Complete genome of a unicellular parasite (Antonospora locustae) and transcriptional interactions with its host locust

Abstract

Microsporidia are a large group of unicellular parasites that infect insects and mammals. The simpler life cycle of microsporidia in insects provides a model system for understanding their evolution and molecular interactions with their hosts. However, no complete genome is available for insect-parasitic microsporidian species. The complete genome of Antonospora locustae, a microsporidian parasite that obligately infects insects, is reported here. The genome size of A. locustae is 3 170 203 nucleotides, composed of 17 chromosomes onto which a total of 1857 annotated genes have been mapped and detailed. A unique feature of the A. locustae genome is the presence of an ultra-low GC region of approximately 25 kb on 16 of the 17 chromosomes, in which the average GC content is only 20 %. Transcription profiling indicated that the ultra-low GC region of the parasite could be associated with differential regulation of host defences in the fat body to promote the parasite’s survival and propagation. Phylogenetic gene analysis showed that A. locustae, and the microsporidian family in general, is likely at an evolutionarily transitional position between prokaryotes and eukaryotes, and that it evolved independently. Transcriptomic analysis showed that A. locustae can systematically inhibit the locust phenoloxidase PPO, TCA and glyoxylate cycles, and PPAR pathways to escape melanization, and can activate host energy transfer pathways to support its reproduction in the fat body, which is an insect energy-producing organ. Our study provides a platform and model for studies of the molecular mechanisms of microsporidium–host interactions in an energy-producing organ and for understanding the evolution of microsporidia.

Data Summary

The whole-genome sequencing datasets from this study have been submitted to the BioProject database of the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/) under accession number PRJNA353563. All transcriptome data were uploaded to the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra) database under the GenBank accession numbers SRR5171247, SRR5171248, SRR5171251-SRR5171254, SRR5171257 and SRR5171258.

Impact Statement.

Antonospora locustae is the only species that has demonstrated strong potential as a bio-insecticide for controlling outbreaks of locusts worldwide. Following the development of high-throughput sequencing, the complete genome of A. locustae was assembled to the chromosome level based on the method used for combining second- and third-generation high-throughput genome sequencing technologies. Complete genomes of other microsporidia may be obtained in the same way, accelerating research into microsporidia. We have reported the most complete genome sequences of A. locustae to date by far, along with detailed gene annotation and a relatively clear explanation of the interaction between A. locustae and its locust host. In-depth analysis of the interactions between A. locustae and its host at different stages showed that A. locustae recognizes its host and transitions in the midgut, regulates host energy transfer processes after entering the fat body, inhibits host melanization of A. locustae to evade the host’s immune system, and then completes its generational change in the host’s fat body cells.

Introduction

Microsporidia are a large group of obligate intracellular parasites of insects and mammals [1–8]. The taxonomic status of this group of unicellular parasites remains controversial, although recent studies have suggested that microsporidia belong to the fungal kingdom. Regardless of their classification, most studies place these parasites in a unique node position between prokaryotes and unicellular eukaryotes, using them as model organisms for evolutionary studies of interactions with eukaryotic hosts [9–11].

Over 1400 species of microsporidia in 200 genera have been identified since Nägeli first isolated Nosema bombycis from an infected silkworm in 1871 [12, 13]. Both invertebrates and vertebrates fall within the host ranges of microsporidia, including honeybee [14], silkworm [15], fish [7], shrimp [16], swine [17], horse [18], cattle [19] and goat [20]. Approximately 10 species of microsporidia cause human diseases [21]. Symptoms of human infection with microsporidia include keratitis, myositis, encephalitis, cholecystitis, hepatitis, osteomyelitis, pulmonary infection and death [22–26].

Microsporidia that infect insects have relatively simple life cycles, providing ideal model systems for studying evolution and molecular interactions with their hosts. Following the initial genome sequencing work with the parasite Encephalitozoon cuniculi, which infects mammals [2], partial genomic sequences have been produced for N. bombycis [3] and Nosema ceranae [27], whose target tissues of infection are primarily the midgut, silk gland and malpighian tube of host insects. In contrast to N. bombycis and N. ceranae, the major organ targeted for infection by Antonospora locustae is the fat body of locusts. A. locustae, which was formerly known as Nosema locustae and is synonymous with Paranosema locustae, has a fairly narrow host range, only infecting locusts [28]. The fat body in insects is functionally equivalent to the liver in vertebrates, and may provide another special model for interactions between microsporidia and their hosts. These previous works facilitated in-depth investigation of the mechanisms of parasite infection and host–parasite interactions, as well as their control.

Numerous studies have investigated the use of A. locustae as a powerful biological control agent against locusts in agriculture since the 1980s, and it is the only species that has demonstrated strong potential as a bio-insecticide for controlling outbreaks of locusts worldwide [29–31]. A. locustae showed great potential during the locust disaster that broke out at the end of last year, in particular. However, few studies have been carried out on the molecular infection mechanism of A. locustae [32]. Such research would be helpful for improving the application of microsporidia as bio-insecticides. Preliminary sequences of the A. locustae genome and transcriptome have been reported, along with limited information on A. locustae genomics [8].

In this study, we report a full and complete genome sequence of A. locustae based on second- and third-generation genome-sequencing technologies. The genome of the parasite consists of 17 chromosomes on which a total of 1857 coding genes have been annotated and mapped. An ultra-low GC region of approximately 25 kb was found in 16 of the 17 chromosomes. Our phylogenetic study based on genomes suggested that microsporidia are a special evolutionary group. A transcriptional study of pathogen-infected and healthy host tissues highlighted the molecular interactions between A. locustae disease and the locust host. Our study provides a platform and model for studies of the molecular mechanisms of microsporidia–host interactions in an energy-producing organ, as well as for understanding the evolution of microsporidia.

Methods

A. locustae sample preparation and DNA isolation

The stocks of A. locustae spores used in the experiments were routinely maintained in the Key Laboratory of Biological Control, Ministry of Agriculture, China Agricultural University. To isolate genomic DNA from the parasite, the spores were propagated in vivo by infecting its natural host, Locusta migratoria, which were routinely maintained in the same laboratory, following procedures described previously [32].

Briefly, colonies of the host locust were maintained at 28–30°C and 60 % relative humidity. For infection, fresh corn leaves coated with A. locustae spores were fed to third instar larvae of locusts for 12 h. Establishment of A. locustae in the fat body of locusts was determined microscopically 15–18 days after infection. The fat body and other control tissues were collected from the infected hosts, and cells were lysed for initial removal of host genomic DNA. After proper filtration, a modified CTAB extraction method was followed to isolate spores of the parasites. The Omniprep DNA extraction kit (G-Biosciences) was then used to extract genomic DNA of the parasite [33]. The isolated DNA was examined for the presence or absence of host DNA contamination, and only host DNA-negative samples were retained for subsequent experiments.

Genome de novo sequencing through Illumina HiSeq

DNA sequencing libraries were constructed according to the manufacturer’s instructions (NEBNext Ultra DNA Library Prep Kit for Illumina). For each sample, 2 µg of genomic DNA was randomly fragmented to <500 bp through sonication (Covaris S220). The fragments were treated with End Prep Enzyme Mix for end repair, 5′ phosphorylation, and dA-tailing in one reaction, followed by ligation to adaptors with a ‘T’ base overhang. Size selection of adaptor-ligated DNA was then performed using the AxyPrep Mag PCR Clean-up kit (Axygen), and fragments of ~410 bp (with an approximate insert size of 350 bp) were recovered. Each sample was then amplified via PCR for eight cycles using P5 and P7 primers, with both primers containing sequences that anneal with the flow cell for bridge PCR and the P7 primer containing a six-base index to allow for multiplexing. The PCR products were cleaned up using the AxyPrep Mag PCR Clean-up kit (Axygen), validated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA), and quantified with a Qubit2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA).

Libraries with different indices were multiplexed and loaded in an Illumina HiSeq instrument according to the manufacturer’s instructions (Illumina, San Diego, CA, USA). Sequencing was carried out using a 2×150 paired-end (PE) configuration; image analysis and base calling were conducted with HiSeq Control Software+OLB+GAPipeline-1.6 (Illumina) on the HiSeq instrument. The average final read depth in the assembly was 2000×. The sequencing results were processed and analysed using GENEWIZ.

Third-generation de novo sequencing of the A. locustae genome using the PacBio RSII SMRT platform

To assemble the A. locustae genome, 15 µg of high-quality genomic DNA was randomly interrupted using an ultrasound method (Covaris S220) to obtain double-stranded fragments of approximately 5–10 kb, and DNA fragments of more than 2 kb were recovered. The end of the DNA fragment was ligated to the linker of the hairpin structure. Library construction was carried out using the commercial SMRTbell library method. Sequencing of the library was performed using the PacBio RSII SMRT system [34]. Assembly of the PacBio reads was conducted using PBcR WGS-Assembler 8.2 software [35–40]. The average final read depth of the assembly was 200×.

Assembly and annotation of genomic data

Based on clean data optimized with the Illumina HiSeq platform, Velvet (version 1.2.10) software was used for k-mer analysis, a de Bruijn diagram was constructed based on the overlapping relationship between k-mers, and the segmented contig sequence was spliced. SSPACE (version 3.0) was used to compare the reads obtained from sequencing of all libraries to the contig sequences obtained using the pairing relationship between paired-end reads and insert size distances, as well as to support primer shifts and design of primers for PCR experiments and further assembly of contig sequences to the scaffold sequence. Finally, using GapFiller (version 1–10) software, all reads from all libraries were aligned to the scaffold sequence. Gaps in the scaffold sequence were complemented by the aligned reads, and thus the scaffold sequence could be extended. Finally, we obtained a scaffold sequence with a lower ratio of unknown bases, N, and increased sequence length.

The PacBio off-machine data obtained using the PacBio RSII SMRT platform were assembled using the assembly software wgs-assembler (version 8.3) to obtain preliminary assembly results. Based on the preliminary assembly results, Illumina HiSeq sequencing data were simultaneously imported with the PacBio assembly data, and this assembly was corrected using the calibration software Quiver (version 1.1.0) to obtain the final assembly result.

The methods used for gene prediction in the genomes as follows. Glimmer gene prediction software was primarily used for the prediction of single-exon genes [41]. De novo gene prediction was carried out using Augustus software [42], combined with existing transcriptome data for mapping of the genome, analysed with StringTie software, and finally, more accurate gene predictions were obtained. Annotation of coding genes was performed by comparison with the Nr database of the National Center for Biotechnology Information (NCBI). Functional annotation of these genes was performed using the GO database [43], pathway annotation was carried out using the KEGG database [44] and systematic classification of proteins encoded by the genes was performed using the Clusters of Orthologous Groups (COG) database [45]. Scanning for transfer RNAs in the genome was mainly performed using rRNAscan-SE software with default parameters [46], and ribosomal RNAs were identified using RNAmmer software [47].

RNA isolation and sequencing

Locusts were infected with A. locustae as described above. Transcriptome sequencing was divided into four groups, with each group consisting of two independent repeated samples, and a total of eight independent samples. The midgut of healthy locusts (NM), the infected midgut (M), the fat body of healthy locusts (NF) and the fat body of infected locusts (F) were analysed, with numbers representing different individuals. On the 15th day after infection, the fat bodies and midguts of living locusts were collected. All samples were immediately frozen in liquid nitrogen, and the remaining tissues were ground and diluted with deionized water to determine whether the infected group was actually infected with A. locustae and the intensity of infection. Only locusts with the same infection intensity were used for RNA extraction and sequencing. In addition, the healthy group of locusts was tested as a control. Locusts in this group were only used for RNA extraction and transcriptome sequencing after passing the infection test.

One microgram of total RNA was used for library construction. For transcriptome sequencing, we used the NEBNext UltraTM RNA Library Prep kit from Illumina according to the manufacturer’s instructions. After mixing the various index-labelled libraries, 2×150 bp double-end sequencing (PE) was performed according to the Illumina HiSeq 2500/3000 (Illumina) instrument instruction manual, using the HiSeq Control Software provided by HiSeq and the OLB+GAPipeline−1.6 (Illumina) program to obtain sequence read data. Four uninfected samples provided approximately 6.0 Gb of data per sequencing library; four infected samples provided approximately 8.0 Gb of data per sample library (including locust and microsporidium).

Data analysis

Clean data for subsequent analysis were obtained by removing the linker and low-quality sequences from the raw data (Pass Filter Data) using the second-generation sequencing data quality statistical software Trimmomatic (v0.30) [48, 49]. The filtered clean data were analysed against the locust genome and the A. locustae genome sequenced in this study [50]. The reads obtained from A. locustae RNA sequencing were applied to density statistical analysis of each chromosome using Hisat2 (v2.0.1) software with the default parameters for short-read comparison [51, 52]. BUSCO was applied to evaluate the assembly completeness by identifying a set of highly conserved microsporidia orthologues in the assembly [53]. For the analysis, gene annotation was performed with Augustus, and the analysis for homology and positive matches was performed with HMMER 3 [54]. The expression levels of each A. locustae chromosome were calculated. Gene expression levels were analysed using the fragment per kilo bases per million reads (FPKM) method with rsem software (V1.2.6). Differential gene expression analysis was performed using DESeq2 in Bioconductor software (V1.6.3). Screening for differentially expressed genes was based on expression level changes that were greater than twofold with a false discovery rate ≤0.05. Statistical analyses were performed on upregulated and downregulated genes to identify significant differences. All statistical t-tests (and nonparametric tests) followed by two-tailed comparison tests were performed using GraphPad Prism version 6.00 for Windows (GraphPad Software, Inc., La Jolla, CA, USA).

For the genomic data of A. locustae, Saccharomyces cerevisiae, Kazachstania naganishii, Babesia bigemina, Babesia motasi, Encephalitozoon cuniculi and Encephalitozoon hellem, paralogous and syntenic collinear blocks were characterized using the MCScanX strategy [55]. Briefly, proteomic sequence data were obtained using the blastp algorithm to generate blast outputs, which were imported into MCScanX software to generate collinearity outputs. Then, a circle plotter program was employed to graph the paralogous and syntenic collinear blocks with the collinearity outputs.

For phylogenetic analysis, the microsporidian protein sequences of frataxin were retrieved from GenBank databases. Orthologous sequences were identified using blastp searches at an E-value cutoff of 1E−20, using A. locustae frataxin proteins as queries. Each group of proteins was aligned using the MAFFT program (version 7) with the E-INS-I algorithm [56], and ambiguous regions in the aligned sequences were removed with TriMal [57]. Maximum-likelihood phylogenetic trees were estimated in phyML 3.0 software [58] using the best model of amino acid substitutions estimated with mega 6 and the Regrafting (SPR) branch-swapping algorithm.

Results

DNA sequencing of the A. locustae genome

Genomic DNA from A. locustae were prepared successfully without contamination from host DNA (Fig. S1, available in the online version of this article). Subsequent DNA sequencing using the Illumina HiSeq II platform generated a total of 87 379 454 reads with a bidirectional read length of 150 bp, GC content of 42.53 % and uniform distribution (Fig. S2). Statistical analysis of the original data (Figs S3 and S4) showed that all reads were of good quality, as indicated by the error rate Q20 (<1 %)=95.58 %, Q30 (<0.1 %)=91.42 % and N=6.99/Mb. The data were cleaned and optimized using Trimmomatic software, yielding a total of 58 316 056 reads of 144.70 bp on average, with GC content=41.77 %, Q20=99.70 %, Q30=98.63 % and N=0.77 per million bases. In parallel, PacBio RSII SMRT third-generation high-throughput genome sequencing (GENEWIZ) was used to generate a total of 160 777 sequence reads with an average length of 3918.61 bp per read, N50=5237 bp and GC content=41.57 %.

The complete genomic sequence of A. locustae, determined to comprise 3 170 203 nucleotides and encode 1857 predicted genes (Table S1), was successfully assembled de novo using the clean sequence data generated from the PacBio RSII SMRT platform supplemented with sequence data from the Illumina HiSeq II platform. A total of 17 scaffolds, ranging from 88.763 to 388.82 kb, were identified. Each of these scaffolds was assigned to a chromosome of A. locustae, from chromosome 1 to 17 (Table S2). The assembly completeness of the A. locustae genome was evaluated with benchmarked universal single-copy orthologue (BUSCO), for a total of 600 genes, and the HMMER 3 homology (reference genome: E. cuniculi) search revealed 85 % complete single-copy orthologues (C), <1 % complete duplicated orthologues (D), <1 % fragmented orthologues (F) and 14 % missing (M) from the universal orthologue microsporidia database (Fig. S5). Among the genes predicted in the A. locustae genome, 1755 are single-exon genes, accounting for 94.5 % of all genes found. By contrast, only 102 genes were found to contain multiple exons, accounting for 5.5 % of the entire genome (Table S3). A brief parameter comparison of the A. locustae genome with other available microsporidian genomes is included in Table 1. The assembled sequence was submitted to GenBank. Using a combination of gene ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) and Eukaryotic Orthologous Groups (KOG) pathways (Tables S4–S6, Figs S6–S8), we successfully constructed a complete chromosome map for the genome of A. locustae (Fig. S9). The gene annotation and the locations of predicted genes are summarized in Fig. 1 (See Table S7 for details).

Table 1.

Comparison of genome information between A. locustae and other microsporidia

Genomic features	A. locustae	N. bombycis	N. ceranae	E. cuniculi	E. intestinalis	E. bieneusi	Octosporea bayeri	Spraguea lophii	Bursaphelenchus xylophilus
Chromosome no.	17	18	nd	11	11	≥6	nd	15	6
Genome size (Mbp)	3.2	nd	7.7	2.9	2.3	6	≤24.2	6.2–7.3	63–75
Assembled (Mbp)	3.2	15.7	7.9	2.5	2.2	3.9	13.3	4.98	74.6
No. of scaffolds/contigs	17	1605	5465	11	137	1646	41 804	1392	1231
N50 (bp)	183 675	57 394	2902	nd	nd	2349	nd	5923	1158
Genome coverage (%)	100	100	90	86	96	64	55	70–80	nd
G+C content (%)	41.6	31	27	48	41.4	26	26	20	40.4
Predicted ORFs	1857	4458	2614	1999	1833	3804	2174	2543	18 074

nd, no data.

Fig. 1.

A circular representation of the complete genome of A. locustae. DNA sequencing revealed that the genome size of A. locustae is 3 170 203 base pairs, with a total of 1857 predicted coding genes distributed on 17 chromosomes. The outermost circle shows chromosome size (kb) and the distribution of KEGG pathways, as indicated with colour-coded bars (see the colour bar for key) for each chromosome. The second circle from the outside shows the variation in GC content for each of the 17 chromosomes, characterized by a sharp decrease in GC content near the centre of each chromosome. The third circle from the outside represents the distribution of coding genes on the positive strand (red) and negative strand (green) of DNA, respectively. The non-coding RNA (ncRNA) detected is shown in the fourth circle. Information about long-fragment repeat sequences is represented in the fifth circle, and genomic long-fragment repeat sequences are indicated on the innermost circle.

Features of the A. locustae genome

Compared with the genome of E. cuniculi (Fig. S10), each A. locustae chromosome contains an ultra-low GC region of approximately 25 kb, except for chromosome 9, which has a contrasting ultra-high GC content (Fig. 2A). No coding genes were identified in any of the ultra-low GC regions. A scatter plot of GC content showed two similar Poisson distributions, with a small number of sequences exhibiting lower GC content (approximately 20 %) and the rest exhibiting normal (approximately 45 %) GC distribution (Fig. S11a,b), indicating that the genome of A. locustae may have two distinct forms of organization, which has not yet been observed in other microsporidian species (Fig. S12).

Fig. 2.

Distribution of GC content and variations in transcriptional density along each chromosome of A. locustae in the fat body and midgut of the host. (a) The distribution of GC content on each chromosome of A. locustae. (b) The density distribution of mRNA transcripts of A. locustae in the locust fat body. The ordinate shows the value of log₂ for the depth distribution of sequences on the chromosomes; the abscissa indicates the length of the chromosome. (c) The density distributions of transcripts of A. locustae in the midgut of the locust.

Analysis of repetitive sequences in the A. locustae genome resulted in the identification of a total of 298 simple tandem repeats. The number of low complexity repeats and microRNA was 86 and 45, respectively. Only one long terminal repeat was found in the whole genome, and no microsatellite DNA sequences were found (Table S8).

Transcriptional analysis of non-coding RNAs in the ultra-low GC regions showed a noticeable difference between the fat body and midgut (Fig. 2a). In the fat body, transcript levels in the ultra-low GC regions were relatively low (Fig. 2b), but the levels of the same DNA regions were significantly higher than those in the midgut for the majority of chromosomes (Fig. 2c). This finding indicates that non-coding RNA in the ultra-low GC content region is relatively highly expressed in the fat body. While the density of microsporidia in the midgut was very low after infection, that in the fat body was extremely high (Fig. 3d). As no coding genes were identified in the ultra-low GC regions, we speculated that these regions have an evolutionary effect that is currently unknown. Given that A. locustae carries out schizogamy in the host fat body, including cell division, we hypothesize that the ultra-low GC content regions may be associated with cell division and schizogamy.

Fig. 3.

Statistical analysis of transcripts. (a) The ratio of reads mapped to the genome of transcripts between diseased and healthy locusts. Comparison of the percentage of reads in midgut tissues (P=0.3604) and fat body tissues (P=0.0182) of diseased and healthy locusts to the locust genome. (b) Comparison of the percentage of the reads mapped to A. locustae genomes in the fat body and midgut tissues of diseased and healthy locusts (P=0.0142). (c) Principal component analysis (PCA) of A. locustae (left) and locust (right) transcripts. (d) A. locustae spore was detected in the midgut/fat body of the locust after inoculation. The t-test was used for determination of significance, *P<0.05.

Evolutionary analysis

Genomic and protein sequences of A. locustae were compared with those of several other single-cell organisms to assess genetic synteny and collinearity using the MCScanX method [55]. Organisms within the same taxonomic class generally showed markedly higher degrees of homology, while those in different classes showed little or almost no homology (Fig. 4a, b), and those in the microsporidian family showed greater homology (Fig. 4c). Within the microsporidia family, although genes from different host species showed high variability and some microsporidia exhibit gene loss and inversion, most genes in the microsporidian genomes still have good collinearity (Fig. 4d). Systematic analysis of frataxin, a key single-exon gene in the A. locustae mitosome, also found in other organisms such as fungi, prokaryotes, protozoa, etc. [59], provided an insight into the evolutionary status of the parasite, which is located at the base of the fungal evolutionary tree (Fig. 4e). In addition, the observation that single-exon genes occupy a predominant portion of the A. locustae genome is similar to observations in prokaryotic organisms (Fig. S13a, Table S3). This evidence suggested that A. locustae, and the microsporidian family in general, are more primitive eukaryotes.

Fig. 4.

Homology and collinearity analysis of genomes based on all predicted genomic protein sequences and a maximum-likelihood (ML) phylogenetic tree of the frataxin gene. (a) Homology and collinearity analysis between microsporidia and yeast. A. locustae with S. cerevisiae (GenBank ID: GCF_000146045.2) (left); S. cerevisiae with K. naganishii (GenBank ID: GCF_000348985.1) (right). (b) Homology and collinearity analysis between microsporidia and protozoa. A. locustae with B. bigemina (GenBank ID: GCF_000981445.1) (left); B. bigemina with B. motasi (GenBank ID: GCF_000691945.2) (right). (c) Homology and collinearity analysis between A. locustae and E. cuniculi (GenBank ID: GCF_000091225.1) in the microsporidia family (left). Homology and collinearity analysis between E. cuniculi and E. hellem (GenBank ID: GCF_000277815.2) in the family Encephalitozoon (right). (d) The order of genes from partial genomes among several representative species of microsporidia. Arrows of the same colour represent homologous genes and the direction of the arrow indicates the direction of the encoded protein. (e) An ML phylogenetic tree of frataxin was constructed with phyML 3.0 using the best model of amino acid substitutions as determined with mega 6.0 and the Regrafting (SPR) branch-swapping algorithm. Numbers indicate the corresponding levels of bootstrap support; values below 70 are hidden. Branch lengths are drawn to scale as noted below. Details: Trichuris trichiura (GenBank ID: CDW51770.1); Trichuris suis (GenBank ID: KHJ45876.1); Danio rerio (GenBank ID: NP_001076485.1); Mus musculus (GenBank ID: NP_032070.1); Homo sapiens (GenBank ID: NP_000135.2); Ochotona princeps (GenBank ID: XP_012784301.1); Apis cerana (GenBank ID: PBC30404.1); Pararge aegeria (GenBank ID: JAA83326.1); Culex quinquefasciatus (GenBank ID: XP_001864042.1); S. cerevisiae (GenBank ID: ONH78464.1); Roseateles depolymerans (GenBank ID: ALV05461.1); Oblitimonas alkaliphila (GenBank ID: AKX59165.1); Haemophilus influenza (GenBank ID: KIS34558.1); Buchnera aphidicola (Aphis glycines) (GenBank ID: ALD15510.1); Escherichia coli K-12 (GenBank ID: CQR83218.1); Spraguea lophii 42_110 (GenBank ID: EPR79861.1); A. locustae CLX; N. bombycis (GenBank ID: ABW91182.1); Nosema apis BRL 01 (GenBank ID: EQB60682.1); E. hellem ATCC 50504 (GenBank ID: XP_003886726.1); Encephalitozoon romaleae SJ-2008 (GenBank ID: XP_009263955.1); E. cuniculi GB-M1 (GenBank ID: XP_965969.1); B. bigemina (GenBank ID: XP_012769760.1); Cryptosporidium parvum Iowa II (GenBank ID: XP_625594.1); Leishmania major strain Friedlin (GenBank ID: XP_001683860.1); Leishmania infantum JPCM5 (GenBank ID: XP_001466138.1); Trypanosoma brucei (GenBank ID: AAX69885.1); Trypanosoma cruzi (GenBank ID: RNC58480.1); K. naganishii CBS 8797 (GenBank ID: XP_022462458.1); Eremothecium gossypii FDAG1 (GenBank ID: AEY99209.1).

Transcriptomic profiling of A. locustae in host tissues

We calculated the differential locust genes in healthy and diseased locusts, and also the biological process of differential genes in A. locustae and locusts. The results showed that the number of fat body and midgut transcripts varied greatly between diseased and healthy locusts; in general, after A. locustae infected the locust, there were significantly more differential transcripts for fat body mobilization than for the midgut. At the same time, the expression level of transcripts in the diseased locust showed a downward trend (Fig. S13b-d). As for the biological process of differential genes, we can see that it mainly participates in the metabolic process (Fig. S13e).

The transcriptome profiles of A. locustae differed significantly in the fat body and midgut during the middle and late stages after infection (Fig. 3a, b), suggesting that A. locustae is present in the locust and exercises different functions affecting the host in these two kinds of tissue. In addition, we found from principal component analysis (PCA) that the main components of the transcripts in the fat bodies of diseased and healthy locusts differed, while those in diseased and healthy locust midguts did not differ significantly (Fig. 3c). Thus, the transcriptome in the host’s fat body responded strongly, further illustrating the main site affected by A. locustae is the fat body. As a control, the differential gene expression of A. locustae in the midgut and fat body of healthy locusts (NF-VS-NM) was 0 (Fig. S13a).

Based on GO terms and KEGG and KOG pathway analysis (Table S9), the upregulated microsporidian genes in the midgut include LRR receptor-like protein kinase, adenylyl cyclase and a large variety of leucine-rich repeat units and MEIS1 transcription factors, which are mainly involved in activating parasite–host membrane surface signalling pathways, thereby promoting microsporidian invasion of the host through the midgut. However, we did not detect differences in the expression of microsporidian polar tube proteins in different locust tissues, suggesting that a high level of polar tube expression should occur in the intestinal tract outside the midgut. In addition, during the late infection stage, the load of microsporidia in the fat body was much larger than that in the midgut (Fig. 3d), indicating that the fat body is the site where A. locustae eventually multiplies. These findings provide useful information on the molecular relationships driving spore reproduction.

Interactions between A. locustae and its host

Analysis of transcripts from the midgut of diseased locusts showed that a huge number of host genes were activated in response to A. locustae infection as compared to the uninfected healthy midgut (Table S10). The microsporidian spore load in the midgut was controlled at a low level in the host midgut (Fig. 3d), and this control was correlated with abundant expression of antimicrobial peptides and other defence genes, such as peroxiredoxin and amine oxidase. Although it appeared that A. locustae could inhibit the melanization pathway, microsporidia can only survive temporarily in the midgut before being carried to fat body cells through vesicle transport (Fig. 5, Tables S9 and S10).

Fig. 5.

Simplified life cycle of infection by A. locustae and critical interactions with its locust host at the level of gene transcription.

After the pathogen entered the host fat body, the spore load in the fat body increased greatly compared to that of the midgut (Fig. 3d), and this change was correlated with obvious inhibition of melanization compared to that in the healthy fat body. In particular, several critical phenol oxidases and peroxisome proliferator-activated receptors in the locust were inhibited, reducing melanization in the locust and also enabling immune escape by the parasite, aiding its survival and proliferation (Fig. 5, Tables S9 and S11).

Discussion

The genome of A. locustae was sequenced by the Marine Biological Laboratory (USA) in 2002 with first-generation sequencing technology, obtaining approximately 648 contigs with a total size of approximately 2.1 Mb. However, few reports on the molecular biology of A. locustae were based on this sequence, and incompleteness of the dataset may have been an important limiting factor. In this study, an elaborate genome map of A. locustae was obtained through a combination of second- and third-generation sequencing technologies. A total of 17 chromosomes and 1857 genes were obtained, with a total size of approximately 3.2 Mb. The features of this microsporidium have been characterized. The 25 kb structure in the ultra-low GC region of the A. locustae genome was unique in that it is present in 16 of the 17 chromosomes sequenced, and this phenomenon has not been previously reported in other microsporidian species. We suspect that the sequence of the ultra-low GC region could represent the microsporidian centromere region, which is associated with cell division and schizogamy. Some centromere-related genes have been identified in the A. locustae genome, such as gene18 encoding the centromere-associated protein NUF2 and gene538 encoding the centromere-associated protein HEC1. These two proteins are involved in cell cycle control, cell division and chromosome partitioning. In addition, we found putative centromere/microtubule-binding protein 5 encoded by gene1155, which is homologous to the family Encephalitozoon, and centromere protein F encoded by gene1349, which is homologous to that of Propithecus coquereli. However, there are currently no reports on the centromere region of microsporidia related to proliferation, division, or the regulation of gene expression. Comparison with the midgut to determine whether the non-coding RNA encoded by the ultra-low GC content region of microsporidia in the fat body transcriptome is related to energy metabolism or immune evasion in the host–parasite interaction remains to be conducted.

The taxonomic status of microsporidia is a controversial topic. The level of conserved genes is close to that of fungi, representing either a basal branch or sister group [60]. Genomic and protein sequences of A. locustae were compared with those of several other unicellular organisms to identify genetic synteny and collinearity using the MCScanX method, and the results showed that A. locustae has high homology within the microsporidia group, while those in different classes showed little or essentially no homology. Within the microsporidia, despite large variation in genes among different host species and the presence of gene loss and inversion in some microsporidia, most genes in the microsporidian genome exhibit good collinearity. Additionally, our evolutionary analyses on the important mitosome gene frataxin showed that the microsporidia evolved side by side with fungi and prokaryotes, each as an independent group. Single-exon genes occupy a predominant portion of the A. locustae genome, similar to observations in prokaryotic organisms. These findings provide some evidence that A. locustae, and the microsporidian family in general, are more primitive eukaryotes.

Through a combination of genome and transcriptome sequencing, a striking picture of the intensive interactions between parasite and host has been revealed. A. locustae proliferates in the fat body, and turns glucose into pyruvate through a series of reactions in the mitosome residual of the mitochondria (Fig. S14). Due to the lack of related enzymes in the mitosome, pyruvate may undergo decarboxylation through the action of pyruvate dehydrogenase E1 component (PDH-E1) [2]. By comparing the transcriptome before and after infection of a host locust, we found that A. locustae increased the expression of trehalose-6-phosphate synthase in the sugar metabolism pathway after infection, indicating accelerated glucose metabolism, which may lay the foundation for evasion of host immunity and rapid reproduction. In addition, expression of RAB5 in A. locustae increased after infection (Fig. S14), indicating an increase in vesicle transport of parasite spores, similar to that of macromolecules [2]. The elevated levels of otsA and RAB5, involved in energy metabolism and material transportation, serve as a sign of intensive interactions between A. locustae and the locust.

This study found that A. locustae could systematically inhibit essential pathways in the locust, including phenoloxidase PPO, the TCA cycle, the glyoxylate cycle and PPAR pathways in the host fat body, making it difficult for the host to melanize the pathogen (Fig. 5). In addition, A. locustae activated the locust energy transfer pathway, which transports ATP produced by the locust to the microsporidium, and ADP produced by the microsporidium was transported back to locust cells for reuse in the synthesis of ATP, meaning that the microsporidia could use energy substances in the locust fat body for reproduction. A. locustae uses surface protein recognition to identify the membrane proteins of the midgut, constructs a polar tube to pierce the midgut cells and transports the cytoplasm into the midgut. The peroxisome in the locust was inhibited, and therefore A. locustae was not cleared by host cells. On the other hand, to combat microsporidian infection, the complement system and coagulation cascades are activated in the host, systematically inhibiting the proliferation of A. locustae in the midgut [61]. The expression level of cytochrome P450 in the midgut increased correspondingly, and some microsporidia were eliminated [62]. Increased expression of locust GSK3β was considered to be beneficial to the locust in its fight against A. locustae [63]. After entering the fat body, the MAPK pathway of the locust was inhibited, which reduced the immune level in the locust fat body [64].