A synthetic promoter for multi-stage expression to probe complementary functions of Plasmodium adhesins

Malaria is caused by apicomplexan parasites of the genus Plasmodium. During its life cycle Plasmodium switches continuously between two hosts, an insect vector and tetrapod vertebrates, such as reptiles, birds and mammals. During this cycle, the parasites develop in at least two environments under very different conditions regarding growth and development. Plasmodium parasites can replicate asexually within hepatocytes and red blood cells as well as extracellularly within oocysts at the midgut wall of Anopheles mosquitoes. They also undergo sexual development in the vertebrate and gamete formation within the mosquito blood meal (Fig. 1). To sustain these developmental progressions in different environments the parasite expresses both housekeeping genes and stage-specific sets of genes. Stage-specific gene expression is under tight transcriptional control and these genes are often essential in the respective life cycle stage (for example: Ménard et al., 1997; Sultan et al., 1997; Dessens et al., 1999; Kehrer et al., 2016; Kumar et al., 2016; Dankwa et al., 2016; Elsworth et al., 2014; Van Dijk et al., 2001; Mueller et al., 2005; Yuda et al., 1999). Genetic modifications of human and rodent malaria parasites are performed during the intraerythrocytic stage. Genes with non-essential functions for the blood stage can be analysed across the life cycle. Yet, the study of gene function after the stage for which a gene is essential is difficult. To allow analysis of gene function at ‘later’ stages a number of methods have been developed to regulate gene expression, i.e. by using anti-sense RNA (Gardiner et al., 2000), inducible promoters (Meissner et al., 2001; Pino et al., 2012) as well as the stage-specific knockout of genes using the flipase-flipase recombination target (Flp-FRT) (Carvalho et al., 2004; Combe et al., 2009) or the Cre-Lox recombination system (Collins et al., 2013). Another method named ‘knocksideways’ does not interfere with gene expression but allows the study of essential genes through guided mislocalisation of target proteins upon addition of specific compounds (Birnbaum et al., 2017). Most of these methods rely on addition of RNA or different compounds (e.g. tetracycline, rapamycin) to initiate protein knockdown or mislocalisation of the target protein. This works efficiently in cell culture but is difficult to achieve in mosquitoes, as the compounds can show toxic side effects and their bioavailability differs from those in in vitro culture settings. With the stage-specific expression of Flp-FRT, excision of the desired coding sequence was sometimes incomplete (Combe et al., 2009).

To overcome these limitations, it has become increasingly popular to directly swap the promoter of the gene of interest in order to bypass developmental defects in specific life-cycle stages (Fig. 1). This method was initially used to investigate genes essential for blood-stage parasites, which could not be deleted (Laurentino et al., 2011; Sebastian et al., 2012; Siden-Kiamos et al., 2011). By placing a blood-stage-specific promoter in front of these genes, their function in later stages of the life cycle could be investigated. Similarly, genes with an essential function during early mosquito stages were investigated in late mosquito stages by placing them under the control of a promoter that allows sexual development and infection of the mosquito midgut but drives no expression in later stages (Kehrer et al., 2016). Transcription in Plasmodium is tightly regulated, which makes this promoter-swap system very reliable (Reid et al., 2018). Yet, the number of available promoters regarding all stages and different expression strength is currently limited. So far, investigation of later stages, in which the importance of the specific gene is unknown, has been attempted by using two different strategies. First, small genes can be expressed as a second copy driven by a different promoter, which leaves the main expression timing untouched (Voss et al., 2016; Sato et al., 2016). Second, replacement of the main promoter region with a different promoter of the desired expression profile (Kehrer et al., 2016; Siden-Kiamos et al., 2011; Laurentino et al., 2011; Santos et al., 2015). Here, we generate a ‘synthetic’ promoter by combining cis-regulatory elements of two different stage-specific promoters to test if this approach can be used to tailor promoters across the life cycle. We describe promoters made in this fashion as ‘synthetic’ because their DNA sequences cannot be found in the Plasmodium genome although the promoters are assembled from endogenous promoter sequences. By using in silico analysis of highly expressed genes, cis-regulatory elements have been first identified in the promoters of stage-specific transcribed genes (Young et al., 2008; Kaneko et al., 2015; Yuda et al., 2009, 2010; Westenberger et al., 2010; Santos et al., 2017). The assignment of the identified elements to specific life-cycle stages indicates that these elements function as enhancers that are bound by particular transcription factors, probably by members of the apicomplexan Apetala2-like transcription factor (ApiAP2) family (Balaji et al., 2005; Campbell et al., 2010; Modrzynska et al., 2017). Importantly, it has been previously shown that disruption of the cis-regulatory elements can completely abrogate gene expression (Yuda et al., 2010).

Synthetic promoters have been mostly explored in bacteria, yeast and plants, because they are easily amenable to screen for promoter activity (Rushton, 2016; Kang et al., 2016; Zhou et al., 2017). In mammalian systems, complexity of promoter regulation increases with tissue-specific and development-specific expression (Roy and Singer, 2015). The general promoter architecture is relatively universal, whereas transcription factors and associated cis-regulatory elements as well as core promoters are not (Wittkopp and Kalay, 2011). In Plasmodium only a few general components have been identified (Gopalakrishnan and Lopez-Estrano, 2010; Ruvalcaba-Salazar et al., 2005) and only recently transcription start sites (TSS) of genes that are transcribed during the blood stage of the parasite have been mapped (Adjalley et al., 2016). Additionally, high-quality RNAseq data (Otto et al. BMC Biol, 2014) allows to predict TSSs.

To test if cis-regulatory elements alone are sufficient to drive stage-specific gene expression, we first generated a synthetic promoter that triggers expression in two stages of the mosquito by combining different elements from ookinete- and sporozoite-specific genes to create a new synthetic promoter named Spooki 1.0, for expression during the sporozoite and ookinete stages (Fig. 2; Table S1). We also generated a second promoter named Spooki 2.0 mediating increased expression in late mosquito stages. After initial characterisation of the promoters by using a fluorescent reporter, we made use of Spooki 1.0 and 2.0 to probe for complementary functions of the ookinete-specific adhesin circumsporozoite and TRAP-related protein (CTRP) and the sporozoite-specific adhesin thrombospondin-related anonymous protein (TRAP) in absence of each other. Swapping the ctrp promoter with that of trap ablates CTRP expression in ookinetes, leading to impaired midgut invasion and absence of insect stages (e.g. sporozoites). In addition, the long coding sequence of the ctrp gene (1905 aa) complicates gene manipulations but favours promoter exchange studies. Generated parasite lines expressing either TRAP or CTRP in the respective stage revealed that TRAP can complement the function of CTRP in gliding motility of ookinetes but not ookinete development into oocysts, whereas CTRP cannot complement TRAP function in sporozoites. Although these results suggest that engineering of a synthetic promoter functioning in mosquito stages can be achieved, the knowledge about gene regulation in general and cis-regulatory elements in particular, is still limited in Plasmodium, constraining the transferability of our approach to all parasite stages.

To design a stage-transcending promoter for gene expression in both the ookinete stage and the complete lifetime of the sporozoite, we employed the promoters of two highly expressed stage-specific genes. Initial analysis of the upstream regions of the genes encoding circumsporozoite protein (CSP) and the circumsporozoite and TRAP-related protein (CTRP) revealed that both sequences contain a number of cis-regulatory elements that had previously been described as sporozoite, ookinete or sexual development specific (Yuda et al., 2010, 2009; Westenberger et al., 2010; Young et al., 2008) (Table S1). As the exact extent of the promoter is unknown for both genes, we refer to the 1200 bp upsteam of the start codon as the promoter. The element TAGCTA, which has previously been identified to correlate with ookinete-specific gene expression (Kaneko et al., 2015), is present 6× in the promoter of ctrp but absent in the promoter of csp. By contrast, three previously reported sporozoite-specific elements (Yuda et al., 2010; Young et al., 2008) are only present in the promoter of csp: CATGCA (5×), TGCATG (3×) and TGCATGCA (3×), the latter representing the fusion of the two shorter ones. This suggests that ApiAP2-Sp – an ApiAP2 that has been characterised in the murine malarial parasite P. berghei – recognises all three elements (Yuda et al., 2010). These motifs correlate with the essentiality of the ApiAP2 genes AP-O and AP-Sp that drive ookinete- and sporozoite-specific gene expression in these respective stages (Modrzynska et al., 2017). We also found a remarkable number of imperfect ookinete- and sporozoite-specific elements that contained a single mismatch. Of those altered motifs, we found seven ookinete-specific elements in the promoter of ctrp and 17 sporozoite-specific elements in the promoter of csp (Table S1). In addition, we searched for elements previously associated with sexual development, and identified the motifs AAGACA (9×) (Young et al., 2008) and TGTANNTACA (1×) (Westenberger et al., 2010) containing single mismatches in the promoter of ctrp but no completely matching element. In the promoter of csp we identified the completely matching element TGTNNACA (1×) (Westenberger et al., 2010) but no motifs with single mismatches (Table S1).

To design our promoter, we used the ctrp promoter as template and edited the sequence as follows. Based on the expression profile of both proteins, we considered the sexual development-specific elements with single mismatches in the promoter of ctrp to be relevant but excluded the completely matching element in the promoter of csp. We incorporated 20 sporozoite-specific elements of the csp promoter into the promoter of ctrp, keeping their relative position in respect of the start codon ATG (Fig. 2A; Table S4). The designed promoter was named Spooki 1.0 in reference to the expected expression pattern in sporozoites and ookinetes.

To evaluate the expression pattern of Spooki 1.0, a parasite line named Spooki 1.0^mCherry was generated that expresses mCherry under the control of Spooki 1.0 in a transcriptionally silent region of chromosome 12 (Fig. 2A; Fig. S1). Live-cell imaging of Spooki 1.0^mCherry ookinetes, oocysts, salivary gland (SG) sporozoites and liver stages revealed that mCherry is expressed in all four stages to varying degrees (Fig. 2B; Fig. S2A). Although all parasites were fluorescent, we found that the amount of mCherry in individual parasites was highly variable. In addition, we observed early oocysts that had not yet initiated sporogony, which showed no mCherry. This indicates a shift to later expression compared to that of CSP, which is already expressed in early oocysts (Simonetti et al., 1993; Kumar et al., 2014; Stone et al., 2015). The amount of mCherry in liver stages was low and only detectable 24 h but not 48 h after infection (Fig. S2A). To elucidate the expression profile of mCherry in Spooki 1.0^mCherry parasites in more detail, we used quantitative reverse transcriptase PCR (qRT-PCR) on different parasite stages. We prepared cDNA samples of ookinete cultures 5 h, 10 h, 15 h and 20 h after setting up the culture as well as of midgut sporozoites (MGS; day 12 post infection), haemolymph sporozoites (HLS; day 14 post infection), salivary gland sporozoites (SGS; day 17 post infection) and liver stages (42 h post infection in vivo, 48 h post infection in vitro) (Fig. 2C; Fig. S2B). Subsequently, we analysed the transcriptional profile of mCherry in comparison to csp, ctrp and trap. While transcripts of csp and ctrp in ookinetes showed comparable levels, mCherry under the control of Spooki 1.0 was transcribed at a slightly higher level. Since trap is a sporozoite-specific gene, transcription levels in ookinetes were very low (Fig. 2C). In sporozoites, gene expression of csp was strongest followed, as expected, by trap (Swearingen et al., 2016). Expression of mCherry was notably lower than trap but still ∼13-fold higher than that of ctrp, which was barely detectable. This shows that the sporozoite-specific cis-regulatory elements were sufficient to confer expression at the sporozoite stage but at a much lower level than the endogenous csp promoter. In liver-stage samples transcript abundance of all genes relative to each other remained comparable to sporozoites but dropped by 50–70% compared to 18sRNA (Fig. S2B).

Because of the relative moderate expression pattern of Spooki 1.0 in sporozoites compared to that of the csp promoter we revisited the original in silico identification of cis-regulatory elements that we used as basis of our motif analysis within the promoters of ctrp and csp in order to design Spooki 1.0. Thereby, we identified putative cis-regulatory elements that had previously been shown to have statistical associations within specific gene clusters of coexpressed genes (Young et al., 2008). We used the search function of Plasmodium Genomics Resource (PlasmoDB) to count motif abundances in the promoters of all ookinete and sporozoite-specific genes (Fig. S3). All motifs were more abundant in P. berghei than in P. falciparum, as expected based on differences in AT content of the promoter region. However, this – surprisingly – showed that the abundance of all motifs in ∼1000 ookinete- and ∼1500 sporozoite-specific genes was very similar to the abundance of these motifs in the promoters of all genes, indicating that promoter sequences show no statistical association between present motifs and expression patterns in general. The same was true when using more stringent selection criteria, resulting for P. falciparum and P. berghei in the top 51 and 50 proteins of sporozoites, respectively, as well as the top 59 and 57 transcribed genes for ookinetes, respectively. In this highly abundant set of sporozoite-specific proteins, four of the genes originally used to identify the sporozoite-specific cis-regulatory elements were included (Young et al., 2008). Nevertheless, the ookinete-specific motif TAGCTA bound by AP2-O was slightly more abundant in the 5′UTR of the most transcribed 1000 ookinete-specific genes and all sporozoite-specific motifs were highly enriched in the 5′UTR of the 18 sporozoite-specific genes in which they had originally been identified in (Young et al., 2008). These seemingly conflicting results are difficult to interpret. One possibility is that Young et al. used only the highly expressed genes as known at the time in their selection. Alternatively, the genes expressed at the ookinete or sporozoite stage might not all be specifically expressed but might be shared between most stages.

Revisiting the used promoter sequences provided also an explanation for the slightly higher gene expression mediated by Spooki 1.0 in ookinetes compared to that of the wild-type (wt) ctrp promoter. While analysing our mutants containing the synthetic Spooki 1.0 promoter, the genome sequence of the ctrp promoter was updated by PlasmoDB (Aurrecoechea et al., 2009), now comprising an additional 5th repeat of 80 bp (Table S4). This repeat contains the AP2-G2 binding site for GTTG[T/C], which was shown to be important for gene silencing (Yuda et al. 2015; Modrzynska et al., 2017). According to these new data, the promoter of ctrp contains five of the GTTG[T/C] motifs, whereas the sequence we used to design Spooki 1.0 only contained four. Additionally, we unintentionally replaced one AP2-G2 motif with a cis-regulatory element of csp. Thus, Spooki 1.0 contains only three instead of five full repeats. This might have led to gene expression being less downregulated by AP2-G2, resulting in the increased transcriptional activity of Spooki 1.0 we observed in comparison to the promoter of ctrp in ookinetes.

Nevertheless, we next tested if Spooki 1.0 can functionally replace a sporozoite-specific promoter. To do so we choose the trap promoter. Lack of TRAP leads to a severe phenotype in sporozoites as they fail to enter salivary glands and do not perform any directed movement (Sultan et al., 1997; Münter et al., 2009). We generated a complemented line named trap(−):Spooki 1.0^trap that expresses TRAP under the control of Spooki 1.0 within the trap locus but in absence of its native promoter region (Fig. S4). This line was expected to express TRAP in addition to CTRP in ookinetes as well as displaying a restored TRAP expression in sporozoites (Fig. 3A). Counting of oocysts revealed no significant difference in numbers compared to those of wt (Fig. 3B), indicating that the additional expression of TRAP in ookinetes had no negative effect on oocyst formation. Compared to two control lines, wt and Spooki 1.0^mCherry, we found similar numbers of trap(−):Spooki 1.0^trap sporozoites in the salivary glands of infected mosquitoes (Fig. 3C), suggesting that trap expressed from the Spooki 1.0 promoter rescues the key functions of TRAP in the mosquito. Detection of TRAP using western blot analyses of salivary gland sporozoites that express trap through the wt or trap(−):Spooki 1.0^trap promoter revealed that TRAP protein levels were restored in trap(−):Spooki 1.0^trap parasites albeit at a much reduced (∼80%) level compared to wt (Fig. 3D). This reduction matched the observed difference in mCherry transcript levels in qRT-PCR data of Spooki 1.0^mCherry sporozoites compared to trap transcripts (Fig. 2C). Immunofluorescence analysis on salivary gland sporozoites of trap(−):Spooki 1.0^trap parasites showed the restored expression of TRAP with the expected localisation to the micronemes (Wengelnik et al., 1999; Kehrer et al., 2016; Klug and Frischknecht, 2017) (Fig. 3E). However, circular gliding motility of salivary gland sporozoites in vitro was strongly reduced to ≤4% of the observed sporozoites when compared to wt sporozoites (72%) (Fig. 3F,G). Most trap(−):Spooki 1.0^trap sporozoites did not attach to the substrate and floated in the medium. This suggests that reduced levels of TRAP impair sporozoite adhesion, which was also strongly reduced in trap(−) sporozoites isolated from the haemolymph (Münter et al., 2009; Hegge et al., 2010). Nevertheless, this dramatic decrease of in vitro motility did not influence infectivity of sporozoites after intravenous injection into mice (Fig. 3H). Similarly, three out of four mice bitten by mosquitoes and infected with trap(−):Spooki 1.0^trap sporozoites gave rise to infections without any differences in prepatency when compared to wt infections (Fig. 3H; Table S2). This suggests that highly reduced TRAP protein synthesis has little effect on in vivo infectivity of sporozoites but strongly reduces sporozoite migration in vitro.

TRAP and CTRP share a conserved cytoplasmic and C-terminal domain, and a TRAP chimera expressing the C-terminus of CTRP functions similar to wt TRAP (Heiss et al., 2008). In the extracellular domain TRAP features one integrin-like I-domain and one thrombospondin type-I repeat (TSR), whereas CTRP features six of each (Morahan et al., 2009). We next tested if the two proteins could complement the function of each other at the respective life cycle stage. Thus, we exchanged the promoter of ctrp in trap(−) parasites with the Spooki 1.0 promoter (Fig. S5). The generated parasite line trap(−):Spooki 1.0^ctrp was expected to express CTRP in both ookinetes and sporozoites but no TRAP at either stage (Fig. 4A). Oocyst numbers in mosquitoes infected with trap(−):Spooki 1.0^ctrp were comparable to wt (Fig. 4B) indicating that the introduced sporozoite-specific cis-regulatory elements do not have a negative effect on the expression of CTRP – as already suggested by expression profiling (Fig. 2C). CTRP was expressed in trap(−):Spooki 1.0^ctrp but not in wt midgut sporozoites, as evaluated by RT-PCR (Fig. 4C). Mosquitoes infected with trap(−):Spooki 1.0^ctrp had significantly higher numbers of sporozoites floating in the haemolymph compared to mosquitoes infected with wt (Fig. 4D). These trap(−):Spooki 1.0^ctrp haemolymph sporozoites were unable to migrate in vitro (Fig. 4E) and did not infect salivary glands (Fig. 4F). In addition, we intravenously injected 40,000 trap(−):Spooki 1.0^ctrp haemolymph sporozoites because it had been shown that 5000 injected haemolymph sporozoites of wt cause reliable infections in mice (Sato et al., 2014). However, this did not result in a blood stage infection in mice infected with trap(−):Spooki 1.0^ctrp (Table S2). We could not evaluate the localisation of CTRP due to the lack of working antibodies; however, immunofluorescence assays with trap(−):Spooki 1.0^ctrp midgut sporozoites showed no TRAP-specific signal (Fig. 4G). This suggests that CTRP cannot replace any TRAP function in sporozoites when transcriptionally induced through Spooki 1.0.

We next tested if TRAP expression in ookinetes in the absence of CTRP can rescue motility and invasion of the midgut epithelium. We generated the line trap(−):ctrp(−):Spooki 1.0^trap (Fig. S6), which expresses no CTRP but TRAP in ookinetes and sporozoites controlled by Spooki 1.0 (Fig. 5A). Generation of this line was achieved by replacing the ctrp gene with a codon-modified version of the trap gene under control of Spooki 1.0 in trap(−) parasites (Fig. S6); note that the endogenous promoter of ctrp is also absent in trap(−):ctrp(−):Spooki 1.0^trap parasites to exclude any effects on gene expression. The expression of TRAP in absence of CTRP was verified by RT-PCR, which revealed only transcription of trap but no transcription of ctrp in trap(−):ctrp(−):Spooki 1.0^trap ookinetes (Fig. 5B). The expression of TRAP in ookinetes was also evaluated through western blot analyses, and protein expression was only observed in trap(−):ctrp(−):Spooki 1.0^trap but not in wt ookinetes (Fig. 5C). The presence of TRAP in trap(−):ctrp(−):Spooki 1.0^trap ookinetes was also shown by immunofluorescence with a TRAP-specific antibody, which showed apical and/or peripheral localisation. In contrast, wt ookinetes treated with this antibody showed no specific signal (Fig. 5D). trap(−):ctrp(−):Spooki 1.0^trap ookinetes were moving at a speed similar to that of wt ookinetes (Fig. 5E; Movie 1). However, no oocysts were detected in mosquitoes infected with trap(−):ctrp(−):Spooki 1.0^trap in two independent feeding experiments (Fig. 5F). This suggests that TRAP complements CTRP function in ookinete motility in vitro but not CTRP function for midgut invasiveness.

The aim of this study, i.e. to generate a synthetic promoter that drives gene expression in two distinct stages of the Plasmodium berghei life cycle, was accomplished with Spooki 1.0. Parasites expressing mCherry under control of Spooki 1.0 display protein expression in ookinetes, oocysts as well as sporozoites, and TRAP expression driven by Spooki 1.0 rescues the phenotype of trap(−) parasites. Nevertheless, the observed expression driven by Spooki 1.0 was rather moderate compared to the expression conferred by the trap (∼20% of protein expression) and csp promoters (∼1.4% of transcript abundance). Therefore, we aimed to improve the expression strength of Spooki 1.0 by shifting upstream two cis-regulatory elements of the sporozoite-specific motif TGCATGCA, which interacts with AP2-Sp (Yuda et al., 2010) (Fig. 6A). This shift was done in respect to the estimated TSS of the ctrp promoter, a site that initiates transcription further upstream compared to initiation through the csp promoter (5′UTR in csp; 284 bp, 5′UTR in ctrp; 870) (Silvie et al., 2014; Otto et al., 2014). In addition, we restored one cis-regulatory element of the AP2-G2-specific motif GTTG[T/C] that was affected in Spooki 1.0 by one of the shifted sporozoite-specific motifs (Fig. 6A). The resulting sequence was termed Spooki 2.0 and synthesised as described for Spooki 1.0. To estimate the expression strength of Spooki 2.0, we generated the parasite line Spooki 2.0^mCh that expresses mCherry under control of Spooki 2.0. Measuring the fluorescence intensity in Spooki 1.0^mCh and Spooki 2.0^mCh ookinetes revealed that both promoter sequences display no significant difference in protein expression at the ookinete stage (Fig. 6B). In contrast Spooki 2.0^mCh salivary gland sporozoites dissected 17 days post infection showed significantly higher (∼7-fold) fluorescence intensity compared to that of Spooki 1.0^mCh salivary gland sporozoites of the same age (Fig. 6C). This difference even increased to ∼12-fold in salivary glands dissected 20 days post infection, indicating earlier expression timing and enhanced expression of Spooki 2.0 at the sporozoite stage (Fig. 6D). Encouraged by these results, we generated the parasite line trap(−):Spooki 2.0^ctrp to probe whether the expected enhanced expression of CTRP can rescue TRAP deficiency in sporozoites (Fig. S5). Yet, as observed for trap(−):Spooki 1.0^ctrp, gliding motility was also absent in trap(−):Spooki 2.0^ctrp haemolymph sporozoites (Fig. 6E). However, sporozoite numbers in the salivary glands increased ∼10-fold compared to trap(−):Spooki 1.0^ctrp but stayed at a very low level (∼100–200 sporozoites per mosquito; Fig. 6F).

To date, only four (Fig. 1) endogenous promoters were available to bypass stage-specific gene expression of genes in Plasmodium. Here, we extended this limited toolbox with the first synthetic promoters – Spooki 1.0 and Spooki 2.0 – that drive expression in ookinetes, sporozoites and the early liver stage. We generated Spooki by combining cis-regulatory elements occurring in the promoters of ctrp and csp. These cis-regulatory elements are enriched in a stage-specific manner and were implied to be important for expression of CTRP in ookinetes and CSP in oocysts, sporozoites and early liver stages. Spooki 1.0 was generated by insertion of predicted sporozoite-specific elements found upstream of the csp gene into the promoter of the ctrp gene. Spooki 1.0 consists of ∼85% ctrp promoter and ∼15% csp promoter, and we anticipated expression levels to be at least according to these proportions. However, our qRT-PCR data showed that, in ookinetes, transcription driven by Spooki 1.0 is ∼7-fold higher compared to endogenous ctrp, and ∼1.5-fold higher compared to endogenous csp (Fig. 2C; average expression at 5 h, 10 h, 15 h and 20 h after culture set-up).

In the sporozoite stage (Fig. 2C; average of all four samples, i.e. MGS purified, MGS not purified, HLS, SGS) Spooki 1.0-driven transcription was only 1.4% compared to that of csp, hence, much lower than anticipated. In comparison, the level of endogenous transcription of trap was 4.8%, and ctrp was 0.1% of the level of endogenous csp transcription. Therefore, we would have anticipated that trap expression driven through Spooki 1.0 only yielded ∼28% of TRAP compared to wt level. Consistent with this, quantification of TRAP by western blotting of trap(−):Spooki 1.0^trap sporozoites revealed that TRAP expression through Spooki 1.0 is ∼20% of wt. We noted also that Spooki 1.0-driven transcription of mCherry is barely detectable in liver stages, with only a weak fluorescent signal observable 24 h post infection. CSP has been detected throughout most of liver stage development, possibly due to the stability of the protein at the plasma membrane rather than continued transcriptional activity (Sacci et al., 2006; Vaughan et al., 2012; Silvie et al., 2014). These results directly indicate that gene regulation through cis-regulatory elements is not as well understood as previously anticipated (Young et al., 2008). A similar complexity has also been recently shown in detail during the transition from the blood stage to ookinetes (Modrzynska et al., 2017). The abundance of motifs, their orientation and context in respect to each other as well as to potential, yet unidentified, motifs play important roles in regulating expression timing and strength – as shown for viral promoters and binding sites of nuclear hormone receptors (Näär et al., 1991; Moolla et al., 2002). However, sensitivity to motif and orientation-dependent alterations are highly variable between different species and different loci, which makes predictions very difficult (Lis and Walther, 2016). By contrast, abrogation of gene expression by disruption of the key element in Plasmodium species is possible – as has been previously demonstrated (Yuda et al., 2010), and our data show that induction of expression to adequate levels by introduction of the key elements alone is not straightforward. Indeed, it could be that ookinete- or sexual development-specific cis-regulatory elements repress transcription at the sporozoite stage, or that introduced elements replaced unknown sporozoite-specific suppressive motifs. However, no repressive function of the introduced sporozoite-specific cis-regulatory elements in Spooki were observed for the ookinete stage (Fig. 2C; Fig. 4B). In addition, the introduced changes in the promoter sequence might effect mRNA stability as well as translation, even if transcription in general is not altered (Silvie et al., 2014; Turque et al., 2016). Nevertheless, RNAseq data suggest that the TSS of ctrp in P. berghei is more than 870 bp upstream of the start codon (Otto et al., 2014), which places most of the sporozoite-specific elements in Spooki 1.0 downstream of the TSS and probably renders them functionally silent.

Following these first results we generated a second promoter – Spooki 2.0 – using information predicting the TSS within the ctrp and csp promoter (Otto et al., 2014; Silvie et al., 2014). To achieve reduced expression at the ookinete and increased expression at the sporozoite stage, we shifted the two main sporozoite motifs such that their position relative to the TSS in Spooki 2.0 is as in the csp promoter. Additionally, we reconstituted one AP2-G2 motif that had been replaced by one of the sporozoite-specific motifs in Spooki 1.0. As expected, relocalisation of the two sporozoite-specific motifs significantly increased the expression of mCherry in the sporozoite stage, especially in mature salivary gland sporozoites; however, we did not observe any change in protein expression in ookinetes. We observed also that the measured fluorescence intensities were significantly different between salivary gland sporozoites of day 17 and day 20, indicating that protein expression guided by Spooki 1.0 and Spooki 2.0 continues even after invasion of the salivary glands. This result explains the broad variation in expression of mCherry observed initially, which we interpreted as stochastic differences in gene expression from cell to cell. Whereas the latter explanation definitely influences the expression pattern, the effect is enhanced in salivary gland sporozoites because of the consistent invasion of the glands over time. Since salivary gland invasion takes place over several days, the observed sporozoites represent a mosaic of cells comprising cells of different age (i.e. time spent within the salivary gland) and of different protein (mCherry) content.

Deletion of csp, the gene that is expressed most strongly in sporozoites, inhibits sporozoite formation (Ménard et al., 1997). Deletion of trap, the second most strongly expressed gene, leads to non-infectious sporozoites that fail to migrate (Sultan et al., 1997; Münter et al., 2009). Reducing expression of CSP to 20% of wt levels leads to malformed and non-infectious sporozoites (Thathy et al., 2002). Since Spooki 1.0-driven transcription of mCherry at the sporozoite stage was approximately 80% lower than that of trap, we investigated whether Spooki 1.0-driven transcription of trap causes any measurable phenotypes in motility or infectivity. Complementation of trap(−) parasites with trap under control of Spooki 1.0 revealed complete reconstitution of salivary gland invasion and infectivity to mice (Fig. 3C,H; Table 1; Table S2). However, productive gliding motility in vitro was hugely diminished. Importantly, most sporozoites would not adhere to the glass surface and, hence, were unable to initiate motility. This reduction is likely to be due to the reduced expression of TRAP in trap(−):Spooki 1.0^trap sporozoites, which is only ∼20% of wild-type as observed by western blotting. Yet, this amount of TRAP appears to be sufficient for productive invasion of the salivary glands and infectivity to mice. How can parasites that are unable to attach and move on a flat surface be fully infectious? This is likely to be due to the fact that a 2D surface does not provide enough adhesion sites as a 3D environment and, indeed, similar effects have been observed for mammalian cells (Lämmermann et al., 2008) as well as for sporozoites (Bane et al., 2016; Moreau et al., 2017; Douglas et al., 2018) and Toxoplasma tachyzoites (Whitelaw et al., 2017; Gras et al., 2017). Formation of adhesion sites is likely to be dependent on the interplay between TRAP-family proteins and the underlying actin cytoskeleton (Münter et al., 2009; Hegge et al., 2012; Bane et al., 2016; Quadt et al., 2016; Whitelaw et al., 2017). Small alterations of the macromolecular complexes needed for sporozoite activation can, thus, have distinguishable effects in vitro, while not translating into a measurable effect in vivo (Moreau et al., 2017). The lower expression of TRAP might, thus, lead to less-stable adhesion complexes that decrease the capacity of the sporozoite to adhere to a flat substrate and, hence, to migrate in vitro. Also, during gliding in vivo, cells are exposed to a variety of different ligands (Perschmann et al., 2011), which might lead to enhanced microneme secretion (Carruthers and Sibley, 1999; Carruthers et al., 1999a,b; Carey et al., 2014; Bane et al., 2016) and, thus, compensate for the lowered gene expression.

To test if Spooki 1.0-driven transcription of ctrp can rescue the motility and infectivity phenotype of trap(−) parasites, we replaced the native promoter of ctrp with Spooki 1.0 in the trap(−) background. The resulting parasite line readily formed oocysts in infected mosquitoes, indicating that the enhanced transcription of ctrp by Spooki 1.0 does not affect the fitness of ookinetes. Also, sporozoite numbers were within the normal range. However, we found no motile sporozoites in the haemolymph and no sporozoites within the salivary glands. Furthermore, injection of haemolymph sporozoites into mice did not lead to infection (Table S2). Replacing Spooki 1.0 with Spooki 2.0, did not significantly change these results. Although we observed an increase in the number of salivary gland sporozoites, these reflected only 1% of wild-type infections. Gliding motility in the 2D in vitro assay was also not observed for CTRP expression under the control of Spooki 2.0 in absence of TRAP.

Considering the complete absence of gliding, efficient salivary gland invasion and infectivity to mice of trap(−):Spooki 1.0^ctrp and trap(−):Spooki 2.0^ctrp parasites suggests that there are differences in ligand recognition or outside-in signalling after ligand recognition (Song et al., 2012) between TRAP and CTRP. However, we cannot strictly exclude that CTRP is not correctly expressed, mis-trafficked or misfolded in sporozoites. The replacement of the C-terminus of TRAP with that of CTRP has been previously shown to be functional and, thus, correctly trafficked in sporozoites (Heiss et al., 2008). However, the sequences present in TRAP, which have previously been shown to be important for micronemal trafficking in Toxoplasma gondii (Di Cristina et al., 2000), were not part of this C-terminal swap. CTRP might also be co-trafficked by yet unknown proteins – as is the case in Toxoplasma gondii for MIC2 by M2AP (Rabenau et al., 2001). Alternatively CTRP might be trafficked into different subsets of micronemes in sporozoites that are not secreted during gliding or CTRP might require a chaperone that is only expressed in ookinetes for correct folding of the six I- and TSR-domains. Hence, expression of chimeric proteins, e.g. a TRAP featuring all six I-domains of CTRP, might be a way forward in dissecting which parts of these proteins confer potentially different functionality in the different host environments faced by the ookinete and sporozoite.

We investigated if the Spooki 1.0-driven expression of TRAP in ookinetes can complement CTRP function. Interestingly, we observed that ookinetes that express trap under the Spooki 1.0 promoter in absence of CTRP were still able to glide at speeds comparable to wild-type ookinetes in vitro (Fig. 5E, Movie 1). Since ookinetes that lack the ctrp gene are not able to migrate (Yuda et al., 1999; Dessens et al., 1999; Templeton et al., 2000; Ramakrishnan et al., 2011), this shows that TRAP is sufficient for ookinetes to perform gliding motility. However, ookinetes expressing TRAP were unable to establish an infection in mosquitoes in several independent feeding experiments (Fig. 5F). This could indicate that TRAP is unable to recognize ligands on the midgut epithelium. Also, TRAP might be less stable in the harsh environment of the midgut lumen, while CTRP evolved to allow traversal of the peritrophic matrix. Again, the expression of protein chimeras might be a way towards a better understanding on how these two adhesins function in their different environments.

We designed Spooki 1.0/2.0 as prototypic synthetic promoters that drive gene expression at different parts of the Plasmodium life cycle. We were able to tune timing of Spooki 1.0 or 2.0-driven expression to the stages we intended it to be active, while being inactive elsewhere along the life cycle. However, we did not predict the strong expression in ookinetes and the weak expression in sporozoites of Spooki 1.0. However, it might be possible to create a promoter that yields higher gene expression at the sporozoite stage by creating a ‘reverse Spooki’ that incorporates cis-regulatory elements of the ctrp promoter into the csp promoter. But the ideal expression strength depends on the gene on which the promoter swap will be performed. Although promoters that enable strong gene expression might be useful in some cases, expression guided by Spooki 1.0/2.0 should be sufficient for genes that are consistently expressed at levels lower than csp. Thus, our work provides a step towards the direction of tuning promoters but, at the same time, cautions about the general applicability of our strategy across the parasite life cycle. Nevertheless, we think that, with increasing knowledge about ApiAP2s and their DNA-binding specificities, this approach will allow us to generate more types of attenuated parasites for experimental vaccination studies. For example, one could generate a promoter that is switched on from the blood stage until the early liver stage and then switches off. This would enable the deletion of a large number of house-keeping genes and, hence, providing an array of parasite lines that are attenuated at different time points depending on severity or essentiality, which could ultimately be important for fine-tuning an attenuated parasite strain for humans (Kumar et al., 2016; Matuschewski, 2017). Yet, our results clearly suggest that the design of such a set of promoters is still a long way off.

To identify the abundance of specific cis-regulatory elements we used the search strategy option of PlasmoDB (www.plasmodb.org; version 11–37). We used a RNAseq data set to select the top 20 percentile of ookinete expressed genes (Otto et al., 2014) and all available mass spectrometry data to identify sporozoite-specific proteins (Lindner et al., 2013; Swearingen et al., 2016). Next, we transformed all genes by orthology to the respective species or obtained the orthologues of P. berghei and P. falciparum. Subsequently, DNA motif search was used to identify each motif on either strand of the 1000 bp of 5′ UTR upstream of the start codon. To calculate the random chance of motif appearance, we determined the abundance of each bp in the 1000 bp upstream regions of all genes of P. falciparum and P. berghei (Pf A: 41.88% T, 44.18% C, 6.83% G, 7.10%; Pb A: 39.83% T, 40.16% C, 9.82% G, 10.19%). The chance P of a given motif GCTA of length k within n bases is P=2−(1−G·C·T·A)^(n−k)−(1−pG·pC·pT·pA)^(n−k) where G indicates the species-specific chance for the base G to occur, and pG indicates the chance of the pairing base of G, i.e. C.

As basis for a synthetic transcriptional unit we looked for promoters that guide transcription either exclusively in ookinetes, or in oocysts and sporozoites. We decided to work with the promoter of the circumsporozoite protein (CSP) that results in the strongest expression known in sporozoites (Le Roch et al., 2003; Rosinski-Chupin et al., 2007; Swearingen et al., 2016), and the promoter of the circumsporozoite and TRAP-related protein (CTRP) that is exclusively expressed in ookinetes (Otto et al., 2014; López-Barragán et al., 2011). We defined the promoter as 1.2 kb upstream of the start codon of the ctrp and csp genes and screened both sequences for cis-regulatory elements specific for gene expression in ookinetes, sporozoites and during sexual development (Young et al., 2008; Westenberger et al., 2010; Yuda et al., 2009, 2010). As both csp and ctrp had originally been included in the seed groups for sporozoite and sexual stage expression, respectively (Young et al., 2008), it was expected to identify their respective cis-regulatory elements. Since single mismatches in DNA-binding motifs have resulted in slight reduction of expression or complete downregulation in Plasmodium (Yuda et al., 2010), we decided to look also for cis-regulatory elements that contain one mismatch because these might still function as low-affinity binding sites. For the generation of a synthetic transcriptional unit that combines properties of the csp promoter and the ctrp promoter by shuffling the identified elements we had to use a template sequence that already contained essential regulatory sequences as, for example, the promoter region. This was necessary since information about the composition of promoter elements and additional regulatory sequences that are part of the promoter is still restricted in Plasmodium, and is mainly limited to blood stages (Helm et al., 2010; Adjalley et al., 2016; Zhu et al., 2016). Therefore, the design of a synthetic promoter from scratch would be difficult. The promoter of csp has been previously used in overexpression studies, which revealed specific effects of an actin sequestering protein due to the extremely high expression levels (Sato et al., 2016). We decided to use the promoter of ctrp as basis for the insertion of sporozoite-specific cis-regulatory elements from the csp promoter, expecting expression levels comparable to those of ctrp but lower expression levels than those of csp, which we hoped would allow us to address the function of genes essential at the sporozoite stage without running into potential problems with overexpression. For the insertion of the sporozoite-specific elements into the ctrp promoter, we set the following rules: elements were inserted by mutating the sequence used as basis, this way inserted motifs have the same distance to the start codon compared with those under normal conditions; since it is unknown whether the context of the introduced motifs has an impact on protein binding, we mutated also the three neighbouring bases on each side of the introduced element to mimic the native environment; when there was an overlap between a sporozoite- and ookinete-specific element the sequence was left unchanged, as we did not want to reduce gene expression in ookinetes. This way seven sporozoite-specific motifs without mismatch and 13 sporozoite-specific motifs with mismatch of the csp promoter were introduced into the promoter of ctrp. The sequence of Spooki 1.0 (Table S4) was synthesised by using GeneArt (Invitrogen) and cloned into EcoRI and NdeI restriction sites in front of the fluorescence marker mCherry within the Pb262 vector (Deligianni et al., 2011; Kooij et al., 2005) that was used for additional integration of genes into a locus on chromosome 12. For the generation of Spooki 2.0 two cis-regulatory elements with the sporozoite-specific motif TGCATGCA were shifted upstream to place them at the same distance to the TSS of ctrp (putative start of 5′UTR 870 bp upstream of ATG) as they were normally positioned to the TSS in the csp promoter (putative start of 5′UTR 284 bp upstream of ATG). In addition, we restored one cis-regulatory element with the AP2-G2-specific motif GTTG(T/C) that was affected by one of the shifted sporozoite-specific motifs in Spooki 1.0. The sequence of Spooki 2.0 was synthesised and cloned as described for Spooki 1.0. The final vectors were linearised (PvuI), purified (High Pure PCR Product Purification Kit, Roche) and transfected into P. berghei ANKA schizonts according to standard protocols (Janse et al., 2006).

Parasites lacking trap, termed trap(−) were generated using a vector generously provided by PlasmoGem (PbGEM-107890) (Pfander et al., 2011). Positive transfectants were cloned as described below (see below under ‘Generation of isogenic parasite lines’) and clonal lines were treated with 5-fluorocytosine (1 mg/ml) to select for parasites that lost the resistance cassette yfcu-hdhfr (Braks et al., 2006). Negatively selected parasites were cloned a second time and, subsequently, used for phenotypic characterisation and as recipient line for further transfections.

The synthetic promoters Spooki 1.0/2.0 were cloned into the Pb262mCherry vector using EcoRI and NdeI, giving rise to the vectors Pb262Spooki 1.0/2.0^mCherry. To exchange the promoter of ctrp with the Spooki 1.0/2.0 transcriptional units, the vectors Pb262Spooki1.0/2.0CTRP were generated. To do this the beginning of the ctrp coding region was amplified with primers P1106 and P1105 and cloned into Pb262Spooki1.0 or Pb262Spooki2.0 using NdeI and EcoRV. The region upstream of the ctrp promoter was amplified with primers P1107 and P1108 and cloned into both vectors using KasI and EcoRV. The resulting vectors Pb262Spooki1.0/2.0CTRP were sequenced and linearised with KasI for transfection into Plasmodium berghei ANKA (PbANKA) trap(−), resulting in trap(−):Spooki 1.0^ctrp and trap(−):Spooki 2.0^ctrp. The resulting parasite lacks the original 1.607 bp upstream of the ctrp coding region and contains an insertion of 7.813 bp ending with the Spooki 1.0 or Spooki 2.0 transcriptional unit.

To generate a parasite line expressing trap under control of Spooki 1.0, the transcriptional unit was excised from the vector Pb262Spooki1.0CTRP using NdeI and BssHII and cloned into the vector Pb238cmTRAP, containing a codon-modified version of the trap gene. A region upstream of the promoter of trap was amplified using primers P1232 and P1233 and cloned upstream of Spooki 1.0 using SalI and BssHII. The resulting vector Pb262Spooki1.0TRAP was sequenced, linearised with SalI and KpnI and transfected into trap(−), resulting in trap(−):Spooki 1.0^trap. trap(−):Spooki 1.0^trap parasites that lack he original 1.531 bp upstream of the trap coding region and contain the codon modified trap gene under the control of the Spooki 1.0 transcriptional unit within the trap locus.

To generate a parasite line expressing trap under the control of Spooki 1.0 while lacking ctrp, we used the intermediate vector Pb262Spooki1.0TRAP to generate Pb262Spooki1.0TRAP-CTRPKO. The 3′UTR of ctrp was amplified with primers P1327 and P1328 and the sequence upstream of the ctrp promoter was amplified with the primers P1329 and P1330. The resulting PCR products were purified and combined by PCR using P1327 and P1330 and cloned into the vector using KpnI and SmaI. The final vector Pb262spooki1.0TRAP-CTRPKO was sequenced, linearised with BamHI and transfected into PbANKA trap(−), giving rise to trap(−):ctrp(−):Spooki 1.0^trap. The resulting parasite line lacks the original 1607 bp upstream of the ctrp coding region as well as the entire coding region of ctrp and contains the codon modified trap gene under the control of the Spooki 1.0 element within the ctrp locus.

Clonal parasite lines were generated by serial dilution of parental populations obtained from transfections. Donor mice were injected intraperitoneally with 150–200 µl frozen parasite stocks of parental parasites and allowed to grow under selective pressure (0.07 mg/ml pyrimethamine within the drinking water). Once a parasitaemia of 0.5–1.0% was reached, the blood of infected mice was withdrawn by cardiac puncture and diluted in PBS to 0.8 parasites per 100 µl. To generate clonal populations 8–10 naive NMRI mice were injected with 100 µl each. Parasites were allowed to grow until a parasitaemia of 1.5–2.0% was reached. The blood of infected mice was withdrawn, parasites were purified for the isolation of genomic DNA with the Blood and Tissue Kit (Qiagen Ltd) and parasite stocks were generated and stored in liquid nitrogen.

Frozen parasite stocks of isogenic populations were thawed and intraperitoneally injected (∼200–300 µl) in naive NMRI mice. Parasites were allowed to grow until a parasitaemia of ≥1.5% was reached before mice were bled by cardiac puncture. The obtained blood was used for a fresh blood transfer of 20 million blood stage parasites each into two naive NMRI mice by intraperitoneal injections. Three days post transfer the gametocytaemia of both mice was determined by examining the number of exflagellation centres in a drop of blood after 10 min incubation at room temperature. When at least one exflagellation centre per field of view was observed, mosquitoes were allowed to feed on the mice. Mosquitoes used for blood feeding were at least three but maximal seven days old and starved overnight prior to the day of the planned feeding. Mice were anaesthetised with a mixture of ketamin and xylazin (87.5 mg/kg ketamine, 12.5 mg/kg xylazine) and placed with the ventral side on the mosquito cage for ∼20 min. To enhance the bite rate during this time, mice were covered with paper tissues. Infected mosquitoes were kept at 80% humidity and 21°C in a climate chamber until analysed.

Twenty million blood stage parasites were intraperitoneally injected into naive mice that had been pre-treated with phenylhydrazin once 72–24 h before transfer (100 µl of a 6 µl in 1 ml PBS dilution). In order to reduce asexual blood stage parasites drinking water of mice was exchanged on day three post infection against water supplemented with 30 mg/ml sulfadiazin. Four days post infection blood was withdrawn by cardiac puncture and immediately transferred into 10 ml ookinete medium (RPMI-1640, 25 mM HEPES, 300 mg/l, L-glutamine, 10 mg/l hypoxanthine, 50,000 units/l penicillin, 50 mg/l streptomycin, 2 g/l NaHCO₃, 20.48 mg/l xanthurenic acid, 20% foetal calf serum (FCS) pH 7.8) and incubated for 20 h at 19°C.

For RT-PCR samples ookinete cultures were stopped at certain time points (at 5, 10, 15 and 20 h), RBCs were lysed with 170 mM NH₄Cl for 2–5 min on ice and the resulting parasite pellet was washed 1× with PBS and resuspended in 1 ml Trizol for RNA isolation.

Ookinete gliding assays were performed by using a drop of culture in ookinete medium on a glass slide covered with a cover slip. Images were taken at a frame rate of 20 s for ∼10 min on a Zeiss Axiovert microscope (25× objective, NA 0.8). Image analysis was performed with the manual tracking tool of ImageJ/FIJI (Schindelin et al., 2012).

To perfom gliding assays, the salivary glands of 20−30 infected mosquitoes were isolated and pooled in 50 µl RPMI in a plastic reaction tube (Eppendorf). Isolated salivary glands were disrupted with a pestle to release sporozoites and purified using gradient centrifugation on top of 2.5 ml 17% w/v Accudenz^® solution at 2500 g for 20 min (Kennedy et al., 2012). Purified salivary gland sporozoites were resuspended in 100 µl RPMI and transferred into a 96-well plate with optical bottom. Subsequently, the sporozoite solution was mixed with an equal volume of RPMI containing 6% bovine serum albumin (BSA) and centrifuged for 3 min at 80 g (Heraeus Multifuge S1). Sporozoites were imaged with an Axiovert 200 M (Zeiss) fluorescence microscope and a 25× (NA 0.8) objective. Image series were acquired at one frame every 3 sec for 5 min and analysed with ImageJ. Sporozoites were classified as moving when they were able to move in at least one full circle within 5 min. All sporozoites that behaved differently (i.e. wavers, attached or floating) were categorised as non-moving (Hegge et al., 2010).

To isolate haemolymph sporozoites, ∼20 infected mosquitoes were immobilised by incubation on ice for 10 min. Afterwards mosquitoes were placed on a microscope slide and the last segment of the abdomen was removed. Subsequently, a hand-made Pasteur pipette filled with RPMI was pierced into the side of the thorax of the mosquito. By injecting RPMI the mosquito haemocoel was flushed and the drained solution was collected on a piece of parafilm. The solution was then transferred into a plastic reaction tube (Eppendorf) and centrifuged for 5 min at 9400 g (Thermo Fisher Scientific, Biofuge Primo). The supernatant was discarded and sporozoites resupended in 100 µl RPMI. The sporozoite solution was transferred into a 96-well plate and gliding assays were performed as described previously for salivary gland sporozoites.

Midguts or salivary glands of infected mosquitoes were dissected in RPMI in a plastic reaction tube (Eppendorf). Subsequently, sporozoites were released from the mosquito tissue by mechanical rupture with a pestle. Salivary gland sporozoites in solution were pipetted in wells of a 24-well plate that contained round coverslips. Sporozoites within wells were mixed with the same volume of RPMI containing 6% bovine serum albumin (BSA) to reach a final concentration of 3% BSA per well. Activated salivary gland sporozoites were allowed to glide for 20–30 min before the supernatant was discarded and the cells were fixed in 4% paraformaldehyde (PFA) for 1 h at room temperature or at 4°C overnight. In contrast midgut sporozoites were kept within plastic reaction tubes (Eppendorf) and pelleted by centrifugation for 3 min at 9400 g (Thermo Fisher Scientific, Biofuge Primo). Midgut sporozoites were directly fixed in PFA without previous activation with BSA. All samples were washed 3× with PBS before cells were blocked (PBS containing 2% BSA) or blocked and pemeabilised (PBS containing 2% BSA and 0.5% Triton-X-100) for 1 h at room temperature or at 4°C overnight. Note that midgut sporozoites were centrifuged for 3 min at 9400 g (Thermo Fisher Scientific, Biofuge Primo) after each incubation or washing step. Blocked or permeabilised sporozoites were incubated with primary antibody solutions (anti-CSP mAb 3D11 (Yoshida et al., 1980), 1:10 (cell culture supernatant); rabbit anti-TRAP, 1:100 diluted) for 1 h at room temperature in the dark and, subsequently, washed 3×with PBS. Note that the anti-TRAP antibody was generated against the peptide with the sequence 5′-AEPAEPAEPAEPAEPAEP-3′ by Eurogentec. Antibodies against the same epitope(s) had been generated previously were shown to specifically recognize TRAP in immunofluorescence assays and on western blots (Ejigiri et al., 2012). Afterwards, samples were incubated with solutions of secondary antibody (Cy5 goat anti-mouse, 1:500; AlexaFluor 488 goat anti-rabbit, 1:500 diluted; Hoechst 33342 1:1000 dilution of a 10 mg/ml stock solution in DMSO) for 1 h in the dark. Subsequently, samples were washed 3× with PBS. Samples of midgut sporozoites were resuspended in 50 µl PBS and carefully pipetted on microscopy slides. Cells within the solution were allowed to settle for 10–15 min before the remaining solution was removed and samples were sealed with round cover slips and 7 µl mounting medium (Thermo Fisher Scientific, ProLong Gold Antifade Reagent). Samples with salivary gland sporozoites that had already been fixed on round cover slips were carefully dried on a paper tissue before they were mounted on microscopy slides, as described previously. Samples were kept at room temperature overnight and then stored at 4°C or were directly examined using a spinning disc confocal microscope (Nikon Ti series) or an Axiovert 200 M (Zeiss). Images were acquired with a 60× objective (CFI Apo TIRF 60× H; NA 1.49) or a 63× objective (NA 1.4). Ookinetes were cultured as described above and antibody staining was perfomed in solution, similar to that of midgut sporozoites.

To image oocysts and sporozoites, midguts and salivary glands of infected mosquitoes were placed in a drop of RPMI or PBS on a microscope slide, covered with a cover slip and sealed with paraffin. Imaging of ookinetes was perfomed similarly using ookinete medium, while red blood cells were lysed before – as described for ookinete cultures. For the generation of liver stages HepG2 cells were seeded in glass-bottom Petri-dishes (MatTek corporation) and infected with salivary gland sporozoites in 100 µl of complete DMEM for 2 h. Afterwards cells were washed with PBS and cultivated in complete DMEM supplemented with an antibiotic-antimycotic coctail (Gibco, Thermo Fisher Scientific), and imaged in the presence of the cell-permeable DNA dye Hoechst 33342 at 24 and 48 h post infection. All images were acquired by using an Axiovert 200 M (Zeiss) microscope (63× objective, NA 1.4).

To measure fluorescence intensities of Spooki 1.0^mCh and Spooki 2.0^mCh ookinetes, culturing and imaging was performed as described above. Spooki 1.0^mCh and Spooki 2.0^mCh sporozoites were dissected from salivary glands at day 17 and day 20 post infection, and activated in RPMI containing 3% BSA prior to imaging. Fluorescence intensities were acquired with an Axiovert 200 M (Zeiss), using a 63× Objective (NA 1.4) and a Coolsnap_HQ2 camera (Photometrics). Differential interference contrast (DIC) and Hoechst staining were imaged additionally to verify sporozoite viability and focus. The mCherry signal was acquired by exposing cells for 400 ms with an excitation wavelength of ∼546 nm and detecting emission light with a wavelength of 575–640 nm. For each viable sporozoite and ookinete, a square of 10×10 pixels was analysed covering the central part of the cell to determine the mean fluorescent intensity. In addition, the local background fluorescence intensity was measured (also 10×10 pixels) for every cell analysed, to correct for uneven illumination and differences in sample quality.

Sporozoites were isolated from salivary glands and lysed in RIPA buffer. On each lane, ∼20,000 cells were loaded; after protein separation blotting onto a nitrocellulose membrane was done using the BioRad Transblot turbo system. Blocking was done with 5% milk powder dissolved in PBS containing 0.1% Tween 20. Ookinetes were cultured as described above and cells were lysed in RIPA buffer. Of the culture, ∼2% was loaded per lane. Detection of proteins was performed using the following antibodies: rabbit anti-TRAP: 1:300; mouse anti-CSP (mAb 3D11) 1:200, (Yoshida et al., 1980); mouse anti HSP70 (Wiser and Plitt, 1987); goat anti-mouse HRP 1:10,000 (GE Healthcare); goat anti rabbit HRP 1:10,000 (BioRad). Protein expression levels were quantified using ImageJ, measuring the sum of all pixels (corrected with local background) and normalised using the relative loading control CSP or HSP70.

Total RNA of ookinetes, sporozoites and liver stages was isolated using the TRIzol reagent according to the manufacturer's protocol (Thermo Fisher Scientific). Ookinetes and sporozoites were centrifuged for 2 min at 9400 g (Thermo Fisher Scientific, Biofuge Primo) in a plastic reaction tube (Eppendorf), subsequently the pellet was resuspended in a small amount of residual volume and dissolved in 1 ml of TRIzol. If possible RNA isolation was performed with ≥1 million sporozoites or one complete ookinete culture. For in vitro liver stages, 48 h after addition of sporozoites, cells were washed 1× with PBS and subsequently removed using a cell scraper (Greiner Bio One; Germany) after the addition of 1 ml of TRIzol. For isolation of in vivo liver stages, 42 h after intravenous injection of sporozoites, mice were anaesthetised with a mixture of ketamin and xylazin (87.5 mg/kg ketamine, 12.5 mg/kg xylazine) and the liver was perfused with 10 ml of PBS through the portal vein. The liver was extracted, quickly washed in PBS, cut into several pieces and homogenised with a TissueRuptor (Qiagen) in 4 ml TRIzol in a 50 ml falcon tube. Isolated RNA was digested with DNase using the Turbo DNA-free^TM Kit (Invitrogen) and cDNA synthesis was performed subsequently using the First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). DNase digest and cDNA synthesis were performed according to the manufacturers protocols. Primers used for PCR amplification of 18 s RNA, mCherry, csp, ctrp and trap transcripts using Taq polymerase are given in Table S3.

For quantitative PCR, SYBR Green PCR Master Mix (Thermo Fisher Scientific) was used with the ABI7500 thermo cycler (Applied Biosystems). Reaction volume was 15 µl with 0.4 µl cDNA, 7.5 µl master mix and a primer concentration of 0.67 µM in technical triplicates in AB-1100 Thermo-Fast 96 PCR Detection Plates (Thermo Fisher Scientific) using QPCR SEAL optical clear film (VWR International GmbH).

Experiments with animals were perfomed according to FELASA and GV-SOLAS guidelines. All experiments were approved by the responsible German authorities (Regierungspräsidium Karlsruhe, Germany).

Plasmodium parasites were cultured in Naval Medical Research Institute (NMRI) mice that were obtained from JANVIER LABS. Experiments to estimate the infectivity of sporozoites were performed with C57Bl/6 mice from Charles River Laboratories. All experiments were performed only in 6–8 week old females. Transfections and genetic modifications were done in the Plasmodium berghei ANKA strain (Vincke and Lips, 1948) or in a strain derived from this isolate (trap(-)).

Data were statistical analysed using GraphPad Prism 5.0 (GraphPad, San Diego, CA). P-values are given in the legends to the respective graphs, P<0.05 was considered statistically significant.

We thank Markus Ganter as well as Ross Douglas for comments on the manuscript, Miriam Reinig and Christian Sommerauer for rearing Anopheles stephensi mosquitoes, and Dietmar Mehlhorn for technical assistance. D.K. was a member of the Hartmut Hoffman-Berling International Graduate School (HBIGS).