Paratransgenic manipulation of a tsetse microRNA alters the physiological homeostasis of the fly’s midgut environment

Tsetse flies are vectors of parasitic African trypanosomes, the etiological agents of human and animal African trypanosomoses. Current disease control methods include fly-repelling pesticides, fly trapping, and chemotherapeutic treatment of infected people and animals. Inhibiting tsetse’s ability to transmit trypanosomes by strengthening the fly’s natural barriers can serve as an alternative approach to reduce disease. The peritrophic matrix (PM) is a chitinous and proteinaceous barrier that lines the insect midgut and serves as a protective barrier that inhibits infection with pathogens. African trypanosomes must cross tsetse’s PM in order to establish an infection in the fly, and PM structural integrity negatively correlates with trypanosome infection outcomes. Bloodstream form trypanosomes shed variant surface glycoproteins (VSG) into tsetse’s gut lumen early during the infection establishment, and free VSG molecules are internalized by the fly’s PM-producing cardia. This process results in a reduction in the expression of a tsetse microRNA (miR275) and a sequential molecular cascade that compromises PM integrity. miRNAs are small non-coding RNAs that are critical in regulating many physiological processes. In the present study, we investigated the role(s) of tsetse miR275 by developing a paratransgenic expression system that employs tsetse’s facultative bacterial endosymbiont, Sodalis glossinidius, to express tandem antagomir-275 repeats (or miR275 sponges). This system induces a constitutive, 40% reduction in miR275 transcript abundance in the fly’s midgut and results in obstructed blood digestion (gut weights increased by 52%), a significant increase (p-value < 0.0001) in fly survival following infection with an entomopathogenic bacteria, and a 78% increase in trypanosome infection prevalence. RNA sequencing of cardia and midgut tissues from paratransgenic tsetse confirmed that miR275 regulates processes related to the expression of PM-associated proteins and digestive enzymes as well as genes that encode abundant secretory proteins. Our study demonstrates that paratransgenesis can be employed to study microRNA regulated pathways in arthropods that house symbiotic bacteria.


Introduction
Tsetse flies (Glossina spp.) are obligate vectors of pathogenic African trypanosomes throughout 37 countries in sub-Saharan Africa [1]. These parasites belong to the genus of Trypanosoma, which includes the important human pathogens T. b. rhodesiense and T. b. gambiense (the causative agents of human African trypanosomiasis, or HAT) and non-human animal pathogens T. b. brucei, T. congolense and T. vivax (the causative agents of animal African trypanosomiasis, or AAT). Both AAT and HAT are fatal if left untreated [2]. Current disease control methods include vector control to reduce population size and chemotherapeutic treatment of infected people and domesticated animals [3]. A more complete molecular understanding of tsetse-trypanosome interactions will facilitate the development of novel control strategies, such as reducing or eliminating the fly's capacity to transmit trypanosomes.
The tsetse-specific stages of the trypanosome life cycle begin when the fly ingests a bloodmeal that contains mammalian stage bloodstream form (BSF) parasites. Upon ingestion by tsetse, BSF parasites differentiate into insect adapted procyclic forms (PCF) in the lumen of the fly's midgut [4,5]. PCF parasites then bypass the fly's peritrophic matrix (PM) barrier in the anterior midgut and replicate within the ectoperitrophic space (ES, the region between the PM and the midgut epithelia) [6][7][8]. As part of their development from BSF to PCF parasites, the BSF trypanosomes shed their abundant surface coat antigens, known as variant surface glycoprotein (VSG) into the fly's midgut lumen. Free VSG is transiently internalized by cells of tsetse's PM-producing cardia (also known as proventriculus) [9,10]. This process reduces the expression of genes that encode PM associated proteins and digestive enzymes, and modulates the expression of several microRNAs, including a drastic reduction in the expression of tsetse microRNA 275 (miR275) [10].
miRNAs are small (~23 nucleotides) non-coding RNAs that regulate many important physiological processes. miRNAs often suppress gene expression by guiding the Argonaute (AGO) protein to bind with its target mRNA, which induces the miRNA induced silencing complex (miRISC) and leads to post-transcriptional repression or degradation of the target mRNA [11][12][13]. miRNAs can also upregulate gene expression by inducing translational activation [14,15]. When the expression of miR275 was experimentally reduced in tsetse's cardia and midgut through the provisioning of synthetic anti-miR275 antagomirs (antagomir-275) or VSG purified from BSF trypanosomes, formation of the fly's PM was impaired. This process disrupted blood meal digestion and enhanced the ability of trypanosomes to establish an infection in the fly's midgut [10]. In the mosquito Aedes aegypti, miR275 similarly influences midgut blood digestion and fluid excretion by regulating the expression of its target gene SERCA (sarco/ endoplasmic reticulum Ca2+ adenosine triphosphatase) [16,17], but the mRNA target of miR275 in tsetse remains unknown.
Tsetse flies house a consortium of symbiotic microbes that mediate numerous aspects of their host's physiology [18,19]. One of these is the facultative endosymbiotic bacterium Sodalis glossinidius, which resides extra-and intracellularly within multiple tsetse tissues, including the midgut, salivary glands, and reproductive organs [20]. Sodalis can be cultivated and genetically modified in vitro, and recolonized into tsetse's gut via a blood meal [21,22]. Reintroducing recombinant Sodalis (recSodalis) does not elicit immune responses that would induce any fitness cost [22,23]. Per os provisioned recSodalis remains only in the gut [22]. 'Paratransgenic' tsetse flies that house recSodalis have been successfully used to deliver anti-trypanosomal nanobodies [24][25][26]. Paratransgenesis has also been used to deliver dsRNA for gene silencing in kissing bugs [27,28] and in the malaria mosquito Anopheles gambiae [29,30], and to express intronic miRNA (using a recombinant densovirus) in the mosquito Aedes albopictus [31]. However, paratransgenic expression of small RNA antagomirs to knockdown miRNA expression in tsetse fly has not been reported to date. Herein we engineered Sodalis to paratransgenically express three tandem antagomir-275 repeats (3xant-miR275) in tsetse's cardia and midgut environments, and then used this experimental system to investigate the mechanism(s) by which miR275 regulates the physiological homeostasis of the fly's gut environment. We found that paratransgenic flies presented multiple phenotypes that are associated with the production of a structurally compromised PM barrier and/or disrupted gut homeostasis. Our novel paratransgenic expression system can be applied to further study functions of microRNAs that are involved in the tsetse-trypanosome interaction, thus advancing our understanding of parasitedeployed strategies to manipulates its host physiology. Additionally, this method could be broadly applied to other arthropod systems where a host interacts with microbes (especially with non-model systems where host genetic manipulation can be difficult), which could be particularly useful to study pathogen-host interactions in the field of vector biology.

Tsetse fly and bacterial cultures
Tsetse flies (Glossina morsitans morsitans, Gmm) were reared in the Yale University insectary at 25˚C and 70% relative humidity (RH), and received defibrinated bovine blood every 48 h via an artificial blood feeding system. Wild-type Sodalis glossinidius morsitans were isolated from surface-sterilized Gmm pupae and plated on Difco Brain Heart Infusion Agar (BD Biosciences) plates that were supplemented with 10% bovine blood (BBHI). Clonal Sodalis populations were subsequently maintained in vitro in Bacto Brain Heart Infusion (BHI) medium (BD biosciences) at 26˚C, 10% CO 2 .

Generation of recSodalis strains
To generate recSodalis, two constructs ( Fig 1A) were made using a modified pgRNA-bacteria plasmid (NEB, Addgene plasmid # 44251). This plasmid, which encodes an ampicillin resistance cassette, was originally designed to express short guide RNAs for CRISPR application and is thus well suited for expressing small RNAs [32]. An additional endonuclease cut site Sbfi was built into the original pgRNA plasmid backbone so as to include an RNA terminator sequence in the modified plasmid. Two pairs of two complementary single-stranded oligonucleotides (oligos) that encode either three copies of the miR275 antagomir (3xant-miR275) or the scrambled miR275 control (Scr-275) were synthesized at Yale Keck Oligo Synthesis Resource ( Table 1). The two antisense single-stranded oligos each contains a bulge (not perfectly complementary) at the linker sequence to increase the effectivity of the sponge construct [33]. Each strand encodes SpeI and Sbfi restriction endonuclease cut sites, were annealed at 95˚C for 5 min, cooled to room temperature for 30 min and stored at -20˚C for future use. Both pgRNA and the double stranded miRNA-encoding oligos were subjected to restriction endonuclease treatment by SpeI and Sbfi at 37˚C for 2 h. The oligos were then ligated into pgRNA using T4 DNA ligase (NEB), and the constructs were propagated in E. coli DH5a cells. All purified plasmid constructs were sequenced at Yale's Keck Sequencing Laboratory to confirm their structure.

Establishment of paratransgenic tsetse flies
To generate paratransgenic tsetse flies, two groups of teneral female flies (newly emerged unfed adults) were given two consecutive blood meals (separated by 1 day) containing either Sgm 3xant-miR275 or Sgm Scr-275 (10 6 CFU/mL each in the first two blood meals) and ampicillin (50 μg/mL). After a third blood meal (no recSodalis, no ampicillin), 8-day old paratransgenic flies were used in the experiments described below.

Gentamicin exclusion assay and quantification of recSodalis
Gentamicin is unable to cross the eukaryotic cell membrane and hence only kills extracellular bacteria [35]. Cardia and midgut tissues were dissected from 8-day old paratransgenic and incubated in sterile 0.85% NaCl supplemented with 100 μg/mL gentamicin. Controls were incubated in the sterile NaCl in the absence of gentamicin. Tissues were agitated on a shaking platform at room temperature for 1 h and washed 4 times in 500 μl sterile 0.85% NaCl. After the 4 th wash, tissues were rigorously homogenized in sterile 0.85% NaCl. 50 μl of lysate from each treatment was plated onto BHI Agar plates supplemented with 10% blood and 50 μg/mL ampicillin. After 7 days of incubation at 26˚C, 10% CO 2 , colonies on each plate were counted as described in [22]. Multiple colonies were randomly selected for colony PCR (with primers targeting the inserted section of the pgRNA plasmid) and subjected to sequencing to confirm they housed the correct plasmid construct.

Dual luciferase reporter assay
To clone the 3xant-miR-275 into psiCheck-2 (Promega), two complementary single-stranded oligos that encode 3xant-miR-275 and XhoI and NotI restriction endonuclease cut sites were synthesized at Yale Keck Oligo Synthesis Resource ( Table 1). The complementary oligos were annealed at 95˚C for 4 min and cooled to room temperature for 30 min. The psiCheck-2 vector and the doubled stranded miRNA-encoding oligos were subjected to XhoI and NotI treatment at 37˚C for 2 h followed by inactivation at 65˚C. The oligos were then ligated into the double digested psiCheck-2 plasmid using T4 DNA ligase (NEB) at room temperature for 2 h, and the constructs were propagated in E. coli DH5a cells. All purified plasmid constructs were sequenced at Yale's Keck Sequencing Laboratory to confirm their structure. The psiCheck-2 vector containing the 3xant-miR275 sequence is hereafter referred to as psiCheck-2 3xant-miR275 . For transfection, Drosophila S2 cells (Invitrogen) were maintained at 28˚C in Schneider Drosophila medium supplemented with 10% heat inactivated FBS. We co-transfected 100 ng of psiCheck-2 3xant-miR275 and the synthetic tsetse miR275miScript miRNA mimic at 100 nM (Qiagen) or with AllStars Negative Control (Qiagen) into S2 cell lines using Attractene Transfection reagent following the manufacturer's protocol (Qiagen). A "no miRNA" treatment with only psiCheck-2 plasmid and transfection reagent was also conducted. Dual luciferase reporter assays were performed 48 h post transfection using the Dual Luciferase Reporter Assay System following the manufacturer's protocol (Promega). The renilla (primary reporter) luciferase signal was normalized to the firefly (internal control) luciferase signal. Each treatment was conducted triplicate.

Quantitative real-time PCR
Quantitative real-time PCR (qPCR) was used to quantify the expression levels of miR275, noncoding small nuclear RNA (snRNA) U6, and saliva-associated genes in our paratransgenic flies (described in section 2.3 above). Tsetse cardia, midgut and salivary glands were microscopically dissected 24-48 h after the third blood meal. Total RNA was extracted from pools of 5 cardia, 5 midgut or 10 salivary glands (as one biological replicate) using Trizol reagent [36]. RNA was cleaned and purified using an RNA Clean and Concentrator Kit with in-column DNase treatment (Zymo Research). RNA quality and quantity was quantified using a NanoDrop 2000c (Thermo Scientific). 100 ng of cardia, 500 ng of midgut or 100 ng of salivary glands RNA, respectively, was then reverse transcribed into cDNA using the miScript II RT kit (Qiagen 218160) followed by qPCR. For each sample, two technical replicates were used. Relative expression (RE) was measured as RE = 2 -ddCT , and normalization was performed using U6 gene expression as a reference. Primers for amplifying miR275, saliva-associated genes and the reference gene are listed in S1 Table. qPCR was performed on a CFX96

Tsetse whole gut weight measurements
Individual guts from 8-day old paratransgenic flies (n = 20 per group) were dissected 24 h after their last blood meal and weighed with a digital scale as an indicator for blood digestion.

Serratia infection assay
8-day old paratransgenic individuals were fed a blood meal containing 10 3 CFU/mL S. marcescens strain Db11. Thereafter, all flies were maintained on normal blood and their mortality was recorded every other day for 14 days. Details of the Serratia infection assay are provided in [6,9,10] as well as in the discussion section.

Trypanosome infection prevalence
The 8-day old paratransgenic flies were challenged per os with a blood meal containing 10 7 CFU/mL Trypanosoma brucei brucei strain 503 supplemented with 0.9 mg/mL of cysteine. Thereafter, the flies were maintained on normal blood meals for two weeks. Their guts were dissected and microscopically examined to determine trypanosome infection status.

mRNA library construction and RNA sequencing
Two groups of paratransgenic flies (Gmm 3xant-miR275 vs. Gmm Scr-275 ) were generated as described in Section 2.3. All flies were dissected 36 h after the third blood meal; 10 individual cardia or 5 individual midgut were pooled as one biological replicate and stored in -80˚C prior to RNA extraction, a total 3 biological replicates per treatment were used. Total RNA was extracted using Trizol reagent according to the manufacturer's protocol (Invitrogen), followed by RNA Clean and Concentrator Kit and in-column DNase treatment (Zymo Research). RNA quality and quantity were quantified using a bioanalyzer. All 12 mRNA libraries were prepared and sequenced (pair-ended) at Yale Center for Genome Analysis (YCGA) using Illumina NovaSeq system.

RNA-seq data processing
RNA-seq raw reads were uploaded to FastQC (v. 0.11.9, www.bioinformatics.babraham.ac.uk/ projects/) for quality check, and then trimmed and filtered to remove ambiguous nucleotides and low-quality sequences. The reads were mapped to Glossina morsitans morsitans reference genome [37] using HISAT2 v2.1.0 with default parameters [38,39]. We then used the function 'htseq-count' in HTSeq v0.11.2 [40] to count the number of reads mapped to the genes annotated in the reference genome (version GmorY1.9 at Vectorbase.org) with option "-s reverse". Reads that were uniquely aligned to Gmm transcripts were used to calculate differential gene expression using EdgeR package in R software [41]. Significance was determined using EdgeR General linear models, corrected with a False Discovery Rate (FDR) at p < 0.05. The differentially expressed (DE) genes were uploaded to VectorBase (http://beta.vectorbase.org) for gene ontology (GO) enrichment analysis using the built-in web tool GO Enrichment analysis. REVIGO was used to remove the redundant GO terms [42].

Replicates and statistics
Biological replicates were obtained from samples derived from distinctly repeated experiments. Details about sample sizes and statistical tests used for data analyses in this study are indicated in the corresponding figure legends.

Successfully developed the paratransgenic expression system
To knock down expression of tsetse miR275, we designed two expression constructs that encode 1) 3xant-miR275 to knockdown miR275, and 2) a scrambled miR275 sequence (Scr-275) that served as the control. Individual clonal populations of wild-type Sodalis (Sgm WT ) were transformed with one of the plasmids and are henceforth designated Sgm 3xant-miR275 and Sgm Scr-275 ( Fig 1A). We then colonized individual groups of newly emerged (teneral) adult tsetse per os with either Sgm 3xant-miR275 or Sgm Scr-275 , thus generating paratransgenic tsetse cohorts designated Gmm 3xant-miR275 (treatment) and Gmm Scr-275 (control), respectively. During the development of the paratransgenic lines, we supplemented the first two bloodmeals with ampicillin to suppress the Sgm WT population, which provided the antibiotic-resistant recSodalis populations a selective advantage over the indigenous antibiotic susceptible WT cells.
We performed gentamicin exclusion assays to confirm that the recSodalis successfully invaded tsetse cardia and midgut cells. Gentamicin cannot penetrate eukaryotic cell membranes, and thus treatment with this antibiotic effectively eliminates the extracellular bacteria but leaves the intracellular population intact [35]. We incubated separate cardia and midgut tissues dissected from 8-day old paratransgenic flies in either gentamicin (treatment) or PBS (control). Tissues were subsequently rinsed, homogenized, and plated on BBHI plates supplemented with ampicillin. We recovered 214 (± 54.0) and 9.7x10 5 (± 9.6x10 4 ) gentamicinresistant CFU from the cardia and midgut tissues, respectively ( Fig 1B). Sequencing of the transformation plasmid from several bacterial clones confirmed their identity as either Sgm 3xant-miR275 or Sgm Scr-275 . These findings indicate that recSodalis was successfully internalized by tsetse cardia and midgut cells where they were protected from the antibacterial effects of gentamicin. Additionally, significantly more recSodalis cells were present within midgut cells than cells of the cardia organ. We similarly quantified the Sgm 3xant-miR275 and Sgm Scr-275 present in the no gentamicin control groups (cardia, 684 ± 90, p = 0.002; midgut, 2.0x10 6 ± 1.1x10 5 , p < 0.0001) (Fig 1B), and found that 31% and 49% of recSodalis present in the gut were intracellular within cardia and midgut tissues, respectively. These data also indicated that our recSodalis successfully reside within tsetse's gut at a density similar to that of indigenous Sgm WT in age-matched flies [22]. Thus, we demonstrated that recSodalis successfully colonized tsetse's gut where they reside within cells that comprise the fly's cardia and midgut tissues.
To test the binding efficacy of the antagomirs expressed by 3xant-miR275 to tsetse's mature miR275, we performed a dual luciferase reporter assay. We cloned the 3xant-miR275 construct into the multiple cloning site located in the 3'-UTR region of the reporter gene (renilla) in the psiCheck-2 vector (psiCheck-2 3xant-miR275 ). When miR275 binds to the sponge construct cloned in the 3'UTR region of the reporter gene (which initiates the RNA interference (RNAi) process), we expect the renilla transcript to be degraded, and the renilla Luciferase signal to be decreased. The psiCheck-2 vector also contains a firefly reporter in the expression cassette that is designed to be an intra-plasmid transfection normalization reporter. Thus, the Renilla luciferase signal is normalized to the firefly signal to standardize between different biological samples. We measured luciferase activity in three different experiments: 1) psiCheck-2 3xant-miR275 + synthetic miR275 mimic, 2) psiCheck-2 3xant-miR275 + synthetic AllStars Negative Control, and 3) psiCheck-2 3xant-miR275 alone, and we found that the relative luciferase activity (renilla/ firefly) was significantly suppressed in experiment 1 compared to experiments 2 and 3 (p < 0.05 and p < 0.0005, respectively; Fig 1C). In other words, in the presence of synthetic miR275 mimic, the luciferase activity was significantly repressed, which indicated that our sponge construct was successful when tested in vitro using an insect cell line. This outcome demonstrated that the miR275 effectively binds to the miR275 sponge and initiates the RNAi process with its associated mRNA.
To confirm the knockdown effect of miR275 levels in vivo, we used qPCR to quantify the relative expression of miR275 in Gmm 3xant-miR275 (treatment) and Gmm Scr-275 (control) individuals. Using multiple biological samples (each of which contained 5 dissected tissues pooled per sample) to reduce variability, we confirmed that the expression level of miR275 was significantly reduced in the midgut of the treatment group compared to that of the control group (p < 0.05; Fig 1D). However, our qPCR results did not consistently reveal a significant reduction of miR275 levels in the cardia organ of treatment versus control paratransgenic tsetse ( Fig 1E).

Gmm 3xant-miR275 gut physiological homeostasis is compromised
We demonstrated that recSodalis successfully invaded tsetse cardia and midgut tissues, and that miR275 was knocked down in the midgut of Gmm 3xant-miR275 . We next sought to determine if midgut physiological processes, such as blood meal digestion and PM functional integrity, were impaired in Gmm 3xant-miR275 flies in a manner similar to what was observed when tsetse miR275 [10] and mosquito Ae. aegypti miR275 [16] were depleted through the use of synthetic miR275 antagomirs. We compared the weight of midguts from 14 individual 8-day old Gmm 3xant-miR275 and Gmm Scr-275 flies 24 h after their last blood meal. We observed that guts from Gmm 3xant-miR275 individuals weighed significantly more (8.37 ± 0.64 mg) than did those from Gmm Scr-275 controls (4.03 ± 0.56 mg) (p < 0.001 ; Fig 2A), thus indicating that blood digestion and/or excretory processes (diuresis) were greatly disrupted in Gmm 3xant-miR275 .
We next employed a highly sensitive Serratia infection assay to test whether PM structural integrity was compromised in paratransgenic Gmm 3xant-miR275 compared to Gmm Scr-275 flies. We observed that 22% of Gmm 3xant-miR275 individuals survived for 19 days following per os challenge with Serratia. Comparatively, 0% of Gmm Scr-275 control flies survived this challenge (p < 0.0001; Fig 2B). These data indicate that paratransgenic-mediated repression of miR275 expression impairs tsetse's gut physiology and results in the production of a functionally compromised PM barrier, similar to what we had observed using synthetic antagomirs provided per os in a single bloodmeal [10].
Trypanosome infection establishment success in tsetse's midgut inversely correlates with the structural integrity of the fly's PM [6,43]. We next evaluated trypanosome infection outcomes in the midgut of Gmm 3xant-miR275 relative to Gmm Scr-275 control individuals to further confirm that paratransgenic expression of miR275 sponges interferes with the efficacy of tsetse's PM structure. We provided 8-day old adult paratransgenic flies a blood meal containing cysteine, which inhibits trypanolytic antioxidants present in the tsetse's midgut [9,44], and 10 7 T. b. brucei/mL of blood. Thereafter, the flies were maintained on normal blood meals for two weeks and subsequently dissected and microscopically examined to determine their midgut infection status. We found that significantly more Gmm 3xant-miR275 individuals (49%) hosted trypanosome infections in their gut than did their Gmm Scr-275 counterparts (11%) (p < 0.0001; Fig 2C). The higher parasite infection prevalence we observed in Gmm 3xant-miR275 individuals further signifies that the functional integrity of tsetse's PM is significantly compromised when miR275 sponges are paratransgenically expressed in the fly's midgut.

Global gene expression profiling in paratransgenic cardia and midgut
Our paratransgenic expression system has confirmed prior phenotypes that we observed following per os administration of synthetic antagomir-275, including a significant reduction of miR275 expression in the midgut and modified phenotypes associated with compromised gut physiological homeostasis such as dysfunctional digestive processes and compromised PM functional integrity. Additionally, we observed higher trypanosome infection prevalence in the midgut of Gmm 3xant-miR275 compared to Gmm Scr-275 flies. To obtain a broader understanding of the molecular mechanisms and pathways that are regulated by miR275, we performed global transcriptomic profiling in cardia and midgut tissues that were harvested from paratransgenic Gmm 3xant-miR275 relative to Gmm Scr-275 controls. All flies were age matched and inoculated per os with their respective recSodalis strains in their 1 st and 2 nd blood meals. For both comparisons each biological replicate (n = 3) contained pooled midguts (n = 5) or cardia (n = 10) tissues from 8-day old adults 36 h after their third blood meal. A total of 12 mRNA libraries were sequenced, and the total reads and uniquely mapped reads from each are summarized in S2 Table. We generated multi-dimensional scaling (MDS) plots to understand the overall gene expression differences between the biological replicates and treatment groups. We found that all three replicates within each treatment group clustered closely together as did all control group replicates (Fig 3A and 3B). When comparing gene expression differences in the cardia, we found that 265 genes (out of a total of 6101) were differentially expressed (DE; FDR < 0.05), with 99 (1.6%) and 166 (2.7%) up-and down-regulated in Gmm 3xant-miR275 relative to that of Gmm Scr-275 control individuals, respectively ( Fig 3A). When comparing gene expression differences in midgut samples, we found that 283 genes (out of a total of 5540) were DE (FDR < 0.05), with 116 (2.1%) and 167 (3.0%) up-and down-regulated in the midgut of Gmm 3xant-miR275 relative to Gmm Scr-275 individuals, respectively ( Fig 3B).

Gene Ontology (GO) enrichment analysis in the paratransgenic cardia and midgut
We next applied GO enrichment analyses to acquire broad insights into the functional contributions of the DE genes we identified. In the 99 up-regulated transcripts of Gmm 3xant-miR275 cardia relative to controls, enriched GO terms included chitin binding in the molecular function category, whereas in the 166 down-regulated transcripts, enriched GO terms included iron binding, heme binding, adenosine deaminase activity, and hydrolase and peptidase activity (Fig 4A and S1 Dataset). In the 116 upregulated transcripts of Gmm 3xant-miR275 midguts relative to controls, enriched GO terms included catalytic activity, oxidase activity and peptidase activity in the molecular function category, while in the downregulated transcripts, enriched GO terms included ribosome and cellular component biogenesis in biological processes (Fig 4B and S1 Dataset).

Analysis of DE genes in the cardia from Gmm 3xant-miR275 vs. Gmm Scr-275 control
Given that our phenotypic analysis indicated that miR275 is involved in blood digestion and PM barrier function (Fig 2), we first evaluated the DE genes whose products are likely associated with these functions. Among the genes whose putative products have been identified as PM structural proteins through proteomics analysis of the PM [45], we found that tsetse EP, midgut trypsin (GMOY007063) and choline acyltransferase were significantly down-regulated, while serine type endopeptidase (GMOY009757), pro1 and GmmPer12 were up-regulated in Gmm 3xant-miR275 relative to Gmm Scr-275 controls (Fig 5A and S2 Dataset). Among the secreted products localized to the PM, we found several digestive enzymes, serine proteases (Sp), trypsin and peptidases for which transcript abundance was significantly reduced in the treatment group (Fig 5A and S2 Dataset). The reduction in the production of these gene products may account for the impaired blood digestion we noted in Gmm 3xant-miR275 individuals. The downregulation of several genes whose products are associated with the PM, such as tsetse EP, midgut trypsin, Sp (GMOY006839), Sp15, and choline acyltransferase, were also noted from trypanosome-infected flies where PM functions were also compromised [9]. Tsetse EP protein is localized to the midgut, PM, and hemolymph [46,47]. The gene that encodes this protein is immune responsive, as its expression level was upregulated in response to bacterial challenge [46]. Furthermore, when tsetse EP was depleted via RNAi, trypanosome infection prevalence significantly increased [47]. Interestingly, the expression of chitinase (GMOY005519) and chitin binding protein (GMOY011054) was significantly upregulated in the cardia of Gmm 3xant-miR275 individuals. Different from other arthropod vectors, such as mosquitoes and sandflies, adult tsetse flies have type II PM, which is continuously secreted by cells located within the cardia. The PM is composed of a lattice of chitin fibrils cross linked by glycoproteins (Peritrophins) that contain chitin binding domains (CBD) [48]. Chitin is an extracellular polysaccharide that can be enzymatically hydrolyzed by chitinases [49]. Prior studies on trypanosome-infected cardia [9] and midguts [50] also indicated upregulated expression of chitinases, which likely resulted in compromised PM integrity. The reduction in PM associated gene expression, and the upregulation of the putative chitin degrading products, may contribute to the loss of PM integrity observed in paratransgenic Gmm 3xant-miR275 .
With respect to blood digestion processes, we detected 10 transcripts involved in heme binding and detoxification processes that were downregulated in Gmm 3xant-miR275 compared to controls (Fig 5B and S2 Dataset). Among these putative products were cytochrome (CYP) P450 enzymes, which belong to a superfamily involved in insect metabolism, detoxification and insecticide resistance in many different species [51], as well as several CYPs regulated by Plasmodium [52] and trypanosome [53] infections. Heme in the blood can induce oxidative damage to insect tissues [54] and the presence of heme binding proteins in Ae. aegypti PM suggest the structure exhibits a detoxification role [55].
Among the transcripts encoding transporters and/or transmembrane channel proteins that would be involved in secreting, trafficking and absorbing digestive products, we detected 12 that were downregulated and 10 that were upregulated in Gmm 3xant-miR275 relative to controls (Fig 5C and S1 Dataset). These up and down-regulated genes encode functions that involve transporting nutrients such as sugar and amino acids (e.g., major facilitator super family sugar transporter, glucose transporter, Slif and minidiscs), ions and water (e.g., Na/phosphate cotransporter, calcium channel, Kir family member, magnesium transporter, and aquaporin), and organic compounds (e.g. folate transporter). Annexin and Innexin are both upregulated in Gmm 3xant-miR275 . Annexin belongs to a large calcium dependent membrane binding protein family and the functions range from receptors of proteases in the gut epithelium to inhibitors of blood coagulation [56]. Plasmodium ookinetes use annexin for protection or to facilitate their development in the mosquito gut [57]. Annexin is upregulated in trypanosome-infected salivary glands (SG) [53]. Innexin proteins form gap junction channels and play critical roles in cell-to-cell communication in a variety of physiology activities [58]. Innexin 2 is a target gene of the Wingless signaling pathway in the proventricular cells in Drosophila [59]. One innexin was DE upon trypanosome infection in the tsetse fly, Glossina fuscipes fuscipes [60].
Lastly, six DE genes in Gmm 3xant-miR275 flies encoded products associated with embryogenesis and imaginal cell proliferation. Among these genes, forkhead and wing blister (Wb) were downregulated, while imaginal disc growth factor (Idgf), GMOY004790 (homologous to integrin in Md), wingless (Wg), and Wnt6 were upregulated (S2 Dataset). Idgf is involved in extracellular matrix formation in insects and participates in critical physiological activities such as larval and adult molting and wing development [63]. The wingless pathway is an intracellular signaling network; Wg signaling in Drosophila involves embryonic epidermis and wing imaginal disc [64]. Interestingly, Wg expression was reduced when tsetse miR275 was knocked down using the synthetic antagomir treatment [10], contrary to our data presented here using the constitutive silencing approach, which shows higher levels of Wg.

Analysis of DE genes in the midgut from Gmm 3xant-miR275 vs. control Gmm Scr-275
Similar to our analysis with the cardia, we first analyzed DE genes that are associated with PM components and digestive enzymes in Gmm 3xant-miR275 midgut transcriptomes. Among previously identified PM products [45], we found 7 that were upregulated in Gmm 3xant-miR275 midguts, including pro2, pro3, Sp6, choline acetyltransferase, chitin deacetylase, midgut trypsin (GMOY007063), and a serine type endopeptidase (GMOY9757) (S3 Dataset). In addition, we also identified several digestive enzymes, including trypsin, proteases and peptidases that were upregulated in Gmm 3xant-miR275 midguts relative to the controls (Fig 6A and S3 Dataset). Pro3, Sp6 and serine type endopeptidase (GMOY009757) were upregulated in response to T. brucei gambiense (Tbg) infection [50]. Higher levels of Chitin deacetylase, a hydrolytic enzyme that catalyzes the acetamido group in the N-acetylglucosamine units of chitin [65], could contribute to a compromised PM, similar to what we report for chitinase expression in the paratransgenic cardias above. The increased midgut weight we observed in Gmm 3xant-miR275 flies could reflect a dysfunctional gut enzyme production and/or altered enzyme transport in response to the compromised PM integrity. Among the twenty genes encoding transporters and/or transmembrane channel proteins DE in the midgut (Fig 6B and S3 Dataset), two (GMOY012503 and GMOY010388) were also identified DE in the cardia of Gmm 3xant-miR275 . In addition to transporters, we noted 7 DE genes, including down regulated members of CYP p450, ubiquitin ligase and up regulated nitric-oxidase synthase (NOS) that are associated with heme binding and oxidative response (Fig 6C and S3 Dataset). The ubiquitin ligase and a heme binding protein (GMOY001150) were also down regulated in the cardia of Gmm 3xant-miR275 . Ubiquitin ligase and CYP p450, which are associated with insecticide resistance and metabolism of natural or xenobiotic products in many insect species [66], have been linked to toxin metabolism following a blood meal in An. gambiae [67]. CYP p450-4g1 is also DE (FC>2) in response to Tbg infections in the Gmm midgut [50]. NOS is responsible for producing cellular nitric oxide, which is trypanocidal [68]. NOS expression is down regulated in trypanosome-infected SGs [53] and cardia [9,69], and VSG-treated cardia as well [10] Among the SG preferential genes that are dramatically reduced in Gmm 3xant-miR275 cardia, we detected five that were expressed in the midgut: salivary C-type lectin (GMOY000466), Ag5, secreted peptides (GMOY007065 and GMOY007077) and TTI. However, only the salivary Ctype lectin was DE in the midgut and upregulated in Gmm 3xant-miR275 relative to controls.

Paratransgenic knockdown of miR275 is not observed in tsetse's salivary glands
We observed the significant downregulation of 19 SG preferential genes in the cardia transcriptome from Gmm 3xant-miR275 versus Gmm Scr-275 flies. Because per os provisioned recSodalis is restricted in the gut tissue not in the hemolymph [22], we tested whether miR275 is expressed in the SG (Fig 7A). We anticipated that the miR275 knockdown effects would be restricted to the gut and not impact gene expression levels in other organs. To confirm this, we investigated whether paratransgenic knockdown of miR275 in tsetse's gut induces a systemic response that results in the knockdown of these genes in the fly's SGs. We first dissected the SG organ from Gmm 3xant-miR275 paratransgenic flies and tested the miR275 expression levels. We subsequently monitored the expression of Adgf3 (GMOY012374), Adgf5 (GMOY012375) and SGP1 (GMOY012268), which are abundantly expressed in tsetse's SGs [53,61,62] and significantly downregulated in Gmm 3xant-miR275 cardia. We found that none of the three SGpreferential genes were significantly reduced in the SG of Gmm 3xant-miR275 individuals despite being significantly down-regulated in the cardia (Fig 7B-7D). These results suggest that the effect of the paratransgenic knockdown could be restricted to tsetse's gut tissues where recSodalis reside, and is not likely to impact gene expression at the systemic level.

Discussion
We developed a paratransgenic expression system using tsetse's endosymbiont Sodalis to experimentally modify miR275 transcript abundance in tsetse's gut and to investigate the resulting physiological impact. Specifically, we engineered Sodalis to express miR275 sponges (3 tandem antagomir-275 repeats), and demonstrated that the recSodalis successfully colonize tsetse's cardia and midgut where they invade resident epithelial cells. We then demonstrated that the miR275 sponges successfully bind miR275, which results in posttranslational knockdown in vitro. We detected a significant reduction of miR275 levels in the midgut of paratransgenic tsetse expressing miR275 sponges, although we could not reproducibly demonstrate its reduction in the cardia organ. The paratransgenic flies displayed several robust phenotypes that are similar to those of miR275 depletion via synthetic antagomir-275, including obstructed blood meal digestion, compromised PM functional integrity, and susceptibility to parasite infection, all of which reflect impaired physiological homeostasis within the gut environment. Our transcriptomic studies further identified new molecular pathways heretofore unknown to be influenced by tsetse miR275, including the regulation of abundant secretory proteins functioning in vasoconstriction, platelet aggregation, coagulation, and inflammation or hemostasis. Our study is the first to use paratransgenesis as a strategy to constitutively modify the

PLOS PATHOGENS
expression of a microRNA in midgut tissue where the endosymbionts reside. It is efficient, cost effective, and minimally invasive compared to feeding and/or injecting synthetic antagomirs, and as such, this approach serves as an efficacious alternative to investigate microRNA related functions in the tsetse fly gut. This strategy can similarly be employed in any arthropod that houses genetically modifiable commensal gut symbionts that reside within host cells [70,71].
Several experimental approaches are available to modify miRNA expression in vivo. Chemically synthesized, cholesterol bound antisense oligonucleotides (antagomirs) are currently most commonly used. These single stranded oligos bind their complementary endogenous miRNA, thus preventing it from interacting with its target mRNA, which inhibits downstream protein production [72]. While synthetic antagomirs interact exclusively with their complimentary miRNA, they must be administered repeatedly and often in large doses for long-term effect, their uptake by cells can be inefficient, and they are difficult to target to specific tissues [73]. Transgenic expression of miRNA sponges is another widely used method, which can provide effective and specific inhibition of miRNA seed families (the conserved sequences among miRNAs) [33]. This method, which involves the insertion of multiple, tandem antagomirs into the germline, has been successfully used to constitutively deplete miRNA abundance in mosquitoes in a tissue specific manner via the use of tissue specific promoters [17,[74][75][76]. Because all embryonic and larval development occurs within the uterus of female tsetse [77], the generation of transgenic fly lines using traditional germline modification approaches has not been possible. To overcome this impediment, we developed the paratransgenic expression system described herein to constitutively express miR275 sponges in tsetse's gut.
We consistently observed three phenotypes that are associated with modified tsetse midgut physiological homeostasis in our Gmm 3xant-miR275 flies compared to Gmm Scr-275 controls. These phenotypes all correlate with the presentation of a structurally compromised PM, and they are similar to the phenotypes that we observed previously when synthetic antagomir-275 was administrated to tsetse. Specifically, we observed that Gmm 3xant-miR275 flies presented significantly heavier gut weights, significantly higher survival rates upon challenge with an entomopathogen, and significantly stronger vector competence, as compared to Gmm Scr-275 controls. Increased midgut weight is indicative of impaired blood meal digestion and/or excretion, and this phenotype was similarly observed following treatment of Ae. aegypti [16] and tsetse [10] with synthetic miR275 antagomir. In hematophagous insects, the PM mediates blood digestion by regulating the flux of digestive enzymes from their site of production in the midgut epithelium into the blood bolus-containing gut lumen [78,79]. Our study also demonstrated that significantly more Gmm 3xant-miR275 flies survive in the presence of entomopathogenic Serratia than do Gmm Scr-275 control flies, further indicating that PM functional integrity is compromised in the former group of flies. Serratia marcescens strain Db11 is an entomopathogenic bacterium [80] that can kill tsetse when provided in the bloodmeal. Specifically, flies with an intact PM fail to immunologically detect Serratia, which allows the bacterium to rapidly proliferate in the gut lumen, translocate into the hemolymph and eventually to kill the tsetse and other insects [6,9,10,[80][81][82][83]. Conversely, when PM structural integrity is compromised, the bacterium is quickly detected by tsetse's midgut epithelium and eliminated by the fly's robust antimicrobial immune response. The Serratia infection assay thus serves as a highly sensitive indicator of tsetse's PM structural integrity [6]. Lastly, we observed a higher trypanosome infection prevalence in Gmm 3xant-miR275 flies compared to Gmm Scr-275 controls. This outcome is similar to what observed in flies exposed to anti-PM RNAi (dsRNA targeting pro1, pro2 and chitin synthase) [6] as well as in flies that were provisioned a blood meal containing a purified trypanosome coat protein (sVSG), which interferes with PM related gene expression in the cardia through the reduction of miR275 [10]. Taken together, our results confirm that interference with miR275 expression in the cardia and midgut of Gmm 3xant-miR275 flies results in the modified gut environment we noted in this study.
Herein we repeatedly observed phenotypes that correspond with a depletion of miR275 expression in tsetse's cardia. However, despite these findings, we were unable to quantify a significant reduction in expression of the microRNA in tsetse's cardia (although we could in the fly's midgut). This outcome may be accounted for by one or several reasons. First, the concentration of paratransgenically expressed miR275 relative to the concentration of the binding sites may have reduced the inhibitory effect of the miRNA sponges [73]. Prior investigations demonstrated that tsetse miR275 is highly abundant in the cardia compared to the midgut tissues [10]. Thus, our depletion effect could have been diluted in the cardia organ where miR275 are highly abundant. This outcome is further exacerbated by the conspicuously low number of recSodalis that colonized cells of tsetse's cardia in comparison to the midgut. More experiments are required to optimize the uptake of recSodalis by cells of tsetse's cardia organ. Moreover, qRT-PCR can be an inaccurate method for quantifying the abundance of functional miRNAs, especially in the organ where the miRNAs are highly abundant such as tsetse's cardia. The procedure measures the total amount of miRNAs and doesn't distinguish between functional miRNAs and non-functional ones. Thus, qRT-PCR can quantify the amount of extracellular miRNA released from Trizol-lysed cells, and this represents a physiologically irrelevant population of miRNAs [84]. Combined with the robust phenotypic changes and differential expression of blood digestion and PM related genes, we believe that our paratransgenic knockdown was successful at the functional level.
Our transcriptomic analyses of cardias and midguts from paratransgenic tsetse revealed several interesting insights into the broader functions of miR275 that are related to trypanosome infection. First, with regard to the genes that are associated with PM and digestion, midgut GO enrichment analysis indicated that downregulated genes in Gmm 3xant-miR275 flies included an enriched population of transcripts that encode proteins involved in ribosome biogenesis and cellular component biogenesis. This suggests that protein synthesis is obstructed in the midguts of Gmm 3xant-miR275 flies, which could reflect the compromised PM structure and disrupted digestion we observed in these fly's guts. GO enrichment analysis of upregulated cardia specific genes indicated that genes in Gmm 3xant-miR275 flies included a group of enriched transcripts that encode proteins involved in chitin metabolism and chitin binding processes. Chitinase produced by parasites degrades the sand fly and mosquito PM, which promotes Leishmania [85] and Plasmodium [86] transmission, respectively. The genome of African trypanosomes does not encode a chitinase gene. However, chitinase is a proteinaceous component of tsetse's PM, and infection with trypanosomes induces chitinase expression in the fly's cardia [9,45,87] and gut [10]. These findings suggest that parasites may facilitate their transmission through the fly by transiently upregulating cardia/gut chitinase expression, thus degrading PM chitin fibrils and reducing the structure's ability to serve as a barrier. We also observed that several genes encoding digestive enzymes were downregulated in the cardia of the Gmm 3xant-miR275 . Similarly, miR275 and digestive enzyme-encoding genes (e.g., those encoding trypsin and trypsin-like proteins) were down-regulated in tsetse's cardia following trypanosome exposure [9,10]. In Ae. aegypti, gut-specific depletion of miR275 results in reduced expression of its target gene SERCA, as well as reduced digestive enzyme secretion, disrupted gut microbiota homeostasis and compromised gut actin cytoskeleton integrity. Notably, under these circumstances, protein levels of late trypsin, a late-phase digestive protease in female mosquitoes, are significantly reduced [17]. This outcome likely accounts for the altered midgut phenotypes observed in miR275 knockdown mosquitoes. However, tsetse SERCA does not contain orthologous miR275 binding site motifs, and SERCA levels are not differentially expressed in Gmm 3xant-miR275 compared to Gmm Scr-275 flies. These characteristics suggest that the target of tsetse miR275 may not be SERCA, and a currently unknown pathway (s) regulates the secretion of the above-mentioned proteins in tsetse's gut.
Notably, in this study the expression of some PM-associated genes was the opposite of what was observed previously when tsetse miR275 was knocked down via synthetic antagomir-275 [10]. In the previous study, expressions of pro1-3 were significantly downregulated in the cardia samples after the per os provisioning of either synthetic antagomir-275 or sVSG. Conversely, in this study, pro1 in the cardia, and pro2 and pro3 in the midgut were significantly upregulated in Gmm 3xant-miR275 compared to Gmm Scr-275 flies. However, in a different study of trypanosome-infected tsetse cardia pro1 is no significant different and pro2-3 are downregulated, and the downregulation effect of pro genes by provisioning sVSG in the cardia is transient [9]. This finding suggests that the pro genes regulation might be different based on parasite infection status. The observed disparity in pro gene expression by miR275 can be explained by the possibility that the synthetic antagomir produces a one-time reduction in miR275 expression that causes a different physiological response in the fly compared to that when miR275 is constitutively suppressed in paratransgenic flies (recSodalis continuously produces miR275 sponges). However, further investigation is required to acquire a more complete understanding of the miR275 regulatory network.
Our transcriptomic results consistently showed miR275 functions in reducing the expression of secretory enzymes and similarly impairing secretory and digestive pathways. Nineteen saliva-associated proteins were among the putative secretory products that were dramatically reduced in the cardia of the Gmm 3xant-miR275 individuals. Interestingly, seventeen of these genes were reduced in trypanosome-infected salivary glands [53,61,62], but it remains to be seen if this reduction is also mediated by lower miR275 levels in infected salivary glands. Previous transcriptomic analyses with parasite-infected cardia indicated that 9 of these saliva protein-encoding transcripts [Adgf3, Ag5, Tsal1, Tsal2 (GMOY012360), SGP1, tsetse thrombin inhibitor (TTI), salivary secreted protein (GMOY012067) and two secreted proteins (GMOY003214 and GMOY007077)] are detected in tsetse's cardia, but only four of them [Ag5, Tsal2, TTI and one of the secreted proteins (GMOY007077)] are differentially expressed upon trypanosome infection [9]. Moreover, our earlier transcriptomic analysis of trypanosome-challenged tsetse guts (48 h post provisioning of a parasite containing bloodmeal) found that fourteen of these genes [Tsal1, TTI, SGP1, GRP2, 5' Nuc, both Tsal2s, Adgf1, Adgf2, Adgf3, Adgf5, salivary secreted protein and two secreted peptides (GMOY003214 and GMOY012286)] are significantly reduced relative to unchallenged controls [10]. All of these SG preferential gene products were previously detected in tsetse saliva and thought to be essential for the fly's ability to successfully blood feed [88]. In addition to being major constituents of saliva, Adgf, TTI and 5'Nuc are associated with anticoagulant functions in tsetse's saliva and gut, suggesting a role in digestive processes [60,61,[89][90][91], while Ag5 is a major allergen involved in hypersensitivity reactions in the mammalian host [92]. The reduction of these saliva-associated anticoagulants in infected flies causes increased probing and biting behaviors, which in turn increases the transmission potential of the parasite to multiple hosts [61]. The significant reduction in expression of genes in the Adgf family was also very interesting. Adgf is a secreted enzyme that converts extracellular adenosine into inosine by deamination and is important in anti-inflammation, tissue damage and resistance to bacterial infection in Drosophila [93][94][95]. A Adgf is expressed by immune cells to regulate the metabolic switch during bacterial infection in Drosophila, and the downregulation of Adgf increases extracellular adenosine and enhances resistance to bacterial infection [93]. The loss of Adgf can induce intestinal stem cell proliferation in Drosophila [95]. As evidenced by reduced Adgf gene expressions in trypanosome-challenged tsetse guts, the downregulation of Adgf genes might be triggered by initial infection of trypanosomes to release anti-inflammatory response and/or to repair any damaged tissues. Interestingly, Matetovici et al (2016) [53] noted significantly reduced expression of genes that encode saliva-associated products in the SGs of flies that house trypanosomes in their midgut but not yet in their SGs. This finding is suggestive of a molecular dialogue between the organs, and a possible anticipatory response of the SG environment prior to the parasites infecting the tissue, which may be mediated by miR275 levels in these tissues. Given that these genes encode secreted proteins, their strong reduction in paratransgenic tsetse further supports the role of miR275 in trypanosome infection, possibly through regulation of secretory pathways.
Arthropod-borne diseases impose a debilitating global public health burden. Due to the lack of effective vaccines capable of preventing the majority of these diseases, and the increasing resistance of vector arthropods to pesticides, alternative approaches for disease control are urgently needed. Paratransgenic systems have been applied in efforts to reduce vector competence in mosquitoes [29,30,96,97], kissing bugs [98,99], sand flies [100] and tsetse flies [24][25][26]101]. This technology has many benefits, including the absence of a reliance on inefficient germline modification procedures [85], and the fact that modified symbionts exert no fitness cost on their insect hosts [23] and can potentially spread through wild vector populations via vertical transmission [102]. Additionally, paratransgenically expressed microRNAs cost significantly less than do their synthetically produced counterparts. Our study is the first to use this system to explore the function of an arthropod vector microRNA in relation to disease transmission processes. This system can be easily applied to study the function of other tsetse miR-NAs and for future research aimed at experimentally interfering with the physiological homeostasis of tsetse's midgut environment with the intent of interrupting trypanosome transmission through the fly. This study also expanded our knowledge of the relationship between tsetse miR275 and the regulation of key physiological processes such as blood digestion, PM integrity, and gut environment homeostasis in tsetse. Our transcriptomic data revealed functions regulated by miR275 affecting tsetse's secretory pathways. These findings provide a foundation for discovering the target of tsetse miR275 in future studies.
Supporting information S1