Transcriptome assembly and differential gene expression of the invasive avian malaria parasite Plasmodium relictum in Hawaiʻi

Abstract The malaria parasite Plasmodium relictum (lineage GRW4) was introduced less than a century ago to the native avifauna of Hawaiʻi, where it has since caused major declines of endemic bird populations. One of the native bird species that is frequently infected with GRW4 is the Hawaiʻi ʻamakihi (Chlorodrepanis virens). To achieve a better understanding of the transcriptional activities of this virulent parasite, we performed a controlled challenge experiment of 15 ʻamakihi that were infected with GRW4. Blood samples containing malaria parasites were collected at two time points (intermediate and peak infection stages) from host individuals that were either experimentally infected by mosquitoes or inoculated with infected blood. We then used RNA sequencing to assemble a high‐quality blood transcriptome of P. relictum GRW4, allowing us to quantify parasite expression levels inside individual birds. We found few significant differences (one to two transcripts) in GRW4 expression levels between host infection stages and between inoculation methods. However, 36 transcripts showed differential expression levels among all host individuals, indicating a potential presence of host‐specific gene regulation across hosts. To reduce the extinction risk of the remaining native bird species in Hawaiʻi, genetic resources of the local Plasmodium lineage are needed to enable further molecular characterization of this parasite. Our newly built Hawaiian GRW4 transcriptome assembly, together with analyses of the parasite's transcriptional activities inside the blood of Hawaiʻi ʻamakihi, can provide us with important knowledge on how to combat this deadly avian disease in the future.

much of the endangerment attributable to avian malaria. Malaria is an infectious disease caused by single-celled eukaryotic parasites in the genus Plasmodium, which are transmitted by mosquitoes. In 1826, the bird-biting mosquito Culex quinquefasciatus was introduced to the Hawaiian islands with ships (Hardy, 1960), though it was not until the late 1930s that malaria parasites were discovered in the blood of native birds (van Riper et al., 1986). The parasites were identified as the broadly distributed mitochondrial lineage GRW4 of Plasmodium relictum (Figure 1a) (Beadell et al., 2006). Because the Hawaiian avifauna likely evolved for millions of years in the absence of malaria parasites (Fleischer et al., 1998;Lerner et al., 2011), many native bird species, and nearly all species of Hawaiian honeycreepers, do not possess much natural resistance or tolerance against the disease.
Despite the urgency in understanding how P. relictum affect the endemic avifauna of Hawaiʻi, we know almost nothing of the parasite's transcriptional activities inside its hosts. Malaria parasite gene expression levels in birds have previously been evaluated in two species: Plasmodium ashfordi and Plasmodium homocircumflexum (Garcia-Longoria et al., 2020;Videvall et al., 2017). The first study found that P. ashfordi gene expression did not differ between peak and decreasing parasitemia stages in Eurasian siskins; instead, 28 transcripts showed differential expression depending on which host individual the parasites infected (Videvall et al., 2017). Similarly, the study evaluating P. homocircumflexum found transcriptional differences between hosts; however, this approach evaluated differences across bird species (crossbills and starlings) (Garcia-Longoria et al., 2020).
While these previous results provide valuable information, the sample sizes were limited to three-four individuals (and two time-points for P. ashfordi) and thus require further investigation. In addition, European birds with evolved resistance to malaria are not able to provide accurate estimates of how P. relictum behaves in the blood of native bird species of Hawaiʻi, which have evolved without the parasite.
In this study, we sequenced and built the first transcriptome assembly of P. relictum (lineage GRW4). Using a controlled infection experiment, we evaluated parasite gene expression levels during two time points in the blood of 15 native, high-elevation Hawaiʻi ʻamakihi. We specifically aimed to evaluate whether parasite gene expression differs (a) between intermediate and peak infection stages, (b) between birds that survived and birds that died from malaria, (c) between a mosquito inoculation method and blood injection method, and finally (d) among different host individuals.

| Experimental design
We captured 20 individuals of Hawaiʻi ʻamakihi (Figure 1b Forest Reserve. This high-elevation region is predominantly malaria free where birds are unlikely to encounter malaria parasites in the wild. The birds were kept in individual cages in a mosquito-proof aviary, subjected to natural light, and provided a diet of nectar, fruit, and vegetables. Prior to the experiment, all birds were screened with nested PCR (Lapointe et al., 2016), ELISA (Woodworth et al., 2005), and microscopy to ensure no individual carried hemosporidian infection. The P. relictum GRW4 isolate KV115 was used, originally obtained from a wild ʻapapane (Himatione sanguinea) at Kilauea Crater in Hawaiʻi Volcanoes National Park in 1992. It was passaged once in a canary (Serinus canaria) prior to being glycerolized, then frozen and stored in liquid nitrogen. The same isolate has been used in previous experimental studies (Atkinson et al., 2000(Atkinson et al., , 2013. Prior to the experiment in this study, the isolate was thawed, deglycerolized, and passaged in canaries an additional four times. Birds were acclimated for a minimum of four weeks before being randomly assigned to one of three treatment groups: control, inoculation by mosquitoes, or inoculation by blood injection. Ten birds were infected through exposure overnight to the bite of a single infected Culex quinquefasciatus (mosquito inoculation group). The mosquitoes had been infected by a single canary that was inoculated with P. relictum GRW4. Five birds were experimentally infected by subinoculation in their pectoral muscle with 150 μl of infected blood solution sourced from the same canary individual that infected the F I G U R E 1 (a) Image of Plasmodium relictum GRW4 on a Giemsa-stained blood smear seen through a microscope. Red blood cells are pictured, each containing an elongated nucleus in dark purple color. The pink round shapes within some of the cells constitute the parasites. Image by Carter T. Atkinson. (b) The host species, Hawaiʻi ʻamakihi (Chlorodrepanis virens). Photograph by Loren Cassin-Sackett

| Parasitemia quantification
We measured intensity of parasitemia using a quantitative PCR ( At each sampling period, we also collected samples for RNA sequencing (30 μl whole blood in 210 μl of RNAlater) that were stored for 24 hr at 0°C and then at −20°C until RNA extraction. Blood smears were prepared, air-dried, fixed with methanol, and stained with 6% buffered Giemsa for 1 hr. They were then examined by microscopy to determine the proportion of asexual and sexual parasites in 100-200 infected erythrocytes. Blood smear examination was performed without prior knowledge of experimental group.
Based on prior experimental studies (Atkinson et al., 2000(Atkinson et al., , 2013, ʻamakihi were classified as fatalities when their parasitemia levels exceeded 20%, food consumption fell below 5 ml of nectar over the prior 24-hr period, and individuals appeared moribund. Five birds were classified as fatalities, removed from the experiment, and treated with oral chloroquine (10 mg/kg) to reduce risk of dying without intervention. Despite these efforts, four of these five birds died within a few weeks of chloroquine treatment.

| RNA extraction and sequencing
RNAlater was separated from blood by centrifugation, and RNA from approximately 20 μl of packed red blood cells was subsequently extracted using Dynabeads mRNA Direct Kit (Invitrogen), a poly-A tail binding bead-based approach that captures mRNA. We converted mRNA to first-and second-strand cDNA using SuperScript

| Transcriptome assembly
We assembled the transcriptome of P. relictum GRW4 using Trinity Parameters in STAR were set to be optimized for Plasmodium parasites (Baruzzo et al., 2017), slightly modified to fit our data (Table S2). The GRW4 transcriptome was subsequently de novoassembled with Trinity's genome-guided approach using the STAR-mapped reads to help guide the assembly process. Trinity's genome-guided transcriptome assembly method uses aligned reads partitioned according to locus, followed by de novo assembly at each locus (Haas, 2020). This method is distinct from typical genome-guided approaches because transcripts are constructed de novo and the provided genome is only being used as a substrate for grouping overlapping reads into clusters (Haas, 2020).
Maximum intron size was set to 4,000 based on Plasmodium genomes (Aurrecoechea et al., 2009) Campana et al., 2020). This search (e-value < 1e-10) resulted in 22 contigs matching potential avian rRNA sequences, which were all removed from the assembly. We subsequently used TransDecoder (v. 5.5.0) (Grabherr et al., 2011) to identify open reading frames and blastn against P. relictum DONANA05 (SGS1like) coding sequences for gene annotation. transcripts with mode set to "intersection-nonempty". To evaluate the percentages of parasite sequences in whole blood, we performed an additional read mapping procedure using HiSat2 against the GRW4 transcriptome, but this time we used the full set of unfiltered sequence reads (including host-derived sequences). The proportion of total reads mapping against GRW4 showed a strong correlation with estimated parasitemia (Pearson's correlation test: r = .70, p = 4.52e-06; see Figure S1), meaning that parasitemia intensity was a good predictor of P. relictum sequence depth.  and Plasmodium ovale (n = 26; 0.2% overall).

| Parasite gene expression does not differ between infection stages
We quantified expression levels of all P. relictum transcripts in samples containing sufficient numbers of parasites (>200 SQ values; n = 20).
The most highly expressed transcripts originated from genes previously documented as having the highest expression levels in other Plasmodium transcriptomes , for example, elongation factor 1-alpha and 2, histone H4 and H2A, heat shock protein 70, and alpha tubulin 1 (Table S3). Evaluation of total transcriptome expression showed no clustering of samples based on similarity of parasitemia intensities (Figure 3), demonstrating the read normalization method removed potential biases associated with sequencing depth and parasite load.
Comparing the two time stages of infection (peak and intermediate) also showed few differences in parasite transcript expression, with the exception of two transcripts upregulated during the peak infection stage (Figure 4). These transcripts belonged to a conserved Plasmodium gene with unknown function (PRELSG_0814200; Wald statistic = 3.9, q = 0.06) and a gene coding for DNA-directed RNA polymerases I, II, and III (PRELSG_1105700; Wald statistic = 3.8, q = 0.06). We further found no differences in transcript expression between parasites in hosts that were classified as fatalities compared to parasites in hosts that survived the disease. Testing the effect of inoculation method (mosquito vs. blood injection) resulted in one significant transcript coding for gamete antigen 27/25 (PRELSG_0014900; Wald statistic = −4.1, q = 0.04), which had slightly higher expression in the parasites that had been injected with blood inoculation at the beginning of the experiment.     in blood smears across host individuals or between host infection stages. It is possible that GRW4 is regulating the expression of certain genes to better respond to different host individuals; however, further studies are needed to evaluate the precise mechanism behind this pattern.

| D ISCUSS I ON
The most significant transcripts showing expression differences among host individuals belonged to the multigene families fam-e and fam-h. These subtelomeric gene families are not present in laveranian Plasmodium species but are expanded in avian Plasmodium genomes: fam-e comprises four gene copies and fam-h 49 gene copies in P. relictum (Böhme et al., 2018). Fam-e has been discovered in the genome of Plasmodium vivax (Carlton et al., 2008), while fam-h is believed to be specific to avian Plasmodium (Böhme et al., 2018). They appear related to the P. falciparum protein families RAD and PHIST, which bind the virulence factor PfEMP1 and remodel host erythrocytes (Oberli et al., 2014;Warncke et al., 2016). Several studies have found differential gene expression of PHIST during the P. falciparum life cycle and among different parasite isolates (Eksi et al., 2005;Rovira-Graells et al., 2012). Interestingly, we also found one retrotransposon with differential expression levels among host individuals.
Transposable elements are not present in mammalian Plasmodium, but they have been found in avian Plasmodium genomes (Böhme et al., 2018). It has been suggested that transposable elements like retrotransposons were present in genomes of ancestral apicomplexa and subsequently lost (Roy & Penny, 2007). The transcript in our study matches the intact Plasmodium gallinaceum Ty3/Gypsy LTR retrotransposon (PGAL8A_00410600), which has a continuous open reading frame. Because almost nothing is known about this particular retrotransposon, it is difficult to speculate why it is differentially expressed in P. relictum; however, we note that studies of Entamoeba have found differential transposon expression between strains (Macfarlane & Singh, 2006).
In conclusion, our results enable an improved understanding of the transcriptional activities of malaria parasites in birds, and the assembled transcriptome of P. relictum GRW4 will become a valuable genetic resource in the long-term quest to better characterize the biology and evolution of this invasive Plasmodium lineage.

ACK N OWLED G M ENTS
We are grateful to M. Renee Bellinger and two anonymous reviewers for helpful comments on the manuscript. This study was

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.