Role of host cell traversal by the malaria sporozoite during liver infection

To investigate the interactions of Plasmodium sporozoites with KCs and ECs in the liver sinusoids, the three cell types were differentially labeled and their dynamic interplay was examined in the liver of mice using intravital laser spinning-disk confocal microscopy. We used P. berghei sporozoites constitutively expressing RedStar fluorescent protein (RFP⁺; Sturm et al., 2009). ECs were visualized using flk1-gfp transgenic C57BL/6 mice (Xu et al., 2010), which express GFP in ECs including in the liver sinusoids. The fluorescence of the thin EC cytoplasmic processes sharply delineated the sinusoidal lumen, thus permitting us to define the exact sites and moments of sporozoite crossing (Fig. 1). KCs were labeled using Alexa Fluor 647–conjugated anti-F4/80 monoclonal antibody injected intravenously in the flk1-gfp mouse 30 min before sporozoite injection (Fig. 1 A, left). The F4/80 specificity was confirmed in vivo by depleting KCs with clodronate (Van Rooijen and Sanders, 1994), which completely abolished KC staining in the sinusoids (Fig. 1 A, middle). F4/80 labeling also colocalized with fluorescent beads taken up by phagocytic cells (Fig. 1 A, right) and with weakly GFP⁺ myelomonocytic cells in the liver of lys-gfp transgenic mice (Fig. 1 B). Importantly, neither the anti-F4/80 antibody nor GFP expression in ECs impaired P. berghei sporozoite infectivity (not depicted), and thus presumably did not alter sporozoite, EC, or KC behaviors in vivo.

After intravenous injection of ∼3 × 10⁵ RFP⁺ sporozoites in F4/80-labeled flk1-gfp mice, individual sporozoites were imaged in the left liver lobe in a volume of 125 × 125 × 40 µm³ (8–10 confocal Z-stacks) until a crossing event was observed, i.e., a sporozoite (pseudo-colored in green) translocating from the sinusoidal lumen delineated by the GFP fluorescence (pseudo-colored in white) into the black parenchyma (Fig. 1 C). 60 such crossing events were recorded, of which 68% (41/60) occurred at sites overlapping or closely adjacent to a F4/80-positive area, noted as KC⁺ events. At least ∼32% (19/60) of the sporozoite crossing events occurred in F4/80-negative areas, i.e., without interaction with KCs at the crossing site, KC⁻ (Fig. 1 C). We also observed that sporozoites of P. yoelii, another rodent-infective Plasmodium species, were able to cross the sinusoidal barrier in a zone lacking KCs (unpublished data). This indicated that Plasmodium sporozoites might cross the barrier not only through KCs but also through ECs.

When we analyzed the GFP⁺ ECs matching the site of sporozoite crossing during events in the KC⁻ group, we frequently observed a rapid, specific, and significant decrease in fluorescence intensity, a phenotype we called EC fading (ECfad⁺), which occurred during or just before the complete passage of the sporozoite through the EC (Fig. 2 A). To determine whether the ECfad⁺ phenotype resulted from sporozoite traversal of ECs, we characterized the phenotype in vitro using primary liver sinusoid flk1-gfp (GFP⁺) ECs and RFP⁺ sporozoites by dynamic microscopy. When a primary liver sinusoidal EC monolayer was incubated with sporozoites, a specific fading of EC fluorescence was frequently observed, triggered by sporozoite contact with, and presumably traversal of, ECs (Fig. 3 A), similar to the ECfad⁺ phenotype observed in vivo. To test whether the ECfad⁺ phenotype correlated with a loss of host cell membrane integrity and GFP leakage, we imaged the interactions in the presence of propidium iodide (PI), a membrane-impermeant molecule that only highly fluoresces after entry into cells and binding to nucleic acids. As shown in Fig. 3 B, the EC fading induced by sporozoites was concomitant with the increase in PI fluorescence in the EC nucleus. This suggested that sporozoites were breaching the EC plasma membrane, allowing the simultaneous entry of PI into, and leakage of GFP from, traversed ECs. To further confirm this, the EC monolayer was incubated with CT-deficient GFP⁺ SPECT2⁻ sporozoites (Amino et al., 2008) or control GFP⁺ sporozoites (Ishino et al., 2006), and sporozoite–EC interactions were imaged over 90 min by dynamic microscopy. SPECT2 is a parasite protein that contains a perforin-like domain and is essential for the sporozoite CT activity (Ishino et al., 2005). Although control sporozoites caused the fading of up to 30% of ECs, with a linear and direct relationship between the frequency of faded ECs and the ratio of sporozoite to ECs (Fig. 3 C, yellow arrowheads and graph), the GFP⁺ SPECT2⁻ sporozoites were unable to induce any detectable EC fading in the same conditions (Fig. 3 C). We conclude that the ECfad⁺ phenotype is a consequence of the sporozoite membrane wounding ability and that, in vivo, the ECfad⁺ phenotype is a signature of sporozoite traversal of ECs.

Intravital imaging revealed that ∼53% of the sporozoite crossing events (32/60) were ECfad⁺, i.e., ultimately depended on EC traversal. Of these, 16.7% were ECfad⁺ KC⁻ (Fig. 2, A and D), corresponding to crossing events by direct EC traversal, whereas 36.6% were ECfad⁺ KC⁺ (Fig. 2, B and D), corresponding to crossing events by EC traversal after interactions with a KC. Finally, because prior observations were made after intravenous injection of sporozoites, an inoculation route which bypasses the natural first steps of the sporozoite in the skin and might thus impact sporozoite behavior in the liver sinusoids, we also imaged sporozoites after inoculation by mosquito bites. Although only few crossing events (n = 8) could be recorded, reflecting the small number of sporozoites inoculated by natural transmission, most were associated with an ECfad⁺ phenotype (Fig. 2, C and E) similar to that induced by sporozoites injected intravenously. Therefore, although these data are too limited to allow an estimation of the proportion of crossing events by EC traversal after sporozoite inoculation in the skin, they show that the ECfad⁺ phenotype is not an artifact of the intravenous inoculation route. Sporozoites, including those inoculated in the skin, can use CT to cross the liver sinusoidal barrier via EC traversal.

We next focused on the crossing events that involved KCs (KC⁺) but did not depend on EC traversal (ECfad⁻), which represented 32% of the total crossing events observed. To monitor KCs during such ECfad⁻ KC⁺ crossing events, we established a PI wounding assay (Tavares et al., 2013) for in vivo detection of KC wounding. We first validated the use of PI as a marker of cell wounding in vivo using ECs as target cells. PI was injected subcutaneously in an flk1-gfp mouse, RFP⁺ sporozoites and anti-F4/80 antibody were injected intravenously, and the proportions of crossing events causing the ECfad⁺ phenotype and the incorporation of PI in the EC nucleus, called ECPI⁺ phenotype, were compared. As exemplified in Fig. 4 A, most ECfad⁺ events were also ECPI⁺ (16/17), which validated PI incorporation as a marker of cell wounding in vivo. We then used the PI intravital assay to quantify the proportion of ECfad⁻ KC⁺ crossing events that wounded KCs and caused PI incorporation in its nucleus, noted as KCPI⁺ phenotype. The majority (∼75%) of the ECfad⁻ KC⁺ events were associated with a KCPI⁺ phenotype (Fig. 4, B and D). These ECfad⁻ KC⁺ KCPI⁺ events were also ECPI⁻ (Fig. 4 B), confirming the absence of EC traversal. These events were thus compatible with sporozoites crossing the barrier by KC traversal.

The remainder (∼25%) of the ECfad⁻ KC⁺ events, representing ∼8% of the total crossing events, did not cause PI incorporation in the contacted KC, KCPI⁻ (Fig. 4, C and D), or in ECs (ECPI⁻). These crossing events are compatible with the invasion-transcytosis variant of the gateway model but could also correspond to parasites crossing the barrier between adjacent KCs and ECs.

Finally, ∼15% of the total crossing events were associated with an ECfad⁻ KC⁻ phenotype (Fig. 5 A), indicating sporozoites crossing at a distance of KC without traversing an EC. This phenotype is compatible with a paracellular pathway, i.e., between two ECs, or with a transcellular pathway, i.e., through ECs but without loss of membrane integrity. The existence of CT-independent crossing paths is in agreement with the findings that CT-deficient sporozoites, e.g., SPECT⁻, SPECT2⁻, and CelTOS⁻, which are all defective in CT activity (Ishino et al., 2004; 2005; Kariu et al., 2006), can all induce a blood infection, albeit with reduced efficiency. To further investigate CT-independent crossing, we imaged GFP⁺ SPECT2⁻ sporozoites in KC-depleted mice, in which SPECT2⁻ sporozoites recover normal infectivity after intravenous injection (Ishino et al., 2005; Fig. 5 B). GFP⁺ SPECT2⁻ or GFP⁺ control sporozoites were injected intravenously into mice and the numbers of sporozoites present in the liver parenchyma 2 h after injection were compared by intravital imaging (Fig. 5 C). Similar numbers of control and SPECT2⁻ sporozoites had reached the parenchyma 2 h after injection, confirming that the CT activity is not essential for sporozoite crossing the sinusoidal barrier and that CT-independent crossing mechanisms are operational in sporozoites. We also imaged crossing events by GFP⁺ SPECT2⁻ sporozoites in KC-depleted (Fig. 5, D and E) as well as KC-harboring (lys-gfp, not depicted) mice, which revealed sporozoites translocating apparently through ECs characterized by the intracellular movement of endocytic vesicles containing BSA. However, in the absence of appropriate intercellular junction markers, we could not discriminate between paracellular and transcellular routes.

Using the intravital PI assay, we imaged F4/80-labeled KCs interacting with RFP⁺ sporozoites gliding inside sinusoids of flk1-gfp mice. First, we evaluated the percentage of KC wounding, indicated by the appearance of a KCPI⁺ phenotype after a transient interaction with a gliding sporozoite. Transient interactions were defined by the superposition of KCs and sporozoite fluorescent signals or by KC extensions surrounding the parasite for at least 2 min (Fig. 6, A and B). The majority (∼60%) of such short-lived interactions caused a KCPI⁺ phenotype (Fig. 6, A and C). The 40% of KCs that remained PI⁻ were indeed not wounded rather than not stained as a result of low PI concentration because other cells were frequently observed becoming PI⁺ in the same microscopic field (Fig. 6 B). When CT-deficient GFP⁺ SPECT2⁻ sporozoites were imaged in the same conditions, PI incorporation into the nuclei of interacting KCs was not observed (Fig. 6 D). This showed that while gliding in the sinusoids, and irrespective of barrier crossing, sporozoites transiently interact with, and frequently wound KCs.

We then evaluated the fate of KCs traversed by sporozoites using a double wounding assay. To detect KC traversal, we injected GFP⁺ sporozoites and PI in F4/80-labeled C57BL/6 mice and scored sporozoite–KC interactions that resulted in the KCPI⁺ phenotype. After 1 h, we injected SYTOX green and assessed its incorporation in the nuclei of traversed KCs (KCPI⁺, Fig. 6 E). SYTOX is a fluorescent dye frequently used to detect dead cells; it does not penetrate live cells and its fluorescence increases upon nucleic acid binding. Of the nine traversed KCs identified by PI incorporation, seven incorporated SYTOX after 1 h (Fig. 6 F). This suggests that most KC traversal events permanently alter KC plasma membrane integrity and presumably kill traversed KCs.

The ∼2.5-fold increase in parasitemia that was reproducibly obtained after injection of sporozoites into clodronate-treated compared with untreated mice (Fig. 5 B, P < 0.05) suggested a negative impact of KCs on parasite infectivity. Because KCs play an important role in removing microorganisms from the blood circulation (Helmy et al., 2006), we tested whether sporozoite CT activity could play a role in resisting clearance by KCs.

For this, we analyzed the fate of CT-deficient GFP⁺ SPECT2⁻ sporozoites interacting with KCs. GFP⁺ SPECT2⁻ and GFP⁺ control sporozoites were injected intravenously into C57BL/6 mice harboring F4/80-labeled KCs and imaged in the liver for 150 min. After this time, most control sporozoites had left the microscopic field (∼48%) or invaded the parenchyma (∼45%). In contrast, only ∼14 and 3% of the GFP⁺ SPECT2⁻ sporozoites had left the observation field and invaded the parenchyma, respectively (Fig. 7 A). As expected, most (∼83%) of the CT-deficient sporozoites were trapped in lasting interactions with KCs; importantly, these interactions started from the beginning of the imaging period (Fig. 7 B). In contrast, only ∼10% of control sporozoites displayed similar lasting interactions with KCs, which only started to occur ∼30 min after injection, possibly involving sporozoites that had exhausted their CT capacity (Fig. 7, A and B). For both populations of early KC interactions with SPECT2⁻ and late KC interactions with control sporozoites, sporozoite fluorescence gradually decreased with time until undetectable levels (Fig. 7, C and D). Therefore sporozoite CT helps evading clearance by KCs.

Lastly, we compared the fate of GFP⁺ SPECT2⁻ and GFP⁺ control sporozoites inside primary KCs in vitro. Primary KCs were purified, incubated with control or SPECT2⁻ sporozoites for 1 h, and samples were stained using a double labeling assay to discriminate internalized and extracellular zoites (Fig. 7, E and F). More than 90% of the SPECT2⁻ and ∼30% of control sporozoites were located inside cells (P < 0.0001, Fig. 7 E). Samples were also stained with antibodies to the lysosome marker LAMP1 (Fig. 7, F and G). Almost all intracellular SPECT2⁻ sporozoites colocalized with LAMP1 staining, suggesting they were inside phagolysosomes, and frequently displayed dim green fluorescence. In contrast, only half of control, internalized sporozoites were associated with LAMP1 staining, confirming that WT sporozoites can also be phagocytosed by KCs. Presumably, the other half corresponds to traversing sporozoites or sporozoites internalized inside a parasitophorous vacuole. We conclude that sporozoite CT confers resistance to phagocytosis by KCs and likely other phagocytic cells.

This study provides the first comprehensive picture of how malaria sporozoites infect the liver and reach their final destination inside a hepatocyte. Our spinning-disk confocal imaging approach using flk1-gfp mouse harboring F4/80-labeled KC allowed a clear delineation of the sinusoids and a precise definition of the sporozoite crossing sites. It demonstrated that KCs do not constitute a mandatory gateway for Plasmodium sporozoites and revealed the multiplicity of P. berghei sporozoite crossing mechanisms. Most crossing events (∼77%, n = 60) were associated with CT activity. Of these, the majority occurred through ECs (53%), sometimes also involving a KC presumably double lining the traversed EC, whereas some CT events also occurred specifically through KCs (∼24%). Approximately a quarter (23%) of the crossing events were CT-independent, presumably accounting for the residual capacity of CT-deficient mutants to cross normal sinusoidal barriers (Ishino et al., 2004, 2005; Kariu et al., 2006). These CT-independent crossing events might correspond to a paracellular pathway, between two ECs or between an EC and a KC, as has been suggested for Toxoplasma tachyzoite transmigration of cultured cell monolayers (Barragan et al., 2005). Alternatively, sporozoites could use a transcellular pathway through the formation of channels inside ECs, like those generated by extravasating leukocytes (Carman and Springer, 2004), or use and manipulate the fenestrations of liver sinusoidal ECs (Warren et al., 2007).

Another unresolved question was the role of the CT activity of the sporozoite during liver infection. Much emphasis has been put on the positive impact of hepatocyte traversal on parasite infection via the release of HGF from traversed hepatocytes, which was proposed to act on neighboring invaded cells and to facilitate parasite development (Mota and Rodriguez, 2004). Conversely, it was also proposed that WT, but not CT-deficient sporozoites, induced the expression of iNOS via activation of NF-κB, causing an approximately twofold decrease in parasite development inside hepatocytes (Torgler et al., 2008). In any event, these opposed effects in artificial in vitro conditions were minor and cannot account for the >20-fold and >100-fold reduction in the in vivo infectivity of P. berghei CT-deficient mutants after intravenous and subcutaneous injection, respectively (Ishino et al., 2004, 2005; Kariu et al., 2006; Amino et al., 2008). Therefore, the primary role of CT in the liver appears to be unrelated to hepatocyte traversal.

Here, we imaged a dual function of sporozoite CT in the liver. First, WT sporozoites wounded and, presumably, traversed KCs while gliding in the sinusoids, whereas the vast majority of the CT-deficient sporozoites were rapidly trapped and subsequently degraded inside KCs. This vital function of CT in avoiding parasite clearance is also likely to help the sporozoite resist attacks from other phagocytic cells than KCs, as visualized for granulocytes in lys-gfp mice (unpublished data). Second, CT was associated with ∼80% of the crossing events through ECs and/or KCs. However, the contribution to sporozoite infectivity of CT-dependent crossing of the sinusoidal barrier is more difficult to assess. CT-deficient sporozoites reached the parenchyma as efficiently as controls in the absence of KCs, showing that CT-independent mechanisms are sufficient to allow efficient barrier crossing. Nonetheless, in the presence of KCs the availability of CT-dependent crossing mechanisms might offer an advantage by reducing the time spent in the sinusoids and decreasing the risk of clearance by phagocytic cells. We also reported previously an important role of sporozoite CT in the skin after sporozoite deposition during natural mosquito transmission, in allowing normal sporozoite progression through the dermis and access to skin vessels (Amino et al., 2008). Together, these intravital imaging data suggest a dual role of the CT activity of the sporozoite, in allowing progression through cellular barriers at multiple steps of the sporozoite journey from the skin to the liver parenchyma and escape from host cell phagocytic activity along the journey. In comparison, hepatocyte traversal appears to add little to parasite infectivity. That sporozoites first traverse several hepatocytes before invading the final one might simply result from the time necessary to repress the CT activity, which would otherwise rupture the parasitophorous vacuole membrane and prevent successful invasion.

Sporozoite CT might have additional effects, for example on adaptive immunity. Indeed sporozoites traverse hepatocytes, KCs, and ECs, which can efficiently present antigens (Ebrahimkhani et al., 2011). The shedding of parasite surface antigens in these traversed cells might induce host tolerance or immunity to the corresponding antigens, possibly presented directly by traversed cells if they recover from CT, or via other antigen-presenting cells if they die. Our data indicate that most KC wounding/traversal events result in permanent alteration of KC plasma membrane integrity, and thus likely in KC death, which might affect mounting of a protective immune response. The repeated KC traversal events during the extensive sporozoite gliding in the liver sinusoids might also liberate antiinflammatory mediators such as IL-10 and TGF-β from traversed KCs (Klotz and Frevert, 2008), which might negatively impact antiparasite immunity.

The notion that the CT activity is important to the sporozoite at many steps of its progression from the skin to the final hepatocyte makes the proteins involved in the process attractive targets for antibody-mediated inhibition strategies. High levels of antibodies against SPECT2 correlate with protection from malaria episodes caused by P. falciparum in children (Crompton et al., 2010). Moreover, immunization using CelTOS, another important protein involved in CT, protects rodents from challenge with sporozoites by a combination of antibody and cell-mediated effector responses (Bergmann-Leitner et al., 2011). Many parasite proteins have been described as being involved in the CT phenotype, e.g., SPECT (Ishino et al., 2004), PL (Bhanot et al., 2005), TLP (Moreira et al., 2008), and GEST (Talman et al., 2011), in addition to CelTOS (Kariu et al., 2006) and the perforin-like SPECT2 (Ishino et al., 2005). Along with antibodies to the CS protein, which inhibit sporozoite motility in the dermis (Vanderberg and Frevert, 2004) and presumably elsewhere, antibodies targeting the CT-associated factors would likely add in preventing sporozoites from reaching their final destination in the liver.

We used P. berghei ANKA clones expressing the GFP under the control of the hsp70 regulatory regions (GFP-Control; Ishino et al., 2006), the RFP under the control of eef1-α regulatory regions (line L733; Sturm et al., 2009), and the CT-deficient parasite SPECT2⁻, expressing the GFP under the control of the hsp70 regulatory regions (GFP-SPECT2⁻; Amino et al., 2008).

C57BL/6 and Swiss mice were purchased from Elevage Janvier. Flk1-gfp C57BL/6 transgenic mice (Xu et al., 2010) were maintained by breeding heterozygous males with C57BL/6 females. All animal experiments were approved by the Animal Care and Use committee of Institut Pasteur and were performed in accordance with European guidelines and regulations.

Anopheles stephensi mosquitoes (Sda500 strain) were reared in the Centre for Production and Infection of Anopheles (CEPIA) at the Pasteur Institute using standard procedures. Mosquitoes were fed on infected mice 1–2 d after emergence and kept as previously described (Amino et al., 2007). Infected mosquitoes used for transmission experiments (18–25 d after the infectious blood meal) were deprived of sucrose for 1 d before experimentation to enhance the rate of mosquito bite. For intravenous injections and for in vitro experiments, sporozoites were isolated from infected salivary glands 18–25 d after the infectious blood meal.

Mice were anesthetized with a mixture of ketamine and xylazine (ketamine: 125 mg/kg body weight; xylazine: 12.5 mg/kg body weight). Animals were prepared for intravital microscopy of the liver as described previously (Thiberge et al., 2007). In brief, once deeply anesthetized, a small left-ventral abdominal incision was made to expose the left liver lobe. The exposed lobe was moistened with a drop of phosphate buffer saline and immobilized on a coverslip with cyanoacrylate glue. The animal was placed on a motorized stage of an inverted Axiovert 200 microscope (Carl Zeiss) and covered with a heating blanket at 37°C. Images were acquired in the Imaging Platform at the Pasteur Institute (PFID) using a spinning disk confocal system (Ultraview; Perkin Elmer) composed of a Confocal Scanner Unit (CSU22; Yokogawa) with a camera (EMCCD C-9100; Hamamatsu) and 4 Diode Pumped Solid State Laser (excitation wavelengths: 405, 488, 561, and 647 nm) controlled by Volocity (Perkin Elmer). Usually continuous images were acquired using an LCI Plan-Neofluar 25×/0.8 Imm Corr DIC objective (Carl Zeiss) for up to 10 min and covering 40 µm in depth (8–10 Z-stacks). Image files were processed using ImageJ (National Institutes of Health). Liver sinusoids were visualized by the specific expression of GFP in the EC of flk1-gfp mice or by the intravenous injection of 50 µg Alexa Fluor 555 BSA in C57BL/6 mice. To visualize KC, 4 µg Alexa Fluor 647 anti–mouse F4/80 antibody (clone BM8; BioLegend) was intravenously injected 30 min before observation. The F4/80 labeling specificity was confirmed in vivo by depletion of phagocytic cells with clodronate liposomes (10 µl/g) injected intravenously 2 d before the antibody. Phagocytic cells in the sinusoids were detected 30 min after intravenous injection of 5 µl of a 2% suspension of red fluorescent (580⁄605) carboxylate-modified microspheres (0.1 µm; Molecular Probes). Sporozoites were injected either intravenously (∼3 × 10⁵) or by mosquito bite (anesthetized mice were placed on top of a mosquito cage containing 50–100 mosquitoes for 20 min). To detect host cell wounding by sporozoites mice were subcutaneously injected with PI (5 mg/kg in PBS; Invitrogen) according to Tavares et al. (2013).

GFP⁺ ECs were isolated from flk1-gfp mice according to Klugewitz et al. (2002). In brief, the livers of anesthetized mice were perfused by the portal vein with 2 ml of a 37°C prewarmed digestion medium (RPMI 1640 supplemented with 5% FBS [Cambrex], 2 mg/ml collagenase IV [Sigma-Aldrich], and 0.2 mg/ml DNase [Sigma-Aldrich]). The livers were then removed, cut in small pieces, and incubated in the digestion medium at 37°C for an additional 15 min. Cell suspension was passed through a 100 µm cell strainer and subjected to a density gradient centrifugation with 26% Optiprep (Axis Shield). GFP⁺ ECs were further purified by magnetic cell sorting using anti-CD146 microbeads (Miltenyi Biotec). The cells were plated in a collagen I (Invitrogen)–coated 8-well µ-Slide (Ibidi) and incubated in RPMI 1640 supplemented with 10% FBS (Cambrex), 1% penicillin (Invitrogen), and 1% streptomycin (Invitrogen) at 37°C, 5% CO₂. After overnight culture, cells debris was removed and cells were used for experiments with sporozoites 2 d after isolation.

KCs were isolated from flk1-gfp mice according to Smedsrød and Pertoft (1985). In brief, the liver cell suspension prepared as described above was layered on a two-step Percoll gradient (bottom and overlaying cushions density of 1.066 g/ml and 1.037 g/ml, respectively) and centrifuged at 800 g for 15 min at 4°C. The KC-enriched fraction, extending throughout the lower Percoll cushion, was collected and washed with PBS. The cells were plated in 8-well µ-Slides (Ibidi) and incubated in RPMI 1640 supplemented with 10% FBS (Cambrex), 1% penicillin (Invitrogen), and 1% streptomycin (Invitrogen) at 37°C, 5% CO₂ for 30 min. Non-adherent cells were removed and cells were used for experiments with sporozoites the day after.

GFP-EC traversal activity of sporozoites was visualized using the spinning-disk system equipped with a xy-automated stage. Control (GFP-Control or RFP⁺) or CT-deficient (GFP-SPECT2⁻) fluorescent sporozoites were incubated with primary GFP-ECs in RPMI 1640 supplemented with 10% FBS (Cambrex), 1% penicillin (Invitrogen), and 1% streptomycin (Invitrogen) at 37°C, 5% CO₂. In some experiments, the assay was performed in the presence of 10 µg/ml of PI (Invitrogen). Usually continuous images from 10–15 different fields were acquired for up to 90 min.

The lower limit of the tolerance interval containing 95% of the fluorescence values of the control population was calculated with 95% of confidence using the R software (R Core Team; http://www.R-project.org/). Values below the tolerance limit were considered statistically significant. Categorical variables were compared using two-tailed Fisher’s exact test using GraphPad. Differences in the parasitemia and in the distribution of lasting interaction were analyzed using two-tailed Student’s t test and Wilcoxon rank-sum test, respectively. P values <0.05 were considered statistically significant.

We thank S. Shorte and the Dynamic Imaging Platform (PFID-Institut Pasteur) for the support with confocal microscopy; C. Bourgouin and the Center of Production and Infection of Anopheles (CEPIA-Institut Pasteur) for mosquito rearing; and G. Milon, G. Lauvau, and T. Graf for providing the lys-gfp mice.

This work was supported by funds from Pasteur Institute, the French National Research Agency (grant no. ANR-10-JCJC-1302-PlasmoPEP), the French Government’s Investissement d’Avenir program, Laboratoire d’Excellence “Integrative Biology of Emerging Infectious Diseases” (grant no. ANR-10-LABX-62-IBEID), the European Community’s Seventh Framework Programme (Evimalar, FP7/2007-2013, grant no. 242095), Natixis, and the Leukaemia and Lymphoma Research Funding (UK). J. Tavares was supported by a fellowship from the Fundação para Ciência e Tecnologia (SFRH/BPD/48340/2008) and P. Formaglio by a Ph.D. fellowship from the Direction Générale de l’Armement.

The authors have no competing financial interests.