MaRNAV-1 infection of Plasmodium vivax is associated with increased parasite transmission and host inflammatory responses

Abstract

Matryoshka RNA virus 1 (MaRNAV-1) is an RNA virus recently identified in Plasmodium vivax infected samples, but definitive evidence that it infects the parasite and influences malaria pathogenesis remains unknown. Here, we demonstrate that MaRNAV-1 is an intracellular virus infecting P. vivax across various stages of its lifecycle, including blood stages, sporozoites, and liver stages. Viral prevalence varied geographically between Cambodia and Ethiopia, and sequence analyses revealed substantial nucleotide polymorphisms but a high level of protein conservation among viruses. MaRNAV-1 presence and load were positively associated with parasite transmission potential, as reflected by increased gametocyte abundance and higher oocyst prevalence and intensity in mosquito infection assays.

In the patients, higher MaRNAV-1 loads were associated with elevated body temperature and increased concentrations of inflammatory cytokines, including IFN-γ, IP-10, IL-6, IL-10, IL-1RA, and VEGF, independently of parasitemia. Consistent with this inflammatory profile, MaRNAV-1 loads were lower in asymptomatic infections compared to symptomatic malaria cases. Transcriptomic analyses further revealed differences in host immune-cell composition, including higher proportions of innate immune populations in virus-positive infections.

Together, these findings demonstrate that MaRNAV-1 is a genuine parasite-infecting virus that is associated with increased parasite transmission potential and host inflammatory responses. Our study broadens the conventional view of host-pathogen interactions in malaria by revealing complex virus-parasite-host relationships.

Introduction

Viruses are ubiquitous biological entities capable of infecting all forms of life, from bacteria to eukaryotes, including parasitic protozoa. Viruses infecting parasites were first described several decades ago, when virus-like particles (VLPs) were observed in Plasmodium, Entamoeba and Leishmania by transmission electron microscopy^1–3. The first well-characterized protozoan-infecting virus was discovered in Trichomonas vaginalis, the causative agent of trichomoniasis^4,5.

Despite these early observations, viruses infecting Plasmodium species remain poorly characterized. Five Plasmodium species primarily cause human malaria: P. falciparum, P. vivax, P. ovale, P. malariae and P. knowlesi. Of these, P. vivax is the most geographically widespread human malaria parasite and represents a major cause of malaria morbidity, particularly in Asia, South America, and the Horn of Africa⁶. Although historically considered responsible for relatively benign disease, increasing evidence indicates that P. vivax infections can cause severe complications and mortality^7,8

Recently, a bi-segmented narnavirus was identified by metatranscriptomic in P. vivax infected blood samples⁹. This virus, termed Matryoshka RNA Virus 1 (MaRNAV-1), was detected across multiple RNA-seq datasets from P. vivax-infected blood samples. By contrast, it was absent in datasets derived from P. falciparum, P. malariae, P. ovale and P. knowlesi infections^9,10. More recently, an unrelated virus was reported in association with P. knowlesi infections¹¹, suggesting that different Plasmodium species may harbor distinct viruses.

This association of MaRNAV-1 with P. vivax infections raises intriguing questions about the virus’s host range and its potential role in parasite biology and pathogenesis. Viruses infecting protozoan parasites can profoundly influence host-parasite interactions^12–18. For example, double-stranded RNA viruses infecting Leishmania species have been shown to exacerbate disease severity by triggering inflammatory responses in the human host¹⁹. Similarly, viruses infecting Trichomonas and Giardia species can alter parasite virulence and treatment outcomes²⁰. In other systems, viruses may reduce the pathogenicity of their host organism, a phenomenon known as hypovirulence, which has been described for several mitoviruses infecting fungi^21,22. These examples illustrate how parasite-associated viruses can influence both parasite biology and host disease manifestations. The evidence suggesting that malaria parasites themselves harbor viruses, therefore, introduces an additional level of biological complexity into host-parasite interactions that remains largely unexplored.

MaRNAV-1 belongs to the family Narnaviridae, one of the simplest groups of RNA viruses. Narnaviruses are non-enveloped viruses with compact single-stranded RNA genomes composed of one or two segments. Most known narnaviruses infect fungal hosts and replicate in the cytoplasm without forming classical virions^23,24. Their transmission is thought to occur primarily through vertical inheritance during host cell division, allowing long-term persistence within host lineages, although horizontal transmission has also been reported^24,25.

The genome of MaRNAV-1 comprises two RNA segments (Supp. Fig 1). The first segment (S1) encodes an RNA-dependent RNA polymerase (RdRp) (GenBank accession No. MN860568.1) that shares conserved motifs with the polymerase of fungal narnaviruses⁹. The second segment (S2) encodes two overlapping open reading frames (ORFs, denoted as S2_ORF1 and S2_ORF2), possibly encoding proteins with no homology with annotated proteins⁹ (GenBank accession No. MN860569.1).

Despite the recent discovery of the cooccurrence of MaRNAV-1 and P. vivax and the description of its wide geographic distribution¹⁰, it remains unclear whether this virus truly infects the parasite (as opposed to coinfecting patients), and whether it influences the parasite biology or host responses.

In this study, we characterize MaRNAV-1 in P. vivax infections. Using imaging approaches, population surveys of clinical isolates from Cambodia and Ethiopia, and analyses of parasite and host data, we examined the localization, prevalence, and potential biological consequences of this virus.

Results

MaRNAV-1 infects P. vivax cells

To determine whether MaRNAV-1 infects P. vivax parasites rather than being a coinfection of the patients, we employed three complementary approaches. First, we performed single-molecule inexpensive fluorescence in situ hybridization (smiFISH) to visualize viral RNA transcripts from both genomic segments (S1 and S2) in P. vivax infected red blood cells from clinical samples that were RT-qPCR positive for MaRNAV-1. Confocal microscopy of blood-stage parasites revealed distinct fluorescent signals for both S1 and S2 that co-localized with the parasite 18S rRNA, demonstrating intracellular localization of the virus within P. vivax (Fig. 1A). No viral RNA was detected in adjacent uninfected red blood cells.

Fig. 1. Localization of MaRNAV-1 in P. vivax parasites revealed by smiFISH and immunofluorescence assays.

A. Detection of MaRNAV-1 RNA by smiFISH in blood-stage P. vivax parasites. From left to right: bright field; DNA stained with Hoechst (blue); P. vivax 18S rRNA labeled with FLAP-Y-Alexa Fluor 488 (green); MaRNAV-1 S2 RNA labeled with FLAP-Y-Cy3 (yellow); MaRNAV-1 S1 RNA labeled with FLAP-Y-Cy5 (red); merged image. Viral RNA signals co-localize with parasite 18S rRNA within infected erythrocytes. No viral signal was detected in adjacent uninfected red blood cell (bottom panel). Scale bar, 5 μm. B. Immunofluorescence localization of the MaRNAV-1 RNA-dependent RNA polymerase (RdRp) across P. vivax life-cycle stages. From top to bottom: blood stage parasites, sporozoite, and liver stage parasites. DNA was stained with Hoechst (blue). In the blood-stages and sporozoites, RdRp was detected using rabbit anti-RdRp antibodies followed by goat anti-rabbit Alexa Fluor 488 secondary antibodies (green). In liver-stage parasites, the parasitophorous vacuole membrane was labeled using mouse anti-UIS4 monoclonal antibodies detected with goat anti-mouse Alexa Fluor 488 secondary antibodies (green), while RdRp protein was detected with goat anti-rabbit Alexa Fluor Plus 594 secondary antibodies (red). Scale bars: 5 μm (blood stage and sporozoite), 30 μm (schizont), and 5 μm (hypnozoite).

Second, we used immunofluorescence assays (IFA) to localize the viral RdRp. RdRp signal was detected in all examined stages of P. vivax, including blood stages, sporozoites, liver schizonts, and dormant hypnozoites (Fig. 1B). For both blood and liver stages, a fluorescent signal was detected only in P. vivax-infected cells and not in surrounding uninfected ones (Fig.1B). Negative controls made of P. falciparum blood stages or using pre-immune IgGs showed no detectable signal Supp. Fig. 2). Together, these complementary imaging approaches provide direct evidence that MaRNAV-1 infects P. vivax cells with expression of viral proteins at different stages of the parasite life cycle.

Third, in 20 patients with P. vivax infections positive for MaRNAV-1, we monitored the presence of viral and parasitic RNA by RT-qPCR over the course of antimalarial treatment with artesunate. Over 4–5 days, both viral RNA and parasitic RNA were cleared with similar kinetics (Supp. Table 1). Importantly, the concomitant disappearance of viral RNA following parasite clearance further indicates that the virus does not persist independently in peripheral blood outside the parasite.

MaRNAV-1 prevalence in P. vivax clinical isolates is associated with symptom presentation and geographic origin.

Having confirmed that MaRNAV-1 infects P. vivax cells, we next evaluated its prevalence among P. vivax clinical isolates. RT-qPCR screening of RNA extracted from blood samples of 220 symptomatic P. vivax-infected patients seeking treatment in Cambodia detected MaRNAV-1 in 56% of samples (123/220) (Fig. 2A). In contrast, among asymptomatic P. vivax-infected individuals identified through active case detection in malaria-endemic communities in Cambodia, MaRNAV-1 detection rate was significantly lower (31%, 49/156, p < 0.0001, Fig. 2A). There was no difference in P. vivax parasitemia (determined by microscopy examination of thick smears) between MaRNAV-1 positive and negative samples among symptomatic infections (median, IQR: 5685 parasites/μL, 3645–11305 and 5755 parasites/μL, 3232–12180, respectively; p = 0.8; Supp. Fig. 3A), while parasitemia (determined by RT-qPCR) was lower in MaRNAV-1 negative asymptomatic infections compared to MaRNAV-1 positive ones (2.2 arbitrary units, AU, 0.77–4.62 and 3.97 AU, 1.53–7.16, respectively; p=0.02, Supp Fig. 3B). This suggests that some asymptomatic infections could have been misclassified as MaRNAV-1 negative due to low parasite density. To account for differences in parasitemia, we expressed MaRNAV-1 load as the ratio of viral S2 RNA to P. vivax 18S rRNA. Using this normalized measure, MaRNAV-1 load was significantly lower in asymptomatic infections compared to symptomatic ones (Mann-Whitney, p < 0.0001, Fig. 2B).

Fig 2. Prevalence and virus load of MaRNAV-1 in Plasmodium vivax infections from Cambodia and Ethiopia.

A. Proportion of MaRNAV-1 positive and negative P. vivax infections in asymptomatic and symptomatic individuals from Cambodia and Ethiopia. The prevalence of MaRNAV-1 among P. vivax infections was higher in Ethiopia than in Cambodia, in both asymptomatic and symptomatic groups (Fisher’s exact test, p <0.0001). B. Comparison of MaRNAV-1 load, expressed as the ratio of viral S2 RNA to P. vivax 18S rRNA, between asymptomatic and symptomatic infections in Cambodia and Ethiopia. Virus load was significantly higher in symptomatic infections in both countries (Mann-Whitney, p < 0.0001 and p = 0.004, respectively). **: p <0.01, ***: p <0.001, ***: p <0.0001, and ns: non-significant.

To confirm and extend our findings in a distinct malaria-endemic setting, we evaluated the prevalence of MaRNAV-1 among P. vivax infections in Ethiopia. Among symptomatic patients, MaRNAV-1 was detected in 74% of samples (75/102), a prevalence significantly higher than that observed in symptomatic infections from Cambodia (Fisher’s exact test, p = 0.003, Fig.2A). Similarly, prevalence among asymptomatic infections was higher in Ethiopia (61%, 46/75) than in Cambodia (Fisher’s exact test, p < 0.0001, Fig.2A). Although the same trend toward higher prevalence in symptomatic compared with asymptomatic infections was observed in Ethiopia, this difference did not reach statistical significance (Fisher’s exact test, p=0.1, Fig.2A). As in Cambodian samples, MaRNAV-1 load was significantly lower in asymptomatic infections compared to symptomatic ones (Mann-Whitney, p = 0.004, Fig. 2B).

MaRNAV-1 presence and load are positively correlated with gametocyte abundance and enhanced transmission to mosquito vectors

To understand the impact of MaRNAV-1 infection on parasite biology, we analyzed bulk RNA-seq data generated from whole-blood samples collected from 126 Cambodian patients seeking treatment for symptomatic P. vivax malaria and enrolled in a clinical trial reported elsewhere²⁶. Among these, 70% (88/126) were positive for MaRNAV-1, and virus transcript abundance spanned a wide range of virus loads (from 10⁻⁷ to 10⁻³, expressed as the ratio of MaRNAV-1 reads to reads mapping to P. vivax) (Supp. Data 1). The S2 segment appeared to be consistently expressed at a higher level than the S1 segment (Mann-Whitney, p<0.0001, Fig. 3A) and the MaRNAV-1 transcripts were overall highly expressed and, particularly the S2 segments, often as abundant as the most highly expressed P. vivax genes (Fig.3A & Supp. Data 2).

Fig 3. MaRNAV-1 load is associated with gametocyte abundance and increased mosquito transmission potential.

A. Expression levels of MaRNAV-1 segments S1 and S2 (TPM, transcripts per million) relative to the 25th and 75th percentiles of P. vivax gene expression within each sample. Each dot represents an individual patient sample (n = 126). B. Distribution of blood-stage parasite composition in Cambodian P. vivax infections positive for MaRNAV-1, inferred by RNA-seq deconvolution. Each column represents one sample, ordered by increasing MaRNAV-1 load (low to high). Colors indicate parasite developmental stages. C. Positive correlation between gametocytemia and MaRNAV-1 load, both measured using RNA-seq data (Spearman correlation, R = 0.2, p = 0.023, n = 126). D. Relationship between oocyst prevalence and gametocytemia in mosquitoes following membrane feeding, stratified by virus infection status. Lines represent fitted probabilities with 95% confidence intervals (CI). E. Relationship between virus load and oocyst prevalence. The fitted line shows the predicted probability of infection with 95% CI. F. Oocyst intensity (number of oocysts per infected mosquito) according to virus status. Boxplots show the median and interquartile range, with individual data points overlaid. G. Relationship between oocyst intensity and virus load, stratified by gametocytemia level (tertiles: low: 0–1581 parasites/μL, medium: 1581–5133 parasites/μL, high: 5133–22284 parasites/μL). The fitted line represents predictions with 95% CI. Across panels D to G, points represent individual mosquitoes. Continuous predictors are displayed on a log₁₀ scale.

Since P. vivax infections are often asynchronous and contain various proportions of different developmental stages, we estimated the proportion of each stage (rings, trophozoites, schizonts, and male and female gametocytes) in every sample using gene expression deconvolution²⁷. MaRNAV-1 was detected in samples with variable proportions of the different developmental stages (Fig.3B). Interestingly, the virus load was positively correlated with the proportion of both male and female gametocytes (Fig.3B, Supp. Fig. 4). While MaRNAV-1 load was not correlated with parasitemia (Spearman correlation, R = 0.033, p = 0.7107, Supp. Data 1), it was with absolute gametocytemia (Spearman correlation, R = 0.2, p = 0.023) (Fig. 3C and Supp. Fig. 4). This correlation was even stronger when analyses were restricted to MaRNAV-1-positive samples (Supp. Fig. 4). Conversely, trophozoite proportions were negatively correlated to MaRNAV-1 load (Supp. Fig. 4).

Given the association of MaRNAV-1 with gametocytes, we examined whether MaRNAV-1 was associated with transmission to mosquito vectors. We analyzed the results of membrane-feeding assays using Anopheles dirus mosquitoes performed on 33 blood samples that were used for RNA-seq (13 MaRNAV-1 negative and 20 MaRNAV-1 positive). Virus-positive samples resulted in 12 infectious feeds (i.e., at least one successfully infected mosquito) out of 20 (60%), compared to 5 out of 13 (38%) for virus-negative samples. Both gametocytemia (${\chi }_1^2$= 10.5, p = 0.001) and viral presence (${\chi }_1^2$= 4.9, p = 0.027) increased the likelihood of mosquito infection (Fig. 3D). Higher viral load was also associated with increased oocyst prevalence in mosquitoes (${\chi }_1^2$= 5.9, p = 0.01, Fig. 3E), regardless of gametocyte density (virus load by gametocytemia interaction: ${\chi }_1^2$= 0.01, p = 0.9). In addition, mosquitoes infected with virus-positive P. vivax exhibited a ~ 7-fold increase in mean oocyst count compared to virus-negative isolates (mean ± se oocyst number in mosquitoes with virus-positive P. vivax: 60.2 ± 3.54 vs. 8.62 ± 1.26, ${\chi }_1^2$= 11.8, p < 0.001, Fig. 3F). Consistently, higher virus load was associated with increased oocyst intensity (${\chi }_1^2$= 7.8, p = 0.005, Fig. 3G), while there was no significant association with gametocytemia (${\chi }_1^2$= 2.8, p = 0.09). There was a significant negative interaction between gametocytemia and virus load on oocyst intensity (<${\chi }_1^2$= 5.4, p = 0.02, Fig. 3G). Overall, these results indicate that MaRNAV-1 load is associated with higher P. vivax transmission, in particular at low gametocytemia.

To validate these findings, we first determined if a similar correlation between gametocytemia and virus load was observed in a larger dataset of samples. We quantified Pvs25 and Pvs47 transcripts, markers of female and male gametocytes, respectively, by RT-qPCR (relative to human RPS18) in 152 samples with MaRNAV-1 load detected and quantified by RT-qPCR (expressed as S2 abundance relative to P. vivax 18S rRNA). Both Pvs25 and Pvs47 relative abundance were correlated with MaRNAV-1 loads (R = 0.32, p< 0.0001 and R=0.32, p< 0.0001, respectively, Supp. Fig. 5).

Then, we examined additional mosquito-feeding experiments using blood samples from 87 patients for which MaRNAV-1 load was calculated by RT-qPCR (including the 33 patient’s samples already analyzed by RNA-seq). Of those, 48 samples were negative for MaRNAV-1 while 39 were positive. The results obtained were confirmed, with (i) higher infectiousness (virus-positive samples resulted in 20 infectious feeds out of 39, compared to 13 out of 48, for virus-negative samples), (ii) higher oocyst prevalence, and (iii) higher oocyst intensity for virus-positive samples, whether gametocytemia covariate was calculated based on Pvs25 (Supp. Fig. 6), or Pvs47 transcripts (Supp. Fig. 7).

Overall, these analyses performed on two separate cohorts using different analytical tools for MaRNAV-1 and gametocytemia quantification (RNA-seq or RT-qPCR) show consistently a positive association between MaRNAV-1 and oocyst prevalence and intensity, independently of gametocytemia.

MaRNAV-1-P. vivax association presents hallmarks of a stable intracellular interaction

To investigate whether MaRNAV-1 forms a stable association with P. vivax, we analyzed genomic and transcriptomic data from paired parasite and viral sequences. First, we sequenced the whole genome of P. vivax isolates that had been analyzed by RNA-seq and for which we had determined the MaRNAV-1 infection status (53 MaRNAV-1 positive and 33 MaRNAV-1 negative samples). To evaluate whether MaRNAV-1 infection was associated with specific parasite lineages, we reconstructed a neighbor-joining tree using whole-genome single nucleotide polymorphisms data from the P. vivax isolates. MaRNAV-1-positive samples were distributed across the parasite population and did not cluster within particular lineages, indicating that virus infection occurs across genetically diverse P. vivax backgrounds in this population, rather than being restricted to specific parasite genotypes (Fig. 4A).

Fig. 4. Genomic analyses indicate a stable association between MaRNAV-1 and genetically diverse P. vivax parasites.

A. Phylogenetic relationships among monoclonal P. vivax isolates (n=86). The neighbor-joining tree was reconstructed from pairwise nucleotide differences based on whole-genome SNP data. Red circles indicate MaRNAV-1-positive samples, while blue circles indicate MaRNAV-1-negative samples. The distribution of virus-positive samples across the tree indicates that MaRNAV-1 infection occurs across genetically diverse parasite backgrounds rather than being restricted to specific parasite lineages. B. Tanglegram comparing phylogenetic trees inferred for MaRNAV-1 genomic segments S1 (left) and S2 (right). Maximum likelihood (ML) trees were inferred independently for each segment. Lines connect corresponding viral sequences from the same sample across the two trees. Tip labels represent individual sample IDs. Node support values are displayed as SH-aLRT/UFboot, with supported nodes defined as SH-aLRT ≥ 80% and UFboot ≥ 70% (red dots). Branch lengths represent nucleotide substitutions per site, and both trees were midpoint rooted prior to visualization. A single discordant connection (in red) suggests a putative reassortment event between S1 and S2 segments. C. Differential gene expression analysis of P. vivax parasites comparing samples with high versus low MaRNAV-1 loads (top and bottom quartiles, n=22 each group). The volcano plot shows log₂ fold change versus −log₁₀ adjusted p-values. Blue dots indicate down-regulated genes in high virus load samples, while red dots indicate up-regulated genes.

Next, we examined the sequence diversity of the MaRNAV-1 S1 and S2 segments. Analysis of MaRNAV-1 nucleotide sequences, obtained after de novo transcript assembly from RNA-seq data, revealed high sequence conservation across both viral genome segments. The S1 segment exhibited substantial nucleotide variability (91.3%−100% identity) but was markedly more conserved at the protein level (96.5%−100% amino acid identity) while segment S2 showed higher nucleotide conservation (94.1%−100% identity) and high amino acid conservation (ORF1: 96.6%−100%; ORF2: 96.3%−100%) (Supp. Data 3 to 7).

We then assessed within-host virus diversity among the 63 samples for which we were able to reconstruct entire viral genomes. Among these, 27 isolates were polyclonal, containing multiple P. vivax genotypes. Interestingly, 12 out of these 27 polyclonal P. vivax infections showed evidence of infection with multiple viral strains while all monoclonal P. vivax infections (36/36, Fisher exact test, p<0.0001) were associated with a single MaRNAV-1 strain, suggesting that within-patient viral diversity reflects parasite infection complexity.

Given the presence of multiple viruses in the same infection, we then tested whether the two segments were systematically co-inherited or whether reassortment could occur between MaRNAV-1 segments S1 and S2. Using the subset of samples containing a single viral strain, we estimated a low reassortment rate of 0.004–0.005 events per year. Consistent with this, comparison of rooted phylogenetic trees for S1 and S2 using a tanglegram approach identified only a single putative reassortment event (highlighted in red, Fig. 4B). These results suggest limited segment exchange, although this observation will require confirmation using larger and more geographically and temporally diverse genomic datasets.

Finally, we tested whether MaRNAV-1 infection affected P. vivax gene expression. After accounting for differences in parasite developmental stages, we compared the transcriptional profiles of the 38 samples with no detectable MaRNAV-1 reads to those of the 38 samples with the highest virus loads. No genes were differentially expressed between these groups (Supp. Data 8). When restricting the analysis to samples with the most extreme virus loads (below the 25^th percentile versus above the 75th percentile, n=22 each group), only 8 genes were differentially expressed, indicating a very limited impact of MaRNAV-1 on the parasite transcriptome (Fig.4C, Supp. Data 9). Among these, two genes of unknown function (PVP01_0519800; PVP01_0701700) were downregulated, while six were upregulated, including a PHIST-domain-containing exported protein (PVP01_0522700), a putative ELM2 domain-containing protein potentially involved in transcriptional regulation or chromatin remodeling (PVP01_0206500), a putative acyl-CoA synthetase (PVP01_0409900), a putative nucleoside diphosphate hydrolase (PVP01_1205700), and two additional genes of unknown function (PVP01_0103900; PVP01_0414100). Altogether, these results indicate that MaRNAV-1 has little, if any, impact on P. vivax gene expression profiles.

MaRNAV-1 elicits antibody responses in P. vivax infected individuals

Next, we evaluated the effect of MaRNAV-1 on the host immune response. We first assessed whether MaRNAV-1 elicits humoral immune responses in infected individuals. IgG responses against viral proteins (RdRP and S2_ORF1) were measured by ELISA in plasma samples from P. vivax-infected individuals in Cambodia and Ethiopia and compared with malaria-naïve controls. Detectable antibody responses against MaRNAV-1 RdRp were observed in 17% (94/552) and 29% (51/176) of P. vivax infected individuals from Cambodia and Ethiopia, respectively, and against S2_ORF1 in 28% (153/552) and 41% (73/176) of individuals from Cambodia and Ethiopia respectively (Fig. 5A and 5B). Seropositivity rates were higher in Ethiopian individuals compared to Cambodian for both RdRp (p=0.0008) and S2_ORF1 (p=0.0008) (Fig. 5A and 5B). These results are consistent with the higher proportion of MaRNAV-1 positive P. vivax-infected detected in Ethiopia compared to Cambodia (Fig. 2A). Overall, seropositivity rates were significantly higher for the S2_ORF1 protein (31%, 226/728) than for RdRp (20%, 145/728, χ² =23.7, p < 0.0001).

Fig.5. MaRNAV-1 elicits an antibody response in P. vivax-infected individuals.

A-B. Antibody responses measured by ELISA (optical density, OD) against MaRNAV-1 RdRp (A) and S2_ORF1 (B) antigens in naïve individuals and P. vivax-infected patients from Cambodia and Ethiopia. The red dashed line indicates the seropositivity threshold. The number of seropositive and seronegative individuals in each group is indicated. Differences between groups were assessed using chi-square tests. C-D. Antibody responses against MaRNAV-1 RdRp (C) and S2_ORF1 (D) stratified by MaRNAV-1 infection status (virus-negative vs virus-positive) in P. vivax-infected individuals. The red dashed line indicates the seropositivity threshold. The number of seropositive and seronegative individuals in each group is shown. Differences in seroprevalence between groups were assessed using Fisher’s exact tests. In all panels, violin plots show the distribution of OD values with individual data points overlaid.

We next analyzed samples for which both viral detection and ELISA data were available (n = 232), all from Cambodia. Seropositivity rates were significantly higher in individuals with MaRNAV-1 positive P. vivax infections (RdRp: 25%, 30/120, S2_ORF1: 41%, 49/120, Fig. 5C) compared with those with virus-negative infections (RdRp: 13%, 15/112, S2_ORF1: 22%, 25/112, p=0.0307 and p=0.0030, respectively, Fig. 5D). These results suggest that either some infections were misclassified as MaRNAV-1 negative and/or that some individuals were previously infected with a MaRNAV-1 positive infection.

Altogether these analyses show that MaRNAV-1 proteins, especially S2_ORF1, are immunogenic during P. vivax infection and elicit measurable antibody responses.

MaRNAV-1 is associated with increased host inflammatory signaling during P. vivax malaria

Having shown that MaRNAV-1 elicits an antibody response, we next investigated whether MaRNAV-1 was associated with clinical correlates in P. vivax-infected patients, while correcting for parasitemia as covariate. Clinical data were available for the 126 Cambodian patients enrolled in a clinical trial and for whom RNA-seq data were generated and presented above. The clinical correlates analyzed included: hemoglobin, white blood cells and platelet counts, body temperature, and C-reactive protein levels.

When all 126 patients’ data were analyzed, none of these clinical parameters were significantly associated with infection status whether using MaRNAV-1 as a binary variable (presence/absence) (Supp. Table 2) or using virus load as a continuous variable (Supp. Table 3). Investigating MaRNAV-1 positive individuals only, body temperature was significantly associated with virus load (FDR-adjusted p = 0.01), parasitemia (FDR-adjusted p = 0.04) and their interaction (FDR-adjusted p = 0.04) (Supp. Table 4). These findings indicate that, at low parasitemia, MaRNAV-1 load is associated with increased body temperature, whereas the effect of parasite burden on body temperature dominates at higher parasitemia (Fig. 6A, Supp. Fig. 8).

Fig 6. Association between MaRNAV-1 infection and host responses during P. vivax malaria.

A. Predicted body temperature as a function of MaRNAV-1 virus load and parasitemia (MaRNAV-1-positive samples only). Generalized additive models were used to model nonlinear relationships. Curves represent predicted temperature across log-transformed virus load, stratified by parasitemia categories (low: 107–4190 parasites/μL, medium: 4227–10119 parasites/μL, high: 10147–39353 parasites/μL). Shaded areas represent 95% confidence intervals. B. Predicted immune cell composition according to MaRNAV-1 infection status. Stacked area plots show the model-predicted relative proportions of immune cell populations inferred from human RNA-seq data using deconvolution and Dirichlet regression across increasing parasitemia levels. Separate panels show predictions for virus-negative and virus-positive infections. C. Association between MaRNAV-1 virus load and plasma IFN-γ levels. The plot shows log-transformed IFN-γ concentrations as a function of log-transformed virus load. The fitted curve represents the model’s nonlinear relationship (gam), with shaded areas indicating 95% confidence intervals.

To further explore host immune responses, we examined association between MaRNAV-1 and the proportion of different immune cell subsets estimated by gene expression deconvolution. When MaRNAV-1 infection was analyzed as a binary variable, overall immune cell composition was associated with viral infection status (p =0.008), parasitemia (p < 0.001), and their interaction (p = 0.001) (Fig. 6B, Supp. Table 5 & Supp. Data 10). Immune cells of MaRNAV-1-positive patients were relatively enriched in several innate and inflammatory cell populations, (neutrophils, monocytes, and resting NK cells), and cells of the adaptive immune system (naive B cells and naive CD4 T cells). When virus load was analyzed as a continuous variable among virus positive samples, no significant association with immune cell composition was detected, whereas parasitemia effect remained significant (Supp. Fig. 9A & Supp. Table 5). These results indicate that the presence of MaRNAV-1, rather than its load, is associated with differences in immune cell composition. These results contrast with the gene expression analysis of immune cells that did not show any differentially expressed genes following RNA-seq deconvolution when comparing samples MaRNAV-1-positive and negative (Supp. Data 11). Even when restricting our analysis to MaRNAV-1-positive samples and comparing those with the lowest and highest virus load (below 25^th vs. above 75^th percentile) we identified only 16 genes differentially expressed (14 down-regulated and 2 up-regulated in the high MaRNAV-1 load group) suggesting a modest detectable impact of MaRNAV-1 on immune cell gene expression (Supp. Fig. 9B & Supp. Data 12).

Finally, to investigate host inflammatory responses, we analyzed plasma cytokines using a canonical inflammatory cytokine panel. After excluding analytes below the limit of detection, 26 cytokines were retained for analysis. Comparing cytokine concentrations in samples with or without detectable MaRNAV-1 showed no significant difference for any of the cytokines (Supp. Table 6). Next, we evaluated the correlation between MaRNAV-1 load and cytokine concentrations in MaRNAV-1 positive individuals. IFN-γ was associated with MaRNAV-1 load (FDR-adjusted p = 0.047), but not with parasitemia (FDR-adjusted p = 0.103) (Fig. 6C). Additionally, IP-10, IL-10, IL-1RA, IL-6, and VEGF were significantly associated with both parasitemia and MaRNAV-1 load independently (Supp. Table 7, Supp. Fig. 10). An interaction between virus load and parasitemia was detected (p=0.011) only for IL-6 suggesting that virus load predominantly drives IL-6 secretion at low parasitemia, whereas parasite burden dominates at higher parasitemia (Supp. Table 7, Supp. Fig. 10).

Together, these results indicate that MaRNAV-1 infection is associated with alterations in host inflammatory signaling and immune cell composition during P. vivax malaria.

Discussion and Conclusion

This study provides the first comprehensive characterization of MaRNAV-1, a recently described narnavirus associated with P. vivax infections⁹. Using complementary approaches, we demonstrate that MaRNAV-1 is localized within P. vivax parasites across multiple life stages. These observations provide direct evidence that MaRNAV-is an intracellular parasite-associated virus rather than a coincidental co-infection. This conclusion is further supported by the observation that viral RNA is cleared from peripheral blood following antimalarial treatment with kinetics closely paralleling parasite clearance, indicating that the virus does not persist independently in the host once parasites are eliminated.

Our epidemiological analyses revealed that MaRNAV-1 is highly prevalent in P. vivax infections, with substantial geographic variation between study sites. The virus was significantly more frequently detected in Ethiopian samples than in Cambodian samples, suggesting that viral prevalence may vary across parasite populations or transmission settings. Such differences could reflect variation in parasite population structure, transmission intensity, or historical virus-parasite co-evolution. Phylogenetic analyses further showed that virus-positive infections were distributed across genetically diverse P. vivax lineages, indicating that MaRNAV-1 infection is not restricted to a particular parasite genotype.

Genomic analyses revealed that MaRNAV-1 displays substantial nucleotide polymorphism but high protein conservation, consistent with an important role of the viral proteins, including those encoded on segment S2 that have no homology to annotated proteins. The high conservation of viral proteins and the limited evidence of reassortment observed in our genomic analyses suggest that MaRNAV-1 may maintain a stable, vertically transmitted, intracellular relationship with P. vivax. Similar evolutionary patterns have been reported for other endosymbiotic viruses infecting microbial eukaryotes through vertical transmission such as Totiviruses^18,28.

An intriguing feature of MaRNAV-1 we observed here is that S2 transcripts were more abundant than the S1-derived polymerase reads and S2_ORF1 elicited a higher seroconversion rate in infected individuals compared to RdRp. These observations suggest that the S2 proteins (at least ORF1) may play important roles during viral replication or host-parasite interactions. Although narnaviruses are generally considered capsid-less viruses, the high expression and immunogenicity of S2_ORF1 protein raise the possibility that they may contribute to viral particle stability, parasite cell interaction, or immune recognition. Further functional studies will be required to determine the precise role of these proteins in the MaRNAV-1 life cycle.

A major finding of this study is the association between viral presence and load and parasite transmission potential. We observed that MaRNAV-1 was correlated with increased gametocyte proportion. Gametocyte production is a critical determinant of malaria transmission, and these results suggest that MaRNAV-1 may influence the balance between asexual replication and transmission-stage development in P. vivax. One possible explanation is that viral infection acts as a physiological stress signal within the parasite, promoting gametocytogenesis.

MaRNAV-1 was associated with markedly increased P. vivax infectivity to mosquitoes, whether oocyst prevalence or intensity, independently of gametocytemia. This may result from viral modulation of mosquito immunity, either by suppressing antiparasitic pathways or redirecting resources toward antiviral defenses, thereby reducing parasite clearance. The virus could also alter midgut physiology, through changes in oxidative stress, microbiota composition, or blood meal properties, or partially mask parasite recognition signals by the mosquito immune system. Finally, virus-positive isolates might be intrinsically more infectious. These findings suggest a potential convergence of selection forces between viruses and parasites: by enhancing mosquito infection, the virus may simultaneously favor both its own transmission and that of P. vivax. Although the molecular mechanisms underlying this association remain unclear, these observations raise the possibility that parasite-associated viruses may contribute to variability in malaria transmission dynamics.

Despite these potential effects on parasite transmission, our transcriptomic analyses revealed only minimal effects of MaRNAV-1 on parasite gene expression. Differential expression analyses identified very few P. vivax genes associated with MaRNAV-1, suggesting that MaRNAV-1 infection does not strongly perturb the parasite transcriptional program. This limited transcriptional response may reflect a long-term evolutionary association between the virus and its host, in which viral replication occurs with minimal disruption of parasite cellular processes. Note that here, bulk RNA-seq data were analyzed and may not capture the subtle effects of the virus on P. vivax gene expression, especially given the different developmental stages present in each infection. In addition, since different stages have different transcriptional activity, the profiles obtained from RNA-seq data are biased towards genes expressed in trophozoites and against those expressed in rings, which might be particularly relevant for understanding sexual commitment. Further studies using single-cell RNA-seq would allow circumventing these limitations.

Beyond its potential effects on parasite transmission, our data indicates that the human host recognizes and responds to MaRNAV-1 during P. vivax infection. We detected antibody responses against viral proteins in infected individuals from both Cambodia and Ethiopia, demonstrating that MaRNAV-1 is immunogenic in humans. Seropositivity rates broadly mirrored the geographic prevalence of the virus detected by RT-qPCR, suggesting that host exposure to viral antigens reflects the underlying distribution of virus-positive parasite infections. This is supported by the higher seropositivity rates observed for individuals infected with MaRNAV-1 positive P. vivax compared to individuals infected with virus-negative P. vivax.

Perhaps most importantly, our analyses also indicate that MaRNAV-1 is associated with increased host inflammatory responses during malaria infection, independently of parasitemia. Virus load was associated with increased body temperature and with elevated levels of several inflammatory mediators, including IFN-γ, IL-6, IP-10, IL-10, IL-1RA, and VEGF. Some of these cytokines are key regulators of interferon-driven inflammatory pathways and are known contributors to fever and malaria pathogenesis²⁹. Interestingly, the relationship between MaRNAV-1 load and fever and virus load and the pyrogenic cytokine IL-6 was most pronounced at low parasitemia levels, whereas parasite density became the dominant driver of body temperature and IL-6 production at higher parasitemia. This pattern suggests that MaRNAV-1 may contribute to inflammatory signaling particularly during early or low-density infections. It has been described for many years that the pyrogenic threshold is lower in P. vivax than in P. falciparum and that, at similar parasite densities, P. vivax induces a stronger inflammatory response in infected hosts compared to P. falciparum⁸. MaRNAV-1 may contribute to these differences in malaria physiopathology, potentially acting as an additional pathogen-associated molecular signal alongside parasite-derived molecules such as glycosylphosphatidylinositol anchors and hemozoin^29,30.

In addition, we observed that MaRNAV-1 loads were lower in asymptomatic P. vivax infections compared to symptomatic cases. This pattern is consistent with our observation that higher virus loads are associated with increased inflammatory responses and elevated body temperature. Together, these findings suggest that MaRNAV-1 may contribute to the inflammation associated with symptomatic malaria. Although causality cannot be established from observational data, these results raise the possibility that parasite-associated viruses may modulate the balance between asymptomatic carriage and clinical disease by amplifying host inflammatory responses.

Integration of cytokine profiling with transcriptomic deconvolution further suggests that MaRNAV-1 may influence the composition of circulating immune cells during infection. Virus-positive infections were characterized by relative enrichment of innate immune populations, including neutrophils, monocytes, and natural killer cells, indicating a more activated immune state in MaRNAV-1 infected cases.

One possible mechanistic explanation is that viral RNAs (or proteins) are released during schizont rupture or phagocytosis of infected erythrocytes and may be sensed by host pattern recognition receptors. Single-stranded RNA viruses are known ligands for endosomal receptors such as TLR7, which trigger type I interferon signaling and an anti-viral response. The anti-viral response, characterized by the production of IFN-γ and IP10, might skew the Th balance to a more Th1-driven response, thereby further increasing systemic inflammation. The concurrent induction of anti-inflammatory mediators such as IL-10 and IL-1RA suggests that the host immune responses attempt to balance virus-induced inflammation with anti-inflammatory feedback to limit tissue damage. The association between virus load and VEGF production further raises the possibility that MaRNAV-1 may modulate vascular or inflammatory pathways implicated in malaria pathogenesis^31,32.

Despite these findings, several limitations should be considered. First, most of our analyses were based on bulk transcriptomic and immunological measurements from clinical samples, which limit our ability to resolve parasite or host-cell specific responses to virus infection. Future studies using single-cell transcriptomics of clinical samples could help determine whether MaRNAV-1 directly alters parasite gene expression or host immune signaling at the cellular level. Second, although our analyses reveal associations between virus load, parasite transmission potential, and host inflammatory responses, these observational data do not establish causality. Experimental validation is particularly challenging in P. vivax, which cannot be maintained in long-term in vitro culture or genetically manipulated. However, closely related Plasmodium species that can be experimentally maintained may provide tractable systems to investigate parasite-virus interactions. In particular, the recent identification of a virus infecting P. knowlesi suggests that comparative experimental studies in related Plasmodium species could provide valuable insights into the biological consequences of parasite-associated virus infections. Finally, the geographic variation in MaRNAV-1 prevalence observed between Cambodia and Ethiopia suggests that MaRNAV-1 distribution may vary across parasite populations, highlighting the need for broader genomic surveillance to determine the global diversity and evolutionary history of this virus.

Together, our findings support a conceptual model in which MaRNAV-1 acts as a hitherto hidden component of P. vivax infection, that may simultaneously affect parasite transmission and host inflammatory responses. By enhancing gametocyte abundance and mosquito infection prevalence and intensity, the virus may increase parasite transmission potential, while virus-associated inflammatory signaling may contribute to fever and excessive immune activation in infected individuals.

Our study reveals a previously unrecognized virus-parasite-host interaction in malaria. The presence of an RNA virus infecting P. vivax adds an additional layer of biological complexity to malaria and raises new questions about the evolutionary and clinical consequences of parasite-associated viruses. These findings open a new avenue in malaria research, and further investigation will be required to determine how MaRNAV-1 influences parasite fitness, transmission dynamics, and disease severity in endemic populations.

Methods

Symptomatic patient enrolment, sample collection, and processing

P. vivax-infected blood samples used in this study were collected from treatment-seeking patients presenting with uncomplicated malaria and diagnosed with P. vivax infection by rapid diagnostic test or microscopy at local health facilities. P. vivax mono-infection was later confirmed for all samples using a species-specific cytB PCR³³. In Cambodia, patients were enrolled between 2018 and 2025 in Mondulkiri and Kampong Speu provinces, while in Ethiopia, patients were enrolled in the Arba Minch district between 2023 and 2025.

Following informed consent, venous blood samples were collected prior to treatment by healthcare providers in accordance with national treatment guidelines in each country (chloroquine and primaquine in Ethiopia, artesunate-mefloquine and primaquine in Cambodia). An aliquot of whole blood was immediately stored at −80 °C in TRIzol reagent until RNA extraction. Blood was also used to prepare Giemsa-stained thick and thin smears for microscopy and was subsequently centrifuged to separate plasma from cell pellets, which were stored at −80°C until further use. For a subset of Cambodian patients, blood samples were additionally used for imaging analyses and experimental mosquito infections (see below).

In addition, this study included samples and associated data from patients enrolled in a clinical trial in Cambodia evaluating the therapeutic efficacy of primaquine, as described elsewhere²⁶. At enrolment, complete blood counts, C-reactive protein levels, and body temperature were recorded. Venous blood collected prior to artesunate treatment was used for whole-genome sequencing, RNA-seq, and, in a subset of patients, mosquito infections. Clearance of both MaRNAV-1 and parasites following artesunate treatment was assessed in 20 of these patients by RT-qPCR performed on RNA extracted from capillary blood samples collected every 24 hours, from Day 0 (prior to treatment initiation) to Day 7.

Asymptomatic participant enrolment, sample collection, and processing

P. vivax-infected blood samples from asymptomatic carriers used in this study were collected between 2023 and 2025 from participants enrolled in two longitudinal cohorts one: conducted in Mondolkiri, Cambodia and the other in Arba Minch, Ethiopia. All participants provided informed consent prior to enrolment. Blood samples were processed for TRIzol preservation for subsequent RNA extraction, as well as for DNA extraction and plasma separation. Genomic DNA was extracted using DNeasy kits (QIAGEN) and used for P. vivax detection by species-specific cytB PCR³³.

Experimental mosquito infections and liver stage assays.

Membrane-feeding assays (MFAs) were conducted using laboratory-reared Anopheles dirus, a primary vector in Southeast Asia. Blood samples collected from symptomatic treatment-seeking patients in Cambodia were used directly to feed 5 to 7 day-old female mosquitoes for 1h via an artificial membrane attached to a water-jacketed mini-feeders maintained at 37 °C³⁴. Unfed and partially fed females were discarded while fed A. dirus mosquitoes were maintained at 26°C and 80% humidity and fed a 10% sucrose solution containing 0.05% PABA in distilled water. On the sixth day post-blood meal (dpbm), 50 mosquitoes were dissected, and the oocyst prevalence (the proportion of blood-fed mosquitoes with at least one oocyst in the midgut at 6 dpbm) and oocyst intensity (oocyst count per infected mosquito at 6dpbm) were recorded. On day 15 to 18, individual mosquitoes were dissected to isolate sporozoites. These sporozoites were either fixed with 4% paraformaldehyde (PFA) for imaging or used immediately for liver-stage assays. P. vivax liver stage assays were performed as previously described^35,36. Primary human hepatocytes (purchased from BioIVT) were seeded in 384 wells plate (18,000 cells per well) and 15,000 sporozoites were added 48h later in each well and allowed to infect PHH. Culture medium was changed 24h after infection and then every second day until day 15 post-infection when cells were fixed with 4% PFA.

SmiFISH detection of S1 and S2 MaRNAV-1 RNA

The primary probes and FLAPs (secondary probes with fluorescent tags) were synthesized and purchased from Integrated DNA Technologies (IDT). The primary probes were designed using the Stellaris FISH Probe Designer (biosearchtech.com/stellarisdesigner). Twenty-one primary probes were synthesized for MaRNAV-1 S1 and S2 and 27 were synthesized for P. vivax 18S rRNA. All probe sequences are available in Supp Data. 13. The secondary probes were conjugated to Cy3, Cy5, and Alexa Fluor 488 through 5’ and 3’ amino modifications. The FLAP-Y sequences were used as secondary probes as previously described³⁷. 18S rRNA was labeled with FLAP-Y-Alexa Fluor 488, S1 with FLAP-Y-Cy5, and S2 with FLAP-Y-Cy3. P. vivax-infected red blood cells were enriched using a KCl-Percoll density gradient³⁸, fixed at room temperature in 4% PFA and 0.0075% glutaraldehyde. The cells were washed twice with 1X PBS and hybridized according to Stellaris’ instructions for suspension cells. Image stacks were captured using a confocal microscope (Leica SP8) with a 63X/1.4 NA oil-immersion objective and a sCMOS PCO camera. Subsequent image analysis was performed using Image J software (1.8.0) was used for further image analysis.

Immunofluorescence detection of MaRNAV-1 RdRP

Recombinant RdRp and rabbit polyclonal antibodies were purchased from Genscript. Briefly, the full-length RdRp (NCBI sequence) was expressed in pET30a in E. coli, purified from inclusion bodies and used to immunize New Zealand rabbits. Antibodies were purified by antigen affinity purification, titrated by ELISA and shipped to Cambodia for immunofluorescence assays (IFA). Fixed P. vivax-infected red blood cells or sporozoites were placed on poly-L-lysine coated slides. Cells were permeabilized using 0.3% Triton X-100, then washed with PBS, and incubated overnight with polyclonal anti-RdRp (1:1000). After washing, samples were incubated with goat anti-rabbit Alexa Fluor 488 (1:1000, Thermo Fisher). Samples were washed and DNA was stained with Hoechst 33342 (1:1000 dilution). Slides were mounted with Vectashield prior to imaging. IFA on infected hepatocytes followed a similar procedure with some adaptations. Fixed cells were stained overnight with recombinant mouse-anti P. vivax Upregulated in Infectious Sporozoites 4 (rPvUIS4, 1:10,000) and rabbit anti- RdRp antibody (1:1000). After washing, samples were then stained goat anti-mouse Alexa Fluor^™ 488 and anti-rabbit Alexa Fluor^™ Plus 594 IgG (H+L) secondary antibodies (1:1000, Thermo Fisher). After washing, cell DNA was stained with Hoechst 33342. Stained red blood cells, sporozoites and hepatocytes were visualized using a Lionheart FX Automated Microscope (Biotek^®). Subsequent image analysis was performed using Image J software (1.8.0) was used for further image analysis.

RNA extraction and RT-qPCR detection and quantification of MaRNAV-1, gametocytemia and parasitemia

Total RNA was extracted from TRIzol preserved samples using chloroform. After centrifugation at 12,000 g for 15 minutes, the aqueous phase was collected and mixed with 100 % ethanol. RNA purification was performed at 4 °C using the RNeasy Mini Kit (QIAGEN), according to the manufacturer’s instructions. Residual genomic DNA was digested with DNase I (Promega). Reverse RNA transcription to complementary DNA (cDNA) was carried out using Promega GoScript Reverse Transcriptase Kit. All qPCR were conducted in a 20 μL volume with 0.25μM of primers, 1X EvaGreen SYBR mix, and 1 μL of cDNA. All reactions were performed under the following conditions: 95 °C for 15 minutes, followed by 45 cycles of 95 °C for 15 seconds, 60°C for 20 seconds, and 72 °C for 20 seconds, with a subsequent melting curve analysis. All primers used are listed in Supp. Table 8. MaRNAV-1 load was determined by relative quantification of S1 or S2 normalized to P. vivax 18S rRNA. As S2 transcripts were consistently more abundant than S1, S2 relative quantification was preferred for increased sensitivity. Gametocytemia were determined by relative quantification of either Pvs25 or Pvs47 normalized to human RPS18. Parasitemia for asymptomatic infections was determined by the relative quantification of P. vivax 18S rRNA normalized to human RPS18. For symptomatic cases, parasitemia was determined by microscopy counting of all parasites on Giemsa-stained thick blood films.

RNA extraction, library preparation, and RNA sequencing analysis

RNA was extracted using phenol-chloroform from TRIzol preserved samples. After rRNA depletion and polyA selection (NEB), RNA-seq libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit (NEB). We sequenced all libraries on an Illumina NovaSeq 6000 to generate ~30–414 million paired-end reads of 75 bp per sample. We first aligned all reads from each sample using Hisat2 (v2.1.0)³⁹ to a FASTA file containing the P. vivax P01 and human hg38 genomes, using default parameters and a shorter maximal intron length (--max-intronlen 5000)⁴⁰. We retained only unmapped reads and then de novo assembled them into contigs using the Trinity assembler⁴¹. Minimap was used to map the contigs to the reference sequences of MaRNAV-1 (S1: accession No. MN860568.1, S2: accession No. MN860569.1). Finally, we realigned all reads to a file containing all NCBI and de novo assembled virus sequences using Hisat2, removed PCR duplicates using custom scripts, and calculated the number of MaRNAV-1 reads. Separately, we calculated read counts per gene using gene annotations downloaded from PlasmoDB (P. vivax genes) and NCBI (human genes) and then used subread featureCounts (v1.6.4)⁴². Using these data, MaRNAV-1 load was calculated for these RNA-seq samples as the ratio between the total number of reads mapping to MaRNAV-1 and those mapping to P. vivax.

Read counts per gene were normalized to transcripts per million (TPM) for P. vivax genes and the virus, and to counts per million (cpm) for humans and P. vivax genes separately for differential gene analysis. To filter out lowly expressed genes, only genes expressed at least 10 cpm were retained for further analysis (9493 human and 4934 P. vivax genes, respectively). The edgeR package⁴³ was used for statistical assessment of differential gene expression analysis with a quasi-likelihood negative binomial generalized model. All results were corrected for multiple testing using FDR⁴⁴, and p-value < 0.1 is considered to be significant.

We also estimated the proportion at each developmental stage and the proportion of human immune cell types in individual blood samples from infected individuals. We performed gene expression deconvolution using CIBERSORTx⁴⁵ and a custom signature matrix derived from orthologous Plasmodium berghei genes was used for P. vivax stage deconvolution. We used the total proportion of male and female gametocytes obtained from this deconvolution multiplied by total parasitemia (obtained from thick smear microscopy) to quantify absolute gametocytemia in these RNA-seq samples. To deconvolute human genes expressed profiles, we used a validated leukocytes gene signature matrix which uses 547 genes to differentiate 22 immune subtype⁴⁶.

DNA extraction, whole genome sequencing, and complexity of infection

We extracted parasite DNA from leukocyte-depleted blood samples using the DNeasy blood and tissue kit (Qiagen) and prepared whole genome sequencing libraries using the NEBNext ^R Ultra^™ II FS DNA Library Prep Kit for Illumina NovaSeq 6000 to generate 25–50 million paired-end reads of 100 bp per sample. We used Hisat2³⁹ with default parameters to map the reads to the P01 reference genome⁴⁷ (version 67). Samples with an average coverage greater than 50X were further analyzed. We then estimated whether each sample was monoclonal or polyclonal, using GATK⁴⁸ to call nucleotide variants, excluding telomeric regions and multigene families. We considered only positions with at least 20X coverage in at least 80% of the samples and only polymorphic positions with a maximum of 2 alleles. We then used moimix⁴⁹ to estimate polyclonality, with a F_ws ≥ 0.95 being considered as monoclonal and those F_ws < 0.95 being considered as polyclonal. We then analyzed the relatedness of monoclonal samples by calculating pairwise distances. We also analyzed the relatedness of monoclonal samples by calculating pairwise distances. We used the proportion of those positions without shared nucleotides to generate a distance matrix. We used this matrix to generate a Neighbor-Joining tree in MEGA11⁵⁰.

Genetic analysis of MaRNAV-1 and sequence rearrangement determination

Multiple nucleotide sequence alignments for the MaRNAV-1 S1 and S2 were generated using the MUSCLE algorithm in MEGA11⁵⁰. The phylogenetic trees of the S1 and S2 of the MaRNAV-1 were inferred using a maximum likelihood approach. To improve the quality of alignments, we filtered out poorly aligned regions and gaps using trimAl with an automated method⁵¹. All trees were estimated using IQ-TREE2 (v3.0.1). To assess the robustness of the resulting topologies, we calculated branch support using 1,000 bootstrap replicates with the UFBoot2 algorithm and an implementation of the SH-like approximate likelihood ratio test within IQ-TREE2^52,53. The best-fit model of nucleotide substitution was evaluated using the Akaike information criterion (AIC), the corrected AIC, and the Bayesian information criterion (BIC) implemented in the ModelFinder function in IQ-TREE 2^54,55, while TPM2u+I+G4 was selected according to BIC for downstream phylogenetic analyses. The tanglegram was annotated using the R package phytools (v2.5.2)⁵⁶.

Detection of antibodies against MaRNAV-1 proteins by ELISA

We determined if antibodies against RdRP and S2_ORF1 could be detected in human plasma samples by ELISA. Recombinant RdRp (see above) and recombinant S2_ORF1 were purchased from Genscript and expressed in E. coli. Note that attempts to express and purify S2_ORF2 in high enough yield and purity failed. Maxisorp plates (Nunc Immunoplate, Thermo Scientific) were coated overnight with RdRp protein at 2 μg/ml and with S2_ORF1 at 1 μg/ml. Wells were blocked at 37 °C for 1 hour with blocking buffer (5% non-fat milk in PBS-0.05% Tween-20). Human plasmas were added at 1:50 dilution to the wells for 1h at 37°C. After washing, goat anti-human IgG-HRP was added at a dilution of 1:10,000 for RdRp protein and 1:50,000 for S2_ORF1 for 1 hour at 37 °C. The assay was developed using a TMB substrate solution (BioLegend) and stopped with 0.18 M H₂SO₄. The absorbance at 450 nm was measured using the iMark Microplate Absorbance Reader. Human plasma used were collected from confirmed P. vivax infected individuals from both Cambodia and Ethiopia. Negative controls were made of plasma collected from malaria-naïve volunteers originating from Europe and residing in Phnom Penh.

Multiplex assay determination of plasma cytokines

Cytokine concentration in human plasma samples were determined using a magnetic bead-based multiplex Luminex multi-analyte profiling (MAP) assay (MAGPIX^™ systems) with the Cytokine 30-Plex Human Panel (Thermo Fisher Scientific), following the manufacturer’s instructions. Results were plotted as picograms per milliliter. The samples below the lower limit of detection (LLOD) were replaced with the lowest detectable by the machine, divided by 2 for data analysis.

Statistical Analyses

Mann-Whitney or Kruskal-Wallis tests were conducted for nonparametric comparisons. Fisher’s exact test and Chi-square test were used for proportion comparisons. We used R (version 4.5.1) and RStudio (version 2025.05.1) for modeling analysis. To examine the impact of the MaRNAV-1 on transmission potential, we analyzed oocyst counts from the mosquito membrane-feeding assay. We fitted generalized linear mixed models GLMMS using the glmmTMB package. A negative binomial hurdle model was employed, including both the oocyst prevalence and oocyst intensity. Oocyst prevalence was analyzed using a binomial mixed model with log-transformed parasitemia, log-transformed virus load, and their interaction as fixed effects, and patient identity as a random effect. To investigate the effects of MaRNAV-1 and parasitemia on clinical correlates and host immune response signaling (cytokines, chemokines, and growth factors), we employed linear regression with MaRNAV-1 as a binary variable (present/absent) or generalized additive models (GAM), allowing flexible adjustments for non-linear effects of MaRNAV-1 as a continuous variable, adjusting for parasitemia and their interaction. The interaction between MaRNAV-1 and parasitemia was added to each model unless it was significant in a Type II ANOVA. The approximate effective degree of freedom (edf) and global p values were reported for smooth terms. To determine whether the proportions of immune cell types are affected by MaRNAV-1 and parasitemia, a Dirichlet regression model was applied to jointly analyze the composition of 19 cell types. To account for multiple hypothesis testing and control the false discovery rate (FDR)⁴⁴, p-values were adjusted using the Benjamini-Hochberg method for all explanatory variables across all models. The DHARMa package was used to assess whether the model assumptions were adequately met. The p-values < 0.05 were considered statistically significant.

Supplementary Material

Supplement 1

Supp. Data 1. RNA-seq reads mapping to MaRNAV-1 and to P. vivax.

Supplement 2

Supp. Data 2. Summary data of P. vivax genes expressed as TPM (transcripts per million) for calculating the threshold of the 25^th and 75^th percentiles of P. vivax genes

Supplement 3

Supp. Data 3. MaRNAV-1 S1 nucleotide alignment

Supplement 4

Supp. Data 4. MaRNAV-1 S2 nucleotide alignment

Supplement 5

Supp Data 5. RdRP protein alignment

Supplement 6

Supp Data 6. S2_ORF1 protein alignment

Supplement 7

Supp Data 7. S2_ORF2 protein alignment

Supplement 8

Supp. Data 8. P. vivax DGE analysis of presence and absence of MaRNAV-1.

Supplement 9

Supp. Data 9. P. vivax DGE analysis of low and high of MaRNAV-1 load (below the 25^th percentile versus above the 75th percentile)

Supplement 10

Supp. Data 10. Significant component-level associations identified by Dirichlet regression analysis of immune cell composition

Supplement 11

Supp. Data 11. Results of human DGE analysis comparing the presence and absence of MaRNAV-1.

Supplement 12

Supp. Data 12. Results of human DGE analysis comparing samples with MaRNAV-1 load below the 25^th percentile to those above the 75^th percentile

Supplement 13

Supp. Data 13. Primary probes sequences used in smiFISH including P. vivax 18 rRNA, S1 and S2 segment of the MaRNAV-1.

Supplement 14

Supp. Fig 1. Genomic structure of MaRNAV-1

Supp. Fig. 2. Negative controls for RdRp immunofluorescence staining.

Supp. Fig. 3. Comparison of parasitemia levels between MaRNAV-1 negative and positive samples in symptomatic and asymptomatic Cambodian P. vivax infected patients

Supp. Fig. 4. Associations between MaRNAV-1 load and blood-stage parasite composition and gametocytemia

Supp. Fig 5. Correlation between MaRNAV-1 load with gametocytemia

Supp. Fig. 6. Association between MaRNAV-1 infection status or viral load and transmission of P. vivax to mosquito vectors

Supp. Fig 7. Association between MaRNAV-1 infection status or viral load and transmission of P. vivax to mosquito vectors using Pvs47 as gametocytemia measure

Supp. Fig. 8. Interaction between MaRNAV-1 load and parasitemia on predicted body temperature

Supp. Fig. 9. Predicted immune cell composition and host gene expression associated with MaRNAV-1 load

Supp. Fig. 10. Nonlinear associations between MaRNAV-1 virus load, parasitemia, and circulating cytokine levels

Supp. Table 1. Clearance kinetics of P. vivax parasites and MaRNAV-1 following artesunate treatment.

Supp. Table 2. Association between MaRNAV-1 infection status and clinical correlates in P. vivax-infected patients.

Supp. Table 3. Non-linear associations between MaRNAV-1 load and clinical correlates in P. vivax-infected patients

Supp. Table. 4. Non-linear associations between MaRNAV-1 load and clinical correlates in MaRNAV-1-positive P. vivax-infected patients

Supp. Table 5. Global effects of MaRNAV-1 infection, parasitemia, and their interaction on immune cell composition

Supp. Table. 6. Association between MaRNAV-1 infection status and plasma cytokine levels in P. vivax-infected patients

Supp. Table 7. Non-linear associations between MaRNAV-1 load and cytokine levels in MaRNAV-1-positive P. vivax-infected patients.

Supp. Table 8. Primers used in this study