Exosomes secreted by nematode parasites transfer small RNAs to mammalian cells and modulate innate immunity

In mammalian systems RNA can move between cells via vesicles. Here we demonstrate that the gastrointestinal nematode Heligmosomoides polygyrus, which infects mice, secretes vesicles containing microRNAs (miRNAs) and Y RNAs as well as a nematode Argonaute protein. These vesicles are of intestinal origin and are enriched for homologues of mammalian exosome proteins. Administration of the nematode exosomes to mice suppresses Type 2 innate responses and eosinophilia induced by the allergen Alternaria. Microarray analysis of mouse cells incubated with nematode exosomes in vitro identifies Il33r and Dusp1 as suppressed genes, and Dusp1 can be repressed by nematode miRNAs based on a reporter assay. We further identify miRNAs from the filarial nematode Litomosoides sigmodontis in the serum of infected mice, suggesting that miRNA secretion into host tissues is conserved among parasitic nematodes. These results reveal exosomes as another mechanism by which helminths manipulate their hosts and provide a mechanistic framework for RNA transfer between animal species.

P arasitic nematodes are ubiquitous pathogens of plants and animals, including species that infect over 2 billion people and generally reside in extracellular niches in their hosts. H. polygyrus is a parasite related to human hookworm that naturally infects mice, and is in the same nematode clade as Caenorhabditis elegans 1 . Within the mouse host, the parasite life cycle is exclusively intestinal: following ingestion, the larvae invade the small intestine, moult into adult worms and emerge into the lumen of the duodenum to mate and to produce eggs expelled in the faeces. The infection induces Type 2 innate and adaptive (Th2) immune responses in parallel with a large expansion of regulatory cells that mediate immunosuppressive effects [2][3][4] , some of which have beneficial properties in allergy and auto-immunity 5 . Immune suppression has been shown to be mediated in part by a suite of immunomodulatory proteins actively secreted by the nematodes 6,7 . Given the burgeoning body of data detailing extracellular small RNAs in mammalian systems, and emerging evidence that these can mediate cell-to-cell communication 8 , it is intriguing to think this mechanism could also be used by parasites. Small RNAs derived from bacteria, plants and parasites have been detected in human body fluids [9][10][11][12] ; however, the mechanism by which these are secreted or excreted is unknown, and the meaning of their extracellular existence unclear 13 . We show here that H. polygyrus secretes a specific set of miRNAs and full-length Y RNAs that are stabilized against degradation by encapsulation within vesicles. The vesicles are of intestinal origin and are enriched for homologues of mammalian proteins found in exosomes, including heat shock proteins, tetraspanins and ALIX, a protein associated with exosome biogenesis 14,15 as well as a nematode Argonaute (Ago) protein.
Local administration of the nematode exosomes to mice by the intranasal route suppresses Type 2 innate responses and eosinophilia induced by the allergen Alternaria in vivo. The nematode vesicles are internalized by mouse intestinal epithelial cells in vitro and suppress genes involved in inflammation and immunity, including the receptor for the alarmin IL-33 and a key regulator of mitogen-activated protein kinase (MAPK) signalling, DUSP1. This work identifies exosomes as a new class of immunomodulatory complex produced by helminths and provides the first steps towards a mechanistic framework for RNA-mediated communication between animal species.

Results
Small RNAs in H. polygyrus secretory products. Total RNA was extracted from the secretory products of H. polygyrus and compared with the profile of small RNAs in adult nematodes, eggs and infective larvae. A heterogeneous population of small RNAs o25 nucleotides (nt) was observed in all samples and several additional species between 25 and 30 nt were apparent in the secretion product (Fig. 1a). Small RNA sequencing (o30 nt) identified miRNAs as the dominant class of secreted parasite small RNA (Fig. 1b) and also identified RNA fragments mapping to nematode stem-bulge RNAs, herein referred to as Y RNAs based on their recognized homology to this class of small RNA 16 (Fig. 1b). In contrast, piRNAs (or '21-U' RNAs) were exclusively identified in the adult library (B14% of reads; Supplementary  Fig. 1 and Supplementary Data set 1) with no evidence of secretion. RNAs between 70 and 100 nt were also present in the secreted product (Fig. 1a) and sequencing identified full-length Y RNAs as the major component of this fraction, with two predominant classes of structure identified ( Fig. 1c and Supplementary Fig. 2).
A total of 263 'high confidence' miRNAs were classified from the combined libraries based on representation of reads from both 5p and 3p arms of the hairpin and/or homology with known miRNAs in other nematodes (Supplementary Data set 1). Each of these matches DNA sequences in the H. polygyrus whole genome currently under assembly. Many also show stage-specific expression patterns: for example, miR-35 family members are exclusively expressed in the egg library, consistent with their functions in embryogenesis 17 (Fig. 2a, boxed). The secretory products are dominated by miRNAs with identical seed sites to mouse miRNAs, many of which are ancient: six are shared among Uncharacterized 10% Eumetazoa (lin-4/miR-125 and five miR-100 family members, Fig. 2a, red) and five among Bilatera (miR-79/miR-9, miR-83/ miR-29, miR-263/miR-183 and two let-7 family members, Fig. 2a, blue). In addition, five bantam family members dominate the secretory products along with miR-87 and miR-60, which also are shared among Protostomia (Fig. 2a, green) and three miRNAs that evolved in the nematode lineage: miR-63, miR-239 and miR-77 (Fig. 2a, orange). miR-63 shares an identical seed site to mammalian miR-425, although it is not of common ancestry 18 , Fig. 2b. On the basis of their sequences, many of the secreted parasite miRNAs could therefore hijack existing mouse miRNA target networks if taken up by host cells.
Nematode vesicles are associated with secreted RNA. In mammalian systems, miRNAs have been found in body fluids in association with specific proteins or in extracellular vesicles 8 .
To determine whether these RNAs could be present in vesicles, the H. polygyrus secretory products were ultracentrifuged and quantitative reverse transcription-PCR (qRT)-PCR used to measure miRNA levels in the pellet and supernatant, revealing the majority to be present in the pellet ( Supplementary Fig. 3). Transmission electron microscopy (TEM) identified vesicle-like structures between 50 and 100 nm in diameter in the pelleted material ( Fig. 3a). Label-free quantification of proteins in the vesicles and supernatant by liquid chromatography-electrospray tandem mass spectrometry LC-MS/MS identified 362 proteins, of which 139 were specifically enriched in the vesicle fraction (Po0.05, Fig. 3b, Supplementary Data set 2) including homologues of mammalian proteins present in exosomes: heat shock proteins, Rab proteins, tetraspanins 19 and Alix, which is associated with exosome biogenesis 14,15 (Table 1). The venom allergen-like proteins (members of the CAP superfamily, Pfam00188), which were previously identified as the dominant proteins in the H. polygyrus secretory products 20 , are almost exclusively in the supernatant fraction (Fig. 3b, orange), further demonstrating specificity in the molecular composition of the vesicles and possibly indicating distinct routes of secretion. On this note, nematode intestinal proteins are enriched in the vesicle fraction (Fig. 3b, green and Supplementary Data set 2) and, consistent with an intestinal origin of the nematode exosomes, vesicles of similar size are observed in the intestinal tissue of adult H. polygyrus analysed immediately ex vivo (Fig. 3c). One Argonaute protein was identified in both vesicle and supernatant fractions (Fig. 3b, red) that belongs to the clade of Worm-specific Agos (WAGO). Phylogenetic analysis suggests that homologues to this WAGO are present in many parasitic nematodes but may have been lost in Caenorhabditis (Fig. 3d).  Nematode RNAs are protected from degradation by exosomes.
To determine which RNAs identified in the total secretion product ( Fig. 1) are specifically associated with vesicles, small RNA sequencing of replicate vesicle and nonvesicle (supernatant) fractions of the secretion product was carried out. Results from three biological replicates demonstrate that the parasite miRNAs are enriched in the vesicle fractions (75% of reads compared with 10% in supernatant, which is dominated instead by rRNA and Y RNA fragments, Fig. 4a). This analysis also identified three mouse miRNA homologues: miR-193, miR-10 and miR-200, within the top five most abundant secreted miRNAs. These were ranked much lower in the initial Illumina analysis (Supplementary Table 1) likely because of the sequencing bias of the different kits and platforms 21 , underscoring the importance of comparing both approaches. Overall the three replicates showed the same profile of miRNAs in each vesicle sample (Supplementary Table 2) and neither vesicle nor supernatant contained intact large ribosomal RNA ( Supplementary Fig. 4). Northern blot analysis confirmed the specificity of small RNA biotypes in vesicles versus supernatant, showing miR-100 to be exclusively present in the vesicles and the Y RNA fragment to be exclusively present in the supernatant (Fig. 4b). Notably, on the same blot the full-length Y RNA was detected in the vesicles and both the miRNA and full-length Y RNA were largely resistant to degradation by RNases in untreated samples but became susceptible in the presence of Triton-X-100 (Fig. 4c)   these results demonstrate that mature miRNAs and full-length Y RNAs are secreted by a parasitic nematode and are protected through encapsulation within vesicles of intestinal origin that share similar size and protein composition to mammalian exosomes.
Helminths are well known to suppress pathogenic immune responses in both the gastrointestinal tract and airways 5 . To examine the functionality of the parasite-derived exosomes in vivo, they were administered intranasally in combination with extracts of the allergenic fungus Alternaria, which induces rapid IL-33 release as part of the Type 2 Th2-like innate immune response that leads to lung eosinophilia 22 . Pre-treatment with parasite-derived exosomes before Alternaria extract administration led to a sharp reduction in bronchoalveolar lavage eosinophilia (Fig. 5a), and suppressed expression of the Type 2 cytokines interleukin (IL)-5 and IL-13 by innate lymphoid cells (ILCs; Fig. 5b,c). Neutrophilia, which does not depend on Type 2 cytokines, was undiminished by exosome administration (Fig. 5d). Intriguingly, the overall expression of the IL-33 receptor (also known as ST2) was also suppressed in recipients of exosomes (Fig. 5e).

Internalization of nematode exosomes and RNAs by mouse cells.
To determine whether the nematode-derived exosomes can enter mammalian cells, uptake was examined in mouse small intestinal epithelial cells, a cell type that is naturally in direct contact with H. polygrus in vivo. Exosomes were labelled with the lipid dye PKH67 and incubated with MODE-K cells in vitro. Uptake was analysed by fluorescence-activated cell sorting (FACS) and confocal microscopy. Over 60% of the cells were PKH67-positive after 1 h of incubation with H. polygyrus vesicles compared with 1.5% when incubated with background dye ( Supplementary Fig. 5). These results are unlikely to be due to nonspecific association with the cell membrane as treatment with trypsin did not eliminate the signal ( Supplementary Fig. 5). Confocal analysis confirmed uptake to the cytoplasm and demonstrates that this requires physiological temperature (Fig. 6a). qRT-PCR analysis of the treated cells detects the parasite-specific miRNAs in cells after 20 h of incubation, with no change in the endogenous miR-16 (Fig. 6b). The full-length parasite-derived Y RNA could also be detected by northern blot analysis in cells which were treated directly with exosomes followed by washing (Fig. 6c).
Regulation of mouse genes by nematode exosomes. To determine the function of these vesicles in mouse cells, gene expression analyses were carried out on MODE-K cells following incubation with H. polygyrus exosomes. A total of 128 genes were differentially expressed upon treatment (false discovery rate (FDR) Po0.05). Relatively subtle changes in gene expression were observed (Fig. 7a); however, the most strongly downregulated gene was Dusp1 (also known as MKP-1 in human), a key regulator of MAPK signalling associated with dampening the type 1 pro-inflammatory reaction to Toll like receptor (TLR) ligands. Another gene significantly downregulated by exosomes is Il1rl1 (also known as IL33R in human and so referred to here as Il33r), the ligand-specific subunit of the receptor for IL-33, a key alarmin cytokine required for protection against multicellular parasites, which is produced by innate cells to drive early type 2 immune responsiveness 23 and is suppressed by the exosomes in ILCs in vivo (Fig. 5e). The effects of the exosomes on Dusp1 and Il33r were validated by RT-qPCR and are unlikely to reflect a nonspecific response to vesicle uptake as exosomes derived from mouse intestinal cells showed similar uptake but did not alter Dusp1 and Il33r levels ( Fig. 7b and Supplementary Fig. 5).
There are a number of potential mechanisms that could mediate the decrease in Dusp1 and Il33r levels. The 3 0 untranslated region (UTR) of Dusp1 is highly conserved and contains 7mer binding sites for the parasite homologues of mouse miR-200 (aka miR-8) and let-7 as well as a 6mer site for miR-425 (aka miR-63) in between these sites ( Supplementary Fig. 6). We therefore examined whether the parasite miRNAs could suppress translation of a reporter vector containing the 3 0 -UTR of Dusp1 fused to luciferase. Synthetic parasite miRNAs were transfected into MODE-K cells, resulting in 1.2-to 2.0-fold reduction in luciferase levels for the Dusp1 reporter but not control (Fig. 7c). Notably, transfection of a cocktail of three of the miRNAs (at the same total RNA concentration) resulted in an increased reduction in luciferase activity (3.1-fold). This is consistent with enhanced repressive effects of miRNA sites in close proximity and suggests that secreted parasite miRNAs could work in cooperation to exert maximal effects on host genes. In contrast, the 3 0 -UTR of the IL33R-encoding gene Il1lr1 is not conserved and, although binding sites for some of the secreted miRNAs were identified, we did not observe repression of a Il1rl1 3 0 -UTR reporter by transfection of miR-71, which contained two 7mer sites ( Fig. 7c and Supplementary Fig. 6).
Circulating nematode miRNAs in serum. To establish whether the secreted nematode miRNAs naturally circulate in host tissues in vivo, we examined serum from mice infected with H. polygyrus (which resides in the gut lumen) or the filarial nematode L. sigmodontis (which resides in the pleural cavity). No H. polygyrus miRNAs were detected in the serum; however, a total of 1,188 reads mapped perfectly and unambiguously to the L. sigmodontis draft genome and 761 of these derived from 16 nematode miRNAs (Table 2 and Supplementary Table 3). Although we cannot rule out the possibility that some of the miRNAs in serum could derive from dying worms, the most abundant miRNAs detected are homologues of those found in H. polygyrus exosomes, including miR-100, bantam, miR-71 and miR-263 (Table 2, Fig. 8). These data confirm the in vivo secretion of parasite miRNAs and are consistent with the idea that exosomes and associated RNAs operate locally in the host's body such that their detection in body fluids will be dictated by the life stage and localization of the parasite in the host.

Discussion
In summary, we have shown that nematode parasite-derived miRNAs and Y RNAs are transported into mammalian host cells via exosomes that regulate host genes associated with immunity and inflammation and suppress an innate Type 2 response in vivo. Extracellular vesicles are emerging as a central mechanism for cell-to-cell signalling within mammalian systems, and our report of their secretion by a nematode species is within the setting of vesicle secretion by an increasingly diverse range of pathogens [24][25][26] . We have demonstrated for the first time that nematode-derived RNAs are a key component within exosomes that can be transferred to host cells. Nematodes are ubiquitous pathogens of both plants and animals and we anticipate that RNA secretion is a conserved phenomenon, supported by the fact that we detect miRNAs from the filarial nematode L. sigmodontis in host tissue, consistent with a recent report 12 . In fact, RNA secretion may be a ubiquitous feature across a range of parasites; an initial report suggests that miRNAs are also associated with vesicles in the trematode Dicrocoelium dendriticum 27 .
Given that many of the nematode miRNAs are homologues to mouse miRNAs, it is tempting to speculate that these could tap into existing miRNA regulatory networks in host cells. In support of this, we show with a reporter assay that three of the secreted nematode miRNAs that have identical seed sites to mouse miRNAs can together downregulate DUSP1 through conserved sites in its 3 0 UTR. Many questions remain, however, regarding the mechanism by which the nematode miRNAs can operate in host cells. The exosome is a functional ensemble and immune suppression is likely to require a combination of protein and miRNAs for fusion and gene regulation. It will be challenging, therefore, to pin point the individual contributions of each. For example, a nematode Ago protein is secreted with the miRNAs that may be required for functionality. This Ago belongs to the WAGO clade of Agos that evolved in the nematode lineage. The WAGOs mediate diverse RNA interference mechanisms in nematodes and can operate at epigenetic, transcriptional and post-transcriptional levels 28 ; it is intriguing to now consider how these possibilities could extend to their hosts.
An exciting finding in this study is the fact that the exosomes can suppress an innate Type 2 response in vivo, identifying vesicles as another class of immunomodulator used by the parasite and opening the door to further exploitation of exosomes in a therapeutic context. Our previous work has shown that H. polygyrus-secreted material suppresses IL-33 release and it is likely that a combination of soluble proteins and exosomes together suppress this important pathway 22 . From analyses in vitro we identify Il33r and Dusp1 as host genes directly suppressed by the exosomes. Although DUSP1 has been broadly viewed as an attenuator of immune activation, it is known to preferentially downregulate IL-6 that has recently been shown to promote susceptibility to H. polygyrus 29 , while upregulating IL-10, which acts as a broadly immunosuppressive cytokine 30 . Hence, parasite survival is likely to be favoured by reduced DUSP1 levels. Further, DUSP1/MKP-1 dampens the acute inflammatory response to lipopolysaccharide, promoting macrophage arginase expression over nitric oxide synthase 31 . Hence, parasite repression of DUSP1 could block the induction of arginase, a known mediator of killing of H. polygyrus in the mouse 32 . These possibilities are now being investigated in our laboratories. Our reporter assays suggest that Dusp1 could be directly targeted by the parasite miRNAs; however, we do not observe repression of Il33r when transfecting the parasite miRNA, miR-71, that is predicted to target its 3 0 UTR. It may be that additional parasite-derived RNAs or proteins could regulate Il33r expression, or that the effect operates indirectly through a separate target gene. For example, reduced expression of Dusp1 or other regulators of MAPK signalling could result in GATA-2 phosphorylation, which might inhibit its ability to promote Il33r transcription 33,34 .
Finally, our work has revealed not only secreted miRNAs that are packaged in exosomes but also full-length Y RNAs that are transferred to host cells at an abundance level detectable by northern blot analysis. Y RNAs are not known to function in gene ARTICLE silencing but were recently shown to be packaged into exosomes secreted from dendritic cells 35 and play roles in RNA quality control and DNA replication in humans 36 . Further work is required to understand whether and how each of these classes of secreted parasite RNA can contribute to the capacity of this parasite to manipulate its environment within the host.

Methods
Purification of vesicles from secretion product. For collection of H. polygyrus secretion product, CF1 mice are infected with infective-stage larvae by gavage and adult parasites collected from the small intestine 14 days post infection. The worms are maintained in serum-free media in vitro as described elsewhere 37 ; secretion product is collected every 3 days for a maximum of 3 weeks (samples used here were from the first week of collection) and purified as follows: eggs are removed by spinning at 400 g before filtering of the secretion through 0.2-mm filter (Millipore). Filtered media is then processed following a modified protocol from that described in ref. 38, by being spun at 100,000 g for 2 h in polyallomer tubes at 4°C in a SW40 rotor (Beckman Coulter). Pelleted material is washed two times in filtered PBS at 100,000 g for 2 h. The supernatant is concentrated using Vivaspin 6 5000 MWCO tubes (Fisher) at 5,000 g and washed two times with PBS.
Small RNA library preparation and analysis. For the analysis of small RNAs in the life stages and total secretion products, total RNA was size-selected on 15% denaturing PAGE and libraries prepared from the 18 to 30 nt fraction using Illumina Small RNA preparation kit version 1.5 and sequenced on an Illumina GAIIX instrument in Edinburgh Genomics (http://genomics.ed.ac.uk/). To identify larger RNAs in the secretion product, separate libraries were also prepared for RNA size selected between 60 and 100 nt and sequenced in parallel. For analysis of vesicle and nonvesicle fractions, small RNA libraries were prepared using the TruSeq kit and sequenced on MiSeq platforms, without prior size fractionation of the RNA. All libraries were analysed by first clipping the 3 0 sRNA adapter using cutadapt, searching for at least a six-base match to the adapter sequence. For analysis of small RNAs only reads that contained the adapter were 16-40 nt in length and were present at more than two copies were retained for further analysis. For analysis of RNAs 460 nt in the secretion product, sequences present at 4100 reads in the library (out of 490,614 reads sequenced) were aligned in Clustalw and manually inspected for sbRNA (Y RNA) content in terms of secondary structure and location of a UUAUC motif in the terminal loop as described in 16 (Fig. 1c and Supplementary Fig. 2). The Y fragments in the small RNA libraries (o40 nt) were then identified based on criteria that they aligned to these full-length Y RNAs. The draft genome assembly for H. polygyrus was created with the CLC de novo assembler using two lanes of Illumina GAII data with 50 bp paired-end and 100 bp paired-end reads from Edinburgh Genomics (http://genomics/ed.ac.uk/); the raw and assembled data are available at http://heligmosomoides.nematod.es/. This version was used to map the sequences around small RNA reads to identify structures consistent with miRNA precursors (according to prediction programmes detailed below). Reads matching the genome were aligned to a set of RNA sequences consisting of known H. polygyrus 18S, 28S and 5.8S rRNA sequences (Genbank AJ920355.1, AM039747.1 and DQ408618.1:527-678), 5S rRNA from a closely related species (Trichostrongylus colubriformis, Genbank U32119.1) and Rfam sequences (version 10, obtained from ftp://ftp.sanger.ac.uk/pub/databases/ Rfam/10.0/Rfam.fasta.gz). The best hit with at most two edits was used to classify the reads. Any reads that matched an rRNA or non-microRNA Rfam family were filtered before miRNA analysis. The analysis of piRNAs was carried out with reads that did not match known Rfam classes or miRNAs; initial identification was based on the presence of a 'GUUUCA' between 35 and 65 nt upstream of the 5 0 RNA alignment start site 39 . Inspection of the distribution plot identified the region 42-45 nt upstream of the 5 0 RNA alignment start site as being the key area for subsequent analyses (Supplementary Fig. 1).
Two miRNA prediction programmes were used to identify miRNAs in the data sets: miRDeep2 (ref. 40) and mireap (http://sourceforge.net/projects/mireap/). Both programmes use miRNA biogenesis to model the expected alignment of sRNA reads to a potential miRNA precursor. For miRDeep2, the following default settings were used: (a) requirement that reads match the genome perfectly, (b) removal of reads that match to more than five places in the genome and (c) cutoff -v 1, (d) the '-s option' was employed, using all mature sequences from mirbase (version 19). The default settings of minimum free energy (o À 20 kcal mol À 1 ) and read length (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) were employed. In both programmes, precursor predictions with fewer than 10 reads were discarded. Where multiple precursor loci predicted identical mature miRNAs, only the precursor with the largest number of matching reads was reported.
pCp end labelling and northern blot. For 3 0 end-labelling, total RNA was extracted from the life stages and secretion product using the miRNAeasy kit (Qiagen): 1 mg total RNA was used from life stages and RNA extracted from a volume of secretion product equating to 15 mg protein (the total RNA concentration was too low to detect by nanodrop or qubit). The 3 0 -end labelling was carried out at 4°C overnight in 10 ml using RNA ligase I (NEB) according to the manufacturer's instructions with 3,000 Ci mmol À 1 32 P PcP (Perkin Elmer). Reactions were quenched by the addition of 2 Â loading buffer (8 M urea, 0.5% TBE) and 4 ml run on an 18% PAGE at 350 V for 8 h, which was then visualized by phosphorimaging using a Typhoon Scanner (GE Healthcare). For northern blot analysis, total RNA was extracted from volumes of vesicle and nonvesicle fractions that contained equivalent protein (10 mg) and then separated by denaturing 15% PAGE, transferred to Hybond-N þ membrane (GE Healthcare) and chemically crosslinked as described previously 41 . Blots were prehybridized in PerfectHyb (Sigma) for 1 h at 42°C before overnight incubation with DNA probes (Invitrogen) that were perfectly complementary to the miRNA or Y RNA: miR-100: 5 0 -ACACAA GTTCGGATCTACGGGTT-3 0 , YRNA-5P: 5 0 -ACCCTACGACTCCGGACCA AGCGCG-3 0 , YRNA-3P: 5p-GCGCCGGTCGAGCTTTTGTCGAAGGGAAT-3p, Y RNA-loop: 5p-AAGGGAATTCGAGACATTGTTGATAAC-3p. The probes were labelled with T4 PNK (NEB) and 6,000 Ci mmol À 1 32 P ATP (Perkin Elmer) according to the manufacturers' protocols.
LC-MS/MS. Five micrograms of total protein from the secretion product ultracentrifuge pellet or supernatant were loaded on a 12% Tris-Bis NuPAGE gel (Invitrogen) and electrophoresis carried out for 5 min before in-gel digestion as described in ref.  42 where the total number of Features (that is, intensity signal at a given retention time and m/z) was reduced to MS/MS peaks with the charge of 2, 3 or 4 þ and we only kept the five most intense MS/MS spectra per 'Feature'. The subset of multicharged ions (2 þ , 3 þ and 4 þ ) was extracted from each LC-MS run. For a specific protein, the associated unique peptide ions were summed to generate an abundance value that was transformed using an ArcSinH function required for the calculation of the P value. A total of 362 proteins were identified in either the supernatant or pellet based on requirement of at least two peptides present; of these, 122 were enriched in the supernatant and 139 in the pellet, while the remaining 101 did not show statistically significant enrichment and were detectable in both samples. The within-group means were calculated to determine the fold change and the transformed data were then used to calculate the P values using one-way ANOVA. Differentially expressed proteins were considered meaningful under the following conditions: detected by two or more peptides, with an absolute ratio of at least 1.5 and Po0.05 associated with the protein change. Classification of intestinal proteins is based on homology to proteins identified in other nematodes, described in ref. 44.
TEM. For visualization of the vesicles, the purified ultracentrifuged pellet from H. polygyrus secretion product (100 mg ml À 1 protein concentration) was fixed in 2% paraformaldehyde (PFA), deposited on Formvar-carbon-coated EM grids and treated with glutaraldehyde before treatment with uranyl oxalate and methyl cellulose as described in ref. 38 Flow cytometry and confocal analyses of uptake. Purified exosomes from H. polygyrus or MODE-K cells (measured as 5 mg of total protein) were labelled with 2 mg of PKH67 dye (Sigma) for 5 min at room temperature following the manufacturer's protocol. The staining reaction was stopped by adding an equal amount of 1% bovine serum albumin (BSA), and exosomes were washed in PBS and pelleted by ultracentrifugation (1 h at 100,000 g). A probe solution was prepared with the PKH67 following the same protocol but mixed with PBS solution in the absence of exosomes. MODE-K cells 45 were obtained from Dominique Kaiserlian (INSERM) and grown following the standard protocol in DMEM (Invitrogen) medium supplemented with 10% fetal bovine serum (Invitrogen), 1% penicillin-streptomycin (Lonza), 1% L-glutamine (Lonza), 1% non-essential amino acids/sodium pyruvate (Gibco). These were mycoplasma-free based on testing every 4 weeks. On the day of the experiment, cells were seeded in 24-well plates (1 Â 10 5 cells per well) using advanced DMEM serum-free medium (Invitrogen) supplemented with 1% L-glutamine and subsequently incubated (1 h at 37°C) either in the presence of PKH67-labelled H. polygyrus-derived exosomes (5 mg of total protein) or in the presence of the probe alone. After incubation, cells were harvested, washed twice in FACS buffer (PBS 1 Â , 2.5% FBS, 0.1% BSA, 0.05% NaN 3 ) and finally resuspended in 500 ml of the same buffer. A subset of the samples were then incubated with 50 ul of 0.25% Trypsin/EDTA (Gibco) for 5 min before analysis (indicated in Supplementary Fig. 5). Samples were analysed using the BD FACS Canto II flow cytometer (BD Bioscience). Data files were acquired from the cytometer, with 5,000 events collected for each tube and the data analysis was performed using the FlowJo software (Tree Star Inc.). For confocal analyses, MODE-K cells were seeded on round microscope cover glasses in 24-well plates (2.5 Â 10 4 cells per well) in media described above. Cells were allowed to attach on to the coverslips overnight and the following day shifted to advanced DMEM medium supplemented with 1% L-glutamine. Cells were incubated (1 h at 37 or 4°C) either in the presence of labelled exosomes or probe only. After incubation, medium was aspirated, cells were washed twice in PBS and fixed with 4% PFA, with residual PFA quenched with 50 mM glycine. Slide coverslips were washed extensively in PBS, and nuclei were stained with 4 0 ,6-diamidino-2-phenylindolesupplemented ProLong Fade Gold (Invitrogen) mounting media. Samples were examined on the Leica SP5 II (Leica Microsystems, lasers exciting at 405 and 488, Â 63 objective) using the LAS AP software (Leica). Images were analysed using the Volocity software (Improvision).
Microarray analysis. MODE-K cells were grown in DMEM media as described above and seeded into 24-well plates at 20,000 cells per well. In vivo analysis of exosome function in Alternaria model. BALB/c mice were bred in-house at the University of Edinburgh and accommodated according to Home Office regulations. Female mice were used when they were 6-10 weeks old.
For all experiments presented in this study, the sample size was large enough to measure the effect size. No randomization and no blinding were performed in this study. H. polygyrus exosomes (10 mg) were administered intranasally (under isoflurane sedation) in 50 ml PBS, or 50 ml PBS alone to controls, 24 h before intranasal administration of 50 mg Alternaria extract with a further 5 mg of exosomes. Mice were killed 24 h after Alternaria administration, and bronchoalveolar lavage and lung cell suspensions stained for flow cytometry as described previously 22 . Briefly, cells were counted, then surface stained for Siglecf þ CD11c À (eosinophils) or stimulated with phorbol myristate acetate (PMA) and ionomycin for 4 h in the presence of BrefeldinA and surface stained as negative for lineage markers (CD3/CD4/CD5/CD19/CD11b/CD11c/CD19/GR1) and positive for CD45, ICOS and ST2 (ILC2s), and assessed for staining of IL-5 and IL-13. Samples were acquired on a Becton-Dickinson LSRII flow cytometer. Data were analysed using Prism 6 (Graphpad Prism, La Jolla, CA, USA). Variance within groups was assessed by Brown Forsythe test and data were log-transformed and analysed by one-way ANOVA, with a Tukey's multiple comparisons post test. Unless otherwise indicated, differences are not significant. ****Po0.0001, ***Po0.001, **Po0.01, *Po0.05, N.S. not significant P40.05.
For reporter assays, 15,000 cells were reverse transfected into a 96-well plate with 0.3% lipofectamine (Invitrogen) and 50 ng of each Psicheck reporter in the absence or presence of 50 nM synthetic miRNA mimic (Thermofisher). Luciferase measurements were carried out at 48 h post transfection, using the Dual Glo Luciferase assay system (Promega) and Luminensence measured on a Varioskan plate reader (Thermofisher). Data shown in Fig. 7b represent n ¼ 3 replicates (separate transfection experiments of the MODE-K cell line) measured in parallel to control for consistent Renilla and Luciferase ratios using the same kit; data were analysed by one-way ANOVA, with a Dunnett's post test, ****Po0.0001, ***Po0.001, **Po0.01, *Po0.05.
MicroRNA target prediction. A custom Perl script was used to identify seedmatching sites for the H. polygyrus miRNAs identified by sRNA-Seq. All the 3 0 UTR sequences corresponding to probes on the microarray were scanned, and the results were passed to the TargetScan (v6.2) context scores Perl script. Targets for each miRNA were ranked by 'Context þ Scores'. Conservation scores for relevant 3 0 UTRs (Dusp1 and Il33r) were obtained from the UCSC Genome Browser, using the 'rtracklayer' package to access the 'phastCons60wayPlacental' table 52,53 .
Sequencing analysis of serum from infected mice. F1 mice were infected with H. polygyrus (400 L3 larvae introduced by oral gavage) and serum collected on day 14 post infection. The presence of adult parasites was confirmed by visual inspection of the mouse gut lumen. Six-week-old BALB/c mice were infected with L. sigmodontis (subcutaneous inoculation of 40 L3), and gel-separated serum (BD Microtainer) was collected by arterial exsanguination at 60 days post infection, which was confirmed by detection of adult worms in the mouse pleural cavity and microfilarie in peripheral blood. For library preparation, 200 ml of serum was extracted with the miRNAeasy kit (Qiagen) and libraries generated following the Trueseq protocol and sequenced on the Illumina Rapid HighSeq in Edinburgh Genomics. Data were processed as described above and analysed for perfect alignment to the mouse or L. sigmodontis genome (reads mapping to both genomes were not analysed). Reads that aligned were then categorized by matches to Rfam or prediction as miRNAs with miRdeep2, as described in ref. 11. 50. Smyth, G. K. Linear models and empirical bayes methods for assessing differential expression in microarray experiments.