Single-cell insights: Mosquito Toll activation enhances granulocyte differentiation and immune surveillance via Notch suppression

Significance

Mosquito immune cells (hemocytes) mediate defenses that can eliminate Plasmodium parasites. Understanding the development and function of hemocytes is essential for defining natural barriers to malaria transmission. This study provides a comprehensive single-cell map of Anopheles gambiae hemocytes, revealing how four immune signaling pathways Toll, Immunodeficiency (IMD), Jun-kinase (JNK), and Signal Transducer and Activator of Transcription (STAT)- modulate hemocyte differentiation and function. We characterized seven previously unrecognized hemocyte subpopulations within the granulocyte and oenocytoid lineages, and show that Toll and JNK signaling promote differentiation of precursors to granulocytes, while repressing oenocytoid development, through inhibition of Notch signaling. These findings reveal fundamental mechanisms by which immune signaling regulates mosquito cellular immunity and may offer targets for preventing disease transmission.

Abstract

Activation of the Toll, Immunodeficiency (IMD), Jun-kinase (JNK), and Signal Transducer and Activator of Transcription (STAT) innate immune signaling pathways limits Plasmodium infection in Anopheles gambiae. Hemocytes (mosquito immune cells) restrict malaria transmission by eliminating Plasmodium parasites through complement-mediated lysis and phagocytosis. However, the mechanisms underlying mosquito hemocyte lineage determination and maturation remain poorly understood. Here, we used single-cell transcriptomics to investigate how these immune signaling pathways regulate hemocyte differentiation and function. Analysis of 37,529 hemocytes revealed oenocytoid and granulocyte subpopulations not previously described that express specific molecular markers, as well as the expansion of selective subpopulations by different signaling pathways. Toll and JNK activation trigger terminal differentiation of granulocytes (phagocytic cells) with a dramatic increase in megacytes and reduce oenocytoids by transcriptionally suppressing Notch signaling. Toll activation strongly affects gene expression in most hemocyte clusters, whereas STAT and IMD activation impact only specific clusters. Moreover, Toll activation induces the expression of extracellular matrix-interacting genes in megacytes, enhancing midgut immune surveillance. Collectively, these findings uncover how immune signaling modulates hemocyte differentiation and function in a medically relevant mosquito species, providing insights into the mechanism by which hemocytes hinder Plasmodium infection.

Malaria remains a significant challenge to global public health, with 263 million cases and 600,000 deaths reported in 2023 (1). Anopheles gambiae is a major vector of Plasmodium falciparum malaria in Africa, the parasite that causes the most virulent form of the disease and is responsible for 95% of the mortality, mostly in young African children (1). Four major immune signaling pathways, initially characterized in Drosophila—Toll (2), Jun-kinase (JNK) (3), Immunodeficiency (IMD) (4), and Signal Transducer and Activator of Transcription (STAT2) (5, 6)— are known to mediate antiplasmodial responses in anopheline mosquitoes that limit malaria transmission. However, it is often unclear which organs and/or cell types mediate these responses.

The role of STAT2 signaling in insect immunity was first described in A. gambiae (7) and it mediates antiplasmodial responses to P. falciparum and Plasmodium berghei oocysts (6), as well as defense responses of Anopheles aquasalis to Plasmodium vivax infection (8). The IMD pathway also limits Plasmodium infection in A. gambiae but targets the ookinete stage (9).

JNK signaling mediates apoptosis and nitration of ookinete-invaded midgut cells in A. gambiae females infected with P. berghei (10), which activates the local release of hemocyte-derived microvesicles (11) that promote parasite lysis by the mosquito complement-like system (12). Interestingly, P. falciparum can evade the mosquito immune system by suppressing these responses in invaded midgut cells (13).

Hemocytes mediate innate cellular immune responses in insects. Our current understanding of how immune signaling cascades affect hemocyte differentiation and immunity is mostly based on studies of Drosophila larval hemocytes. Besides their role in activating antimicrobial peptide gene expression, Janus Kinase/Signal Transducer and Activator of Transcription (JAK/STAT), Toll, and JNK signaling also regulate hemocyte proliferation and differentiation in Drosophila larvae. STAT92E signaling regulates plasmatocyte (phagocytic granulocytes) differentiation in the Drosophila larval lymph gland (14), and constitutive activation of JAK/STAT triggers lymph gland hypertrophy, overproliferation of hemocytes, and the formation of melanotic tumors (15). Activation of STAT92E signaling in Drosophila larvae during wasp infection (16) and activation of JNK signaling (17) induce the differentiation of plasmatocytes into lamellocytes (large hemocytes). Similarly, Toll pathway regulates larval hematopoiesis, and overactivation also induces lamellocyte differentiation (18) and melanotic tumor formation (4).

Several studies on the molecular characterization of larval hemocytes from silkworms (19), Drosophila (20), and adult hemocytes from A. gambiae (21), ticks (22), and crustaceans (23) have been conducted. Based on their morphology, hemocytes from A. gambiae females are classified as precursor prohemocytes (PHem) that give rise to oenocytoids, hemocytes involved in pathogen melanization, and phagocytic granulocytes (24). Recently, single-cell transcriptome analysis of adult A. gambiae hemocytes identified several granulocyte subpopulations with distinct features, each expressing unique combinations of molecular markers (21). Additionally, flow cytometric analysis revealed distinct characteristics among hemocyte subpopulations from adult A. gambiae females based on their granularity, ploidy, and phagocytic capacity (25). Furthermore, activation of Toll signaling has been shown to promote the differentiation of granulocytes into megacytes, giant cells with an enhanced response to infection by bacteria and Plasmodium parasites (26).

In this study, we investigated, at single-cell resolution, the effect of modulating four key immune signaling pathways—Toll, JNK, STAT2, and IMD—on the differentiation, gene expression, and immune function of hemocytes from adult A. gambiae females. Seven hemocyte subpopulations were identified by analyzing the transcriptome of over 37,000 hemocytes which were not previously described. We uncovered crosstalk between the Toll and JNK pathways and provided direct evidence that Toll signaling promotes prohemocyte differentiation into several different subpopulations of final effector granulocytes while suppressing the oenocytoid fate. Furthermore, activation of the Toll pathway enhances hemocyte immune surveillance and patrolling activity. These studies provide in-depth molecular insights into the complex interplay between different signaling pathways that coordinate hemocyte differentiation and function to orchestrate an effective immune response.

Results

Detailed Atlas of A. gambiae Hemocyte Subpopulations in Adult Females at a Single-Cell Resolution.

The four major immune signaling pathways in A. gambiae—Toll, JNK, IMD, and STAT2—were either activated or suppressed to investigate their effects on hemocyte differentiation and gene expression in adult female mosquitoes. These immune pathways were modulated by silencing regulatory genes in sugar-fed females through systemic injection of dsRNA. Circulating hemocytes were collected by perfusion 4 d postsilencing, and single-cell and bulk transcriptome profiling were performed (Fig. 1A). The Toll, JNK, IMD, and STAT2 pathways were activated by silencing the signaling inhibitors Cactus, Puckered, Caspar, and SOCS, respectively. Conversely, genes essential for immune pathway activation (Rel1, JNK, Rel2, and STAT2) were silenced to suppress signaling. The silencing efficiency of immune pathway genes in hemocytes ranged between 63% and 82% (SI Appendix, Fig. S1 A and B).

Fig. 1.

Detailed atlas of hemocyte subpopulations in adult A. gambiae females at a single-cell resolution. (A) Schematic of experimental workflow. (B) UMAP of 37,529 single hemocyte transcriptomes identified by integrated Seurat clustering from adult A. gambiae, each cluster represented by a distinct color. (C) Identity of Antimicrobial granulocytes (AM-Gran), Dividing Granulocytes (Div-Gran), Granulocyte cluster 4 (Gran 4), Megacyte, Oenocytoid assigned by “Cell type label transfer” analysis, and cell cycle phases of the hemocyte clusters determined by Cell cycle scoring analysis. (D) Dot plot represents the expression of cluster-specific marker genes, color of the dot represents the average scaled expression, and size indicating % of cells in the cluster expressing the marker gene. The last 5 digits of each gene label are the Vectorbase accession numbers after AGAP0.

Single-cell RNA sequencing libraries from hemocytes were generated using the 10× Genomics Chromium Single Cell platform and analyzed using a previously reported A. gambiae hemocyte transcriptome as a reference (27). To ensure that only high-quality cells were included in the analysis, low-quality cells with more than 0.5 decontX contamination estimate and cells with more than 15% mitochondrial gene expression were removed. Cells with transcribed genes ranging from 250 to 2,500 were included for all samples, to exclude probable doublets, except for Cactus-silenced hemocytes. A different threshold of cells with 250 to 4,000 transcribed genes was used for Cactus-silenced hemocytes, because previous research has shown that this treatment greatly increases the proportion of megacytes (giant granulocytes) (26). In addition, silencing Cactus could also alter the overall transcriptional activity of hemocytes. As a result, this sample contains a substantial number of cells with more than 2,500 transcribed genes (SI Appendix, Fig. S1 C–H). The detailed data analysis pipeline is described in Materials and Methods section. A total of 37,529 high-quality single-cell transcriptomes were obtained and subjected to integrated clustering analysis to derive a comprehensive picture of the transcriptional profile of different hemocyte subpopulations. Quality control metrics of the cells from each treatment before and after filtering are provided in SI Appendix, Fig. S1 C–H.

12 hemocyte clusters with distinct transcriptional profiles were identified (Fig. 1B), and two strategies were implemented to determine their functional characteristics. First, the overall transcriptomic profile of the 12 clusters was compared to that of the five previously described hemocyte subpopulations—Oenocytoids (Oen), AM-Gran, Megacytes (Mega), Div-Gran, and Gran 4—based on the transcriptome of 5,383 cells (21). The Seurat software function “TransferData” assigned the identity of AM-Gran, Div-Gran, and Gran 4 granulocytes in this Uniform Manifold Approximation and Projection (UMAP) clustering analysis with a high prediction score (Fig. 1C). Interestingly, more than one cluster in the current dataset mapped to the previously defined Megacyte and Oenocytoid populations, suggesting the presence of subpopulations of these cell types revealed by this expanded analysis of 37,529 cells that were previously not described (Fig. 1C).

Second, marker gene discovery analysis was also carried out to identify the set of genes characteristic of each hemocyte subtype. The cluster-specific expression level of the top three gene markers and the proportion of cells in the cluster expressing each marker gene are summarized in Fig. 1D. The central cluster of the UMAP includes PHem, which are known to give rise to oenocytoids and granulocyte subpopulations (27). In general, the expression of PHem markers, such as Ficolin, Sphingomyelin phosphodiesterase, and EGF-like domain-containing protein, decreases as oenocytoids and granulocytes undergo terminal differentiation and begin to express specific functional marker genes, which are either absent or have very low expression in PHem (Fig. 1D).

The AM-Gran cluster expresses the antimicrobial genes Defensin1, C-type lysozyme 1, 2, and the UreE-like protein involved in pathogen recognition, while the previously described Megacyte cluster (Mega) expresses high levels of the transmembrane protein AGAP007319 (a newly annotated longer transcript identified in the hemocyte-specific transcriptome annotation that includes the sequence of AGAP007318) (27) (Fig. 1D). We identified an additional cluster in our dataset that shares some transcriptional similarities with the “classic” megacytes, but also expresses high constitutive mRNA levels of cell-adhesion and extracellular matrix (ECM)-interacting proteins, such as Integrin alpha-ps, Dystroglycan, and the heparin sulfate proteoglycan Terribly reduced optic lobes isoform B (Trol), and we refer to it as ECM-interacting Granulocytes (ECM-Gran) (Fig. 1D). There is a cluster between PHem and ECM-Gran, which expresses high levels of putative immune-regulatory genes such as CLIPD1, two Zinc finger domain-containing proteins (AGAP007956 and AGAP001358), Toll, and TOLL-like receptor 5A, that we refer to as Regulatory Granulocytes (Reg-Gran) (Fig. 1 B and D). An additional small cluster directly linked to the PHem expresses multiple retrotransposable element genes, along with the two retrotransposon gag domain-containing proteins (AGAP029366 and AGAP029372), that we refer to as Retrotransposon Granulocytes (RTr-Gran) (Fig. 1 B and D).

Cell cycle scoring analysis was used to confirm the functional state of Div-Gran and identify other subpopulations that may also be at a different stage of the cell cycle. We found that one of the granulocyte clusters expresses high levels of S-phase markers (SPh-Gran) (Fig. 1 B and C), including specific markers involved in DNA replication, such as Proliferating cell nuclear antigen and Ribonucleoside diphosphatase reductase, as well as dUTP pyrophosphatase (Fig. 1D). The Div-Gran cluster expresses high levels of G2/M phase genes, indicative of cells preparing to undergo mitosis (Fig. 1 B and C), with high expression of gene markers involved in cell division such as Aurora kinase, chromatin assembly regulator Barrier-to-autointegration factor B, and Cyclin A (Fig. 1D), in agreement with previous reports (21). The small cluster previously called Gran-4 (27) (Fig. 1C) expresses G-protein-coupled receptors (GPCR) such as the Calcitonin receptor and Methuselah receptor 4 and is referred to as GPCR-Granulocytes (GPCR-Gran) (Fig. 1 B and D).

The Oen-0 cluster is an intermediate stage between PHem and mature oenocytoids that exhibit strong expression of genes associated with the Notch signaling pathway (hairy and enhancer of split/HES and Notch homolog 1) (Fig. 1 B and D). The two oenocytoid subpopulations identified in this study (Oen-1 and Oen-2 clusters) share several markers, with higher expression of Prophenoloxidase 4 (PPO4) in Oen-1, and higher expression of PPO2 and Aldo-1 epimerase in Oen-2. A comprehensive list of marker genes for each hemocyte cluster is provided in Dataset S1.

Trajectories of Prohemocyte Differentiation into Final Effector Hemocytes.

The differentiation path of the hemocyte clusters in the integrated clustering analysis was investigated using Monocle3, which performs pseudotemporal ordering of the cells based on the gradual changes in their transcriptome profile and generates a trajectory of cluster differentiation. Since PHem are precursors of all other hemocytes (27), it is critical to first precisely identify the root of the differentiation pathways. To this end, we reanalyzed the prohemocyte cluster by itself in more detail, and this revealed four distinct subpopulations: Prohemocyte cluster-0 (PHem-0) and three previously undescribed clusters of precursor S-Phase Granulocytes (P-SPh), precursor Megacytes (P-Mega), and precursor ECM-Granulocytes (P-ECM) (Fig. 2A). The PHem-0 cluster is at the center of the P-SPh, P-Mega, P-ECM, and the Oen-0 cluster (Fig. 2A).

Fig. 2.

Trajectory of hemocyte differentiation. (A) UMAP highlighting the transcriptional subprofiles of the PHem cluster in different colors, with the Oen-0 cluster at 0.8 resolution. (B) Pseudotime plot of the hemocyte clusters analyzed with Monocle3, using PHem-0 as root. (C) Schematic summarizing the differentiation trajectory of the hemocyte clusters determined by Moncole3 analysis. The solid black arrow is the trajectory prediction based on the current scRNAseq dataset; the red dotted arrow indicates trajectory prediction based on the previous analysis report, including sugar-fed and blood-fed samples (27). (D) Violin plot showing the expression of Ficolin, Laminin IV, and Lipase in the P-MEG cluster (indicated in green in the UMAP) in dsLacZ, and dsCactus (****p < 0.0001, Wilcox test). (E) Violin plot showing the expression of HES, Notch, and Aminoacylase in the Oen-0 cluster (indicated in purple in the UMAP) in dsLacZ, and dsCactus (****p < 0.0001, Wilcox test). (F) Relative mRNA expression of Notch and PPO4 in hemocytes in Notch-silenced and control dsLacZ mosquitoes, 4 d post dsRNA injection, n = 3 biological replicates, ****p < 0.0001, ***P < 0.001, Error bar SEM, t test. mRNA expression normalized to RpS7. (G) Percentage of live perfused oenocytoids/mosquito from Notch-silenced mosquitoes, along with the control (dsLacZ), N = 20 individual mosquitoes evaluated in two biological replicates each for dsLacZ and dsNotch, with 10 mosquitoes in each replicate. Median value of the % of oenocytoids/mosquito is indicated by the horizontal line, error bar SEM, ****p < 0.0001, Mann–Whitney U test.

These precursor clusters (P-SPh, P-Mega, P-ECM, and Oen-0) express specific marker genes that are not present in PHem-0. Oen-0 and P-Mega exhibit more prominent expression of unique markers in a higher proportion of cells compared to other precursors. Oen-0 cells express genes involved in Notch signaling (Notch homolog 1 and HES) and Aminoacylase (AGAP000679) (SI Appendix, Fig. S2 and Dataset S2), while P-Mega express genes involved in interactions with the ECM) such as Laminin IV, Ficolin (AGAP012916) as well as a specific Lipase (AGAP001652) (SI Appendix, Fig. S2 and Dataset S2). P-ECM hemocytes express the transcription factor Krueppel-like factor 6/7, an alpha domain-containing Integrin (INT0032), and the putative immune-regulatory CLIPD1 (SI Appendix, Fig. S2 and Dataset S2), while expression of specific markers in the P-SPh-Gran cluster is low and is only present in a small proportion of cells (~10 to 30%). The complete list of differentially expressed genes in precursor hemocyte clusters relative to PHem-0 is provided in Dataset S2.

The Monocle-3 trajectory graph with PHem-0 as the root shows that this cluster gradually transitions toward five distinct primary lineages (Fig. 2B). Oenocytoids and AM-Gran derive directly from the PHem-0 cluster. The P-Mega and P-ECM originate from the PHem-0 cluster, which further differentiates, ultimately leading to Megacytes and ECM-granulocytes. The P-SPh cluster deriving from PHem-0 gives rise to SPh- Gran, which presumably differentiates into Div-Gran, based on a previous study (27) (Fig. 2B). A summary of the differentiation trajectories of the different hemocyte subpopulations is shown schematically in Fig. 2C.

Previous studies showed that Toll activation through dsRNA-mediated Cactus silencing (dsCactus) increases the proportion of megacytes (26). Interestingly, expression of the three P-Mega markers (Ficolin, Laminin IV, and Lipase) is highly induced when Toll signaling is activated (Fig. 2D) (****p < 0.0001, Wilcox test). Conversely, expression of the top three Oen-0 markers (Notch homolog 1, HES, and Aminoacylase) is strongly suppressed (Fig. 2E) (****P < 0.0001, Wilcox test). Notch silencing (dsNotch) reduced the expression of PPO4, a gene expressed in oenocytoids, in hemocytes by 86% compared to dsLacZ controls (P = 0.0001, t test) (Fig. 2F). In addition, the median percentage of circulating oenocytoids from individual mosquitoes was significantly reduced (by two-fold) in Notch-silenced females compared to dsLacZ controls (from 10.9 to 5.0%, P < 0.0001, Mann–Whitney U test), providing direct experimental evidence for the importance of Notch signaling in oenocytoid differentiation (Fig. 2G and Dataset S3). Taken together, these results support the hypothesis that Toll signaling reduces differentiation into the oenocytoid lineage by suppressing Notch expression in oenocytoid precursor (Oen-0) hemocytes.

Modulation of Immune Signaling Alters the Hemocyte Population Structure.

The effect of modulating mosquito innate immune signaling pathways on hemocyte subpopulations was investigated by comparing the relative abundance of cells in the different hemocyte clusters with respect to control females injected with dsLacZ (Fig. 3). A significant increase in cell number is shown in red, a decrease in blue, and white indicates no difference in the proportion of cells. Hemocyte clusters with fewer than 50 cells in both the control and experimental samples were excluded from the analysis and are shown in gray (Fig. 3).

Fig. 3.

Effect of immune signaling modulation on hemocyte cluster distribution. Heatmap representing the fold increase (red), decrease (blue), and no change (white) in the relative proportion of each hemocyte cluster analyzed from the single-cell transcriptome data during signal activation and suppression in log₂ Scale. Gray indicates Clusters excluded from the analysis due to low abundance of cells. ****p < 0.0001, ***p < 0.001, **p < 0.005, *p < 0.05, χ² test.

Activation of the Toll pathway (dsCactus) had the strongest effect of all signaling pathways, affecting the proportion of differentiated hemocyte clusters (Fig. 3 and Dataset S4). We found that it not only significantly increased the proportion of Mega cluster (4.23 times; X² test, P < 0.0001), but also three other granulocyte clusters: AM-Gran (5.75 times; X² test, P < 0.0001), Reg-Gran (8.65 times; X² test, P < 0.0001), and ECM-Gran (7.59 times; X² test, P < 0.0001). Conversely, it strongly decreased the proportion of two oenocytoid clusters: Oen-1 (8.1 times; X² test, P < 0.0001) and Oen-2 (7.05 times; X² test, P < 0.0001) (Fig. 3 and Dataset S4). The observed changes in hemocyte subpopulations support the hypothesis that Toll signaling promotes differentiation of effector granulocytes at the expense of the Oenocytoid lineage. Furthermore, a modest increase in the Div-Gran cluster (1.57 times, Χ² test, P = 0.0006) and a slight decrease in the PHem cluster (1.72 times, X² test, P < 0.0001) were also observed (Fig. 3 and Dataset S4). JNK signal activation (dsPuckered) had a similar but less striking effect than Toll, as it also increased the proportion of the Mega (2.43 times, X² test, P < 0.0001) and ECM-Gran clusters (2 times, X² test, P = 0.0128), and decreased the proportion of cells in all oenocytoid clusters (Oen-0 = 2 times, Oen-1 = 2.1 times, and Oen-2 = 3.21 times; X² test, p < 0.0001) (Fig. 3 and Dataset S4).

The effect of STAT2 activation (dsSOCS) was more specific, greatly increasing the proportion of cells in the GPCR-Gran cluster (10.41 times; X² test, p < 0.0001), with a moderate increase in the proportion of ECM-Gran (2.85 times; X² test, p < 0.0001) and Oen-1 clusters (1.4 times, X² test, p = 0.0001). Activation of the IMD signaling pathway (dsCaspar) had no significant effect on the oenocytoid lineage but significantly increased the ECM-Gran (5.92 times; X² test, p < 0.0001) and AM-Gran clusters (3.1 times; X² test, p < 0.0001). Activation of Toll, STAT2, and IMD pathways resulted in a significant decrease in the proportion of PHem (Fig. 3; X² test, p < 0.0001 and Dataset S4), indicating that activation of these signaling pathways promotes differentiation of PHem into specific effector hemocytes.

Suppression of any of the four immune signaling cascades had no significant effect on the proportion of the oenocytoid lineage clusters (Fig. 3). However, inhibiting Toll signaling (dsRel1) increased AM-Gran (3.9 times; X² test, p < 0.0001) and Reg-Gran (2.9 times; X² test, p < 0.0001) clusters (Fig. 3 and Dataset S4). Suppressing JNK signaling (dsJNK) resulted in a decrease in Div-Gran (2 times, X² test, p = 0.0046) (Fig. 3 and Dataset S4), and a mild decrease in SPh- Gran (1.25 times; X² test, P = 0.02) was observed after STAT2 signal suppression (dsSTAT2) (Fig. 3 and Dataset S4), suggesting that JNK and STAT2 signaling may be involved in hemocyte proliferation. Suppressing IMD signaling (dsRel2) increased the ECM-Gran cluster (4.2 times; X² test, P < 0.0001) and slightly decreased the PHem (1.19 times; X² test, P < 0.0001) (Fig. 3 and Dataset S4).

General and Cell-Specific Transcriptional Responses of Hemocytes to Modulation of Immune Signaling.

Bulk sequencing analysis was performed by generating short-read Illumina sequencing cDNA libraries from mRNA of hemocytes subjected to activation or suppression of the four immune signaling cascades to investigate the global impact on the hemocyte transcriptome, using the previously reported in-depth hemocyte transcriptome as a reference (27). Gene Ontology (GO) enrichment analysis was used to elucidate the functional classes of genes differentially expressed in hemocytes. In parallel, cluster-specific differential gene expression (DE) analysis using the scRNAseq data was conducted to establish the transcriptional response of individual hemocyte clusters. In general, activation of signaling pathways had a more prominent effect on hemocyte transcriptome compared to signal suppression (Fig. 4 A–H and SI Appendix, Figs. S3 and S4).

Fig. 4.

Transcriptional response of hemocytes during immune signaling activation. (A) Volcano plot depicting the upregulated (red dots) and downregulated (blue dots) transcripts (Log₂ Fold change >1, adjusted p value < 0.05) in hemocytes during Cactus silencing, (B) Puckered silencing, (C) SOCS silencing, (D) Caspar silencing compared to control (dsLacZ) in bulk transcriptome analysis with the total number of upregulated and downregulated transcripts indicated in the lower right and left corners respectively of each volcano plot. (E) Heatmap representing the upregulated (red) and downregulated genes (blue) in specific hemocyte clusters upon silencing Cactus, (F) Puckered, (G) SOCS, (H) Caspar compared to control ds LacZ (Log₂ Fold change >1, adjusted p value < 0.05) from scRNAseq data analysis. Each column represents an individual hemocyte cluster, each row represents a gene, color intensity represents the Log₂ Fold change. The top 50 upregulated and downregulated genes are shown for Cactus silencing, and the complete list is available in Dataset S7. RTr-gran and GPCR-Gran not included in the analysis due to low abundance of cells, and Reg-Gran had DE genes only upon Cactus silencing (Refer to Dataset S7). Ph. lipase B- Phospholipase B-like, TFIID - Transcription initiation factor IID subunit1, TFIIIB-Transcription factor IIIB 90 kDa subunit, Cytd. deam. -Cytidine deaminase.

Activation of Toll Signaling.

Activation of Toll signaling (dsCactus) significantly induced expression of several immune-related proteins, including Toll, multiple Ficolins, PGRP-LA, immunomodulatory serine proteases, Serpins, TEP3, and Cytidine deaminase, an enzyme that inhibits retroviral replication, as indicated in the bulk transcriptome analysis (Fig. 4A and Dataset S5). Expression of several proteins mediating adhesion and interaction with the ECM, including multiple integrins, such as Integrin alpha-ps (INT6826) and Dystroglycan, as well as the megacyte marker TM7319, also increased significantly. GO analysis of all differentially expressed genes revealed enrichment of genes involved in actin cytoskeleton organization and intracellular vesicular transport (SI Appendix, Fig. S4 A, Left and Dataset S6) and underrepresentation of transcripts from enzymes involved in nucleotide, energy, and amino acid metabolism (SI Appendix, Fig. S4 A, Right and Dataset S6). Cluster-specific analysis showed that activation of the Toll signaling (dsCactus) resulted in a broad increase in expression of some immune-related genes such as Toll, multiple Lysozymes, and TEP3, in most hemocyte clusters (Log₂FC > 1; p < 0.05) (Fig. 4E and Dataset S7); and broad suppression of Aminoacylase, EGF-like domain-containing protein, and PPO6 expression in multiple clusters. In contrast, the decrease in expression of genes involved in Notch signaling (Notch homolog 1 and HES) was very specific to the Oen-0 cluster (Fig. 4E).

Activation of JNK Signaling.

JNK signal activation (dsPuckered) also has a major effect on the bulk hemocyte transcriptome with some similarities, but a milder response than Toll activation with a lesser number of differentially regulated transcripts (Fig. 4B). It increases expression of several immune genes, such as Toll and Toll-like receptors, Fibrinogens, CLIPB17, Metalloproteases, INT6826, and DE-cadherin, along with the megacyte marker TM7319 (Dataset S5). GO analysis showed enrichment of genes involved in regulation of actin cytoskeleton remodeling, signal transduction, intracellular transport, and cell migration (SI Appendix, Fig. S4 B, Left and Dataset S6); and underrepresentation of genes required for amino acid and nucleotide metabolism (SI Appendix, Fig. S4 B, Right and Dataset S6). JNK signal activation (dsPuckered) affected the transcriptome of several hemocyte clusters, but to a much lesser extent than Toll activation. (Fig. 4F and Dataset S7).

Activation of STAT2 Signaling.

In general, activation of STAT2 signaling (dsSOCS) increased expression of transcripts involved in cell proliferation (DNA replication ATP-dependent helicase, HES, 1-D myo-inositol-triphosphate 3-kinase, Phospholipase B-like, and M-Phase Phosphoprotein) (Fig. 4C and Dataset S5), and genes mediating hemocyte migration, hematopoiesis, and cellular stress responses were underrepresented in the GO analysis (SI Appendix, Fig. S4C and Dataset S6).

Interestingly, STAT2 activation (dsSOCS) had a specific effect on gene expression in two hemocyte clusters, AM-Gran and ECM-Gran. Several immune genes, including Cec B, Cec 2, Defensin 1, Serine Protease 24D, and PGRPs (short) were induced in AM-Gran (Log₂FC > 5) (Fig. 4G and Dataset S7). In ECM-Gran hemocytes, STAT2 activation induced expression of genes known to mediate interaction with the ECM in other systems, such as Chitinase, Terribly reduced optic lobes isoform B (Trol), and Dystroglycan, while expression of Laminins and Sparc beta/gamma was suppressed. In addition, expression of genes involved in energy (Cytochromes b and c) and lipid metabolism (Lipophorin and ApoD) also increased in ECM-Gran, while expression of immune-responsive genes such as Nimrod and LRIM16A was reduced (Fig. 4G and Dataset S7).

Activation of IMD Signaling.

Overall, activation of IMD signaling (dsCaspar) had a moderate effect on hemocyte gene expression, resulting in 230 differentially expressed transcripts (Fig. 4D). Expression of two transcription initiation factors (TFIID, TFIIIB) and molecular stress-response chaperones (DnaJ homologs and HtpG) increased, while expression of transcriptional repressors such as Methyltransferase, BAH domain-containing protein, and some metabolic enzymes was reduced (Dataset S5), also indicated in the GO analysis (SI Appendix, Fig. S4D and Dataset S6). IMD activation (dsCaspar) had the strongest effect in the ECM-Gran cluster with increased expression of Chitinase and Actin, and reduced expression of the immune genes NimrodB2 and LRIM16A (Fig. 4H and Dataset S7).

Effect of Suppressing Immune Signaling on Hemocyte Transcriptome.

Bulk transcriptomic analysis revealed that, as expected, suppression of Toll signaling (dsRel1) decreased expression of several immune-related genes such as Cecropin 1, CLIPs (B3,19, A1, and SCLIP1), C-type Lysozyme, and TEPs (TEP1,4 and 6) (SI Appendix, Fig. S3A and Dataset S5), which is also reflected in the GO enrichment analysis (SI Appendix, Fig. S4E and Dataset S6). However, single-cell analysis showed that it also had a unique effect in AM-Gran. Surprisingly, Toll suppression significantly increased expression of several immune genes, such as antimicrobial peptides (AMPs) (Cec B, and Defensin 1), CLIPs (CLIPA14, B8, B15, and B17), Serpins (Serpin 1, 11), and PPO activating factor (SI Appendix, Fig. S3E and Dataset S7) in AM-Gran, indicative of immune activation of this hemocyte subpopulation.

Inhibition of JNK signaling (dsJNK) had a modest effect on the bulk hemocyte transcriptome, affecting expression of only 30 transcripts (SI Appendix, Fig. S3B and Dataset S5), with reduced expression of proteins that interact with the ECM, such as Integrin (INT6826), Dystroglycan, Chitinase, and Matrix metalloprotease (Dataset S5), and only very few genes were differentially expressed in specific hemocyte clusters (SI Appendix, Fig. S3F and Dataset S7).

Suppressing STAT2 signaling (dsSTAT2) affected the expression of more transcripts (230 transcripts) than activating this pathway (dsSOCS, 76 transcripts) in the bulk transcriptome analysis. Expression of some immune genes, such as Toll, TNF-receptor Wengen, and TGF-beta receptor, increased, while that of Ficolin, CLIPs, and PPO9 was reduced (SI Appendix, Fig. S3C and Dataset S5). STAT2 suppression (dsSTAT2) had a specific and striking effect on the ECM-Gran cluster, with increased expression of immune receptors like Toll, Toll 5A, and Serpins (Serpin 7, 9 and 10), LITAF-like1, and LRIM17 (SI Appendix, Fig. S3G and Dataset S7). The decreased expression of several ribosomal proteins (RpL 23, 26, 38, 40, Rps 21) indicates that translational activity is suppressed in these cells (SI Appendix, Fig. S3G and Dataset S7).

Inhibition of IMD signaling (dsRel2) reduced expression of immune defense genes (C-type Lysozyme, Cecropin B, CLIPC7, and Serine proteases), while expression of proteins involved in mRNA translation (multiple large and small subunits of rRNAs) and intracellular trafficking regulators (phosphatidylinositol-4-phosphate 5-Kinase and ankyrin repeat-containing protein) was induced (SI Appendix, Fig. S3D and Dataset S5). Several genes involved in regulating energy metabolism were also enriched in the GO analysis (SI Appendix, Fig. S4F and Dataset S6). IMD suppression (dsRel2) has a strong and specific effect on Oen-2, inducing expression of kinases of the PI3K/Akt pathway (AKT and PDK1), of two transcriptional regulators (Vrille and basic helix–loop–helix protein), a vesicular trafficking protein (TMED10), and a modulator of lipid metabolism (Phosphatidate phosphatase LPIN) (SI Appendix, Fig. S3H and Dataset S7).

Effect of Activating Toll and JNK Signaling on Hemocytes.

Given the significant impact of Toll and JNK activation on granulocyte differentiation and gene expression, their effect on the size and morphology of circulating granulocytes was examined directly by fluorescence microscopy (Fig. 5 A–C). We confirmed the previously reported striking size increase of granulocytes in response to Toll activation (dsCactus) (26) (Fig. 5 D and E), as granulocytes (cells with diameters of 15 to 25 µm) differentiate into megacytes (cells with a diameter >25 µm); with a 31.9-fold increase in the proportion of megacytes in dsCactus (n = 510 granulocytes), relative to dsLacZ controls (n = 543 granulocytes), from 2.39 to 76.47% (X² test, p < 0.0001, Fig. 5 D and E and Dataset S8). JNK activation (dsPuckered) also promoted differentiation of granulocytes to megacytes, with high significance, but lesser magnitude than with Toll signaling, with a 7.28-fold increase in the proportion of megacytes in dsPuckered (n = 708 granulocytes), relative to dsLacZ controls (n = 543 granulocytes), from 2.39 to 17.44% (X² test, p < 0.0001, Dataset S8, Fig. 5 D and E). The median diameter of Megacytes when Toll is activated (37.39 µm, n = 431 megacytes analyzed) was also significantly larger than with JNK activation (30.7 µm, n = 211 megacytes analyzed, p < 0.0001, Mann–Whitney U test) (SI Appendix, Fig. S5A and Dataset S9). Direct counting of live perfused hemocytes using a Neubauer chamber confirmed a significant reduction in the percentage of circulating oenocytoids per mosquito in response to Toll activation, relative to dsLacZ controls from 8.4 to 5.3% (p < 0.0001, Mann–Whitney U test) and from 8.2 to 5.8% with activation of JNK signaling (p < 0.001, Mann–Whitney U test) (SI Appendix, Fig. S5 B and C and Dataset S3).

Fig. 5.

Role of Toll and Jnk activation on hemocytes and midgut immune surveillance. (A–C) Morphology of attached hemocytes from control (dsLacZ), JNK signal-activated (dsPuck), and Toll-activated (dsCact) mosquitoes by phalloidin staining of actin. White arrows indicate megacytes. Green- Actin, Blue- DAPI, (Scale bar- 20 μm) (D) Size distribution and (E) percentage of megacytes in the adherent granulocytes collected from dsLacZ, dsPuck, and dsCact injected mosquitoes, megacyte diameter >25 μm. N represents the total number of granulocytes analyzed from 3 biological replicates, ****p < 0.0001, χ² test. (F and G) RNA in situ hybridization for INT6826 (green) and TM7319 (red), along with Tubulin antibody staining (cyan) of adherent granulocytes and (H and I) megacytes from dsLacZ-injected sugar-fed A. gambiae mosquitoes. Arrows in (F and G) indicate INT6826 expression and undetectable TM7319 expression in granulocytes (Scale bar- 20 μm), and the arrow in (H and I) indicates high TM7319 and negligible INT6826 expression in megacytes (Scale bar- 20 μm). Nuclei stained in blue (DAPI). (J) Relative mRNA expression level of INT6826 and TM7319 in midgut from control and bacteria-fed mosquitoes, 4 to 6 h postfeeding, n = 3 biological replicates, ****p < 0.0001, Error bar SEM, t test. (K) Relative mRNA expression level of INT6826 and TM7319 in midgut from control blood-fed and P. berghei infected blood-fed mosquitoes, 24 h postfeeding, n = 3 biological replicates, ****p < 0.0001, Error bar SEM, t test. (L and M) RNA in situ hybridization for INT6826 (green) and TM7319 (red), along with Tubulin antibody staining (cyan) of adherent hemocytes from dsCactus-injected sugar-fed A. gambiae mosquitoes. Arrows indicate high expression of INT6826 and TM7319 in megacytes. Nuclei stained in blue (DAPI), (Scale bar- 20 μm). (N) Percentage of megacytes expressing high levels of TM7319 and INT6826 in adherent hemocytes from dsLacZ and dsCactus injected mosquitoes, n = 3 biological replicates, error bar SEM, p < 0.005, χ² test. (O) Relative mRNA expression level of INT6826 and TM7319 in midgut from sugar-fed mosquitoes injected with dsLacZ and dsCactus 4 d postinjection, n = 3 biological replicates, ****p < 0.0001, Error bar SEM, t test. mRNA expression normalized to RpS7.

Role of ECM-Gran and Megacytes on Midgut Immune Surveillance.

Under basal conditions (dsLacZ), ECM-Gran hemocytes express Integrin alpha (INT6826) mRNA, while expression of the Megacyte marker (TM7319) cannot be detected (INT6826^h, TM7319^l), and they represent 40% of adherent granulocytes (n = 132 granulocytes analyzed by in situ hybridization, from three biological replicates) (Fig. 5 F and G and Dataset S10). Megacytes (TM7319^h, NT6826^l) are less abundant, representing ~3% of granulocytes (Fig. 5E), and these markers have the opposite expression pattern (Fig. 5 H and I). We have previously shown that when female mosquitoes are fed with bacteria, the number of granulocytes associated with the basal surface of the midgut increases significantly (28). We analyzed expression of these hemocyte markers in midguts isolated 4 h postfeeding from females fed a meal with or without bacteria and found that the presence of bacteria increases mRNA levels of INT6826 by 4.2-fold (n = 3 biological replicates, p < 0.0001, t test), with no difference in mRNA levels of the Megacyte marker TM7319 (Fig. 5J and SI Appendix, Table S1A). A similar response was observed in midguts isolated 24 h after females were fed on a healthy control or on a mouse infected with P. berghei parasites. INT6826 mRNA expression increased by 4.4-fold (n = 3 biological replicates, p < 0.0001, t test) in P. berghei–infected midguts, while mRNA levels of TM7319 did not change significantly (Fig. 5K and SI Appendix, Table S1B). These observations suggest ECM-Gran hemocytes (expressing INT6826) participate in basal immune surveillance and are recruited to the surface of the midgut in response to immune challenges.

Activation of Toll signaling (dsCactus) has a dramatic effect on hemocyte differentiation, increasing the proportion Megacytes in the granulocyte lineage by 32-fold compared to dsLacZ controls (Fig. 5E). We confirmed this increase in Megacytes by in situ hybridization (Fig. 5 L and M) and found that 74.5% of megacytes express high levels of TM7319 mRNA (TM7319^h) (n = 253 granulocytes analyzed, from three biological replicates). We also observed an impressive increase in Megacytes expressing high levels of INT6826 mRNA from 0% in dsLacZ controls (n = 132 granulocytes analyzed, from three biological replicates) to 66.18% after Toll activation (dsCactus) (p < 0.005, χ² test) (Fig. 5N and Dataset S10).

Midguts from naïve sugar-fed females that had not been immune challenged were dissected to investigate the effect of Toll activation on the basal level of hemocyte patrolling of the midgut surface. Interestingly, Toll activation (dsCactus) resulted in a striking increase (29-fold) in TM7319 mRNA levels associated with the midgut relative to dsLacZ controls (p < 0.0001, three biological replicates, t test) and a significant, but more moderate increase (3.6-fold) of INT6826 expression indicating enhanced association of megacytes with the midgut even under normal conditions (p < 0.0001, three biological replicates, t test) (Fig. 5O and SI Appendix, Table S1C).

Finally, we investigated whether dsCactus could also induce expression of markers of other effector granulocytes in Megacytes. We found that, indeed, 48% of TM7319^h Megacytes also expressed high levels of the Reg-Gran marker POI4674 (SI Appendix, Fig. S5 D and E). Similarly, 36% of TM7319^h Megacytes coexpressed high levels of the RTr-Gran marker UP29365 (SI Appendix, Fig. S5 F and G). In dsLacZ controls, Megacytes (0.9 to 1.7%), Reg-Gran (2.7%), and RTr-Gran (4.6%) represent a small proportion of the total granulocytes, but the expression of the molecular markers is very specific (Dataset S10). High TM7319 expression was only present in dsLacZ Megacytes, and all of them were positive for this marker. Furthermore, high expression of TM7319 with any of the other markers was never observed in dsLacZ hemocytes (Dataset S10). These findings indicate that activation of Toll signaling results in broad transcriptional activation and leakiness of genes whose expression is normally restricted to specific hemocyte subpopulations. They also indicate that Toll activation can induce the differentiation of other hemocytes, besides TM7319^h Megacytes, into a giant cell phenotype that is not normally observed in dsLacZ controls.

Discussion

The scRNAseq libraries generated from stabilized hemocytes enabled the recovery of high-quality single-cell transcriptomes from a very large number of hemocytes (37,529 cells) compared to previous reports (21, 29). Integrated clustering analysis, using an in-depth hemocyte transcriptome annotation as a reference (27), identified 12 hemocyte clusters with distinct transcriptional profiles. Careful functional analysis of cluster-specific markers allowed the identification of three subpopulations of oenocytoids (Oen-0, Oen-1, and Oen-2) and three low-abundance effector granulocytes with specialized functions, RTr-Gran, Reg-Gran, and ECM-Gran not previously described (Fig. 1). We also identified proliferating granulocytes at different stages of the cell cycle, with the SPh-Gran cluster specifically expressing markers involved in DNA replication, and Div-Gran expressing high levels of G2/M phase genes, indicative of cells preparing to undergo mitosis. A third cluster, GPCR-Gran, is characterized by the expression of multiple GPCRs. GPCR signaling influences intracellular cAMP levels, which in turn can modulate cell cycle progression by regulating the expression of cyclins and cyclin-dependent kinase inhibitors (30).

Retrotransposon gag domain-containing proteins confer genetic diversity in Drosophila and yeast by regulating the mobilization and activation of transposable elements (31, 32). In Drosophila, they also provide systemic antiviral immunity (33) and serve as chaperones during stress responses (34), bind mRNAs and package them into extracellular vesicles (35). Furthermore, the Toll pathway is critical for limiting viral infections in Drosophila (36) and Aedes (37). The observed expression of several retrotransposon gag domain proteins in RTr-Gran and the striking increase of this cell population upon Toll activation suggest that RTr-Gran may play a specialized role in antiviral immunity and/or in the regulation of retrotransposons.

A detailed analysis of hemocyte differentiation trajectories identified early genes expressed as PHem commit to three major lineages: oenocytoids, proliferating granulocytes, and effector granulocytes (Fig. 2), with gradual induction of cluster-specific marker gene expression as cells undergo terminal differentiation. Activation of Notch signaling, through interactions of Notch with the transcription factor Lozenge, is known to be critical for the development of crystal cells (equivalent of oenocytoids in mosquitoes) in Drosophila larvae (38). The high expression of Notch ligand and HES (a downstream target of Notch signaling) in the Oen-0 cluster suggests a conserved role of Notch signaling in oenocytoid differentiation in adult A. gambiae mosquitoes, in agreement with a recent report showing that silencing Lozenge reduced PPO1, PPO3, and PPO8 expression, as well as the number of cells expressing other oenocytoid markers (29). In vertebrates, the Toll pathway promotes activation of Notch signaling in macrophages by inducing the expression of Notch receptors and HES (39). However, in A. gambiae females, we observed the opposite effect, with Toll signaling strongly reducing the expression of the Notch receptor and HES in the Oen-0 cluster and suppressing PHem differentiation to the oenocytoid lineage, with a significant decrease in both Oen-1 and Oen-2. Furthermore, activation of JNK signaling by silencing the phosphatase Puckered also reduces Notch and HES expression and decreases the proportion of oenocytoids. However, this effect seems to start earlier than with Toll activation, as all oenocytoid subpopulations, including Oen-0 are significantly reduced. The observed differentiation of hemocytes into the megacyte lineage following Toll signal activation is consistent with previous reports (26) (Fig. 2). A concomitant increase in megacytes observed upon JNK activation is not previously described, but the increase in the proportion and cell size of megacytes is less intense than Toll signal activation.

The Toll and IMD signaling pathways are known to regulate AMP expression in Anopheles mosquitoes (40), and the significant increase in AM-Gran upon Toll and IMD activation is probably contributing to the observed overall increase in AMP expression (Fig. 2). However, the observed increase in AM-Gran when Toll signaling was suppressed (dsRel1) is counterintuitive. One can envision that suppressing this pathway may cause an imbalance in the microbiota, indirectly activating other pathways, such as the JAK-STAT pathway, which is known to induce AMP expression (41), and may restore homeostasis by promoting differentiation to AM-Gran. The AM-Gran cluster comprises only a small proportion of hemocytes, and it is hard to imagine that they, by themselves, could mount an adequate systemic immune response. Nonetheless, immunomodulatory effects of AMPs have been recently documented, such as defensins acting as chemo attractants for various immune cells (42). Furthermore, the AMPs produced by AM-Gran may only eliminate a small number of microbes, but in doing so, the immune elicitors they release may serve as a signal to amplify the response of other hemocytes, and/or of the fat body, a major source of AMPs in the hemolymph (43).

The impact of modulating immune signaling on hemocyte gene expression was investigated via two complementary approaches. Bulk transcriptome analysis provides greater depth of coverage, while cluster-specific DEG analysis using scRNAseq data allows investigation of cell-specific responses (Fig. 4).

The stronger effect of activating signaling pathways on hemocytes, compared to suppressing them, may be due to the very low basal levels of immune signaling in the absence of an immune challenge. Bulk transcriptome analysis showed that activation of Toll (dsCactus) and JNK (dsPuckered) had a strong effect on the transcriptional profile of hemocytes, with a large number of differentially expressed transcripts, whereas modulation of STAT2 and IMD signaling had a modest effect with only a small number of differentially expressed transcripts. This was confirmed by the cluster-specific differential gene expression analysis from the scRNAseq data, indicating that Toll activation (dsCactus) triggers a broad transcriptional response, with differential expression of several genes in most hemocyte clusters. In contrast, disrupting Toll signaling (dsRel1) and modulation of the STAT2 pathway (dsSOCS and dsSTAT2) affected gene expression of very specific hemocyte clusters. The similar effect of Toll and JNK activation on hemocyte differentiation and gene expression suggests that there may be crosstalk between the Toll and the JNK signaling pathways.

Integrins are heterodimeric transmembrane proteins composed of alpha and beta subunits that mediate interactions between cells, as well as interactions of cells with the ECM (44). Integrins regulate the migration, adhesion, aggregation, maturation, and even immunoregulation of several immune cells, such as lymphocytes, dendritic cells, NK cells, neutrophils, and macrophages (45), aiding “rolling” leukocytes to adhere firmly to the endothelium (46).

Under basal conditions, ECM-Gran expressing Integrin alpha-ps (INT6826) is found to be 40% of the granulocytes detected by in situ hybridization (Fig. 5). Both bacterial feeding and Plasmodium infection increase the association of ECM-Gran expressing INT6826 with the midgut, while TM7319 mRNAs from megacytes, remain low. In vertebrates, activation of Toll signaling increases integrin expression in B cells (47) and functionally activates integrin in leukocytes (48), which is critical for their adhesion to the vascular endothelium. Global transcriptome analysis revealed a dramatic increase in INT6826 expression in hemocytes during Toll activation. We showed that, besides promoting granulocyte differentiation to megacytes, Toll activation also induces high levels of INT6826 expression in a large fraction of megacytes. Moreover, megacytes during Toll activation show a constitutive association with midgut, even in the absence of an immune challenge, indicative of an enhanced basal level of immune surveillance. This response could explain the highly efficient release of hemocyte-derived microvesicles that mediates a very strong complement-mediated elimination of Plasmodium ookinetes upon Toll activation in A. gambiae females (49).

By investigating the effects of activation and suppression of individual signaling pathways at a single-cell resolution, this study identified several hemocyte subpopulations not previously characterized in A. gambiae adult females, the main vector of P. falciparum in Africa. It also revealed how different signaling pathways affect hemocyte differentiation, gene expression, and immune function. This study provides detailed insights into the specialized functions of previously uncharacterized hemocyte populations and identifies cell-specific molecular markers that will be invaluable for future studies on how different hemocytes contribute to basal immune function, as well as antiplasmodial and antiviral responses, which could be potential targets to disrupt disease transmission.

Materials and Methods

Mosquito Rearing.

A. gambiae G3-CDC strain was used for all the experiments. The mosquitoes were reared at 27 °C, with 80% humidity on a 12-h light-to-dark cycle.

Gene Silencing in Hemocytes by dsRNA Microinjection.

The dsRNAs targeting specific genes were synthesized by in vitro transcription using the MEGAscript RNAi kit (Ambion, Thermo Fisher Scientific, Waltham, MA) following the manufacturer’s protocol. A list of primers used to synthesize the dsRNA and to check the gene expression by qPCR is listed in SI Appendix, Table S2. The silencing efficiency data of the dsRNA in hemocytes are provided in SI Appendix, Fig. S1B. The detailed method is provided in SI Appendix, Materials and Methods.

RNA Isolation from Hemocytes.

Hemolymph was collected from 20 female A. gambiae mosquitoes with modified anticoagulant buffer (95% Schneider’s media with 5% citrate buffer, pH 7.0 to 7.4) and stabilized with RNAprotect Cell Reagent (Qiagen 76526) before cell lysis and RNA isolation. The detailed method is provided in SI Appendix, Materials and Methods.

Hemocyte Preparation for scRNAseq Library Construction.

Hemolymph was collected from the female A. gambiae mosquitoes as described earlier 4 d post dsRNA injection. The hemocytes were stabilized using vivoPHIX (Rapid Labs, RNAssist Ltd, Cambridge, UK, Catalog RD-VIVO-50) immediately after perfusion and later recovered from the stabilizing agent using 75% ethanol before proceeding with the scRNAseq reaction. The detailed method is provided in SI Appendix, Materials and Methods.

scRNAseq Library Construction, Sequencing, and Data Analysis.

The scRNAseq libraries were generated using the Chromium Next GEM Single Cell 3’ Reagent kit v3.1 by 10× GENOMICS following the manufacturer’s protocol. The cDNA was amplified by 13 cycles of PCR and the Dual Index kit was used to label the libraries. The quality of the libraries was monitored using Agilent Tapestation 4200 instrument. Three scRNAseq libraries were pooled together for each paired-end sequencing reaction run on the NovaSeq SP platform for 100 cycles run to target 30,000 reads per cell. The pool balancing of the libraries was done by iSeq QC run before the NovaSeq sequencing run. The detailed description of the scRNAseq data analysis is provided in the SI Appendix, Materials and Methods.

Bulk RNA Sequencing and Data Analysis.

The hemocytes were collected from the mosquitoes 4 d post dsRNA injection followed by RNA isolation as described earlier. The integrity of RNA collected from hemocytes was checked with High Sensitivity RNA Screen Tape Analysis on the Agilent Tapestation 4200 instrument before proceeding with library preparation. Poly-A enrichment workflow method was used to select the mRNAs and NEBNext Ultra II RNA library prep kit was used to generate nondirectional libraries. The libraries were pooled and paired-end sequencing was done using the Illumina Novaseq 6000 instrument. The description of the bulk RNA sequencing data analysis and GO enrichment analysis of differentially regulated genes is provided in SI Appendix, Materials and Methods.

RNA Isolation from Midguts.

The mosquito midguts were dissected in 1× PBS and the tissue without any PBS was transferred to a tube containing 700 µL of TRIzol (Thermo Fisher Scientific 15596026) used for RNA isolation. After dissecting pooling 10 to 15 midguts per sample, the tissue was homogenized with a plastic pestle, and the volume was adjusted to 1 mL and incubated at room temperature for 5 min to allow lysis of the tissue. Total RNA isolation and cDNA synthesis was performed as described before.

Morphology-Based Quantification of Hemocyte Population.

Hemocytes from individual mosquitoes were counted 4 d post dsRNA injection with 10 µL of anticoagulant buffer (60% Schneider’s medium, 30% Citrate buffer, pH 4.5, 10% FBS, final pH of the buffer is 7 to 7.4). The hemolymph was collected using siliconized tips and transferred to a Neubauer-improved disposable hemocytometer (Fisher Scientific, Catalog 22-600-100). The slides were incubated in a humid chamber for 10 to 15 min at room temperature to allow the cells to settle down before counting. The hemocytes from 10 to 12 individual mosquitoes were counted for each experimental condition using a light microscope, and the cell count data from three biological replicates were plotted.

Phalloidin Staining of Hemocytes to Determine the Morphology.

5 µL of hemolymph from each mosquito was collected 4 d post dsRNA injection and allowed to attach on a Cell-Tak coated chamber slide, fixed with 4% paraformaldehyde solution and stained with 1 U Phalloidin Alexa Fluor 488 (Molecular Probes, Thermo Fisher Scientific, Waltham, MA). The detailed method is provided in SI Appendix, Materials and Methods.

Bacterial Feeding and P. berghei Infection.

A bacterial mixture was isolated from the midguts of the A. gambiae G3 from the lab colony (28) and was used to feed 3- to 4-d old mosquitoes. P. berghei strain (ANKA 2.34) expressing GFP was used for mosquito infections. The detailed method is provided in SI Appendix, Materials and Methods.

RNA In-Situ Hybridization of Hemocytes.

Specific RNA probes for the target genes to label the hemocyte subtypes were designed in collaboration of ACDBio. The hemocytes were subjected to ISH using the RNAscope multiplex fluorescent reagent kit v2 assay (catalog 323110, ACDBio) following the manufacturer’s instructions. Tyramide Signal Amplification-based fluorophores Opal 4- color automation IHC kit (catalog NEL801001KT, PerkinElmer, Waltham, MA) was used for the development of fluorescence. The detailed method is provided in SI Appendix, Materials and Methods.

Ethics Approval.

All the animal experimental protocols were approved by the ethics committee of National Institute of Allergy and Infectious Diseases, the NIH Animal Care and User Committee, approval ID ASP-LMVR 5. Public Health Service Animal Welfare Assurance #A4149-01 guidelines were followed according to the NIH Animal Office of Animal Care and Use. Mice were anesthetized before mosquito feeding. Research and experiments with A. gambiae mosquitoes are reported as per ARRIVE guidelines. No humans or human samples were used in this study.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Dataset S10 (XLSX)

Data, Materials, and Software Availability

All the raw fastq files from single-cell and bulk RNA sequencing used in this study are deposited in the Sequence Reads Archive of the National Center for Biotechnology Information under the BioProject accession number PRJNA1312101 (50). All scripts used for data processing and analysis in this study are available in the Zenodo repository at https://doi.org/10.5281/zenodo.18224827 (51). All other data are included in the manuscript and/or supporting information.