Transcriptional heterogeneity and tightly regulated changes in gene expression during Plasmodium berghei sporozoite development

Significance

A better understanding of the molecular mechanisms underlying the development of Plasmodium parasites in Anopheles mosquitoes would facilitate the development of malaria vaccines and of novel strategies to interrupt disease transmission. We characterized, at the single-cell level, the gene expression profiles of sporozoites from the rodent malaria parasite Plasmodium berghei, throughout their development in mosquitoes and upon salivation. Our analyses reveal heterogeneity in gene expression among parasites isolated from the same anatomical location, suggesting that parasite development is asynchronous in mosquitoes. In addition, our results highlight the role of gene expression changes in regulating the ability of sporozoites to remain quiescent in the salivary glands, and their rapid reactivation upon salivation.

Abstract

Despite the critical role of Plasmodium sporozoites in malaria transmission, we still know little about the mechanisms underlying their development in mosquitoes. Here, we use single-cell RNA sequencing to characterize the gene expression profiles of 16,038 Plasmodium berghei sporozoites isolated throughout their development from midgut oocysts to salivary glands, and from forced salivation experiments. Our results reveal a succession of tightly regulated changes in gene expression occurring during the maturation of sporozoites and highlight candidate genes that could play important roles in oocyst egress, sporozoite motility, and the mechanisms underlying the invasion of mosquito salivary glands and mammalian hepatocytes. In addition, the single-cell data reveal extensive transcriptional heterogeneity among parasites isolated from the same anatomical site, suggesting that Plasmodium development in mosquitoes is asynchronous and regulated by intrinsic as well as environmental factors. Finally, our analyses show a decrease in transcriptional activity preceding the translational repression observed in mature sporozoites and associated with their quiescent state in salivary glands, followed by a rapid reactivation of the transcriptional machinery immediately upon salivation.

Malaria is a disease caused by unicellular parasites of the Plasmodium genus that are transmitted to humans by the bites of infected female Anopheles mosquitoes. Socioeconomic development (1–4), combined with the use of extensive entomological controls (5, 6) and improved antimalarial drug development and distribution (7, 8), have significantly decreased the mortality associated with malaria over the last 50 y. However, the disease remains a heavy burden affecting more than 200 million people and responsible for half a million deaths annually (9). An efficient malaria vaccine remains elusive, but encouraging progress has been achieved in recent years with the development and testing of vaccines using recombinant proteins (10, 11) and attenuated parasites (12–14) from the human infective stage, the sporozoite. However, despite renewed interest in Plasmodium mosquito stages for vaccine development, many of the molecular processes regulating sporozoite development remain unclear.

Following the ingestion of male and female gametocytes during a blood feeding, fertilization occurs in the mosquito midgut, producing zygotes that undergo meiosis and develop into motile ookinetes. The ookinete traverses the midgut epithelium and matures into an oocyst. On the basal surface of the midgut epithelium, the sessile oocyst undergoes multiple cycles of DNA replication and cell divisions, which leads to the production of thousands of sporozoites that are released into the mosquito’s hemolymph. The crescent-shaped sporozoites are then transported by the hemolymph and ∼20% of them successfully enter the salivary glands, where they wait to be inoculated into a mammalian host, remaining viable for days to weeks (15). Transmission to the mammalian host occurs when a small number of sporozoites, typically less than 100, are ejected with the mosquito saliva into the dermis during the probing phase of a bite (16, 17). Once in the skin, sporozoites move rapidly to locate blood vessels and enter the blood circulation, which carries them to the liver, where they enter hepatocytes and develop into liver stages. During their migration, from the mosquito midgut to the mammalian liver, sporozoites do not show major morphological changes but go through important developmental changes. For example, sporozoites collected from oocysts or from the hemolymph can cause successful mammalian infections but these sporozoites are, overall, much less infectious than salivary gland sporozoites (18–20). Conversely, sporozoites collected from a salivary gland and injected into the hemolymph show reduced infectivity for the mosquito salivary glands (21). These changes in infectivity are mirrored by changes in mRNA expression and protein abundance that have been characterized for rodent parasites and, to a lesser extent, for Plasmodium falciparum (22, 23). Combined with elegant reverse genetic experiments, these analyses have highlighted some of the molecular processes underlying sporozoite maturation, and revealed key Plasmodium genes involved in sporozoite egress from the oocyst [e.g., PbSERA5 (24), SIAP-1 (25), and PCRMPs (26)], and in the recognition and invasion of the mosquito salivary glands [e.g., TREP (27), MAEBL (28), and CSP (29)]. However, despite these important studies, several outstanding questions remain regarding the regulation and development of Plasmodium sporozoites in mosquitoes. For example, it is not clear whether the maturation of sporozoites is primarily driven by extrinsic (e.g., the location of the sporozoites in the mosquito) or intrinsic factors (e.g., the age of the sporozoites) as these parameters are confounded in most studies. Similarly, we do not know whether all sporozoites at the same anatomical location are identically regulated or whether they mature asynchronously. These questions have been difficult to rigorously address due to technical limitations of available methods. In order to have sufficient amounts of material for analysis, genomic and proteomic studies typically rely on bulk approaches utilizing large numbers of parasites, which only provide “averages” on these populations and mask potential heterogeneity among individual parasites. The advent of single-cell technologies allows us to examine the regulation of individual cells and has already provided exciting new insights on the biology of Plasmodium parasites (30–33). Here, we use the rodent malaria parasite Plasmodium berghei, a parasite extensively used for studying preerythrocytic stages in the mammalian host and for performing forward genetic screens (34–36), to describe the transcriptomes of 16,038 individual sporozoites collected throughout their development, from oocysts to salivary glands, and salivated sporozoites, and analyze the succession of molecular changes, and their variations, occurring during these transitions.

Results and Discussion

Gene Expression Profiling of Individual Plasmodium Sporozoites.

We analyzed P. berghei sporozoites collected throughout their development in Anopheles stephensi mosquitoes, from late-stage oocysts (14 d postinfection [dpi]), mosquito hemolymph (16 dpi), salivary glands (21 to 22 dpi), and forced salivation experiments (21 to 22 dpi) (Fig. 1A). Overall, these sporozoites were derived from a total of 12 independent collections, represented two P. berghei lines, and were isolated from two different An. stephensi colonies (SI Appendix, Table S1 and see Materials and Methods for details).

Fig. 1.

Changes in gene expression during sporozoite development. (A) Schematic summarizing sporozoite populations included in this study and their time of collection in days postinfection (figure created with BioRender.com). (B) PCA showing the relationships among individual P. berghei sporozoites based on their expression profiles. Each dot represents a single sporozoite and is colored according to its collection site (red, disrupted oocysts; blue, hemolymph; green, salivary glands; purple, forced salivation). The black line shows the trajectory of the developmental pseudotime inferred based on the gene expression profiles. (C) Distribution of the pseudotime values (x axis) inferred for each individual sporozoite (each represented by a dot colored as in B) according to their collection site (y axis). Note the overlap between the distributions of each group.

From each sample, we prepared a 3′-end 10× Genomics single-cell RNA-sequencing (scRNA-seq) library (37) and sequenced 128 to 248 million paired-end reads of 75 base pairs (SI Appendix, Table S1). Between 0.6 and 67.2% of the reads mapped to the P. berghei ANKA genome (38), providing, on average, 20.2 million Plasmodium reads per sample. Most of the remaining reads mapped to the An. stephensi genome and represented contamination by mosquito RNA. After stringent quality filters (Materials and Methods), we obtained between 82 and 4,630 single Plasmodium cell transcriptomes from each library, for a total of 16,038 individual sporozoite transcriptomes, each characterized by an average of 1,033 unique reads per parasite (235 to 5,867) (SI Appendix, Table S1).

scRNA-Seq Reveals Heterogeneous and Overlapping Changes in Gene Expression during P. berghei Sporozoite Development.

We characterized the transcriptomes of 16,038 sporozoites: 614 sporozoites collected from disrupted oocysts, 2,147 sporozoites isolated from the mosquito hemolymph, 5,979 sporozoites dissected from mosquitoes’ salivary glands, and 7,298 sporozoites obtained after forced salivation experiments (SI Appendix, Table S1). After excluding the transcripts of genes detected in less than 300 cells, we obtained expression data from 1,763 genes (out of the 5,120 protein-coding genes annotated in the P. berghei genome) for further analysis (Materials and Methods). To examine transcriptomic changes occurring during sporozoite development, we first compared, using principal component analysis (PCA), the gene expression profiles of individual sporozoites from different anatomical sites. Interestingly, while P. berghei sporozoites primarily clustered according to the anatomical site from which they were collected, the gene expression profiles showed extensive variations within, and overlap between, sporozoites isolated from different anatomical sites (Fig. 1B). For example, the gene expression profiles of hemolymph sporozoites (in blue on Fig. 1B) ranged from profiles indistinguishable from those of oocyst sporozoites (in red) to profiles similar to those of salivary gland sporozoites (in green), consistent with the range of phenotypes observed in this population.

To further examine changes in gene expression during P. berghei sporozoite development, while accounting for the apparent heterogeneity of each sample, we estimated the position of each individual sporozoite along a putative developmental trajectory (pseudotime) calculated solely using the gene expression data (39). This unsupervised analysis sequentially organized sporozoites starting with those collected from disrupted oocysts and ending with salivated sporozoites, but with significant overlaps between sporozoites collected at different points of their development (with the possible exception of salivated sporozoites that displayed a more distinct gene expression profile) (Fig. 1C). Some of this transcriptional heterogeneity could be accounted for by slight differences between biological replicates (SI Appendix, Fig. S1). In particular, we observed a shift between the modes of the distributions obtained from mCherry- and GFP-labeled salivary gland sporozoites, which could be due to differences in the times of collection or in the insectary temperatures, or the effect of the fluorophore and/or its level of expression in these genetically modified parasites. However, variations between replicates did not entirely explain the wide range of transcriptional heterogeneity observed among sporozoites collected at the same anatomical site, as distinct gene expression profiles were consistently observed within each biological replicate (SI Appendix, Fig. S1). Furthermore, to evaluate whether the “outlier cells” might represent technical artifacts caused by low signal, we repeated the analysis considering only cells characterized by more than 1,000 unique reads. The gene expression patterns observed were qualitatively similar to those from the entire dataset and confirmed high heterogeneity among sporozoites collected at the same site as well as overlaps between those collected at different sites (SI Appendix, Fig. S2).

Taken together, these analyses supported previous findings from microarray and RNA-seq studies that showed that gene expression changes over the course of sporozoite development (22, 23, 40), resulting in differences between parasites collected from different anatomical locations (since those were also correlated with the time postinfection; Fig. 1). However, the analysis of individual parasites enabled identifying heterogeneity among sporozoites at the same location, which was previously masked in bulk analyses. This heterogeneity could derive from differences in the rate of oocyst maturation and the time since sporozoite egress: P. berghei oocysts, within the same mosquito, develop asynchronously and can reach maturity at different times (41). It is therefore possible that the differences in gene expression observed among sporozoites reflect variations in their time since egress. Alternatively, these findings could indicate that the rate of sporozoite development stochastically varied between individual parasites, resulting in similar gene expression profiles between, for example, fast maturing sporozoites in the hemolymph and slower maturing salivary gland parasites. Finally, this analysis relied on parasites collected from multiple mosquitoes (SI Appendix, Table S1), which could account for some of the transcriptional heterogeneity: While each experiment used mosquitoes from the same colony and of the same age, it is possible that mosquito-specific factors influenced the rate of Plasmodium development and that the sporozoites with unusual transcriptional profiles (either delayed or accelerated) all derive from the same mosquito. Irrespective of the underlying reason for the transcriptional heterogeneity, the overlap observed between the gene expression profiles of sporozoites collected at different sites clearly indicated that Plasmodium development in the mosquito is not solely regulated by signals from their tissue environment: A small fraction of the oocyst and hemolymph sporozoites looked identical, transcriptionally, to salivary gland sporozoites suggesting that the development of the parasites does not entirely depend upon their anatomical location. This observation could explain why a few sporozoites collected from oocysts or hemolymph can induce mammalian infections (18–20): These successful infections would be caused by the minority of oocyst or hemolymph sporozoites with advanced maturation (i.e., those with greater pseudotimes on Fig. 1C).

Tightly Regulated Changes in Gene Expression during Sporozoite Development.

To systematically identify genes differentially expressed during sporozoite development, we tested, for each individual P. berghei gene, whether the expression level was significantly associated with the developmental pseudotime, from oocyst to salivary gland sporozoites. Since salivated sporozoites displayed low mRNA abundance compared to sporozoites collected from other anatomical sites (see below), we excluded these parasites from the differential analysis reported here (but the results using all parasites are shown in Dataset S1 and SI Appendix, Fig. S3). This differential gene expression analysis recapitulated much of our current knowledge of sporozoite biology. Many genes with validated roles in oocyst egress and gliding motility [SSP3 (42)], salivary gland invasion [MAEBL (28), SIAP-1 (25), TREP (27), ICP (43)] or skin passage and development in the liver [UIS4 (44), UIS3 (45), CelTOS (46), UIS2 (47), TRSP (48), GEST (49), PL (50)] were among the most differentially expressed genes and displayed a timing of expression consistent with their known function (Table 1 and Fig. 2). CSP (51) and TRAP (52) showed high and sustained level of expression from oocyst to salivary gland sporozoites, consistent with previous reports (see, e.g., ref. 40), before decreases in their expression (discussed below). In addition to the information on these well-characterized genes, our analysis highlighted the potential role of genes not previously known to be important in sporozoites: For example, Plasmepsin X (PM-X, PBANKA_1222500), which has been shown to be involved in merozoite egress (53), was highly expressed during early sporozoite development, suggesting a possible role in oocyst egress, while bergheilysin (PBANKA_1137000) was expressed later, consistent with a putative role in salivary gland invasion or skin passage (Fig. 3A). Similarly, several genes without functional annotation or predicted domains showed high level of expression and tightly regulated timing of expression (Fig. 3B) consistent with a role in sporozoite development, and it will be exciting to further examine these potential candidates with functional studies. Finally, four AP2 domain transcription factor genes were consistently detected in the scRNA-seq data, including PBANKA_0521700, which was specifically expressed in salivary gland sporozoites (SI Appendix, Fig. S4). The complete list of genes is available in Dataset S1.

Table 1.

List of the 30 genes most differentially expressed according to the sporozoite developmental pseudotime, from oocyst to salivary gland sporozoites

Gene	Annotation	Name	Max expr.	Phenotype (reference)	RMgmDB
PBANKA_1002500	Sporozoite-specific protein S10	S10	8	Liver [Togiri et al. (79)]	SG sporozoites (RMgm-2710)
PBANKA_0901300	Membrane associated erythrocyte binding-like protein	MAEBL	5	SG [Kariu et al. (28)]	SG sporozoites (RMgm-220)
PBANKA_0501200	Early transcribed membrane protein	UIS4	30+	Liver [Mueller et al. (44)]	Liver (RMgm-1934)
PBANKA_1465051	Plasmodium exported protein		30+
PBANKA_1400800	Protein UIS3	UIS3	30	Liver [Mueller et al. (45)]	Liver (RMgm-1449)
PBANKA_1206800	Zinc finger (CCCH type) protein		8		No phenotype (RMgm-374, -3195)
PBANKA_1236200	Conserved Plasmodium protein		10
PBANKA_1349800	Thrombospondin-related anonymous protein	TRAP	Sustained	Motility [Sultan et al. (52)]	SG sporozoites (RMgm-149, -1344)
PBANKA_1433700	Conserved Plasmodium protein		5		Egress (RMgm-312)
PBANKA_1425200	Sporozoite surface protein 3	SSP3	20	Motility [Harupa et al. (42)]	No phenotype (RMgm-3858)
PBANKA_1432300	Cell traversal protein for ookinetes and sporozoites	CelTOS	25	Skin plus liver [Kariu et al. (46)]	Oocysts (RMgm-46)
PBANKA_1006200	Sporozoite invasion-associated protein 1	SIAP1	Sustained	SG [Engelmann et al. (25)]	Egress (RMgm-109, -233)
PBANKA_1349300	Conserved Plasmodium protein		25
PBANKA_1225000	Serine/threonine protein kinase, FIKK family		8	Oocysts [Jaiyan et al. (80)]	No data (RMgm-3295)
PBANKA_1236100	Ribosomal large subunit pseudouridylate synthase		10
PBANKA_1340100	l-Lactate dehydrogenase	LDH	5
PBANKA_1204200	IMP1-like protein	IMP4	0		No data (RMgm-3185)
PBANKA_1422900	Conserved protein		0
PBANKA_0605900	Conserved Plasmodium protein		20
PBANKA_1306500	TRAP-like protein	TREP	0	Motility [Combe et al. (27)]	SG sporozoites (RMgm-145, -159)
PBANKA_1328000	Serine/threonine protein phosphatase UIS2	UIS2	25	Liver [Zhang et al. (47)]	Liver (RMgm-1366)
PBANKA_0209100	Thrombospondin-related sporozoite protein	TRSP	20	Liver [Labaied et al. (48)]	Liver (RMgm-34)
PBANKA_0813000	Inhibitor of cysteine proteases	ICP	8, 25	SG [Boysen et al. (43)], liver [Lehmann et al. (81)]	SG sporozoites (RMgm-1114, -970)
PBANKA_1350500	Parasitophorous vacuolar prot. 3	PV3	8
PBANKA_1312700	Gamete egress and sporozoite traversal protein	GEST	20	Skin [Talman et al. (49)]	Liver (RMgm-667)
PBANKA_0601600	Nuclear export mediator factor	NEMF	18
PBANKA_0403200	Circumsporozoite (CS) protein	CSP	Sustained	Sporozoite development [Menard et al. (82)]	Sporozoites (RMgm-9)
PBANKA_1005200	Conserved Plasmodium protein		18		Liver (RMgm-2722)
PBANKA_0720800	RNA-binding protein NOB1	NOB1	5
PBANKA_1107700	SET domain protein	SET9	5

For each gene, the table indicates the gene name and annotation, the time of the peak of expression (Max Expr., in pseudotime units), whether the gene has been associated with a sporozoite phenotype in the literature (and the corresponding reference), and whether a phenotype is reported in the Rodent Malaria genetically modified Parasites database (RMgmDB). The genes are ranked based on the statistical significance of the association between expression and pseudotime (most significant on top). See SI Appendix, Table S2 for the complete list of genes, the statistical significance, and the results of the association when salivated sporozoites are included in the analysis).

Fig. 2.

Heatmap showing the expression patterns for the 50 most differentially expressed genes during sporozoite development in mosquitoes (from oocyst to mature salivary gland sporozoites). Each row shows the expression of one protein-coding gene. Each column shows the data for one individual sporozoite, organized from left to right according to its pseudotime and colored by its anatomical location in the track above the heatmap (red, oocyst; blue, hemolymph; and green, salivary gland). The heatmap color shows the number of a given transcript per cell from dark blue (not detected) to white and red (most highly expressed). The black arrow at the bottom indicates the corresponding pseudotimes.

Fig. 3.

Novel candidate genes that could play a role in sporozoite development. (A) Changes in expression of Plasmepsin-X (PMX, red) and Bergheilysin (BLN, blue) during sporozoite development. (B) Changes of expression of five conserved unannotated P. berghei genes with distinct expression patterns throughout sporozoite development. Each plot shows the median gene expression of 100 individual parasites (y axis) binned according to their pseudotime (x axis). The plot under each graph summarizes the distribution of the pseudotimes obtained for oocyst (red), hemolymph (blue), salivary gland (green), and salivated (purple) sporozoites.

Despite the clear separation of salivated and salivary gland sporozoites based on their overall gene expression profiles (Fig. 1), few protein-coding genes were consistently transcribed in salivated sporozoites (SI Appendix, Table S2 and discussion below). Two notable exceptions were the early transcribed membrane protein (UIS4, PBANKA_0501200), which reached even higher levels of expression in salivated sporozoites than those observed in salivary gland sporozoites (SI Appendix, Fig. S5A), and an uncharacterized exported protein (PBANKA_1465051; SI Appendix, Fig. S5B), which contains a predicted circumsporozoite-related antigen domain and would be fascinating to functionally evaluate.

Transcription and Translation Are Dynamically Regulated in Sporozoites.

Once they reach the salivary glands, Plasmodium sporozoites can remain quiescent for several days to weeks (15) before being injected into the mammalian host, where they quickly become motile and able to invade host cells (54). A proposed molecular mechanism underlying this quiescent state of salivary gland sporozoites and their rapid reactivation upon salivation is translational repression of the mRNAs encoding the proteins required later in the mammalian host (23, 55–57). This global translational repression results from the phosphorylation of the eukaryotic translation initiation factor 2-alpha by eIK2 (PBANKA_0205800) (55), while the Pumilio protein (PUF2, PBANKA_0719200) binds to the matching phosphatase-encoding transcripts (UIS2, PBANKA_1328000) and blocks their translation (reviewed in refs. 56, 58, and 59). Consistent with this mechanism, we observed a clear peak of expression of eIK2 and PUF2 in salivary gland sporozoites (Fig. 4A) but not in oocyst sporozoites, as has recently been described (23).

Fig. 4.

Regulation of transcription and translation in sporozoite. (A) Changes in expression of Plasmodium genes involved in mRNA stability (SAP1, green) and translation inhibition (eIK2, orange; PUF2, red; UIS2, pink). (B) Changes in relative abundance of mRNA (green) and rRNAs (purple) during sporozoite development. Both the mRNA and rRNA curves show the proportion of RNA at a given pseudotime compared to the maximum observed (fixed arbitrarily at 1). The plot under each graph summarizes the distribution of the pseudotimes obtained for oocyst (red), hemolymph (blue), salivary gland (green), and salivated (purple) sporozoites.

Some of the proteins regulating the long-term storage of the mRNAs in stress granules in female gametocytes have been characterized (e.g., refs. 57 and 60), but these genes do not seem to play a similar role in sporozoites. Indeed, aside from eIK2 and PUF2, none of the proteins involved in translational repression in gametocytes [e.g., DOZI or CITH (61)], nor the orthologous proteins of those associated with stress granules in humans or yeasts (62), were detectable in our dataset. Only the sporozoite asparagine-rich protein 1 (SAP-1), which has been shown to regulate mRNA stability in sporozoites (63, 64), was robustly detected and showed a pattern of expression akin to those of eIK2 and PUF2, although with an earlier peak of expression (Fig. 4A). Interestingly, among the genes expressed at the same time as eIK2, we observed one RNA helicase (PBANKA_1103800) that was lowly but specifically expressed in salivary gland sporozoites, and it will be interesting to test whether it may play a role in translational repression.

Finally, the increased expression of genes involved in translational repression appeared to be preceded by an overall and steady decrease in mRNA transcription (Fig. 4B). Fluctuations in overall mRNA abundance have been described in blood-stage Plasmodium parasites (65, 66), and the decrease in mRNA transcription during sporozoite development is consistent with the hypothesis that mature salivary gland sporozoites are quiescent and their transcriptional activity reduced. Conversely, we observed a twofold to threefold increase in the number of ribosomal RNA (rRNA) molecules synthesized in sporozoites collected by forced salivation compared to the amount observed in oocyst to salivary gland sporozoites (Fig. 4B). While rRNA molecules are not polyadenylated in Plasmodium and should theoretically not pass the polyA selection used during the 10× scRNA-seq library preparation, they are extremely abundant in cells and are often detected in transcriptomic experiments. Indeed, we observed very specific and consistent expression of the 5.8S, 18S, and 28S rRNA genes (SI Appendix, Fig. S6) across the entirety of sporozoite development, with a clear increase upon salivation. This observation would be consistent with the abrupt reactivation of quiescent sporozoites and the restarting of the entire transcriptional/translational machinery upon salivation. In contrast to multicellular eukaryotes that typically have hundreds of copies of rRNA genes in their genome, most Plasmodium parasites only carry four copies of the 18S–5.8S–28S ribosomal unit and the expression of these rRNA genes is developmentally regulated (67, 68): The rRNA genes located on P. berghei chromosomes 5 and 6 are primarily expressed in mosquito stages (S-forms), while the rRNA genes from chromosomes 7 and 12 are the dominant forms in blood stages (A-forms). Analyses of the rRNA sequences that carried sufficient genetic information to reliably differentiate the chromosomal origin of each rRNA transcript (Materials and Methods) indicated that there were approximately twice as many transcripts derived from chromosome 6 (S-form) than the combined number of transcripts derived from chromosomes 7 and 12 (A-forms), while no transcripts specifically derived from chromosome 5 (also S-form) could be detected in our data (SI Appendix, Fig. S7). Interestingly, the A/S proportions remained relatively constant throughout sporozoite development and, in particular, did not change upon salivation, suggesting that the switch to the A-form rRNAs observed in blood-stage parasites occurs at some point during the liver stage development.

Conclusion

Our data provide a comprehensive perspective on the regulation of gene expression accompanying the development of P. berghei sporozoites from midgut oocyst to mature salivary gland sporozoites, and upon salivation. The single-cell analyses reveal a high level of transcriptional heterogeneity among sporozoites from the same anatomical location, which was masked in previous gene expression studies. This transcriptional heterogeneity possibly derives from the asynchronicity of oocyst maturation and sporozoite egress, and indicates that the development of Plasmodium parasites in mosquitoes is, at least partially, regulated by an intrinsic clock and not solely determined by the parasite environment. The data generated also highlight the key role of transcriptional regulation in the quiescence and reactivation of salivary gland sporozoites: The genes underlying the global translational repression in mature parasites are specifically transcribed once the sporozoites reach the salivary glands and this transcription is accompanied by a general decrease in transcriptional activity, while, upon salivation, the increase in the level of rRNA genes is consistent with a restarting of the entire transcriptional machinery.

The precise determination of the timing of expression of each P. berghei gene may enable the identification of novel candidate genes that underlie specific biological processes (e.g., oocyst egress or invasion of salivary gland cells) and improve our understanding of the regulation of these functions in sporozoites. In addition, this study can constitute a framework to identify and manipulate critical determinants of sporozoite development and transmission to mammalian hosts and serve as a foundation for developing new strategies to malaria control.

Materials and Methods

Parasite and Mosquito Strains.

Infections of An. stephensi mosquitoes [Nijmegen strain (69)] were performed with P. berghei ANKA parasites expressing the green fluorescent protein (GFP) (PbGFP_CON) (70) at the insectary of the Laboratory for Malaria and Vector Research of the National Institute of Allergy and Infectious Diseases (NIAID)/NIH. Infections of An. stephensi mosquitoes (Liston strain) were performed with P. berghei ANKA line 159cl1 that constitutively expresses mCherry (54) at the Johns Hopkins Bloomberg School of Public Health.

An. stephensi mosquitoes were reared under standard insectary conditions at 80% humidity, 12/12-h light/dark cycle, and maintained with cotton pads soaked either in 10% sucrose solution or 10% corn syrup solution (Karo; ACH Food Companies).

Animal Handling and Ethics Protocol.

This study was performed following the recommendations from the Guide for the Care and Use of Laboratory Animals of the NIH (71). The animal use was approved by the Johns Hopkins Institutional Biosafety Committee and Johns Hopkins University Institutional Animal Care and Use Committee (MO17H325), and by the NIH Animal Ethics Proposal and registered in the Standard Operating Procedures of Laboratory of Malaria and Vector Research (NIAID/NIH-SOP LMVR 22).

Infection Conditions.

Three- to 4-wk-old Swiss-Webster mice were infected with P. berghei from frozen stocks, and the mouse parasitemia was monitored daily by light microscopy analysis of methanol-fixed blood smear stained with 10% Giemsa. Four- to 5-d-old An. stephensi females were starved for 12 h before being allowed to feed for 30 min on an anesthetized mouse infected with P. berghei with ∼1% parasitemia and an exflagellation rate of 1:10 fields. Only fully engorged mosquitoes were kept and maintained at 18 °C (Johns Hopkins Bloomberg School of Public Health) or 19 °C (NIAID) with 10% sugar solution ad libitum until sporozoite collection.

Isolation of P. berghei-GFP Sporozoites.

To isolate sporozoites from oocysts, 50 female mosquitoes were dissected 14 d after the infected blood meal. Mosquitoes were anesthetized on ice and the midguts were dissected under a stereomicroscope in phosphate-buffered saline (PBS). The midguts were placed in an excavated Petri dish with PBS under a fluorescent microscope, opened, and the oocysts were disrupted with a needle to release the sporozoites. The released sporozoites were collected with a Sigmacote-coated pipette tip (Sigma) and transferred to a low-retention tube (Protein LoBind; Eppendorf).

Hemolymph sporozoites were collected by perfusion of 50 ice-cold anesthetized females at day 16 postinfection. Mosquitoes were gradually injected with 10 µL of PBS into the thorax, and the sporozoites were collected from the flow-through by an incision made in the abdominal wall using forceps. The samples were collected with a Sigmacote-coated tip and transferred to a low-retention tube (Protein LoBind Tube; Eppendorf) on ice.

Sporozoites from salivary glands were collected at 21 dpi. One day before sporozoite collection, salivary gland infection was confirmed by the accumulation of parasite-expressed GFP on the mosquito thoracic cavity under a fluorescent microscope (MZ10 F; Leica). Fifty female mosquitoes were anesthetized on ice and their salivary glands dissected in PBS under a stereomicroscope. The salivary glands were transferred to a low-retention tube (Protein LoBind Tube; Eppendorf) containing PBS, homogenized with a disposable pestle, and kept on ice.

Salivated sporozoites were collected from 60 infected female mosquitoes at day 21 postinfection by forced probing in a Sigmacote-coated tip filled with 10 µL of PBS. Briefly, mosquitoes were anesthetized on ice for 5 min, the wings were removed, and the mosquito proboscis was gently inserted into the tip. The mosquitoes were left to salivate for 30 min, and the salivated sporozoites were pooled in a low-retention tube (Protein LoBind Tube; Eppendorf). After collection, all the samples were homogenized, passed through a 20-µm Pluriselect filter (Cell Strainer) to remove cellular debris, and counted using a hemocytometer (C-Chip; Chemglass Life Sciences). The excess volume was removed by centrifugation at 15,000 × g for 10 min, and the sporozoite final concentration adjusted to 500 to 2,000 sporozoites per microliter in PBS.

Isolation of P. berghei-mCherry Sporozoites from Mosquito Saliva and Salivary Glands.

One day prior to sporozoite collection, mosquitoes were fluorescently sorted and mosquitoes with fluorescence in the salivary gland area were kept for sporozoite harvest. Salivary gland and salivated sporozoites were collected from mosquitoes 21 or 22 dpi. Briefly, mosquitoes were anesthetized on ice for 5 min and immobilized on a glass plate by placing their wings on double-sided tape. The mosquito proboscis was gently inserted into a low-adhesion P10 pipet tip containing 2 µL of immersion oil (ZEISS; Immersol 518F) or 4 µL of PBS. Salivation was induced by applying 1% pilocarpine (Sigma; P6503)/0.2% Tween 20 (Sigma; P1379) in PBS to the mosquito abdomen. Following this, mosquitoes were allowed to salivate for 30 min. Salivated sporozoites were collected from 60 to 78 mosquitoes by pooling pipet tips containing saliva sporozoites in a low-retention 0.6-mL tube (Thermo Fisher; 3446). For sporozoites salivated into oil collection, pooled oil was mixed with 20 µL of PBS and spun at 1,000 × g for 5 min to separate saliva from oil.

Library Preparation and Sequencing.

The number of sporozoites isolated from each sample was crudely determined using a hemocytometer and light microscopy. An estimated 1,000 to 5,000 sporozoites per sample were loaded on a 10× Genomics Chromium controller and 12 individually barcoded 3′-end scRNA-seq libraries were prepared according to the manufacturer’s protocol. Briefly, the 10× 3′-end library preparation includes physical isolation of individual cells in droplets, capture of the polyadenylated mRNAs using oligo-dT priming, and reverse transcription incorporating a barcode specific to each droplet (37). After cDNA synthesis, the emulsion is broken and cDNA molecules are fragmented into ∼300-bp fragments, before the addition of sequencing adapters. Each library was sequenced on an Illumina high-output sequencer to generate a total of 2,296,539,124 75-bp paired-end reads (SI Appendix, Table S1).

Single-Cell Transcriptome Analysis.

A custom analysis pipeline, similar to the Cell Ranger single-cell software (37), was developed to process all scRNA-seq reads (32). Briefly, we identified all reads containing the 10× Genomics barcodes, trimmed of sequences downstream of 3′ polyadenylation tails, and only kept for further analyses reads longer than 40 bp. We mapped all reads using HISAT2 [version 2.0.4 (72)] to the P. berghei ANKA genome (38), allowing for a maximum intron length of 5,000 bp, and to the An. stephensi genome [AsteS1 (73)]. To identify sequences that represent PCR duplicates resulting from the library preparation, we identified reads with identical sequences for the 16-mer 10× Genomics barcode, 12-mer unique molecular identifier sequence, and mapped to the same genomic location of DNA and on the same strand, and only kept one of them. These “unique reads,” of which each represents a unique mRNA molecule present in one parasite, were then further analyzed. We used the 10× Genomics barcodes to assign unique reads to individual parasites and tallied the number of unique reads mapped to each annotated P. berghei gene (i.e., from the annotated transcription start site to 500 bp after the 3′-end).

We then combined all scRNA-seq libraries and used the scran [v1.14.6 (74)] addPerCellQC function to identify and remove outlier cells based on library size and gene count while accounting for variations between libraries (e.g., sequencing depth). Only transcriptomes with more than 200 unique reads were included. Transcripts detected in less than 300 cells (out of 16,038) were excluded from this analysis. We normalized the transcriptomes of these filtered transcriptomes and calculated size factors by deconvolution using the scran functions quickCluster and computeSumFactors. The calculated size factors were used to compute normalized counts per cell via logNormCounts function in scater (v1.14.6) (75) and performed PCA with runPCA. In addition, to evaluate the effect of cells characterized by a small number of unique reads, we reran this analysis using only transcriptomes characterized by at least 1,000 unique reads.

Estimation of the Pseudotime and Differential Expression Analysis.

The dataset analyzed with scran/scatter—raw counts, normalized counts, reduced dimensionality (PCA), and metadata—was then imported to slingshot (v1.4.0) (39) to calculate pseudotime for all individual parasites.

To identify genes that were differentially expressed along the estimated pseudotime trajectory, we used the fitGAM function (k = 7) of tradeSeq [v1.1.19 (76)]. Due to the low mRNA transcript abundance observed in salivated sporozoites, we performed the statistical testing separately with and without data from these experiments (i.e., using only data from sporozoites collected from oocyst, hemolymph, and salivary gland; or using the entire dataset). We tested for differentially expressed genes according to the calculated trajectory using the associationTest function and corrected P values for multiple testing using false discovery rates (77, 78).

Analysis of rRNA Genes.

All scRNA-seq reads were also mapped independently to the four chromosomal locations of the P. berghei genome containing the 18S–5.8S–28S rRNA genes (on chromosomes 5, 6, 7, and 12) using Hisat2 but without allowing for spliced alignments (–no-spliced-alignment mode) to avoid mismapping caused by the high sequence homology among loci. Very few reads mapped to the 5S rRNA gene cluster on chromosome 10, and those were not included in these analyses. We tallied the number of reads unambiguously mapped to each chromosomal location for each individual parasite. To control for variations in the number of informative reads derived from each chromosome, we split the reference sequence of each rRNA genes into 75 bp (with a shift of 1 bp) and determined the proportion of reads, from each gene, mapped unambiguously to its original location.

Data Availability

All sequence data generated in this study have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive under the BioProject ID PRJNA676126.