy, Plasmodium MyoA is considered vital for merozoite invasion of erythrocytes, but no experimental evidence has been reported thus far.
In this report we studied the role of the orthologous MyoA from P. berghei (Margos et al., 2000; Matuschewski et al., 2001) in ookinete motility as an example to test the robustness of the proposed promoter exchange approach.
Rationale for the stage-specific depletion of target genes
The crucial role of MyoA in T. gondii motility (Meissner et al., 2002) suggested that Plasmodium MyoA is refractory to gene deletion, since similar function during merozoite invasion would be essential. Therefore, we designed a strategy to exchange the endogenous MyoA promoter, which is active in all invasive stages, with a promoter that is active in merozoites, but switched off in ookinetes, the next extracellular stage that penetrates the mosquito midgut epithelium (Fig. 1A). By placing MyoA expression under the control of the stage-specific AMA1 promoter we expected to be able to study the PbMyoA loss-of-function mutant in the Plasmodium ookinete.
To verify that the AMA1 promoter is indeed silent in the ookinete but active in merozoites, i.e. during transfection, we examined the transcription profile of MyoA and AMA1 in all motile parasite stages by quantitative reverse transcriptase PCR (qRT-PCR) (Fig. 1B). MyoA steady-state transcript levels were comparable in merozoite and ookinetes, confirming previous results using an anti-MyoA antiserum (Pinder et al., 1998; Chaparro-Olaya et al., 2005; Siden-Kiamos et al., 2006) and proteome analysis (Hall et al., 2005). MyoA transcript levels were reduced in sporozoites, in agreement with expression profiling data (Rosinski-Chupin et al., 2007), but distinct from readily detectable MyoA protein (Margos et al., 2000; Matuschewski et al., 2001; Hall et al., 2005). In contrast to MyoA, AMA1 mRNA levels were very low in ookinetes but abundant in merozoites and sporozoites. We therefore concluded that expression of MyoA under the control of the AMA1 promoter would result in viable asexual parasites but would deplete MyoA in ookinetes.
Complementation of PbMyoA expression by a heterologous promoter
We designed an integration construct that is predicted to result in a promoter exchange (Fig. 1C). A single-cross-over event at the site of linearization within the 5′ region of the MyoA open reading frame leads to an upstream copy that contains the endogenous promoter before a 3′ deleted, non-functional MyoA copy. Downstream of this copy is the positive selectable marker and a functional copy consisting of a full-length MyoA allele, which is now under the control of the AMA1 promoter.
The genotype of the resulting mutant parasite clones, termed ‘MyoA absent in ookinete’ (Mao), was verified by PCR analysis of genomic DNA (Fig. 1D). Allele-specific primers revealed the presence of the expected fragments, indicative of an integration event. Generation of a clonal parasite population indicated that the replacement on the endogenous promoter with the AMA1 promoter did not affect the viability of the recombinant parasites. Consistently, we detected similar levels of MyoA and AMA1 protein in Mao and wild-type mature blood-stage parasites by immunofluorescence analysis (Fig. 1E). We showed that placement of the MyoA gene under the heterologous AMA1 promoter leads to viable parasites, which we subsequently subjected to phenotypic characterization during life cycle progression.
PbMyoA has no role in ookinete formation
As expected by the successful generation of Mao parasites, in vivo growth of asexual parasites was comparable to wild-type parasites (data not shown). Moreover, exflagellation and exflagellation centre formation were also normal (data not shown). Previously, we have shown that MyoA is not detectable in young zygotes (2–5 h after fertilization) (Siden-Kiamos et al., 2006); hence we assumed that depletion of MyoA in these stages will not cause a phenotype.
We next assayed the in vitro formation of ookinetes in the absence of MyoA. For ookinete culture, gametocyte-positive blood from mice infected with Mao and wild-type parasite was seeded and cultured for 24 h. Upon affinity purification the number of ookinetes per 100 µl seeded blood was determined (Fig. 2A). No significant difference in the number of ookinetes could be detected in multiple independent cultures of Mao and WT parasites, excluding a role of MyoA in ookinete morphogenesis. We next quantified gamete-to-ookinete conversion by staining of these cultures with an antibody against the surface protein P28, which stains female gametes, zygotes and ookinetes (Sinden et al., 1987). The numbers of ookinetes and round forms (female gametes and zygotes) were indistinguishable in Mao and wild-type parasites (Fig. S1). The viability of cultured ookinetes was analysed by a live/dead assay, which monitors membrane integrity (Fig. 2B). In this assay, the loss of MyoA had no effect on ookinete viability. In T. gondii, class XIV myosins have recently been implicated in the generation of the inner membrane complex (IMC) (Agop-Nersesian et al., 2009). Our results do not support a possible function of PbMyoA in IMC formation during ookinete maturation, since Mao parasites form viable ookinetes that cannot be distinguished from WT ookinetes.
Ookinete-specific depletion of PbMyoA
Next, we verified the absence of MyoA in the Mao ookinetes. Equal numbers of in vitro cultured WT and Mao ookinetes were analysed on a Western blot, probed with an antiserum against Plasmodium MyoA and, in parallel, using a monoclonal anti-actin antibody as a loading control (Fig. 2C). The MyoA antiserum recognized a band at the expected molecular weight (∼90 kDa) in wild-type ookinetes while no signal was detected in the Mao samples. The actin antibody, on the other hand, detected a band of similar intensity and size (∼42 kDa) in both strains. Immunofluorescence analysis of the wild-type and Mao ookinetes further confirmed the absence of the typical apical staining of MyoA in recombinant parasites (Fig. 2D and E). Despite the lack of MyoA in the mutant, the predominant actin localization in ookinetes was unaffected (Fig. 2D), supporting previously published evidence for an independent transport of the central motor machinery components to the ookinetes apical end (Siden-Kiamos et al., 2006). As expected, AMA1 was detected only at very low intensity in both wild-type and mutant ookinetes (Fig. 2E).
PbMyoA is vital for ookinete motility and malaria transmission to mosquitoes
We have previously characterized ookinete motility in vitro (Siden-Kiamos et al., 2006) and described several modes: (i) stationary flexing and stretching of the cell, (ii) twirling, where the parasite attaches with one end to the substratum and rotates around it own axis, and (iii) productive migration over a distance, or locomotive motility, at an average speed of 6 µm min−1. These three modes of motility can be seen in the same parasite over time (Movie S1). One method for the induction of ookinete motility is the addition of mosquito cells and, hence, all experiments described were performed in the presence of the Aedes aegypti Mos20 cell line (Singh, 1971).
We employed video time-lapse microscopy to visualize different motility patterns of wild-type and Mao ookinetes (Fig. 3 and Movies S1 and S3, and S2 and S4 respectively). Individual affinity-purified ookinetes were imaged and tracked over the full length of the movie. Subsequently their velocity was plotted (Fig. 3A and B and Movies S3 and S4). MyoA-deficient ookinetes did not show any signs of productive gliding locomotion. Enriched ookinetes often aggregated and, as expected, the wild-type parasites rapidly moved out of these aggregates after adding insect cells (Movies S1 and S3). In marked contrast, the Mao mutant did not display any kind of motility (Movies S2 and S4). Even after prolonged incubation with insect cells ookinetes remained in tight aggregates and the number of aggregated versus individual ookinetes was employed to determine the motility index (Siden-Kiamos et al., 2006). As a control, we added the actin inhibitor cytochalasin D, which abolishes motility (Miller et al., 1979; Dobrowolski and Sibley, 1996; Siden-Kiamos et al., 2006) (Fig. 3C). These results confirm that Mao ookinetes are devoid of motility.
Our next aim was to test the mutant parasites ability to progress through the life cycle in vivo, i.e. transmission to mosquitoes. Due to the lack of productive motility, we predicted an impaired capacity to traverse the mosquito midgut epithelium and, hence, a reduced number of sporogonic oocyst. Anopheles stephensi mosquitoes were fed on mice infected with Mao and wild-type parasites respectively. After 10 days, mid-guts were dissected and the number of oocysts determined. Confirming our in vitro motility data, the formation of oocysts was completely abolished in the Mao parasite line (Fig. 4).
Ookinete microinjection restores the defects of Mao parasites
In order to exclude a general defect in oocyst and sporozoite formation we finally tested whether ookinetes develop normally once they reach the mosquito haemocoele. Therefore we bypassed the natural midgut passage by microinjection of ookinetes directly into the mosquitoes' thorax (Weathersby, 1952; Sinden et al., 2002). At day 16 post injection we quantified the number of haemocoele-associated sporozoites (Fig. 5A). Most importantly, Mao parasites formed substantial numbers of sporozoites, and the number of sporozoites per 500 injected ookinetes was indistinguishable to wild-type parasites. Since oocysts are likely not associated with the midgut upon ookinete injection, but scattered all over the mosquito's haemocoele, we could not test for MyoA and AMA1 expression in oocysts. Previous work established that AMA1 is translationally regulated resulting in detectable protein only in mature, salivary gland-associated sporozoites (Srinivasan et al., 2004). Translational repression in Plasmodium depends on elements in the 5′ as well as the 3′ UTR (Braks et al., 2008). Hence, MyoA might also be expressed in Mao oocysts, since it is under control of two distinct elements, i.e. the AMA1 5′ UTR and the endogenous MyoA 3′ UTR. Therefore, we cannot formally exclude a function of MyoA during sporogony. Since MyoA protein cannot be detected in oocysts (Hall et al., 2005), such an additional role is highly unlikely. Therefore, our results support the conclusion that the impaired mouse-to-mosquito transmission of Mao parasites was caused by the defective ookinete motility and not by impaired sporogony.
The expression of MyoA and AMA1 in sporozoites was assessed by immunofluorescence analysis (Fig. 5B). We isolated mature sporozoites from infected salivary glands and the mosquito haemocoele. Localization and intensity of MyoA- and AMA1-specific signals were comparable in WT and Mao mutant sporozoites originating from both compartments. This finding provides additional evidence for release of translational repression of AMA1 in haemocoele sporozoites and supports the notion that sporozoites mature, at least partially, following an endogenous developmental programme (Sinden and Matuschewski, 2005). Most importantly, this analysis also demonstrated presence of Mao sporozoites associated with salivary glands, indicative of their capacity to enter the final organ in the mosquito vector.
We finally analysed the sporozoites' ability to glide by immunofluorescence analysis. While moving, sporozoites shed their mayor surface protein, termed circumsporozoite (CS) protein, which can be visualized with antibodies (Potocnjak et al., 1980). These antibodies detect the sporozoite as well as trails of CS (Fig. 5C). Recombinant and wild-type sporozoites derived from injected ookinetes were able to glide productively (Fig. 5C). Owing to the limited numbers of ookinetes, which can be injected into a mosquito, we obtained only low numbers of haemocoele- and salivary gland-associated sporozoites precluding comprehensive infection experiments with either Mao or WT sporozoites. However, the ookinete injection experiments showed that Mao parasites are able to form sporozoites, when the midgut barrier was bypassed, demonstrating the vital role of PbMyoA for ookinetes' motility.
In this study we employed a simple and straightforward genetic strategy to study the function of Plasmodium genes that are essential in the pathogenic blood stages but shared with other Plasmodium stages. As proof-of-principle we targeted the unconventional class XIV myosin MyoA in the rodent malaria model P. berghei. Using a promoter exchange approach, we were able to permit functional expression of PbMyoA during asexual growth of the malaria parasites, the stage where transfection is performed. We used the AMA1 promoter to drive MyoA expression, which results in the absence of the corresponding gene product in ookinetes. We propose that characterization of numerous vital genes for asexual parasite growth may be feasible employing a similar promoter exchange strategy, provided their shared function in a parasite stage other than asexual blood stages. Hence, this strategy may also establish whether a potential essentiality is restricted to erythrocytic schizogony or recurs during Plasmodium life cycle progression.
While this promoter exchange approach is a simple and time-efficient method to study ‘likely essential’ genes, it strictly depends on the promoter activity that drives the expression of the gene of interest. In our example we took advantage of the AMA1 promoter, which appears to be particularly suited for genes that exert their vital roles in invasion process and, perhaps, also parasite egress. Potential candidates that could be tested by the same targeting approach include the MyoA tail-binding protein (MTIP; Bergman et al., 2003) and proteins that anchor the motor complex to the IMC, e.g. GAP45, GAP50 (Gaskins et al., 2004; Green et al., 2006; Johnson et al., 2007). Other components of the molecular motor complex, particularly actin and its regulatory proteins likely fulfil diverse roles, not restricted to parasite locomotion. Previous work established that the malaria parasite only forms short actin filaments (Schmitz et al., 2005; 2010; Schüler et al., 2005a). Several Plasmodium actin regulators, e.g. actin-depolymerizing factor 1/cofilin, profilin and formin 1 (Schüler et al., 2005b; Baum et al., 2008; Kursula et al., 2008), and actin-related genes, such as ARP1 (Siden-Kiamos et al., 2010), are refractory to gene deletion (reviewed in Sattler et al., 2011). These and other candidate genes that emerge from systematic reverse genetics analyses (Tewari et al., 2010) likely need to be expressed throughout the intra-erythrocytic schizogony. Identification of suitable promoters that are ‘off’ in mosquito or pre-erythrocytic stages awaits detailed expression profiling in different Plasmodium life cycle stages.
In this study, we provide genetic evidence to support the notion of a vital role of MyoA exclusively in gliding locomotion of apicomplexan parasites (Meissner et al., 2002). While we expected the Mao parasites to be disturbed in locomotive motility we were surprised that even non-locomotive forms of motility, such as twirling of the cell, were dependent upon the presence of MyoA. This is a previously unrecognized finding, since no distinct analysis of motility pattern have been reported in the conditional T. gondii MyoA strain, the only described MyoA mutant thus far (Meissner et al., 2002). In the tetracycline-dependent mutant host invasion and subsequent egress were reduced but not abolished, most likely due to an inefficient silencing of the gene. In contrast, we report a complete block of mouse-to-mosquito transmission, a defect that can be fully restored by mechanical bypass of the midgut epithelial barrier. In the present report, we have proven that MyoA in P. berghei is vital for productive motility of the ookinete. Given the shared molecular machinery employed by all invasive Plasmodium stages it is highly likely that this central role of MyoA can be postulated for the remaining invasive stages of the Plasmodium parasite, namely sporozoites and merozoites. The results from our reverse genetics approach lend strong support for the development of specific drugs directed against the MyoA motor. This possibility and the cellular roles of other parasite key molecules can now be addressed employing the genetic strategy presented herein.
Mice were purchased form Charles River Laboratories. All animal work was conducted in accordance with the German ‘Tierschutzgesetz in der Fassung vom 18. Mai 2006 (BGBl. I S. 1207)’, which implements the directive 86/609/EEC from the European Union and the European Convention for the protection of vertebrate animals used for experimental and other scientific purposes. The protocol was approved by the ethics committee of MPI-IB and the Berlin state authorities (LAGeSo Reg# G0469/09).
Quantitative reverse transcriptase PCR (qRT-PCR)
Total RNA was purified employing the RNeasy kit (Quiagen) and reverse transcription was performed using the RETROscript kit (Ambion). qRT-PCR was performed on cDNA preparations from purified late schizonts/merozoites, mid-gut-associated ookinetes (24 h post infection) and salivary gland associated sporozoites (17–22 days post infection), using the ABI 7500 sequence detection system or StepOnePlus™ and Power SYBR® Gene PCR Master Mix (Applied Biosystems), according to the manufacturer's instructions. qRT-PCR was performed in triplicates, with one cycle of 95°C for 15 min, followed by 40 cycles of 95°C for 15 s, 55°C for 15 s and 60°C for 45 s. Data were analysed with the SDS1.3.1 software (Applied Biosystems). Relative transcript abundance was normalized to the expression of heterologous GFP, which is expressed under the constitutive EF1α promoter. The following primers were used for qRT-PCR (Table S1): MyoA_for and MyoA_rev; AMA1_forward and AMA1_rev; and GFP_for and GFP_rev.
Generation of ‘Myosin A absent in ookinetes’(Mao) parasites
A 1.1 kb 3′ deleted PbMyoA fragment was amplified with primers MyoAfor-BamHI and MyoArev-NdeI. All primers are detailed in Table S1. A 1.3 kb region covering the PbAMA1 promoter was amplified using the primers AMA1for and AMA1rev. Importantly, the PbMyoA fragment contains the 3 bp Kozak sequence immediately before the initiation codon. P. berghei transfection was performed as described (Janse et al., 2006). Briefly, 75 µg of plasmid DNA was linearized with HpaI and electroporated into purified schizonts using a Bio-Rad electroporation system. Positive selection was by daily treatment with 25 mg kg−1 pyrimethamine.
Ookinete cultures and viability assay
The ookinetes were cultured from blood of infected Theiler' Original mice using methods described (Siden-Kiamos et al., 2006). In brief, ookinetes were purified using either ammonium chloride lysis of red blood cells or separation using anti-P28 antibody coated Dynabeads. Ookinete conversion was determined by live staining using a Cy3-conjugated antibody of the 13.1 antibody (Sinden et al., 1987).
To assess the viability of mutant ookinetes compared with the wild type, parasites of both lines were stained in the presence of cells with the Live/Dead kit (Molecular Probes). Dead cells stain red because of the leakiness of membranes to the ethidium homodimer, whereas live cells stain green because of esterases that cleave the fluorescent substrate calcein-AM. Although the green staining of motile invasive ookinetes and sporozoites is typically weak, the two populations can easily be separated primarily based on the red fluorescence of non-viable parasites.
For the time-lapse videos purified ookinetes were mixed with A. aegypti Mos20 cells and spotted on a slide under a Vaseline-rimmed coverslip. Pictures were taken every 30 s. under a Leica microscope fitted with a Spot Jr CCD camera from Visitron Systems (Movies S1 and S2), or images were collected every 15 s on an inverted Zeiss Axiovert 200 M microscope with a Zeiss Axiocam HRm camera using Axiovision 4.6 software (Movies S3 and S4). All image series were eventually imported to ImageJ software for analysis. The quantitative assay has been described previously (Siden-Kiamos et al., 2006). In brief, ookinetes were mixed with A. aegypti Mos20 cells in LabTek eight-well chamber slides in Medium M199 (Gibco). The co-cultures were incubated for 20–24 h at 19°C. The medium was then removed and the ookinetes were stained with an antibody against the Pb70 antigen. The ookinetes are usually found in aggregates during the incubation time. By counting aggregated and single ookinetes an estimate of motility can be obtained. A control with the actin inhibitor Cytochalasin D, at a concentration of 10 µM, is included to determine the scattering of ookinetes during seeding.
Injection of cultured ookinetes
Cultured and purified ookinetes were injected into An. stephensi females as described earlier (Weathersby, 1952; Sinden et al., 2002). In brief, female mosquitoes were anaesthetized with carbon dioxide out on a porous polyethylene pad (Flypad, Flystuff), employing a flow regulator (Flowbuddy, Flystuff). Volumes between 69 and 207 nl per female were injected, equivalent to 500–1500 ookinetes. All injections were carried out utilizing a Drummond Nanoject II microinjector.
Sporozoite development in An. stephensi mosquitoes
Anopheles stephensi mosquito rearing and maintenance was carried out under a 14 h light/10 h dark cycle, 75% humidity and at 28°C or 20°C respectively. Natural infection and dissection of An. stephensi mosquitoes were conducted as described previously (Vanderberg, 1975; Sinden et al., 2002). For determination of oocyst numbers, infected mosquitoes were dissected at day 14 after natural infection and isolated mid-guts were subjected to standard microscope examination. After ookinete injection, haemocoele- and salivary gland-associated mutant and wild-type sporozoites were isolated by standard dissection methods and numbers were determined 16 and 20 days post infection respectively.
We thank Markus Meissner (Glasgow University, UK) and Clemens Kocken (BPRC, Rijswijk, the Netherlands) for providing antibodies against MyoA and AMA1 respectively. Oliver Billker (Sanger Centre, UK) helped with ookinete cultures and the kind gift of the Cy-3-conjugated antibody. Chris Janse (Leiden University, the Netherlands) and Andrew Waters (Glasgow University, UK) provided initial support with P. berghei transfection technology. We thank Herwig Schüler (Karolinska Institutet, Stockholm, Sweden) and Olivier Silvie (Université Pierre et Marie Curie, France) for critical reading of the manuscript. This work was supported by the Max Planck Society and the European Commission through the BioMalPar and EviMalaR networks (joint project of partners I.S.-K., R.E.S., C.L. and K.M.).