MicroRNAs regulate innate immunity against uropathogenic and commensal-like Escherichia coli infections in the surrogate insect model Galleria mellonella

Uropathogenic Escherichia coli (UPEC) strains cause symptomatic urinary tract infections in humans whereas commensal-like E. coli strains in the urinary bladder cause long-term asymptomatic bacteriuria (ABU). We previously reported that UPEC and ABU strains differentially regulate key DNA methylation and histone acetylation components in the surrogate insect host Galleria mellonella to epigenetically modulate innate immunity-related gene expression, which in turn controls bacterial growth. In this follow-up study, we infected G. mellonella larvae with UPEC strain CFT073 or ABU strain 83972 to identify differences in the expression of microRNAs (miRNAs), a class of non-coding RNAs that regulate gene expression at the post-transcriptional level. Our small RNA sequencing analysis showed that UPEC and ABU infections caused significant changes in the abundance of miRNAs in the larvae, and highlighted the differential expression of 147 conserved miRNAs and 95 novel miRNA candidates. We annotated the G. mellonella genome sequence to investigate the miRNA-regulated expression of genes encoding antimicrobial peptides, signaling proteins, and enzymatic regulators of DNA methylation and histone acetylation in infected larvae. Our results indicate that miRNAs play a role in the epigenetic reprograming of innate immunity in G. mellonella larvae to distinguish between pathogenic and commensal strains of E. coli.

maintaining pathogen specificity [10][11][12] . Additionally, the secreted TIR domain homologue TcpC is expressed by many UPEC strains and inhibits MyD88 and inflammasome activation 13,14 . ABU strains have also been shown to modulate host gene expression by suppressing RNA polymerase II 7,9 . Surprisingly, the discriminatory host response is not restricted to humans and also occurs in the greater wax moth Galleria mellonella, which has been established as a surrogate insect model host to study human pathogens, including UPEC [15][16][17][18][19][20][21][22][23] . The infection of G. mellonella larvae with UPEC strain CFT073 or ABU strain 83972 at 37 °C resulted in the differential expression of genes encoding TLRs, cytokine-like proteins and antimicrobial peptides (AMPs) 23 . In eukaryotes, gene expression is regulated by epigenetic mechanisms resulting in heritable phenotypes without mutation. We previously found that DNA methylation and histone acetylation were differentially regulated in larvae infected with UPEC and ABU strains, underpinning the reprogramming of innate immunity at the level of transcriptional initiation 23 .
In this follow-up study, we investigated the role of microRNAs (miRNAs), which have the potential to regulate innate immunity at the post-transcriptional level. These non-coding RNAs are 18-24 nucleotides long and are conserved in most eukaryotes. They bind to the 3′ and 5′ untranslated regions (UTRs) of target messenger RNAs (mRNAs), causing translational repression and mRNA decay 24 . They play important role in various infectious diseases, and facilitate the immune response to bacterial infection in insects [25][26][27] . We have constructed microarrays to analyze the expression of conserved miRNAs in G. mellonella larvae during infection with the entomopathogenic bacterium Bacillus thuringiensis and the entomopathogenic fungus Metarhizium robertsii 28,29 . Here, we carried out miRNA sequencing in G. mellonella larvae infected with UPEC strain CFT073 or ABU strain 83972 to investigate strain-dependent expression of novel and conserved miRNAs, to identify the mRNA targets of these miRNAs, and to analyze co-expression of miRNAs and their mRNA targets in infected larvae.

Results
Small RNA deep sequencing of G. mellonella larvae infected with UPEC/ABU strains. The miR-NAs expressed in G. mellonella in response to UPEC and ABU infections were identified by high-throughput sequencing of whole-larvae samples 24 h after infection. The number of raw sequence reads was 62,511,810 for larvae infected with the UPEC strain (hereafter described as UPEC larvae), 53,675,182 for larvae infected with the ABU strain (hereafter described as ABU larvae) and 75,401,198 for the mock-injected controls. The size distribution of the trimmed, high-quality reads ranged from 17 to 30 nucleotides (Table 1) with a peak at 22 nucleotides. We identified 141 unique precursor hairpins and 257 unique mature miRNA sequences, with the greatest number detected in the UPEC larvae, followed by the control larvae and finally the ABU larvae (Table 1). Among the 257 mature miRNAs, 95 appeared to be novel, 148 were conserved and 14 included up to three mismatching nucleotides (Table S1). expression analysis of miRnAs in G. mellonella larvae infected with UPEC/ABU strains. We were able to classify 213 of the 257 mature miRNAs based on their comparative expression profiles in the UPEC/ ABU larvae and uninfected controls. We found that 147 of the miRNAs were expressed consistently in all three groups, but 26 miRNAs were modulated in a single group and 40 miRNAs showed differential expression in pairwise comparisons (Fig. 1, Table S2). Specifically, we identified 5, 18 and 3 miRNAs that were specifically modulated in ABU larvae, UPEC larvae or controls, respectively, and we identified 3, 19 and 18 miRNAs that showed differential expression between ABU larvae and controls, between UPEC larvae and controls, and between UPEC and ABU larvae, respectively.

Discussion
Host susceptibility and innate immunity-related gene expression help determine the outcome of infections caused by UPEC and ABU strains in the urinary tract. In this study, we used the surrogate insect host G. mellonella to show that the differential innate immune response to UPEC and ABU infections is regulated at least in part by non-coding miRNAs. Following miRNA sequencing in ABU and UPEC larvae, we analyzed the strain-dependent expression of novel and conserved miRNAs, and using our microPIECE pipeline based on AGO-CLIP reference datasets from other insect species we predicted miRNA targets 30 . We found that ABU and UPEC infections can trigger the expression of novel and conserved miRNAs to modulate the expression of innate immunity-related www.nature.com/scientificreports www.nature.com/scientificreports/ genes (Figs. 3, S1-S3). ABU infection induces miRNAs that suppress genes related to cell signaling and innate immunity. For example, the expression of gme-new-160-5p inhibits LITAF like immunity related proteins in larvae and reduces the lipopolysaccharide-induced innate immune activation, resulting in improved host survival 32,33 . The post-transcriptional suppression of LITAF may thus favor the long-term survival of ABU in larvae by attenuating the immune response, whereas the expression of LITAF in UPEC larvae encourages strong immune response 23 . Inhibition of LITAF provides resistance to systemic E. coli LPS-induced lethality in mammals. Thus, LITAF is a promising therapeutic target for the treatment of TNF-mediated inflammatory diseases, and we identified a miRNA in G. mellonella that inhibits its expression 33 . ABU and UPEC infections also modulate miRNAs that affect the production of AMPs. For example, UPEC infections induce the expression of gme-138-3p, gme-new-4-5p and gme-new-135-3p while suppressing gme-new-70-3p and gme-new-161-3p, which has the net effect of increasing the synthesis of the most potent AMPs (lysozymes, cecropins, gloverin, galiomycin and moricin), whereas weaker AMPs (anionic peptides, apolipophoricin) are induced following ABU infection 23 . ABU E. coli strain 83972 differs from UPEC strain CFT073 in terms of virulence gene expression 6 . Colonization and long-term survival of ABU E. coli strain 83972 in G. mellonella larvae and most likely also in the human urinary tract is achieved by miRNA-mediated suppression of gene expression constituting strong antimicrobial response. On the other hand, UPEC infection provokes expression of these AMPs by downregulating miRNAs that specifically inhibit their expression. The expression of miRNAs targeting AMP synthesis has also been shown following infection in the insect host Plutella xylostella, and in some human diseases 34,35 . We used the G. mellonella system to discover new miRNAs that target innate immune related proteins important for UPEC infection implicating scope for therapeutic application to treat UTI in humans.  www.nature.com/scientificreports www.nature.com/scientificreports/ Several human pathogens can manipulate host cell antimicrobial responses and evade the immune system by influencing epigenetic mechanisms such as histone acetylation and DNA/RNA methylation 36 . Qualitative and quantitative differences in the responses to UPEC and ABU strains in G. mellonella are also epigenetically regulated in this manner 23 . While many target genes of miRNAs are known, even less information exists as to how miRNAs cooperate with histone acetylation and DNA methylation in the context of host-pathogen interaction. There is limited evidence suggesting that negative correlation between the expression of miRNAs and HDACs or HATs has been associated with infection by human pathogens 37 . Interestingly, our analysis of miRNA targets revealed that both novel and conserved miRNAs can target genes encoding methyltransferases, HATs and HDACs, which are key regulators of DNA methylation and histone modifications. In UPEC larvae, the induction of miRNAs gme-new-106-5p, gme-new-184-5p and gme-new-13a-3p correlated with the downregulation of target genes encoding HATs, which acetylate histones to form open chromatin that favors the expression of immunity-relate genes. In mammals, HATs are degraded by a zinc-dependent metalloproteinase from enteropathogenic and enterohemorrhagic E. coli in order to dampen inflammatory responses 38,39 . We suggest that UPEC also follow an alternative strategy of miRNA-mediated suppression of HAT expression in an infected host. The activity of HATs is also opposed by HDACs, which deacetylate histones and form condensed chromatin that suppresses gene expression. The HDAC sap18 subunit was downregulated in ABU larvae but upregulated in UPEC larvae and here we identified a miRNA (gme-new-134-3p) that targets the mRNA encoding this protein subunit 21 . The induction of this novel miRNA in ABU larvae suggests a post-transcriptional mechanism for the suppression of HDAC sap18. The novel miRNA gme-new-106-5p was upregulated in UPEC larvae, and this targets the mRNA for S-adenosyl-L-methionine-dependent methyltransferase, which is a methyl donor for DNA methyltransferases. We identified miRNAs that target methyltransferase mRNAs in UPEC and ABU larvae, but surprisingly none of them were the DNA methyltransferases responsible for epigenetic regulation. However, insect genomes are sparsely methylated compared to mammals, and the role of DNA methylation in innate immunity is not well understood. The lack of miRNAs targeting mRNAs encoding maintenance or de novo methyltransferases in UPEC and ABU larvae may indicate the limited significance of DNA methyltransferases in the regulation of innate immunity in G. mellonella 40 .
In addition to the strain-specific regulation of miRNAs, we also identified a large number of miRNA candidates that were commonly modulated in UPEC and ABU larvae. The majority of these miRNAs (such as mir-8 family) are conserved among other eukaryotes, for example let-7 and mir-124 have homologous targets in G. mellonella and humans. Gme-let-7-5p targets the HAT KAT2A, and both gme-miR-283-5p and . Differential expression of selected miRNAs and target mRNAs by RT-PCR in G. mellonella larvae infected with ABU and UPEC strains, and in mock-injected controls. The novel miRNA sequences were obtained from small RNA sequencing and their predicted mRNA targets were validated by RT-PCR to confirm differential expression: (A) gme-new-160-5p, gme-new-106-5p, gme-new-147-3p; (B) gene 9, gene 10, gene 11. The relative fold differences indicated for the miRNAs and mRNAs are normalized against gme-miR-133 and elongation factor 1, respectively, as the internal reference controls (****p < 0.0001, *p < 0.05).

conclusion
Here we show that UPEC strain CFT073 and ABU strain 83972 trigger the modulation of host miRNAs in G. mellonella to epigenetically regulate the innate immune response. Many novel miRNAs showed strain-dependent expression in the infected larvae, whereas others were modulated similarly regardless of the infection status or in a similar manner during both infections. We argue that miRNAs determine the different pathogenic potential of the ABU and UPEC strains in G. mellonella, and may regulate different behavior of these strains in the human urinary tract. Taken together, our results emphasize the importance of G. mellonella miRNAs in the regulation of host innate immunity to distinguish between pathogenic and commensal-like E. coli strains.

Methods
Bacterial strains, insects, and culture media. Cultures of UPEC strain CFT073 and ABU strain 83972 were maintained aerobically in lysogeny broth (LB) at 37 °C and on LB agar plates (Carl Roth, Karlsruhe, Germany). For long-term storage, bacteria were frozen at −80 °C in LB supplemented with 30% (v/v) glycerol. G. mellonella larvae were obtained from Fauna Topics Zoobedarf Zucht und Handels GmbH, Marbach am Neckar, Germany. The larvae were reared on an artificial diet (22% maize meal, 22% wheat germ, 11% dry yeast, 17.5% beeswax, 11% honey and 11% glycerin) at 32 °C in darkness. Larvae at their sixth instar stage, each weighing 250-350 mg, were used in all experiments.

G. mellonella injection.
Injection experiments were carried out using logarithmic growth-phase bacterial cultures in 10 ml LB. The bacteria were washed and serially diluted in 0.9% NaCl, and 10-μl aliquots (10 5 colony forming units/ml) were injected into larvae through the left proleg using 1-ml disposable syringes and 20-mm 0.4 gauge needles mounted on a microapplicator, as previously described 42 . Mock injections with an empty needle were carried out as controls. Larvae were considered dead after incubation at 37 °C when they showed no movement in response to touch.
Small RNA isolation, library construction, sequencing and analysis. We used the following criteria to design experiments and selected miRNA sequencing techniques, to minimize sequencing error rates and limitations for not performing cost-intensive multiple miRNA sequencing experiments. First, considering the importance of biological variations in the identification of conserved and highly expressed miRNA, total RNA samples were extracted using the miRNeasy Mini Kit (Qiagen, Hilden, Germany) and pooled from three independent biological experiments comprising five larvae per experiment. Sequencing of miRNAs was performed by LC Sciences (Houston, Texas, USA). Briefly, total RNA was analyzed using a Nandrop spectrophotometer (Thermo Fischer Scientific, Waltham, Massachusetts, USA) and a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, California, USA) and samples with an RNA integrity number (RIN) > 7 as well as 260/280 and 260/230 absorbance ratios > 2.0 were used for cDNA library preparation based on the TruSeq Small RNA Sample Preparation Protocol (Illumina, San Diego, California, USA). Next, the cDNA libraries were purified and sequenced on an Illumina HiSeq. 2500 in the 50-bp single-end configuration, which has the least error rate (below 0.1%) compared to other sequencing techniques. G. mellonella miRNAs were identified using miRDeep2 43 v2.0.0.8 to screen the recently published G. mellonella genome sequence against miRbase 44 release 22. We designated miRNAs from all animals including related insect species Bombyx mori, Manduca sexta, Plutella xylostella, Spodoptera frugiperda (order Lepidoptera), Aedes aegypti, Anopheles gambiae, Tribolium castaneum, Drosophila melanogaster (order Diptera), Apis mellifera, Nasonia vitripennis (order Hymenoptera), and Acyrthosiphon pisum (order Homoptera) for the purpose of miRNAs identification. The miRNA hits were temporarily numbered in ascending order and identical mature sequences were merged to reduce redundancy. The miRNAs with up to three nucleotide mismatches were renamed according to the best match from miRBase. Next, to reduce false positives, we discarded all reads that mapped to rRNAs from further analysis. Briefly, reads with a minimal length of 17 nucleotides were trimmed using cutadapt v1.8.3 and rRNA sequences were filtered out by screening against the Silva rRNA database 45,46 . Mature miRNAs were mapped using bwa v0.7.10-r789 and we calculated the expression level in reads per million (RPM) 46,47 . We used Deseq v1.32.0 to compute an indicative fold change between samples 47,48 and the stem loop structures of 37 novel miRNAs were determined (Table S4). For the expression calculation, we allowed one measured miRNA read, because adapter ligations in library preparation steps, as well as unwanted intra-and inter-RNA bindings of miRNAs can disfavor certain mature miRNAs and may lead to underrepresented sequencing results. We considered results with a fold change of ≤0.5 and ≥1.5 as differentially expressed between two groups of larvae.
Annotation of G. mellonella transcriptome and miRNA target prediction. For miRNA target prediction, the publicly available paired-end RNA sequencing libraries from UPEC and ABU infected G. mellonella larvae (SRX2727976 and SRX2727977) were compared with non-infected larvae (SRX371340). We then aligned the reads from the transcriptome sequencing against the recently sequenced G. mellonella genome (ASM258982v1) 31 using HiSat2 49 v2.1.0 and transformed and sorted the results using samtools 49 v1.6. Assembly was carried out using cufflinks v2.2.1 with standard parameters 50 and TransDecoder v5.0.2 analysis for the annotation of different isoforms 51 . The analysis was enhanced with a BlastP v2.6.0 + search against SwissProt 52 and a hmmer 53 search against Pfam-A with standard parameters 53 . The assembly was evaluated using rnaQUAST 54,55 . We then annotated the transcripts using InterproScan v5.27-66.0.
Targets were predicted using the microPIECE pipeline 30 , which uses AGO-CLIP libraries from other species and performs a trimming, mapping and peak calling on the datasets, like a typical CLIP analysis workflow. Briefly, the microPIECE takes the AGO-CLIP data from a species A and transfers it to a species B. That means it Scientific RepoRtS | (2020) 10:2570 | https://doi.org/10.1038/s41598-020-59407-3 www.nature.com/scientificreports www.nature.com/scientificreports/ transfer the verified miRNA binding region of a mRNA to the homologous transcript of another species. Given a set of miRNAs from species B the pipeline then predicts their targets on the transferred CLIP regions with a much lower false-positive rate, since it relies on evolutionary conserved and experimentally verified binding regions. Identification of orthologous proteins and mapping of the signaling regions of miRNA binding sites to the orthologous transcripts in the species of interest were computed and used for target predictions with miRanda v3.3a. The transfer of known miRNA binding sites was achieved using a set of six publicly available A. aegypti AGO-CLIP libraries, which is the only AGO-CLIP dataset available in the clade of insects 48,56 . From those six libraries, there are three replicates for two conditions. The resulting mRNA target sites that we have identified by using A. aegypti as AGO-CLIP data donor for G. mellonella are expected to be relatively small than a more closely related insect species, but this technique ensures that at least those target sites are already present between A. aegypti and G. mellonella. In addition to the miRanda tool, we used the prediction algorithms RNAhybrid and RNA22 v2 to validate the miRNA target prediction by microPIECE. We followed union over intersection when combining several tools for miRNA-mRNA predictions 57 .

Data availability
The small RNA sequencing data is available at NCBI via the accession number GSE123965. All data are accessible in the Supplementary Information.