Drosophila MOV10 regulates the termination of midgut regeneration

Abstract

The molecular mechanisms by which stem cell proliferation is precisely controlled during the course of regeneration are poorly understood. Namely, how a damaged tissue senses when to terminate the regeneration process, inactivates stem cell mitotic activity, and organizes ECM integrity remain fundamental unanswered questions. The Drosophila midgut intestinal stem cell (ISC) offers an excellent model system to study the molecular basis for stem cell inactivation. Here, we show that a novel gene, CG6967 or dMOV10, is induced at the termination stage of midgut regeneration, and shows an inhibitory effect on ISC proliferation. dMOV10 encodes a putative component of the microRNA (miRNA) gene silencing complex (miRISC). Our data, along with previous studies on the mammalian MOV10, suggest that dMOV10 is not a core member of miRISC, but modulates miRISC activity as an additional component. Further analyses identified direct target mRNAs of dMOV10-containing miRISC, including Daughter against Dpp (Dad), a known inhibitor of BMP/TGF-β signaling. We show that RNAi knockdown of Dad significantly impaired ISC division during regeneration. We also identified six miRNAs that are induced at the termination stage and their potential target transcripts. One of these miRNAs, mir-1, is required for proper termination of ISC division at the end of regeneration. We propose that miRNA-mediated gene regulation contributes to the precise control of Drosophila midgut regeneration.

Introduction

Genetic control and cellular composition of the Drosophila midgut show remarkable similarities to mammalian digestive systems. With powerful genetic tools available in this model, the Drosophila midgut intestinal stem cell (ISC) offers an attractive system to study intestinal homeostasis, regeneration, and disease conditions (Casali and Batlle 2009; Nászai et al. 2015; Guo et al. 2016; Jiang et al. 2016; Li and Jasper 2016). The ISCs are scattered along the basement membrane of the entire length of the midgut (Micchelli and Perrimon 2006; Ohlstein and Spradling 2006). Two types of precursor cells are generated from ISC divisions: the enteroblasts (EBs) and the enteroendocrine precursors (EEPs). These cells give rise to two types of differentiated cells, the enterocytes (ECs) and the secretory enteroendocrine cells (EE), respectively (Biteau and Jasper 2014; Guo and Ohlstein 2015; Zeng and Hou 2015; He et al. 2018; Hung et al. 2020). ECs, visceral muscle, and tracheal cells secrete factors that regulate ISC proliferation and differentiation, collectively serving as a niche for ISCs.

The ISC model provides an excellent opportunity to study how stem cells respond to tissue damage (Buchon et al. 2009; Jiang et al. 2009; Lucchetta and Ohlstein 2012). Gut epithelium acts as a protective barrier against pathogens and damaging agents, and ISCs increase their rate of division in response to tissue damage (Amcheslavsky et al. 2009; Jiang et al. 2009). Once damaged, the differentiated gut epithelium secretes signaling molecules that promote ISC division and stimulate tissue repair. This regeneration can be readily induced in the laboratory using a few different methods. For example, enteric bacterial infection (Pseudomonas entomophila) or feeding dextran sodium sulfate (DSS), another commonly used damage-inducer, activates several pathways to upregulate ISC proliferation. These pathways include Jak/Stat (Buchon et al. 2009; Cronin et al. 2009; Jiang et al. 2009; Lin et al. 2010; Zhou et al. 2013), EGFR (Buchon et al. 2010; Jiang et al. 2011), Wingless (Wg) (Cordero et al. 2012); Lin et al. 2008 #74; Lee, 2009 #73}, and BMP/Decapentaplegic (Dpp) (Ayyaz et al. 2015; Zhou et al. 2015; Tian et al. 2017) signaling.

In addition to the activation of the proliferative capacity of stem cells, their inactivation at the end of regeneration is equally important. The failure to properly terminate tissue regeneration results in the emergence of unwanted cells, leading to abnormal organ size (Miyaoka and Miyajima 2013) and a high risk of cancer (Hsu and Fuchs 2012; Fuchs et al. 2013). However, our knowledge on the molecular mechanisms underlying regeneration termination is limited. One major question is what molecules facilitate the transition from early stages of regeneration (when stem cells actively proliferate and undergo rapid differentiation) to late stages (when mitosis is downregulated and tissue morphology is restored). Put another way, when the regeneration is completed, how does the tissue properly down-regulate ISC proliferation?

Because transcription of the mitogen ligands, which promote ISC division, is only transiently activated to initiate regeneration, cessation of ligand expression results in halting ISC division. However, additional mechanisms are necessary to properly terminate midgut regeneration. Previous studies reported the inhibitory regulation of ISC proliferation by Dpp, a Drosophila homolog of BMPs, during homeostasis and regeneration (Guo et al. 2013; Ayyaz et al. 2015). Thus, Dpp may serve as a safeguard mechanism against ISC overactivation. In addition, a structural modification of co-receptors for these mitogen ligands regulates ISC proliferation and regeneration termination. Many mitogens, including EGFs, Wnt/Wg, and Unpaired-like cytokines (ligands for Jak/Stat signaling), bind to heparan sulfate (HS) and their activity requires a co-receptor, heparan sulfate proteoglycans (HSPGs), on the cell surface. Importantly, the function of these HSPG co-receptors depends on the level and patterns of sulfate groups on HS, which is controlled by HS modifying enzymes, such as sulfotransferases and sulfatases (Nakato and Li 2016). Our previous study showed that Sulf1, a secreted HS-specific sulfatase, is induced at the termination stage and downregulates ISC mitotic activity (Takemura and Nakato 2017). Thus, in addition to transcriptional inactivation of ligands, a post-transcriptional mechanism functions to ensure the precise control of ISC division at the termination stage.

Another post-transcriptional mechanism that may affect regeneration is gene expression control via the microRNA (miRNA)-mediated network. miRNAs are small noncoding RNAs that are approximately 22 nucleotides long (Bartel 2009). It is typically believed that miRNAs bind to a complementary sequence in the 3' untranslated region (UTR) of target mRNAs, leading to their degradation or translational block. An increasing number of studies have identified miRNAs as key regulators of regeneration in various animal models (Yin et al. 2008; Sehm et al. 2009; Eulalio et al. 2012; Mok et al. 2017; Krishna et al. 2019; Pang et al. 2019). In the Drosophila midgut, it is known that several miRNAs regulate cell fate specification (Foronda et al. 2014; Antonello et al. 2015; Chen et al. 2015) and ISC proliferation (Huang et al. 2014) under homeostatic conditions. However, the functions of miRNAs during dynamic midgut regeneration processes are poorly understood.

In this study, we identified sets of genes and pathways that are up/down-regulated at different stages of regeneration using RNA-seq. Among these molecules, we found that a novel gene, CG6967 or dMOV10, is induced specifically at the termination stage. Premature expression of dMOV10 at an earlier stage suppressed ISC proliferation. In addition, ISC division remained active in dMOV10 mutants at later stages of regeneration. These results show that it has a role in the proper shutdown of stem cell division at the end of regeneration. dMOV10 encodes a Drosophila homolog of the mammalian MOV10 family of RNA-binding proteins. Although dMOV10 function is currently unknown, the mammalian MOV10 homologs have been shown to act as modulators of the miRNA gene silencing complex (miRISC) (Meister et al. 2005). We identified direct target mRNAs of dMOV10-containing miRISC, including Daughter against Dpp (Dad). We also provide the transcriptional profiles of miRNAs during midgut regeneration as a resource for further studies on regeneration mechanisms.

Materials and methods

Fly stocks and husbandry

Oregon-R and w¹¹¹⁸ were used as wild-type strains. The following fly strains were used:

tkv⁷ (BDSC #3242), AGO1: GFP^CA06914 (BDSC #50805), vasa-Cas9 (BDSC #55821), esg-GAL4^NP6267 (Kyoto DGGR #113886), MyoIA-lacZ (a gift from Huaqi Jiang), and MyoIA-GAL4^NP0001 (Kyoto DGGR #112001), esg-lacZ^k00606 (Kyoto DGGR #108851), trol: GFP^ZCL1700 (Kyoto DGGR #110807), UAS-AGO1^RNAi (TRiP.HMC03509, BDSC #53293), ap-GAL4, Dad: GFP-FLAG (VK00037, BDSC #42669), UAS-tdTomato (BDSC #36328), UAS-Dad^RNAi (TRiP.HMS01102, BDSC #33759), UAS-LUC-mir-1 (BDSC #41125), UAS-mir-1-sponge (a gift from David Van Vactor), tub-GAL80^ts (BDSC #7108, BDSC #7018), and esg^tsF/O (a gift from Huaqi Jiang). dMOV10^KI.HA, dMOV10^KO.tdTomato, dMOV10¹⁸, UAS-dMOV10, UAS-dMOV10: dADARcd-V5, and UAS-dADARcd-V5 were generated in this study, as described below. Detailed genotypes used in individual experiments are listed in Supplementary Table S1.

Fly stocks were reared on a standard cornmeal fly medium at 25°C except for those containing tub-GAL80ts, which prevents GAL4-mediated UAS transgene expression at 19°C but activates it at 30°C (McGuire et al. 2003). To induce UAS transgenes, the flies were switched to 30°C 1 day before Pe infection.

Generation of fly strains

dMOV10^KI.HA knock-in allele was generated by inserting smGFP-HA at the N-terminus of dMOV10 using CRISPR/Cas9-mediated homology-directed repair as previously described (Takemura and Nakato 2017). sgRNA sequences targeting dMOV10, chosen using CRISPR Optimal Target Finder, were cloned into pU6-BbsI-chiRNA (a gift from Melissa Harrison, Kate O’Connor-Giles, and Jill Wildonger) (Gratz et al. 2013). smGFP-HA and 1.1-kb dMOV10 homologous sequences on either side of the predicted DSB were cloned into the pHD-DsRed-attP backbone (Gratz et al. 2014) and used for homology-directed repair. To avoid retargeting of the sgRNA to the engineered locus, synonymous mutations were introduced into or near the PAM sequence of the homology arms.

dMOV10^KO.tdTomato null allele was also generated by CRISPR/Cas9-mediated genome editing. To generate the repair template, a tdTomato coding sequence was ligated with 1- and 1.3-kb dMOV10 homologous sequences on either side of the predicted DSBs. The resultant fragment was cloned into the pHD-DsRed-attP backbone. dMOV10¹⁸ was generated using two pBac insertions in the dMOV10 locus, pBac{WH}f07741 and pBac{WH}f01447, both of which contain a FRT. dMOV10¹⁸ bears a deletion of 2.4 kb in the dMOV10 coding sequence resulting from FLP/FRT-mediated recombination between the two pBac insertions.

To generate UAS-dMOV10, the dMOV10 coding sequence was amplified by PCR from cDNA clone LD34829 and cloned into pUASg.attB (DGRC #1422). The resultant pUASg.attB-dMOV10 construct was integrated at the ZH-86Fb attP landing site (BDSC #24749) by BestGene Inc.

UAS-dMOV10: dADARcd-V5 was made from a dMOV10: dADARcd-V5 fusion construct. In this fusion gene, dMOV10 CDS and the catalytic domain of dADAR-V5 are separated by a 4x Gly-Gly-Ser linker, which was amplified from pMT-ADARcd-V5. The resultant construct was inserted into the NotI- and XbaI-digested pJFRC7 vector. Both UAS-dADARcd-V5 and UAS-dMOV10: dADARcd-V5 are integrated at the same genomic landing site (M{3xP3-RFP.attP}ZH-86Fb) on the third chromosome using the PhiC31/attP/attB system.

Primers used in this study are available on request.

Induction of regeneration

To induce midgut regeneration, we used Pe infection as described previously (Amcheslavsky et al. 2009; Takemura and Nakato 2017). Three-day-old adult flies were starved for 2 hours and placed in empty vials containing a piece of Whatman filter paper soaked in 5% sucrose solution (control) or Pe overnight culture. The flies were transferred into vials with standard cornmeal medium 22 hours after Pe infection. Female flies were then dissected after 2 hours (day-0) or indicated time points.

To induce expression of UAS-dMOV10 prematurely, we employed the esg^tsF/O Flp-Out system (esg^tsF/O) (Jiang et al. 2009), with the genotype of esg-Gal4 tub-Gal80^ts UAS-Flp act>CD2>GAL4. The esg^tsF/O flies were crossed with UAS-dMOV10 to obtain the progenies (esg^tsF/O-dMOV10). The culture temperature of the esg^tsF/O-dMOV10 animals was shifted to 30°C 1 day prior to the Pe treatment, and measured ISC mitotic activity at day-0. Control samples were obtained by crossing esg^tsF/O with wild-type animals (esg^tsF/O-control). UAS-LUC-mir-1 and UAS-mir-1-sponge were also induced by esg^tsF/O.

Immunohistochemistry

Immunohistochemistry was performed as described previously (Takemura and Nakato 2017). In brief, dissected adult midguts or larval wing discs were fixed in 3.7% formaldehyde at room temperature for 1 hour or 15 minutes, respectively. The samples were incubated with primary antibodies at 4°C overnight, then washed with 0.1% Triton X-100/PBS for 10 minutes three times. The secondary antibodies conjugated with Alexa Fluor 488, 546, 568, or 633 (Thermo Fisher Scientific) were used at 1:500 dilution in 0.1% Triton X-100/PBS and incubated with samples at 4°C overnight. After three 10-minutes washes with 0.1% Triton X-100/PBS, nuclei were stained with 1 µg/ml DAPI (Thermo Fisher Scientific, 62248) and subsequently mounted in VECTASHIELD Antifade Mounting Medium (Vector Laboratories). F-actin was stained with Alexa Fluor 633 Phalloidin at 1:500 (Thermo Fisher Scientific, A22284). Images were acquired using a Zeiss LSM 710 Confocal Microscope and were edited on Fiji (Schindelin et al. 2012) and Adobe Photoshop CS3. Colocalization analysis was performed to obtain Manders' coefficients using JACoP (Bolte and Cordelieres 2006) on Fiji. The following primary antibodies were used: rabbit anti-pH3 (EMD Millipore, 1:1,000), rat anti-HA 3F10 (Roche, 1:200), rabbit anti-HA C29F4 (Cell Signaling, 1:1,000), chicken anti-β-galactosidase (abcam, 1:2,000), rabbit anti-GFP (Thermo Fisher Scientific, 1:1,000), mouse anti-GFP 3E6 (Thermo Fisher Scientific, 1:200), mouse anti-Pros MR1A (DSHB, 1:50), rat anti-Vasa (DSHB, 1:50), mouse anti-Hts 1B1 (DSHB, 1:20), and mouse anti-Armadillo (DSHB, 1:50). When the transverse sections of the midgut were imaged, spacers were placed between the slide glass and cover glass to keep the structure of the midgut epithelium.

Co-immunoprecipitation

S2 cells were transfected with pAW-3xMyc-dMOV10 and pAFW-AGO1 (a gift from Yukihide Tomari) (Kawamata et al. 2009). Co-immunoprecipitation was performed as described previously (Dejima et al. 2013) using anti-Myc-agarose beads (Sigma). The eluted proteins were subjected to immunoblot analysis using mouse anti-Myc (9E10) (1:2,000; Sigma) and rat anti-HA antibodies (3F10) (1:2,000; Roche Applied Science).

RNA-immunoprecipitation

S2 cells were transfected with pAW-eGFP-Dad and pAW-3xMyc-dMOV10 with or without pAWF-AGO1. RNA-Immunoprecipitation was performed as described previously (Kachaev et al. 2017) using mouse anti-Myc (9E10) antibody (1:200) and protein G-sepharose. TRIzol (Thermo Fisher Scientific) is added to sepharose, and total RNA was extracted as described below.

RT-qPCR

RT-qPCR was performed as described previously (Kim and O'Connor 2021). Total RNA was extracted from 50 midgut samples using TRIzol and purified using an RNeasy MinElute Cleanup Kit (QIAGEN). The Superscript III first-strand synthesis kit (Invitrogen) was used to synthesize cDNA, and RT-qPCR reactions were performed in duplicate on each of three independent biological replicates on the LightCycler 480 (Roche) using a SYBR Green kit. Act5C expression was used for normalization. Fold changes were calculated using the ΔΔCt method. Primers used in this study are shown in Supplementary Table S2.

RNA-seq analyses

RNA was extracted from 20 Oregon-R female midguts using TRIzol, treated with RNAse-Free DNase (QIAGEN), and purified on an RNeasy MinElute Cleanup column. Total RNA quality was checked using Agilent 2100 Bioanalyzer System. mRNA libraries were prepared using TruSeq RNA Library Prep Kit v2 (Illumina) and were sequenced on an Illumina HiSeq 2000 platform (Illumina) for 50 cycles in a paired-end run at the University of Minnesota Genomics Center. Two independent biological replicates were used for sequencing, with a read depth that varied between 36 and 60 million reads. Quality control and quality filtering were performed using fastp v0.19.5 with default settings. Transcript abundance was estimated by the quasi-mapping-based mode of salmon v0.11.3. The index for mapping was built from the Ensembl BDGP6 transcriptome sequence and annotation files. Differential gene expression analysis was performed using DESeq2 v1.22.0. P-values were adjusted using the Benjamini-Hochberg method for multiple testing correction. Gene ontology (GO) analysis was carried out using FlyMine v46.1 (Lyne et al. 2007) and redundant terms were removed using GO Trimming v2.0. (Jantzen et al. 2011) with 40% soft trim threshold.

For small RNA-seq, total RNAs were collected from 20 midguts from female Oregon-R flies using TRIzol Reagent and Direct-zol RNA Miniprep (Zymo Research) with on-column DNase I digestion. Three independent biological replicates were prepared for sequencing. RNA quality was assessed using Agilent TapeStation (Agilent). Library preparation and sequencing were performed by GENEWIZ using TruSeq Small RNA Library Preparation Kits (Illumina) and the Illumina HiSeq 4000 platform. Reads were analyzed using miRge version 2.0 and mapped to mature miRNAs and miRNA hairpins, retrieved from miRbase, release 21. Differential expression analysis was carried out using DESeq2 v1.22.2. Target mRNAs of miRNAs were obtained from TargetScanFly v7.2 (Agarwal et al. 2018) and DIANA-microT-CDS v5.0 (Reczko et al. 2012).

TRIBE

TRIBE experiments were performed as described before (McMahon et al. 2016). RNAs were isolated from 20 midguts of esg^tsF/O flies expressing dADARcd-V5 and dMOV10: dADARcd-V5 using TRIzol Reagent (Invitrogen). After treated with DNase I (QIAGEN), RNAs were purified using RNeasy MinElute Cleanup Kit. RNA quality was checked using Agilent TapeStation. Libraries were prepared using TrueSeq Stranded mRNA Library Prep (Illumina) and sequenced on the Illumina HiSeq 2500 High Output 50-bp paired-end run (v4 chemistry) with two biological replicates. Each sample contained at least 56 million reads. Low-quality bases were trimmed using fastp v0.19.5 with default settings. Reads were mapped to the reference genome (UCSC dm6) using STAR v2.6.1c. Duplicated reads were marked using Picard v2.18.20 and ignored in the downstream analysis. A-to-G variants were detected using JACUSA v1.2.4. Called variants met (i) 85% A in control (ii) 85% A or G in control (iii) 20% increase in the proportion of G compared to control.

Data Availability

All mRNA-seq and small RNA-seq data generated in this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) public repository, and they are accessible through GEO accession number GSE154300 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE154300).

Fly strains and reagents are available upon request.

Supplemental Material available at figshare: https://doi.org/10.25386/genetics.14105972.

Results

mRNA-seq analysis of the regenerating midgut

Feeding flies a gram-negative bacterium, Pe, substantially increases ISC division, as monitored by phospho-histone H3 (pH3) staining. After starvation, 3-day old flies were fed Pe for 22 hours, transferred to normal food vials, and dissected after 2 hours (day-0) or at indicated time points (Figure 1A). This treatment led to massive expansion of escargot-positive (esg⁺) cells and thickened midgut epithelium (Vodovar et al. 2005; Shaw et al. 2010); Figure 1B). ISC mitotic activity peaked immediately after Pe infection (day-0, Figure 1, C and D). During regeneration, newly emerged cells pile up, forming a multilayer structure (Figure 1C, upper middle panel), and extensively form filopodia-like cellular extensions into the basement membrane, which becomes thick and disorganized at this stage (Figure 1C, upper right panel). A similar filopodia-like structure has also been observed in ISC progenitors during normal homeostasis (Antonello et al. 2015). Remarkably, the ISC mitotic activity reverts to a normal level within 3–5 days after the infection (Figure 1D), consistent with a previous study (Buchon et al. 2010). By day-5, the midgut completely reforms a normal single-layer epithelium with a remodeled basement membrane (Figure 1C, bottom panels). These observations of morphological changes during the regeneration time course indicate that midgut regeneration requires a dynamic reorganization of the epithelial architecture, in addition to precise control of ISC proliferation. However, it is not fully understood how these processes are regulated.

Figure 1.

Time-course of midgut regeneration. (A) Scheme of bacterial infection. Three-day-old flies were starved for 2 hours and treated with Pe for 22 hours (red bar). The flies were transferred to normal food and dissected after 2 hours (day-0) or at indicated time points. (B) A diagram showing a transverse view of midgut cells (left). ISC, EB, EE, ECs, and visceral muscle are shown in red, orange, blue, gray, and green, respectively. In Figure 1C, midgut confocal images show transverse views of the regenerating midgut. In Figure 3 and the following figures, all midgut confocal images show horizontal optical sections at a basal focal plane (arrowhead). (C) ISCs and the progenitors are marked by esg>tdTomato after Pe infection. Top panels: uninfected control (left, Sucrose, day-0), midgut at day-0 after Pe infection (middle), and a magnified view of esg>tdTomato signal from the image in the middle (right). ISCs and progenitors at day-1 after Pe infection extensively form filopodia-like cellular extensions into the basement membrane (bracket). Bottom panels: Left, at day-3 (left) and day-5 (right) after Pe infection. F-actin is labeled with phalloidin (blue). The basement membrane (BM) is marked by Trol::GFP (green). (D) Time-course of midgut regeneration. The number of pH3-positive cells in the posterior midgut at each time point after Pe infection is shown.

Taking advantage of this stereotypic time course of the midgut regeneration process, we performed a messenger RNA sequencing (mRNA-seq) analysis to identify genes and pathways that are up/down-regulated at different time points after Pe infection. We isolated total RNA from midgut samples of three different conditions: mock-infected flies (Sucrose, day-0), Pe-infected flies after day-0 (initiation stage), and day-3 (termination stage) recovery periods. RNA-seq was performed with two biological replicates for each condition. Principal component analysis (PCA) of gene expression data showed clear distinctions between the conditions (Figure 2A). A heatmap showing subsets of differentially regulated genes is depicted in Figure 2B. In the earlier stage of regeneration (Pe, day-0), 1133 genes are upregulated (“early genes”) (i) and 1161 genes are downregulated (ii) (adjusted P-values <0.05 and |log2 fold change| ≥1; Supplementary Table S3). Consistent with previous reports (Buchon et al. 2009; Jiang et al. 2009), known “tissue-damage” genes are indeed induced at day-0 in our mRNA-seq analysis. These genes include antimicrobial peptide (Diptericin), upd2, upd3, and vein (vn, an EGFR ligand) (Figure 2C). GO term enrichment analysis (FlyMine v46.1, Lyne et al. 2007) identified genes related to DNA repair machinery, cell cycle control, metabolic processes, and stress response (Table 1). GO terms for the upregulated mRNAs also included oogenesis, cytoskeletal regulation, and developmental processes occurring during later embryogenesis, such as neurogenesis or organ development.

Figure 2.

Transcriptional profiles of the regenerating Drosophila adult midgut. (A) Principle component analysis of transcriptional profiles of three conditions (Sucrose, day-0; Pe, day-0; Pe, day-0, n = 2 each) showing a clear distinction among these conditions. (B) A heatmap showing subsets of differentially expressed genes in the regenerating midgut. The log₂ fold changes of expression of subsets of the following are shown: (i) genes upregulated in Pe_day-0, (ii) genes downregulated in Pe_day-0, and (iii) genes upregulated in Pe_day-3. (C) Genes upregulated in response to Pe infection include Dpt (encodes an antimicrobial peptide), upd2, upd3 (JAK–STAT ligands), and vn (an EGFR ligand). The log₂ fold change in expression level relative to uninfected controls is shown. Expression of these genes returns to the normal levels at day-3. (D) Expression of CG6967, cac, and TrpA1 is induced at day-3 after Pe infection. The expression profile of each host gene was determined by RNA-seq and presented in the form of a bar graph. The gene name is indicated in each panel. *P < 0.05; **P < 0.01; ***P < 0.001. P-values were adjusted as described in Materials and Methods section.

Table 1.

Go Term analysis of genes upregulated at day-0 following Pe treatment

Go Term	P-value
DNA metabolic process	2.05E-22
Cellular response to DNA damage stimulus	2.07E-20
DNA repair	2.60E-19
Cell cycle	5.97E-13
Cellular response to stress	7.47E-13
Response to stress	6.11E-12
Cell cycle process	6.94E-12
DNA-dependent DNA replication	2.32E-11
Double-strand break repair	6.54E-11
DNA recombination	9.87E-11
Mitotic cell cycle process	1.05E-09
DNA biosynthetic process	1.28E-06
Nucleotide-excision repair	8.71E-06
Response to stimulus	1.64E-05
Double-strand break repair via homologous recombination	2.03E-05
Cellular response to stimulus	2.23E-05
Nucleic acid metabolic process	3.11E-05
Response to abiotic stimulus	4.95E-05
DNA strand elongation involved in DNA replication	6.24E-05
Reproductive process	4.92E-04
Regulation of cell cycle	7.33E-04
Regulation of cytoskeleton organization	1.03E-03
Nucleobase-containing compound metabolic process	1.14E-03
Reciprocal meiotic recombination	1.22E-03
Oogenesis	1.62E-03
Regulation of developmental process	1.63E-03
DNA damage checkpoint	2.86E-03
Multi-organism process	3.35E-03
DNA integrity checkpoint	4.02E-03
Telomere maintenance	4.25E-03
Nuclear division	4.48E-03
Mitotic DNA integrity checkpoint	5.70E-03
Double-strand break repair via nonhomologous end joining	5.79E-03
Biological regulation	7.20E-03
DNA replication initiation	7.35E-03
DNA duplex unwinding	8.52E-03
Cell differentiation	9.11E-03
Negative regulation of cellular process	9.37E-03
Positive regulation of developmental process	1.07E-02
Microtubule cytoskeleton organization involved in mitosis	1.15E-02
Regulation of microtubule cytoskeleton organization	1.31E-02
Regulation of cellular process	1.34E-02
Base-excision repair	1.34E-02
Positive regulation of cell development	1.40E-02
Pre-replicative complex assembly involved in nuclear cell cycle DNA replication	1.50E-02
Border follicle cell migration	1.50E-02
Negative regulation of cell cycle	1.52E-02
Positive regulation of multicellular organismal process	1.61E-02
Regulation of cell motility	1.81E-02
Regulation of multicellular organismal development	2.07E-02
Response to radiation	2.57E-02
Nuclear chromosome segregation	3.18E-02
Positive regulation of cellular process	3.23E-02
Regulation of cell differentiation	3.28E-02
Cell development	4.23E-02
Regulation of biological process	4.50E-02
G2/M transition of mitotic cell cycle	4.64E-02
Anatomical structure development	4.87E-02

Upregulated genes in the late stage of midgut regeneration

We identified 113 genes that were induced in the termination stage of midgut regeneration (day-3 after Pe infection). Among these “late genes,” 42 genes were significantly upregulated [adjusted P-values < 0.05, log2 fold change ≥1, and log2 fold change_day-0 < log2 fold change_day-3; Supplementary Table S3 and Figure 2B (iii)], with CG6967 showing the highest fold change compared to the mock infection (Figure 2D). Upregulation of CG6967 at the termination stage of regeneration occurred not only with Pe infection (Figure 2D), but also by DSS feeding (data not shown). This suggests that CG6967 is a general termination factor rather than a molecule specific to bacterial infection. CG6967 encodes the Drosophila ortholog of human MOV10 (DIPOT v7.1). Thus, hereafter, we call this gene dMOV10. dMOV10 is uncharacterized, but previous studies on the mammalian MOV10 showed that this class of molecules are critical RNA binding proteins (Meister et al. 2005; Sysoev et al. 2016). Mammalian MOV10 has an RNA helicase activity that unwinds RNA in a 5'-to-3' direction in an ATP-dependent manner (Gregersen et al. 2014). It has been suggested that mammalian MOV10 plays a dual role in the cell. In the nucleus, MOV10 suppresses viral RNAs and retrotranspositions by directly inhibiting cDNA synthesis (Goodier et al. 2012; Skariah et al. 2017). In the cytosol, it binds to the microRNA (miRNA) gene silencing complex (miRISC) and regulates the level of a set of target mRNAs (Meister et al. 2005; Kenny et al. 2014; Skariah et al. 2017).

dMOV10 is induced at the late stage of midgut regeneration

To examine the spatiotemporal control of dMOV10 expression in regenerating midgut, we generated a knock-in allele of dMOV10 (dMOV10-HA) using CRISPR/Cas9 mutagenesis technology. In this strain, spaghetti monster GFP-HA (smGFP-HA) was inserted in frame at the N-terminus of the endogenous dMOV10 gene (Figure 3A). smGFP-HA carries 10 copies of the hemagglutinin (HA) epitope tag inserted into the backbone of fluorescently inactivated superfolder GFP (Nern et al. 2015; Viswanathan et al. 2015), which allows us to monitor endogenously tagged dMOV10 expressed at physiological levels.

Figure 3.

dMOV10 expression in the regenerating midgut. (A) Schematic of the smGFP-HA insertion strategy to the dMOV10 locus through two 1-kb homology arms by homology-directed repair based on CRISPR-Cas9 technology. Insertion of smGFP-HA was confirmed by PCR with two sets of primers (Supplementary Table S2). Gel electrophoresis shows the PCR products with 1,439 bp (green) and 1246 bp (cyan). The expected insertion was verified by Sanger sequencing. (B) dMOV10-HA expression was undetectable in the unchallenged midgut. In this and the following figures, all midgut images show horizontal confocal sections at a basal focal plane (arrowhead in Figure 1B). Nuclei are stained with DAPI (blue). (C) dMOV10-HA expression is low at day-0 after Pe infection. (D) At day-3 after Pe infection, esg>GFP (green) is strongly expressed in ISCs (yellow arrowhead in D') and newly divided ISC-EB pairs (asterisks in D'). ECs are labeled by MyoIA-lacZ (blue). Low levels of GFP remain detectable in newly formed ECs (arrows) but disappear in mature ECs. dMOV10-HA is transiently expressed in newly generated ECs. Scale bars: 25 µm.

We induced midgut regeneration in dMOV10-HA adults by Pe infection and monitored dMOV10 expression by anti-HA antibody staining (Figure 3, B and D). We found that dMOV10 expression is rarely detected in the noninfected midgut (Figure 3B) or very low at day-0 after infection (Figure 3C). At day-3, however, a high level of dMOV10-HA expression was observed (Figure 3D). In Figure 3D', esg>GFP marks ISCs (yellow arrowhead) and newly divided ISC-EB pairs (asterisks). Residual GFP remains in newly formed ECs (arrow) but does not mark other ECs. dMOV10-HA is specifically detected in these nascent ECs as puncta in the cytoplasm (Figure 3D’’). These results indicated that dMOV10 is transiently expressed in newly generated ECs at a late stage of the regenerating midgut. We also observed a similar induction of dMOV10 expression at a later stage of regeneration induced by DSS feeding (data not shown).

The dMOV10-HA allele allowed us to analyze endogenous expression patterns outside of the midgut. Interestingly, we detected dMOV10 expression specifically in the germline stem cells (GSCs) and their immediate daughters in the ovary and testis (Supplementary Figure S1). For example, in the ovary, dMOV10-HA is specifically detected in undifferentiated germ cell populations, GSCs, and cystoblasts. Similarly, in the testis, dMOV10 is expressed in GSCs, goniablasts, and early spermatogonial cells. This transient expression in a population of cells transitioning from the undifferentiated to differentiated state is similar to its transient expression, we observed in newly emerged ISC progenitor cells during regeneration.

Overexpression of dMOV10 at the activation stage inhibits ISC division

To determine if dMOV10 plays a role in controlling ISC proliferation at the late stage of regeneration, we prematurely expressed this gene at the activation stage. As shown in Figure 3D, we have observed that dMOV10 is expressed in newly generated ECs at day-3 after Pe infection. To achieve a reasonable approximation of this expression pattern in nascent ECs, we used esg^ts Flp-Out (esg^tsF/O) (Jiang et al. 2009). This method combines the TARGET system, which provides temporal control of cell-type specific gene expression, and the Flp-Out cassette, which activates a ubiquitous Gal4 driver, with the genotype of esg-Gal4 tub-Gal80^ts UAS-Flp act>CD2>GAL4. esg-Gal4 in combination with tub-Gal80^ts induces expression of Flp recombinase by shifting the culture temperature from 19°C to 30°C. Flp, then, excises the sequence between the two FRT sites of the act>CD2>Gal4 cassette, converting it to the ubiquitously expressed act-Gal4. Using this system, UAS transgenes will be expressed continuously in the ISC progenitors, including newly differentiating ECs.

We induced expression of dMOV10 by crossing esg^tsF/O with UAS-dMOV10 to obtain the progenies (esg^tsF/O-dMOV10). The culture temperature of the esg^tsF/O-dMOV10 animals were shifted to 30 °C one day prior to the Pe treatment, and measured ISC mitotic activity at day-0. Control samples were obtained by crossing esg^tsF/O with wild-type animals (esg^tsF/O-control). We found that dMOV10 expression at day-0 significantly reduced ISC division (Figure 4A). This was also observed for DSS-induced regeneration (data not shown). This finding indicates that dMOV10 has an inhibitory activity on ISC proliferation during midgut regeneration.

Figure 4.

dMOV10 regulates ISC proliferation. (A) Quantification of pH3-positive cells in the midgut overexpressing dMOV10. Using esg^tsF/O, UAS-dMOV10 expression was induced one day prior to the Pe feeding in ISCs and their progenitors, and ISC mitotic activity was measured at day-0 (dMOV10OE). Control animals bear esg^tsF/O without UAS-dMOV10 (Control). Overexpression of dMOV10 reduced the number of mitotic cells. (B) Generation of two dMOV10 null mutant alleles. dMOV10^{KO; tdTomato} (dMOV10KO) was generated by CRISPR/Cas9 mutagenesis technology using a repair template that contains a coding sequence for tdTomato, flanked by 1- and 1.3-kb dMOV10 homology arms. The recombination resulted in the loss of most of the dMOV10 coding sequence, which is replaced with the tdTomato sequence (magenta). dMOV1018, which bears a 2.4-kb deletion in the dMOV10 coding sequence, was generated by FLP/FRT-mediated excision between two piggyBac insertions, pBac{WH}f07741 and pBac{WH}f01447. (C) Quantification of pH3-positive cells in the midgut of wild-type (WT) and dMOV10 null mutant, dMOV10^KO/dMOV10¹⁸. The number of pH3-positive cells were measured at day-0, -3, and -5 after Pe feeding. n.s.>0.05; *P < 0.05; ***P < 0.001. P-values were calculated using the Wilcoxon rank sum test.

Analysis of dMOV10 null mutant alleles during midgut regeneration

We next asked the effects of loss of dMOV10 on ISC proliferation. Toward this purpose, we generated two null mutant alleles of dMOV10. First, using CRISPR/Cas9 mutagenesis technology, the dMOV10 coding region was replaced with tdTomato (dMOV10^{KO; tdTomato}) (Figure 4B left). The other is dMOV10¹⁸, which has a 2.4-kb deletion in the coding sequence generated by FLP/FRT-mediated excision between two pBac insertions, pBac{WH}f07741 and pBac{WH}f01447 (Figure 4B right; (Parks et al. 2004)). We found that homozygous null mutants are viable, which allows us to study dMOV10's role in the adult midgut.

We induced regeneration by Pe feeding in wild-type and dMOV10 mutant (dMOV10^KO/dMOV10¹⁸) midgut and monitored ISC division. In the wild-type control, the number of pH3-positive cells gradually decreased from day-0 to day-3, and returned to the baseline level at day-5 (Figure 4C). There was no significant difference in the number of mitotic cells observed at day-0 between wild-type and dMOV10 mutant, indicating that early regeneration processes occur normally in mutants. At day-3, however, significantly increased numbers of mitotic cells were detected in the dMOV10 mutant midgut compared to wild-type (P < 0.05). The ISC mitotic activity remained high in the mutants, and the difference became even more remarkable at day-5 (P < 0.001). These results indicate that ISCs failed to properly halt division at the termination stage in dMOV10 mutants. We also observed the same phenotype in dMOV10^KO homozygotes and dMOV10¹⁸ homozygotes (data not shown). The severity of phenotype was indistinguishable between the three genotypes, confirming the amorphic nature of both dMOV10^KO and dMOV10¹⁸ alleles.

RISC is required for ISC proliferation during midgut regeneration

The effect of dMOV10 in ISC proliferation suggested that the miRNA pathway has a role in the regenerating midgut. Mammalian MOV10 binds to Argonaute (Ago), a key component of miRISC (Meister et al. 2005). Previous studies suggested that MOV10 is a modulator which affects the activity of miRISC in regulating the levels of a specific set of target mRNAs, rather than acting as a core member of miRISC (Meister et al. 2005; Kenny et al. 2014; Skariah et al. 2017). These studies found that the joining of MOV10 to the complex may either up- or down-regulate target gene products. Drosophila has three Ago homologs: AGO1, AGO2, and AGO3, and AGO1 is responsible for miRNA-RISC assembly (Kobayashi and Tomari 2015).

To test our hypothesis that the miRISC plays a role in the midgut regeneration, we first determined if Ago1 is expressed in the midgut. An analysis using AGO1: GFP^CA06914, a protein trap line with a GFP insertion in the endogenous ago1 locus, showed that AGO1 is abundantly expressed in all cell types in the midgut (Figure 5A). Ago proteins and other key components of the miRNA pathway are known to be enriched in a cytoplasmic compartment called P-bodies (Behm-Ansmant et al. 2006). Consistent with this, AGO1-GFP exhibited cytoplasmic localization with slight enrichment in the perinuclear region (Figure 5A). Based on our mRNA-seq analysis, expression levels of AGO1, AGO2, and dGW182, the major components of miRISC, do not significantly change during the course of regeneration (data not shown). Interestingly, however, co-staining of AGO1-GFP and dMOV10-HA in the midgut bearing both transgenes at day-3 after Pe infection showed that AGO1 expression is higher in dMOV10-expressing cells (Supplementary Figure S2). In these cells, the two proteins are partially, but significantly colocalized throughout the cytoplasm (Manders' Coefficients M1 = 0.811, M2 = 0.72).

Figure 5.

AGO1 is required for ISC proliferation during midgut regeneration. (A) Expression of AGO1 in the posterior midgut as revealed by AGO1: GFP. Nuclei were stained with DAPI. ISCs and EBs are marked by esg-lacZ. (B) Co-immunoprecipitation of dMOV10-Myc and AGO1-FLAG. Myc-tagged proteins were recovered from cell lysate by anti-Myc-agarose beads and eluted with urea. The eluates were analyzed by immunoblotting using anti-FLAG and anti-Myc antibodies. Input, samples incubated with the agarose beads; IP, immunoprecipitation. (C and D) Control midgut (esg^ts>tdTomato) (C) and midgut expressing UAS-AGO1 RNAi (TRiP.HMC03509) in addition to UAS-tdTomato (esg^ts>AGO1^RNAi) (D) obtained from animals cultured for two days after Pe infection at 30°C. EEs are marked by Pros. AGO-GFP expression is lost in Gal4-expressing ISC progenitor cells, marked by tdTomato (yellow arrowheads). (E) Quantification of the number of mitotic cells in the control midgut and the midgut expressing UAS-AGO1 RNAi (TRiP.HMC03509). Scale bar: 25 µm. ***P < 0.001. P-values were calculated using the Wilcoxon rank sum test.

To ask if Drosophila MOV10 has a conserved function in miRISC, we next determined whether dMOV10 physically interacts with AGO1 by co-immunoprecipitation (co-IP) experiments using Drosophila S2 tissue culture cells. We co-transfected S2 cells with expression constructs for Myc epitope-tagged dMOV10 and FLAG-tagged AGO1. dMOV10-Myc was recovered from the cell lysate by immunoprecipitation using anti-Myc antibody agarose beads. We found that AGO1 was co-precipitated with dMOV10 (Figure 5B), showing the ability of dMOV10 to physically interact with AGO1. This is consistent with the information on mammalian MOV10 and suggests that dMOV10 functions with miRISC at the termination stage of midgut regeneration.

To determine if miRISC plays a role in midgut regeneration, we assessed the effect of RNAi knockdown of AGO1 on ISC mitotic activity. Using the TARGET system (esg^ts>), we expressed an UAS-RNAi transgene for AGO1, UAS-AGO1^RNAi (TRiP.HMC03509), for 2 days at 30°C after Pe infection. UAS-tdTomato was used to mark ISCs and the progenitor cells, and AGO1 expression was monitored by AGO1-GFP (AGO1: GFP^CA06914). In control midgut samples (esg^ts>tdTomato), a high level of expression of AGO1-GFP was detected in most cells, including tdTomato-positive progenitors (Figure 5C). In midgut expressing AGO1 RNAi (esg^ts>tdTomato+AGO1RNAi), AGO1-GFP expression was significantly reduced in tdTomato-positive cells, validating the effect of the RNAi transgene (Figure 5D, yellow arrowheads, Supplementary Figure S3A). We first noticed that the number of tdTomato-positive cells was decreased in RNAi-treated samples. AGO1 knockdown midguts showed a 59% reduction in the number of tdTomato-positive cells per area compared to control (data not shown), suggesting impaired ISC division. To measure the ISC mitotic activity, we quantified pH3-positive cells in control and RNAi midgut samples at day-0 after Pe infection. We found that AGO1 knockdown significantly decreases the number of mitotic cells during regeneration (Figure 5E), indicating that the miRNA pathway is required for ISC division at the onset of regeneration. Taken together, these observations suggest that the core members of RISC function in multiple aspects of midgut homeostasis and regeneration. dMOV10 appears to join the complex specifically at the late stage of regeneration to modulate miRISC activity.

Identification of RNA targets of dMOV10-containing miRISC

To elucidate the role of the dMOV10-RISC machinery during midgut regeneration, we identified RNA species that are directly targeted by dMOV10-containing miRISC using a technique, “Targets of RNA-Binding Proteins Identified By Editing (TRIBE),” followed by RNA-seq analyses (McMahon et al. 2016). TRIBE involves the fusion of an RNA-binding protein to the catalytic domain of the Drosophila RNA-editing enzyme ADAR (ADARcd), which catalyzes an adenosine (A)-to-inosine (I) conversion of target RNAs. The RNA-binding region of ADAR is missing from the fusion protein, and therefore its target specificity is determined by the RNA recognition features of the fused RNA-binding protein. We constructed a fusion protein of dMOV10 and ADARcd (dMOV10-ADARcd) and generated a transgenic strain bearing UAS-dMOV10-ADARcd (Figure 6A).

Figure 6.

Identification of dMOV10-binding mRNAs and the role of Dad in midgut regeneration. (A) Schematic of TRIBE analysis for identifying dMOV10-binding mRNAs. The fusion protein of dMOV10 and the catalytic domain of dADAR (dADARcd) was expressed using esg^tsF/O in the midgut. Midgut samples expressing only dADARcd are used as control. ADARcd catalyzes an A-to-I conversion. The target binding specificity of the fusion protein is determined by dMOV10. The A-to-I editing events by the dMOV10-dADARcd were detected by mRNA-seq. (B) A TRIBE site in the Dad 3' UTR. Black box, coding sequence; horizontal lines, 5' and 3' UTRs; red line, TRIBE site. (C) RNA immunoprecipitation. S2 cells were transfected with Dad-GFP and Myc-dMOV10 constructs, with or without an AGO1 cDNA (dMOV10+AGO1 and dMOV10, respectively). dMOV10 was immunoprecipitated with protein G-sepharose fixed with anti-Myc antibody. Control sample was prepared without dMOV10 cDNA. Relative levels of Dad mRNA recovered were measured by RT-qPCR (the control value was set as 1). Dad mRNA was recovered by the immunoprecipitation of dMOV10 in the presence of AGO1. Values are mean±s.e.m. n.s.>0.05; *P<0.05; **P<0.01. The P-value was calculated using one-way ANOVA followed by Bonferroni test. (D) Dad expression in the midgut of a Dad: GFP transgenic construct, PBac{Dad-GFP.FLAG}VK00037. All nuclei are stained with DAPI (blue) and ECs are labeled by MyoIA-lacZ (magenta). Dad: GFP is detectable in a fraction of ECs (green). Scale bar: 25 µm. (E) The effect of Dad RNAi knockdown on ISC division. The number of pH3-positive cells was measured in control (MyoIAts>) and Dad RNAi (MyoIA^ts>Dad RNAi) animals at day-0 after Pe infection. (F) The genetic interactions between dMOV10 and Dad. Control animals bear esg^tsF/O (Control). UAS-dMOV10 expression was induced using esg^tsF/O one day prior to the Pe infection and ISC mitotic activity was measured at day-0, as in Figure 4A (dMOV10OE). dMOV10 was co-expressed with UAS-Dad (dMOV10OE+DadOE) or UAS-Dad RNAi (dMOV10OE+DadRNAi). (G) The genetic interactions between dMOV10 and tkv. The number of pH3-positive cells were quantified in wild-type (WT), dMOV10^KO/dMOV10¹⁸ (dMOV10/dMOV10) and dMOV10^KO tkv⁷/dMOV10¹⁸ + (dMOV10 tkv/dMOV10) at day-0, -3, and -5 after Pe feeding. n.s.>0.05; *P < 0.05; **P < 0.01; ***P < 0.001. The P-value was calculated using the Wilcoxon rank sum test.

We expressed dMOV10-ADARcd using esg^tsF/O in ISCs and their progenitor cells, to permanently mark endogenous dMOV10 target transcripts with the editing event by dMOV10-ADARcd (the A-to-I conversion). The modified RNAs were identified by RNA-seq analysis. Control samples were made by expressing of UAS-ADARcd. The sequence data were mapped to the Drosophila genome and unique sequences with the A-to-I modification, or TRIBE sites, were identified. The TRIBE assay identified 408 dMOV10 targets in the regenerating intestine (Supplementary Table S4). Out of these targets, 101 genes altered their expression at day-0 after Pe infection in the mRNA-seq analysis, whereas only 15 were regulated at day-3 (Supplementary Table S3), showing that dMOV10 binding is correlated with a return to basal expression. The TRIBE-identified targets include a number of cell signaling components (e.g., fz4, Dsor1, babo, Dad, Dome, and Sara) (Supplementary Table S5). In addition, they include actin filament network regulators (Fim and spir) and ECM receptors (Sdc and Dg). The identification of these molecules as dMOV10 targets may reflect that filopodia-like cellular extension formation, basement membrane remodeling, and the restoration of epithelial integrity are functionally important during regeneration.

Dad in midgut regeneration

One of the dMOV10 targets identified by TRIBE was Dad. A TRIBE site was found in the Dad 3'UTR (Figure 6B). We confirmed the binding of dMOV10 and Dad mRNA by RNA-IP experiment using S2 cells. S2 cells were transfected with Dad and Myc-dMOV10 constructs, with or without an AGO1 cDNA. We found that Dad mRNA was co-precipitated with dMOV10 when AGO1 was co-transfected (Figure 6C). This AGO1-dependent binding of dMOV10 to a target RNA strongly suggests that dMOV10 functions in the RISC.

Dad is the Drosophila ortholog of human SMAD7, an inhibitory Smad. Vertebrate inhibitory Smads play crucial roles in the negative regulation of TGF-β family signaling by interacting with type I receptors and inhibiting phosphorylation of R-Smads (Hayashi et al. 1997; Imamura et al. 1997; Nakao et al. 1997). Expression of inhibitory Smads is induced in response to various ligands of the TGF-β family and thus creates a negative-feedback loop for the pathway (Tsuneizumi et al. 1997; Inoue et al. 1998). Dpp/BMP signaling is required for midgut homeostasis and regeneration (Li et al. 2013; Ayyaz et al. 2015; Zhou et al. 2015; Tian et al. 2017). To determine the role of Dad in midgut regeneration, we first examined Dad expression in the midgut using a transgenic strain, Dad: GFP (Dad^{sfGFP, Tag: FLAG}). This strain bears a genomic fragment including the Dad gene with a sfGFP-Tag: FLAG tag cassette inserted at the C-terminal end of the Dad coding sequence (PBac{Dad-GFP.FLAG}VK00037). Since this strain has an extra copy of the Dad gene, we observed Dad: GFP expression in a Dad^P1883/+ heterozygous background. In the midgut, Dad: GFP expression was detected in the nuclei of most ECs (Figure 6D). Punctate signals of Dad: GFP were also observed in the cytosol. The nuclear staining was much weaker in some ECs; however, this could reflect either heterogeneity of EC populations or technical issues of staining. Since dMOV10 is expressed in newly generated ECs, it is likely that dMOV10-containing RISC forms at an early stage of EC differentiation and regulates target genes in differentiating and mature ECs.

Our RNA-seq analysis showed that Dad mRNA is upregulated at day-0 and returns to the normal level at day-3 (Supplementary Table S3). To ask if Dad plays a role in midgut regeneraion, we performed an RNAi knockdown experiment of Dad mRNA. First, we validated the efficacy of UAS-Dad RNAi (TRiP.HMS01102). We confirmed that the expression of UAS-Dad RNAi in the dorsal compartment of the wing disc eliminated Dad: GFP signals in the dorsal cells (Supplementary Figure S3, B and C). We next examined the effect of Dad RNAi on ISC division after Pe infection. Since Dad is strongly expressed in mature ECs (Figure 6D), UAS-Dad RNAi expression was induced in ECs by TARGET using an EC driver, MyoIA-GAL4 (MyoIAts>Dad RNAi). Control samples were obtained from animals with MyoIA-GAL4 and tub-Gal80^ts but no UAS transgene. We found that Dad RNAi knockdown significantly reduced the number of mitotically active ISCs at day-0 after Pe infection (Figure 6E). These results are consistent with the idea that dMOV10 affects ISC division at the termination stage of midgut regeneration via affecting the levels of target transcripts, including Dad mRNA.

We next examined the genetic interactions between dMOV10 and Dad genes (Figure 6F). As shown in Figure 4A, dMOV10 overexpression at day-0 after Pe infection significantly reduced the ISC mitotic activity. We found that co-expression of Dad with dMOV10 partially suppressed this phenotype, showing competing effects of dMOV10 and Dad overexpression (Figure 6F). On the other hand, when we knocked down Dad by co-expressing a Dad RNAi transgene with dMOV10, no significant change was observed from the effect of dMOV10 overexpression alone. These results suggest that dMOV10 overexpression effectively impairs Dad function, and therefore a further reduction in Dad mRNA does not appear to cause a major difference. Thus, Dad mRNA may be one of the key downstream targets of dMOV10 for its function to inhibit ISC proliferation. We also observed that deleting one copy of thickveins (tkv), a BMP Type I receptor, enhanced dMOV10 mutant phenotype (Figure 6G). Together, our results support the idea that dMOV10 regulates ISC division by influencing the Dpp/Dad regulatory module.

miRNA transcriptional profile during midgut regeneration

In Drosophila, miRNAs associate primarily with AGO1 protein to form the RISC (Okamura et al. 2004; Förstemann et al. 2007; Tomari et al. 2007). The miRNA functions as a sequence-specific guide that recognizes target mRNAs via pairing to complementary sites within the 3' UTRs (Lai 2002; Brennecke et al. 2005; Bartel 2009). miRNAs typically reduce the stability or repress translation of these mRNAs, thus downregulating target gene activity. Our finding on the functions of AGO1 and dMOV10 during midgut regeneration suggests that miRNAs play roles in regulating the regeneration process.

To determine differential expression patterns of miRNAs during midgut regeneration, we performed small RNA-seq using RNA samples from different time points of regeneration (Suc_day-0, Pe_day-0, and Pe_day-3). Interestingly, our small RNA-seq analysis did not show any miRNA upregulated in the earlier stage of midgut regeneration (fold change ≥1 and P < 0.05) (Supplementary Table S6). Given that miRISC plays an essential role to promote ISC division (Figure 5), this may suggest that regeneration initiation occurs without upregulated synthesis of novel miRNAs. On the other hand, we identified 6 miRNAs that are significantly upregulated in the later stage of midgut regeneration (“late” miRNAs): mir-1, mir-34, mir-314, mir-996, mir-1007, and mir-10404 (Figure 7 and Supplementary Table S6).

Figure 7.

miRNA transcriptional profile during midgut regeneration. (A) Small RNA-seq analysis showed that expression of 6 miRNA, mir-1, mir-34, mir-314, mir-996, mir-1007, and mir-10404, is induced at day-3 after Pe infection. (B) Effects of loss- or gain-of-function of mir-1 on ISC mitotic activity. Expression of UAS-LUC-mir-1 (mir-1 OE) or UAS-mir-1-sponge (mir-1 SP) was induced using esg^tsF/O one day prior to the Pe infection, and the number of pH3-positive cells was measured at day-0 and -3. esg^tsF/O without UAS transgenes was used as control. *P < 0.05; **P < 0.01; ***P < 0.001. The P-value was calculated using the Wilcoxon rank sum test.

Target sequences of these miRNAs are found in mRNAs that encode “early” regeneration genes. Table 2 summarizes these potential target genes based on the following criteria. First, using TargetScanFly v7.2 (Agarwal et al. 2018) and DIANA-microT-CDS v5.0 (Reczko et al. 2012), we first identified mRNAs that contain target sequences of the five miRNAs in their 3'UTR. Neither program identified any mRNA targets for mir-10404, and thus it is not included in Figure 7 and Table 2. Second, among those, we selected genes that have been identified to be upregulated at day-0 (early genes) in our earlier RNA-seq analysis (Figure 1 and Supplementary Table S3). These genes are induced by Pe feeding at day-0, but their expression levels significantly go down near the baseline at day-3, and therefore are possible targets of these miRNAs.

Table 2.

miRNAs induced at day-3 after Pe infection and their potential target genes Our miRNA-seq analysis identified six miRNAs, mir-1-3p, mir-34-5p, mir-314-5p, mir-996-3p, mir-1007-5p, and mir-10404-5p that were significantly upregulated in the later stage of midgut regeneration (day-3 after Pe feeding). Potential target genes that bear a target sequence for respective miRNA were obtained from both TargetScanFly v7.2 and DIANA-microT-CDS v5.0. The targets listed here are selected from the genes identified to be upregulated at day-0 after Pe infection and to return to the baseline at day-3 in our RNA-seq assay (Supplementary Table S3). We did not identify any mir-10404-5p target genes that fall under these criteria.

mir-1-3p	mir-34-5p	mir-314-5p	mir-996-3p	mir-1007-5p
Act42A	CG30062	alpha-Man-IIb	Atg7	CG31694
Asl	CG3408	arc	brat	dysc
CG10098	CG42240	axed	bsh	Eip78C
CG12948	CG42319	CG10011	CG15097	Ekar
CG16995	dysc	CG10494	CG6752	Hr3
CG32195	hid	CG13516	Dhit	inaE
CG32369	Hr4	CG42240	GlcT	LRP1
CG32939	Loxl1	CG42541	Hr38	spri
CG6330	luna	CG4945	mus81	sqz
cno	mys	CG6330	Nrk	—
DAT	Nrk	CG6428	rho	—
Gli	Plod	CG6841	sug	—
Hsepi	Sin1	crol	tamo	—
l(2)gd1	sqz	Eip75B	—	—
Mrtf	Swim	eys	—	—
Phf7	trbl	Hsp22	—	—
plum	CG30062	MCU	—	—
upd3	CG3408	numb	—	—
—	CG42240	pins	—	—
—	—	plum	—	—
—	—	Rhp	—	—
—	—	RNaseZ	—	—
—	—	skd	—	—
—	—	Slimp	—	—
—	—	sol	—	—
—	—	Swim	—	—
—	—	Syx13	—	—

To ask if some of the genes listed in Table 2 are affected by dMOV10, we performed RT-qPCR experiments using control and dMOV10 mutant midgut samples for 11 genes randomly picked from the list. We found that the levels of most mRNAs we have tested (8 out of 11) were higher in dMOV10 mutants compared to control at day-3 after Pe infection. A few examples are shown in Supplementary Figure S4A. Similar results were also observed in AGO1 RNAi knockdown animals (Supplementary Figure S4B). This is consistent with the idea that dMOV10 is involved in destabilizing these mRNAs. In addition, the target sequences of these six late miRNAs are found in the 3' UTR of dMOV10 target mRNAs identified by TRIBE (Supplementary Table S7). Among 408 dMOV10 target genes, we identified 96 genes that are potential targets of the late miRNAs. These observations suggest that some of the late miRNAs may function together with dMOV10 at regeneration termination.

mir-1 plays a role in midgut regeneration termination

mir-1 is one of the six “late” miRNAs that are upregulated at day-3 after Pe infection. It has a binding site in the 3' UTR of upd3 (Table 2), which is substantially induced in the earlier stage of regeneration and plays a major role to promote ISC division [Figure 1C; (Jiang et al. 2009)]. This suggests the possibility that in addition to transcriptional inactivation of upd3, miR-1 may function to eliminate residual upd3 mRNA to ensure proper regeneration termination.

To test this idea, we examined the effect of loss-of-function of mir-1 using mir-1-sponge. UAS-inducible miRNA “sponges” have multiple copies of sequences that are complementary to a specific miRNA and thus, sequester the targeted miRNAs (Fulga et al. 2015). We expressed UAS-mir-1-sponge using esg^tsF/O to knock down mir-1 and quantified ISC mitotic activity during Pe-induced regeneration. We found that the ISC division rate is significantly higher in mir-1-sponge animals compared to control (Figure 7B). The effect was more substantial at day-3, when mir-1 expression is normally upregulated. Consistent with this loss-of-function phenotype, we also observed that overexpression of mir-1 [using UAS-LUC-mir-1 (Bejarano et al. 2012)] resulted in a significant decrease in the number of pH3-positive cells at day-0 and day-3 compared to control values (Figure 7B). Together, these results indicated that miR-1 contributes to the shutdown of ISC proliferation at the termination stage.

In addition to upd3, the potential targets of the “late” miRNAs include genes that are involved in proliferation and asymmetric division of neuroblasts (neural stem cells; asl, brat, and pins), cell adhesion and epithelial integrity (Swim, crol, Loxl1, and mys), and cell death (CG13516, Sin1, and hid). These observations are consistent with the idea that these biological processes are important during regeneration, and their regulators are transiently activated by tissue damage. Our study suggests that post-transcriptional regulation by the miRNA-mediated system plays an important role in regeneration as a parallel mechanism of transcriptional regulation.

Discussion

Drosophila midgut regeneration involves controlled ISC division and epithelial/ECM remodeling. Although the mechanisms triggering increased mitotic activity of stem cells at the beginning of regeneration are extensively studied, it is poorly understood how stem cell proliferation is downregulated at the termination of regeneration. It is generally accepted that mitogen ligands, which promote stem cell proliferation, are transcriptionally inactivated at the termination stage, leading to slowed stem cell division. Many of these ligands require HSPGs as a co-receptor (HS-dependent factors). Our previous study showed that a specific modification of HS contributes to the shutdown of mitogen signaling at the termination stage (Takemura and Nakato 2017). Thus, a posttranscriptional mechanism plays a role in regeneration termination.

Our current study proposes that another posttranscriptional mechanism, the miRNA-mediated network, provides a new regulatory step for precise temporal control of mitogen signaling, and thus stem cell activity, during regeneration. dMOV10, a regulatory subunit of miRISC induced at the late stage of midgut regeneration, is able to inhibit ISC division when overexpressed at an early stage. Furthermore, dMOV10 is required to properly shut down ISC division at regeneration termination. We identified direct target mRNAs of dMOV10, including Dad. Dad RNAi knockdown significantly suppressed ISC proliferation, consistent with the idea that the dMOV10-miRISC machinery affects ISC division via controlling the level of target transcripts at the termination stage of midgut regeneration (Figure 8). When dMOV10 joins the miRISC complex, it may either activate/inactivate the complex, or modulate target RNA specificities. This leads to up- or down-regulation of target RNAs, which results in the proper control of ISC division. We also identified miRNAs that are specifically induced at the late stage of regeneration. The target sequences of these miRNAs were found in the 3' UTR of various mRNAs that are expressed in rapid response to tissue damage, suggesting that the miRNAs contribute to the proper transition from regenerative to homeostatic phases. For example, mir-1, of which targets include upd3, is required for rapid shutdown of ISC division at regeneration termination.

Figure 8.

Model for dMOV10 function during regeneration termination. The diagram shows a model for the role of the dMOV10-miRISC machinery in midgut regeneration. When dMOV10 joins the miRISC complex at the termination stage of midgut regeneration, it may either activate/inactivate the complex, or modulate target RNA specificities. This leads to the proper control of ISC division and restoration of the epithelial architecture.

A number of our observations, including the binding and colocalization of dMOV10 and AGO1, the AGO1-dependent binding of dMOV10 to a target RNA, and the increased levels of the target mRNAs of the late miRNAs in dMOV10 mutants, support the idea that dMOV10 functions as a RISC component (Figure 8). However, our study does not completely exclude a possibility that dMOV10 has a function independent of RISC. Future studies, such as proteomic analyses of dMOV10-associated proteins and functional assays of putative dMOV10 RNA helicase activity, will be needed to address this issue.

Dad encodes the Drosophila ortholog of SMAD7, an inhibitory Smad. Because TGF-β signaling is a critical factor of neoplastic processes in various organs, SMAD7 is associated with many different types of cancers (Briones-Orta et al. 2011; Stolfi et al. 2013; Hu et al. 2014; Luo et al. 2014). For example, a genome-wide association study has shown that SMAD7 influences susceptibility to colorectal cancer (CRC) (Broderick et al. 2007; Hu et al. 2014). SMAD7 expression is significantly increased in human CRC samples and CRC cell lines, and it contributes to CRC cell proliferation and survival (Stolfi et al. 2014). Interestingly, recent studies have shown that SMAD7 expression is controlled by multiple miRNA molecules, and its dysregulation has been linked to cancer formation and prognosis after therapy (Liu et al. 2017; Nabhan et al. 2017; Schneiderova et al. 2017; Shaker et al., 2017). For example, polymorphisms in miR-375 binding site in the 3' UTR of SMAD7 gene have a significant impact on susceptibility to CRC (Shaker et al., 2017); similarly, polymorphisms in miR-181a binding sites impact the development of acute lymphoblastic leukemia (Nabhan et al. 2017). Our study suggests that Inhibitory Smads also play a role in proper termination of regeneration, preventing tumorigenesis.

This study provides novel resources for studying miRNA-regulated processes in midgut regeneration. Further study may identify additional key target genes controlled by miRNAs in stem cell progenitors in multiple stem cell systems. Given the evolutionarily conserved functions of miRISC and its targets, this line of studies can help build a knowledge foundation for the future development of targeted cancer therapies.