Breed differences in the expression levels of gga-miR-222a in laying hens influenced H2S production by regulating methionine synthase genes in gut bacteria

Background The microbiota in the cecum of laying hens is crucial for host digestion, metabolism, and odor gas production. The results of recent studies have suggested that host microRNAs (miRNAs) can regulate gene expression of the gut microbiota. In the present study, the expression profiles of host-derived miRNAs in the cecal content of two laying hen breeds; Hy-line Gray and Lohmann Pink, which have dissimilar H2S production, were characterized; and their effects on H2S production by regulating the expression of gut microbiota-associated genes were demonstrated. Results The differential expression of microbial serine O-acetyltransferase, methionine synthase, aspartate aminotransferase, methionine-gamma-lyase, and adenylylsulfate kinase between the two hen breeds resulted in lower H2S production in the Hy-line hens. The results also revealed the presence of miRNA exosomes in the cecal content of laying hens, and an analysis of potential miRNA-target relationships between 9 differentially expressed miRNAs and 9 differentially expressed microbial genes related to H2S production identified two methionine synthase genes, Odosp_3416 and BF9343_2953, that are targeted by gga-miR-222a. Interestingly, in vitro fermentation results showed that gga-miR-222a upregulates the expression of these genes, which increased methionine concentrations but decreased H2S production and soluble sulfide concentrations, indicating the potential of host-derived gga-miR-222a to reduce H2S emission in laying hens. Conclusion The findings of the present study reveal both a physiological role by which miRNAs shape the cecal microbiota of laying hens and a strategy to use host miRNAs to manipulate the microbiome and actively express key microbial genes to reduce H2S emissions and breed environmentally friendly laying hens. Video Abstract Supplementary Information The online version contains supplementary material available at 10.1186/s40168-021-01098-7.


Background
The laying hen industry is an important livestock sector that produces eggs as one of the common nutrition sources for human consumption in daily life [1]. Nutrient digestion in laying hens is characterized by inadequate enzymatic hydrolysis in the foregut, followed by further microbial fermentation in the cecum. An increasing number of studies have proven that this 'bacterial organ' plays a vital role in host metabolism, immunity and disease [2][3][4]. Bacteria ferment undigested feed components to generate volatile fatty acids (VFAs), amino acids, ammonia (NH 3 ), hydrogen sul de (H 2 S) and other metabolites [5,6]. As previous studies elucidated that NH 3 and H 2 S are the two main odorous gases in poultry houses and represent a great loss in nutrients and environmental pollution that is a public concern [7,8]. Although, the NH 3 accounts are the largest proportion of odor gas in livestock and poultry facilities, and the accounts of H 2 S is second to that of NH 3 , the odor thresholds of H 2 S is signi cantly lower than NH 3 which also contributes to the bad smell in the farm environment [9,10]. In addition to its adverse effects on air quality, the early study reported that H 2 S cause intestinal diseases in human [11], the recent study also showed that H 2 S induced chicken pneumonia response [12], there is the possibility for H 2 S to damage the health of breeding worker and animals. Therefore, reducing H 2 S besides mitigating the environmental problems will also have a positive effect on the health of animals.
Several recent studies have explored nutritional manipulations, for example, probiotic inclusion or protein reduction in the diet have been used to regulate the gut microbiota to mitigate H 2 S emission in animals [13,14]. However, there are a few disadvantages to nutritional manipulations, for instance, the supplementation of probiotics must be continuous to guarantee the sustained reduction effect of H 2 S emission. In view of this, breeding laying hens with a low-H 2 S emission "cecal microbiota structure" could be a better and permanent measure for H 2 S reduction and environmentally friendly culture in the poultry industry. However, the rst key point in breeding low-H 2 S emission laying hens is to understand the regulatory relationship between the host and its cecal microbiota.
Numerous factors in uence the composition and function of the gut microbiota. For instance, the composition of the gut microbiota is shaped by the host's genetic background and to some extent can be transiently altered by diet, the environment and disease states [15,16]. MicroRNAs (miRNAs) are a group of noncoding RNAs of ~ 22 nucleotides (nt), that are known for their sequence-speci c regulatory function that operates by targeting the 3′ untranslated region of mRNAs in the cytoplasm [17]. Increasing evidence has demonstrated that miRNAs also exist extracellularly and circulate in body uids in the form of exosomes and microvesicles. Secreted miRNAs have been isolated from blood, milk and even stool and urine [18][19][20][21][22]. Some studies have characterized miRNAs as potential markers of tumorigenesis in human stool [23][24][25]. Interestingly, recent studies have also revealed that host-derived miRNAs could serve as an important crosstalk channel between the host and the intestinal bacterial population. Liu found that the host could modulate the gut microbiota through intestinal epithelial cell-secreted miRNAs, which enter gut bacteria and directly regulate bacterial gene expression [26]. Other studies reported that plant-derived miRNAs could be taken up by gut bacteria and shape the gut microbiota [27,28]. Therefore, we hypothesize that there might be cross-regulation between host-derived miRNAs and the cecal microbiota in laying hens. In addition, whether the host-derived miRNAs regulate microbiota abundance or the expression levels of bacterial function genes, which in uence the structure of the microbiota and the metabolism function, and nally lead to H 2 S production differences in different breeds of laying hens, is still unknown.
In our previous study, we found a signi cantly higher daily H 2 S production in Lohmann laying hens than Hy-line Gray laying hens (the daily H 2 S production per kg average daily feed intake was 7.75 and 4.17 mg for Lohmann and Hy-line, respectively) [29]. However, whether the H 2 S emission difference between the two breeds of laying hens was due to host miRNA regulation of the gut microbiota requires further investigation. Therefore, we determined the expression pro les of host-derived miRNAs in the cecal content of these two breeds of laying hens to nd out the signi cantly different miRNAs between the two breeds of laying hens; and then predicted the target relationships between differentially expressed miRNAs and microbial genes related to H 2 S production; nally, the effect of selected targeted miRNA on H 2 S production in laying hens was veri ed by an in vitro experiment. This work may unveil the interkingdom regulation relationships among miRNAs, cecal microbiota and H 2 S production in laying hens, which provide a reference for the breeding of environmentally friendly laying hens. At the same time, miRNAs, which have been proven to regulate the production of H 2 S in the cecum of laying hens, could be used as a safe and clean additive for H 2 S emission reduction in the future.

Results
Identi cation of miRNA pro les in the cecum of laying hens The morphology of exosomes derived from laying hen cecal content was observed using transmission electron microscope (TEM). The exosome-sized (approximately 50-200 nm in diameter) extracellular vesicles were present in the cecal content of Lohmann and Hy-line hens, but the morphological character of exosomes was mostly the same in the two breeds ( Fig. 1A and B). The existing of exosomes elucidated that there was the possible for the further miRNA sequencing to investigate the miRNA differences between the two breeds.
In addition, the KEGG pathway annotations of target genes in the chicken genome of these ten signi cantly expressed miRNAs are shown in Fig. S1. The target genes were mostly enriched in metabolic pathways, neuroactive ligand-receptor interaction, focal adhesion, endocytosis and purine metabolism.

The Expression Of Microbial Genes Related To Hs Production
The sequencing information of the metatranscriptomic was listed in Table S2, and the top 30 KEGG microbial function enrichment pathways were showed in Fig. S2. The 5 pathways of the 30 top pathways were showed the signi cant differences gene enrichment between Lohmann and Hy-line, the microbial genes enrichment abundance of two-component system was signi cantly higher in the Hy-line by the comparison of Lohmann (P < 0.05), and the microbial genes enrichment abundance of amino sugar and nucleotide sugar metabolism, cysteine and methionine metabolism, alanine aspartate and glutamate metabolism and RNA degradation pathways were signi cantly higher in Lohmann by the comparison of Hy-line (P < 0.05). The result of transcriptomics also showed that totally the abundance of 22237 genes were signi cantly different between Lohmann and Hy-line (P < 0.05), in addition based on the combination analysis of the microbial gene enrichment pathway and the signi cantly different genes, we focused on two metabolism pathways involved in H 2 S production, the cysteine and methionine metabolism pathway (map 00270) and the sulfur metabolism pathway (map 00920). Based on the KEGG database recorded, the degradation of the sulfur-containing amino acid, cysteine and methionine, could result in the production of H 2 S. As shown in Fig. 2A, L-cysteine was synthesized from L-serine and sul de under the action of serine O-acetyltransferase and cysteine synthase, moreover, L-cysteine also could be synthesized from L-cystathionine by the regulation of cystathionine gamma-lyase. The L-cysteine degraded into pyruvate and sul te through the regulation of aspartate aminotransferase, and sul te could be the raw material for the production of H 2 S through the sulfur metabolism pathway. In addition, the methionine could be degraded into methanethiol by methionine-gamma-lyase, which could be converted into H 2 S in subsequent processes. Another pathway for methionine degradation was the formation of S-adenosyl-L-methionine by S-adenosylmethionine synthetase without H 2 S production.
Sulfur metabolism was the essential pathway in the cecum of hens which involved in the production of H 2 S. The assimilatory reduction and dissimilatory reduction of sulfate promoted the production of sul te and its eventual conversion into sul de ( Fig. 2A), thus the adenylylsulfate kinase was one of the key enzyme for the production of sul de from sulfate, and it was worth to be tested in order to elucidate the H 2 S production.
Through the analysis of annotation information of KEGG database, the present results indicated that Oacetyltransferase (cysE, EC: 2.3.1.30), methionine synthase (metH, EC:2.1.1.13), aspartate aminotransferase (aspB, EC:2.6.1.1), methionine-gamma-lyase (MGL, EC:4.4.1.11) and adenylylsulfate kinase (cysC, EC: 2.7.1.25) were the key enzymes for the production H 2 S, and the gene expression level of these key enzyme was tested. The expression level of cysE and metH were both signi cantly higher in Hyline hens than that in Lohmann hens (P < 0.05), the expression level of cysE was 0.061 ± 0.004 and 0.048 ± 0.010 in Hy-line and Lohmann respectively, and the expression level of metH was 0.302 ± 0.062 and 0.222 ± 0.036 in Hy-line and Lohmann respectively; but the expression of aspB and MGL were both lower in Hy-line hens than that in Lohmann hens (P < 0.05), the expression level of aspB was 0.081 ± 0.015 and 0.085 ± 0.020 in Hy-line and Lohmann respectively, and the expression level of MGL was 0.011 ± 0.002 and 0.014 ± 0.011 in Hy-line and Lohmann respectively. The present results indicated that there was a higher tendency for synthesis of cysteine and methionine but a lower tendency for their degradation and H 2 S production in the cecum of Hy-line hens (Fig. 2B). In addition, we also found a signi cantly lower expression of cysC in the cecum of Hy-line hens by the comparison of Lohmann (P < 0.05), the cysC expression level was 0.004 ± 0.001 and 0.007 ± 0.001 in Hy-line and Lohmann respectively, the result also indicated the lower sul de production in Hy-line than that in Lohmann hens (Fig. 2B).
Based on the analysis of the regulation enzyme involving H 2 S production, totally, thirteen differentially expressed microbial genes which involved in the encoding of these ve enzymes were found between the two breeds, MGL and cysC only assigned to one gene and one genus, other three enzymes assigned several genes and genera, in addition most of the genes were assigned to the genus Bacteroides (Table 1).

Target Prediction Of Mirnas
Through the combination analysis of differentially expressed miRNAs (Fig. 1C) and differentially expressed genes which related to H 2 S production (Table 1), in total, nine miRNAs could target 9 genes by using miRanda analysis (Fig. 3A). As above mentioned that gga-miR-222a and gga-miR-10a-5p were both high expression level miRNAs in Hy-line and Lohmann, however, the abundance of gga-miR-222a was signi cantly higher in Hy-line by the comparison of Lohmann, and the abundance of gga-miR-10a-5p was signi cantly lower in Hy-lin by the comparison of Lohmann, in addition, our previous study showed that the amount of H 2 S production was obviously higher in Lohmann by the comparison of Hy-line (the daily H 2 S production per kg average daily feed intake was 7.75 and 4.17 mg for Lohmann and Hy-line, respectively) [29], thus we concluded that gga-miR-222a was the potential additive candidate for the reduction of H 2 S production in the cecum of laying hens, it was much worth to select gga-miR-222a for the further investigation rather than the selection of gga-miR-10a-5p. We found that gga-miR-222a could target two genes that associated with methionine synthase, Odosp_3416 (expressed by the bacterium Odoribacter splanchnicus) and BF9343_2953 (expressed by the bacterium Bacteroides fragilis NCTC 9343) (Fig. 3B). The read count of Odosp_3416 was 138 and 225 in Hy-line and Lohmann respectively, and the read count of BF9343_2953 was 75 and 137 in Hy-line and Lohmann respectively, based on above results, we suspected that the regulatory effect of gga-miR-222a on the two gens may be notable. Therefore, the subsequent function veri cation of gga-miR-222a targeted with Odosp_3416 and BF9343_2953 was carried.

Hs Production Investigation In The Fermentation Experiment
After 24 h of fermentation, the amount of the total gas and H 2 S production was tested in each experiment groups. The addition of gga-miR-222a could in uence the total gas and H 2 S production.
The total gas production was 33.75 ± 0.83 mL and 30 ± 0.71 mL in LB (Lohmann intestinal content broth added nothing) and HB (Hy-line intestinal content broth added nothing) respectively, the statistical analysis showed that the total gas production of LB was signi cantly higher than that in HB (P < 0.05), the result potentially indicated that the total gas production ability of Lohmann was higher than that of Hy-line. In present result, we found that the amount of total gas production was signi cantly decreased in LT (Lohmann intestinal content broth added gga-miR-222a, 29.5 ± 0.5 mL) by the comparison of LB (P < 0.05), however there was no signi cantly different between HT (Hy-line intestinal content broth added gga-miR-222a, 28 ± 1.87 mL) and HB, the result elucidated that gga-miR-222a could effectively decrease the total gas production in Lohmann rather than in Hy-line (Fig. 4A). In addition, there was no obvious difference between blank groups and control groups (LC and HC, intestinal content broth added miRNA control), the result suggested that the commercial synthesis gga-miR-222a was credible.
The effect of gga-miR-222a on the H 2 S production was investigated, and the result was showed in the statistical analysis showed that the amount of H 2 S was signi cantly higher in LB by the comparison of HB (P < 0.05), the result indicated that the H 2 S production ability of Lohmann was higher than that of Hy-line. It was worth to note that gga-miR-222a addition signi cantly decreased the amount of H 2 S production in both fermentation broth of two breeds (P < 0.05), the amount of H 2 S production was 81.553 ± 11.95 µg and 71.152 ± 8.94 µg in LT and HT respectively. By the comparison of blank groups, gga-miR-222a addition decreased H 2 S production 22.88% and 26.33% in Lohmann and Hy-line intestinal content broth respectively. The present result suggested that there was the potential ability for gga-miR-222a to decrease the H 2 S production of the intestinal content.
The chemical indexes of the fermentation broth of each group are shown in Table 2. The concentration of soluble sul de (S 2− ) was signi cantly lower in HB (12.32 ± 0.98 µg/g) than that of LB (15.06 ± 0.94 µg/g) (P < 0.05), in addition, the S 2− concentration also signi cantly lower in the gga-miR-222a addition groups (LT, 12.50 ± 0.50 µg/g and HT, 9.84 ± 1.44 µg/g) by the comparison of blank groups (LB and HB) (P < 0.05). The concentration of methionine in the fermentation broth was also con rmed due to methionine was the crucial amino acid to donate the sulfur for the form of H 2 S through the microbial metabolism.

Bacterial Abundance And Gene Expression In Fermentation Broth
DNA was extracted from fermentation broth by using a QIAamp PowerFecal DNA Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions, brie y 1 mL fermentation broth was centrifuged at 20,000 × g for 1 min at 4℃ to collect the precipitate (approximately 200 mg) for the DNA extraction, and the DNA puri cation following steps was referenced the protocol of manufacturer. RNA extraction of fermentation broth followed the protocol of RNeasy® PowerMicrobiome™ Kit (Qiagen, Hilden, Germany), brie y, 1-2 mL fermentation broth was centrifuged at 20,000 × g for 1 min at 4℃ to collect the precipitate (approximately 200-250 mg) for the RNA extraction, and the RNA puri cation following steps was referenced the protocol of manufacturer.
The DNA was used to quantify the relative abundance of Odoribacter splanchnicus and Bacteroides fragilis NCTC 9343. Brie y, the primers of these two bacteria was designed by using the software Primer 3, the sequences of the two bacteria was referenced the 16S rRNA sequencing in the web of NCBI, the details of the primers were showed in Table S5, the primer of bacterial16S rRNA was referenced previous study [45]. q-PCR was used to con rm the relative abundance of the two bacteria, the q-PCR reaction steps were followed the protocol of SYBR® Green PCR Kit (SYBR, Japan) ( Table S6). The relative abundance of the two bacteria was calculated as 2 △Ct , where △Ct represents the difference in the Ct value for the 16S rRNA gene minus that for the genes [46].
Extracted RNA was reverse transcribed into cDNA by using the PrimeScript™ RT reagent kit (TaKaRa, Kusatsu, Japan). The cDNA was used to quantify the expression of Odosp_3416 and BF9343_2953. The primers were designed using the NCBI website with the total bacterial 16S rRNA gene as the reference gene (Table S5). q-PCR was used the con rm the relative expression level of the two genes, the q-PCR reaction steps were followed the protocol of SYBR® Green PCR Kit (SYBR, Japan) with a little modi cation (Table S7). The relative expression level of the two genes was calculated as 2 △Ct , where △Ct represents the difference in the Ct value for the 16S rRNA gene minus that for the genes [46].

In vitro bacterial growth measurements
The anaerobic bacterium Bacteroides fragilis NCTC9343 were cultured at 37ºC by inoculating 40 mL aliquots of anaerobic basal medium (Becton Dickinson and Company, Lincoln Park, USA) and then grown anaerobically in an anaerobic chamber (Mitsubishi Gas Chemical Company, Inc. Tokyo, Japan). gga-miR-222a and the control mimic were supplied in the culture at a concentration of 2 µM. (RiboBio, Guangzhou, China). Growth was monitored as absorbance at 600 nm once per hour for up to 24 h with a spectrophotometer. The cultured bacterial cells were collected at 10 h and used for BF9343_2953 gene expression measurement with the Bacteroides fragilis 16S rRNA gene as the reference gene. The concentrations of methionine in culture medium at 10 h were tested as above mentioned.
In situ hybridization detection of the uptake of gga-miR-222a The bacterial cells of Bacteroides fragilis NCTC9343 were centrifuged at 12,000 × g and washed twice with ice cold PBS. Then, the cells were xed in 4% PFA/0.25% glutaraldehyde. A 5'-DIG and 3'-DIG dual labeled probe for gga-miR-222a was used for in situ hybridization. The detection of the uptake of gga-miR-222a by bacteria was imaged using a Thermo Fisher Talos L120C transmission electron microscope Thermo (Fisher Scienti c, MA, US).

Statistical analysis
The data of the comparison of fermentation incubation indexes, the comparison of the relative abundance of miRNA, bacteria and genes, gas production, H 2 S production and growth curve were examined by analysis of variance (ANOVA) with Statistical Package for the Social Sciences (SPSS) software, version 22.0. Signi cant differences between the means were determined by Tukey's test. Differences were considered signi cant at P 0.05.
The metatranscriptomic results of each sample was analyzed by HTSeq software, and the model used was union, the number of genes in different expression levels and the expression level of individual genes were statistically analysed. In general, the value of FPKM is 0.1 or 1 as the threshold for determining whether genes are expressed. The software DESeq was used for normalization of the read counts from analysis of genes expression levels [47].
The expression level of miRNA was calculated by the TPM formula (normalization read counts= (readCount*1,000,000)/libsize)), libsize was the sum of the read count of all miRNAs.

The Investigation Of Gga-mir-222a Effectiveness
In order to further understand whether the effects of gga-miR-222a were really in uenced the methionine synthetase genes carried bacteria, the culture medium experiment was applied to reach the goal. Because of the relative abundance of Odoribacter splanchnicus was signi cantly lower than that of Bacteroides fragilis NCTC9343 in both breeds (Fig. S3), thus Bacteroides fragilis NCTC9343 was selected for the following experiment. At 8 h, the bacterium Bacteroides fragilis NCTC9343 reached the logarithmic phase, and at 10 h, the strain reached a plateau. In the gga-miR-222a addition group, this miRNA signi cantly improved 29.04% of the abundance of Bacteroides fragilis NCTC9343 at 10 h by the comparison of blank (P < 0.05) (Fig. 6A). In addition, gga-miR-222a addition also signi cantly enhanced 2.23-fold of the expression level of gene BF9343_2953 of Bacteroides fragilis NCTC9343 by the comparison of control (P < 0.05) (Fig. 6B), and the concentration of methionine in the medium also was signi cantly increased 38.71% with the addition of gga-miR-222a by the comparison of blank (P < 0.05) (Fig. 6C).
To determine whether gga-miR-222a could be taken up by Bacteroides fragilis NCTC9343 and then play a series of regulatory functions inside the cell, we measured bacterial internalization of gga-miR-222a by in situ hybridization and TEM, by the comparison of control (Fig. 7A), the exogenous gga-miR-222a was selectively absorbed by Bacteroides fragilis NCTC9343 in the gga-miR-222a addition group (Fig. 7B).

Discussion
Although there is a known association among host genetic background, cecal microbiota structure and odor production by laying hens [30,31], the potential mediators of this relationship remain unclear.
Recently, some ndings demonstrated that mammalian secreted miRNAs could regulate the expression of bacterial genes [26,32]. Here, we presented the rst insight into the characterization of miRNAs derived from the cecal content of laying hens and found that gga-miR-222a could reduce the production of H 2 S by regulating the expression of cecal microbial methionine synthetase genes in the cecum of laying hens.

Differential expression of cecal microbial genes led to dissimilar H 2 S production between the two breeds
In a previous study, we found that Hy-line hens exhibited lower H 2 S production than Lohmann hens as a result of different microbiota structures related to H 2 S production in the cecum [29]. However, due to the limitation of 16S rRNA sequencing, we did not annotate and identify pathways and genes related to bacterial sulfur metabolism. Transcriptomic sequencing can be more accurate than other methods to elucidating the functional makeup of a microbial community and allowing us to characterize potential miRNA interactions across the microbiome and transcriptome. The gene expression of the cecal microbiota was characterized for H 2 S production related pathways using the metatranscriptome. The synthesis of cysteine and methionine requires the participation of sulfur, and related decomposition is accompanied by the release of sulfur [33]. Higher expression of serine O-acetyltransferase and methionine synthase but lower expression of aspartate aminotransferase and methionine-gamma-lyase in the Hy-line cecal microbiota community indicated that the Hy-line hens had a stronger ability to utilize sulfur for the synthesis of cysteine and methionine than the Lohmann hens.
Dissimilatory sulfate reduction is the exclusive sulfate reduction pathway for most sulfate-reducing bacteria (SRB), but assimilatory sulfate reduction can be carried out by most bacteria in the gut [34,35]. The metatranscriptome showed that there was no signi cant difference in the expression of dissimilatory sulfate reduction pathway related genes. The reason could be due to the abundance of SRB was low in the animal gut (approximately 0.028-0.097%) [14]. Low abundances of gut SRB led to a low and unobvious differential expression of related genes. However, for the assimilatory sulfate reduction pathway, the gene expression of adenylsulfate kinase in the Hy-line hens was signi cantly lower than that in the Lohmann hens, indicating a more powerful transformation of sulfate to sul de in the latter case.
Here, we found that the differentially expressed microbial genes in sulfur related metabolism pathways were the reason for dissimilar H 2 S production between the Lohmann and Hy-line hens, but whether the host speci cally regulates microbial genes by some cross-regulation factors is not clear. In this study, we identi ed cecal miRNAs and found that they could directly regulate speci c bacterial gene expression and affect gut microbial growth to affect H 2 S production in laying hens.
The miRNAs in the cecal content differed between the two breeds miRNAs have not been previously characterized in the cecal content of laying hens. First, we demonstrated that microvesicles existed in the cecal content of Lohmann and Hy-line hens, but only 288 known miRNAs were sequenced. Owing to the bacterial RNA sequence accounted for the main proportion of total RNA in the cecal content of laying hens and some miRNAs may be degraded by the high temperature and high uric acid cecum environment [36], the number of types and abundances of sequenced miRNAs were relatively low. Only 10 miRNAs were differentially expressed between the Lohmann and Hy-line hens. The highly conserved and homologous characteristics of miRNAs may lead to a high similarity in miRNA types and abundances between two breeds [37]. Most of the chicken genome targets of these miRNAs were enriched in metabolic pathways, neuroactive ligand-receptor interaction, focal adhesion, endocytosis and purine metabolism, but were not enriched in pathways related to cancer occurrence and disease formation, indicating that these miRNAs did not have a potentially negative effect on the host's normal life activities. This means that the application of these miRNAs to odor reduction may be harmless to the host itself.
Host-derived miRNAs targeted the genes of the cecal microbiota of laying hens miRNA binds with mRNA to perform its regulatory functions. We predicted the possible target relationships between differentially expressed miRNAs and differentially expressed genes related to H 2 S production. It was found that gga-miR-222a had a target relationship with the methionine synthetase genes Odosp_3416 and BF9343_2953 (expressed by Odoribacter splanchnicus and Bacteroides fragilis NCTC 9343, respectively). Therefore, gga-miR-222a may be a host regulator that affect H 2 S emission in laying hens by regulating the production of methionine, a sulfur-containing amino acid. In vitro fermentation experiment and bacterial culture showed that gga-miR-222a could upregulate the expression of the genes Odosp_3416 and BF9343_2953, and increase the abundance of Bacteroides fragilis NCTC 9343 in the logarithmic growth period (10 h), which resulted in a higher concentration of methionine but lower H 2 S production and soluble sul de concentration in fermentation broth and bacterial medium. The concentrations of gut soluble sul de are positively correlated with the release of H 2 S [38]. The decrease H 2 S production and soluble sul de concentration showed that gga-miR-222a reduced H 2 S emission in laying hens.
The host-derived miRNA gga-miR-222a in uenced H 2 S emission in laying hens Interestingly, we found that gga-miR-222a played a positive role in regulating the expression of the Odosp_3416 and BF9343_2953 genes, which was different from the results of most studies, which suggest that miRNA always inhibits the transcription of mRNA or directly degrades the sequence of the mRNA after binding with mRNA [39,40]. However, some studies have shown that miRNA not always played a negative regulation on mRNA [26,27,41]. How miRNA regulates the expression of genes and affects bacterial growth may rely on the function of the genes targeted by the miRNA and the site at which miRNA binds mRNA. Binding between miRNA and bacterial transcripts of 16S rRNA, yegH, RNaseP and β-galactosidase genes upregulates the expression of these genes and promotes the growth of bacteria [31,32]. In this study, we found that gga-miR-222a played a similar role in the regulation of methionine synthetase gene expression and promoted the abundance of bacteria in bacterial medium, especially in the logarithmic phase. After in situ hybridization, we found that exogenous gga-miR-222a could be selectively up-taken by Bacteroides fragilis NCTC 9343, indicating an intracellular crossregulation role of gga-miR-222a. However, the increase of bacteria abundance in fermentation broth was not signi cant except for a slight rise after gga-miR-222a treatment. The reason for this discrepancy may be that the intestinal environment is more complex and there are many interfering factors, such as interactions among various microorganisms. This was the why conducting bacterial growth experiment in a pure culture environment was necessary.

Conclusions
In conclusion, the present study found host-derived miRNAs in the cecum of laying hens for the rst time and the expression pro les of miRNAs were different between different breeds. It was also demonstrated that gga-miR-222a regulated the expression of H 2 S production related genes (Odosp_3416 and BF9343_2953) to affect the production of H 2 S in laying hens. Meanwhile, gga-miR-222a could enter Bacteroides fragilis NCTC 9343, which increased its abundance in the logarithmic period. Therefore, different pro les of host-derived miRNAs in different breeds of laying hens could affect the production of H 2 S trough the gene expression regulation in the H 2 S production related bacteria. Regulation of H 2 S production in the cecum of laying hens by host miRNAs such as gga-miR-222a provides the possibility that if these miRNAs could be incorporated into the breeding of laying hens, they could provide a certain reference value for the selection of low odor yield and environmentally friendly laying hen breeds.

Animals and feeding
Approximately one hundred Hy-line Gray laying hens and one hundred Lohmann Pink laying hens were hatched and fed together at a local hatchery. To eliminate the confounding effects that might be caused by diet, age, weight and feeding environment. Thirty Hy-line Gray laying hens and thirty Lohmann Pink laying hens at 28 weeks of age with similar weights (1.70±0.02kg and 1.71±0.02kg, respectively for Hyline and Lohmann) were selected and moved into twelve respiration chambers in an environmentally controlled room for a daily H 2 S production measurement for the two breeds [29]. Water and the commercial-type laying hen diet were fed to birds ad libitum (Table S3), and a 12-h light cycle at 24°C room temperature management schedule was used. At the end of the experiment, all birds were euthanized by cervical dislocation, and then the cecum was ligated at both sides and removed from the gastrointestinal tract. The contents were aseptically collected into an Eppendorf tube containing Bacterial Protect RNA reagent (Qiagen, Hilden, Germany) at an approximate 1:1 ratio (w/v), and immediately frozen at liquid nitrogen and stored at -80°C until analysis.

Animal ethics statement
All animal experiments were approved by the Animal Experimental Committee of South China Agricultural University (SYXK2014-0136). All experimental steps were performed to decrease animal suffering as much as possible. After the experiment, the bodies of laying hens were incinerated.
The determination of exosomes in the cecum contents The exsomes puri cation referenced Liu, brie y, cecal contents from laying hens were suspended in PBS to 30 mg/ml, spun down at 10,000 ×g for 5 min to remove debris and then ltered through a 0.2 µm lter and the ltrates were observed by Thermo Fisher Talos L120C transmission electron microscope (Thermo Fisher Scienti c, MA, US) [26].
Extraction and analysis of miRNA in the cecum of laying hens Total miRNA was extracted by using mirVana™ miRNA Isolation Kit (Austin, TX, USA) according to Liu [26]. Brie y, approximately 100 mg cecal content was mixed adequately with 600μL 1×DPBS, and the mixture was left at room temperature for 30 min, and then mashed to complete suspension. Then 600 μL acid-phenol: chloroform was added, and the samples were vortexed for 60 sec and then centrifuged for 15 min at 10,000 ×g to separate organic phases. The aqueous phase was recovered, and 1.25 volumes of 100% ethanol was added to the aqueous phase for nal miRNA isolation. For each sample, a lter cartridge was placed into one of the collection tubes (supplied by the kit), and the sample was pipetted onto a lter and centrifuged for 90 sec at 10,000 ×g, and then the ow-through was discarded. The lter was washed with 700 μL miRNA Wash Solution 1 and then washed three time with 700/500/250 μL Wash Solution 2/3 (supplied by the kit). Finally, the lter was transferred into a fresh collection tube, and 50 μL nuclease-free water was applied to the center of the lter. The lter was incubated at room temperature for 10 min, then centrifuged for 5 min at 8000 ×g to recover miRNA and then stored at -80℃. FastQC was applied to obtain clean reads from the raw data by removing the joint sequences, low-quality fragments, and sequences <18 nucleotides (nt) in length. miRDeep2 was used to align the clean sequences to the miRBase database sequences (http://www.mirbase.org/). . Sequences were quality ltered and poor-quality bases of raw reads were removed by using Cutadapt (v1.9.1) software. A 10 bp window was moved across each sequence, and nucleotides in windows with a mean quality score < 20 were removed; reads with "N" bases (>10%) and lengths below 75 bp were discarded; primer sequences and adaptor sequences were also removed. Next, rRNA, tRNA and host reads were ltered using BWA (v 0.7.5).
Putative mRNA reads were then assembled using the Trinity (v2.1.1) de novo assembler. Gene annotation was performed by searching against a protein non-redundant database (NR database), and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analysis was conducted for gene function classi cation.
After the comparation of transcript pro les between the two breeds, we focused on three pathways related to H 2 S production, including cysteine and methionine metabolism, sulfur metabolism and butyrate metabolism, and the expression of microbial genes in these pathways between the two breeds was compared.

Target prediction of differentially expressed miRNAs
After the exploration of miRNA pro les in the cecal content of laying hens, the signi cantly different expression miRNAs between the two breeds were used for the target prediction analysis of microbial signi cantly different expression genes which related to H 2 S production. The target relationship between Bacterial mRNAs and miRNAs were identi ed by using miRanda (http://www.microrna.org). Furthermore, a prediction of host genome genes which was targeted by the 10 different expression miRNAs were also conducted by miRanda, and all the target genes were determined to be enriched by KEGG analysis.
In vitro fermentation experiment The in vitro was referenced Menke and Steingass [42], brie y, thirty Hy-line Gray and thirty Lohmann Pink laying hens at age of 28 weeks were sacri ced respectively, and the caeca were ligated immediately. The cecal contents in the same breed group were pooled, then thoroughly mixed with the fermentation buffer solution which was pre-heat at 39°C as the fermentation broth. After the air in syringe was eliminated from the head-space, approximately 10mL fermentation broth (FB) was added to a 100 mL gas syringe with 0.2g of the substrate. Three groups with different treatments for each breed were designed, the blank group (10mL FB+0.2g substrate+1mL pure water), control group (10mL FB+0.2g substrate+1mL control mimic at a nal concentration of 2µM) and treatment group (10mL FB+0.2g substrate+1mL gga-miR-222a mimic at a nal concentration of 2µM) ( Table. S4). The miRNAs applied in present study were purchased from Guangzhou RiboBio Co., Ltd. (Guangzhou, China). Then, these syringes were sealed with clips and placed in incubator and rotated at 42°C, 60rpm for 24h.
At the end of incubation, the syringes were put on the ice to stop the fermentation, the gas production was recorded as the volume of head-space of syringe and the gas was also injected into a gas collection bag for H 2 S analysis. Ten milliliters of fermentation broth was sampled and stored at -80℃ for chemical analysis. The quantity of H 2 S of the gas sample and the concentrations of soluble sul de (S 2-) of the fermentation broth were determined using the methylene-blue colorimetric method, brie y, the adsorption liquid (per 1000mL, 3CdSO 4 ·8H 2 O 4.3g, NaOH 0.3g, ammonium polyvinyl phosphate 10g) was mixed with gas (10mL adsorption liqiud) and fermentation broth (adsorption liquid : fermentation broth, V/V=9:1), then followed the steps mentioned in previous studies [43,44]. The pH value was determined using a pH meter (INESA Scienti c Instrument, Shanghai, China) [29]. The concentration of sulfate radicals (SO 4

2-
) was determined using the turbidimetric method [29]. The concentrations of VFAs were determined using high-performance liquid chromatography [29]. The concentrations of methionine in the fermentation broth were tested by using automatic amino acid analyzer (Sykam, Munich, Germany), brie y, 1mL fermentation broth was mixed with 10mL hydrochloric acid solution (6mol/L), 3-4 drops of phenol liquid were added to the above mixture solution, then put the tube which contained the mixture solution on ice for 3-5min. The tube which contained the mixture solution was oxygen-free by the treatment of nitrogen ushing, then put the tube into air oven at 110℃ for 22h. After cooling the mixture solution at room temperature, the mixture solution was constant volume as 50mL after ltering by using lter paper. The 15mL ltering solution was dried at 40-50℃, and the deposition was washed with the deionized water twice and dried again. Then the deposition was diluted by 1mL 0.02mol/L hydrochloric acid solution, after ltering 0.22µm membrane the concentration of methionine was tested by automatic amino acid analyzer.
Bacterial abundance and gene expression in fermentation broth DNA was extracted from fermentation broth by using a QIAamp PowerFecal DNA Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions, brie y 1mL fermentation broth was centrifuged at 20,000×g for 1 min at 4℃ to collect the precipitate (approximately 200mg) for the DNA extraction, and the DNA puri cation following steps was referenced the protocol of manufacturer. RNA extraction of fermentation broth followed the protocol of RNeasy ® PowerMicrobiome TM Kit (Qiagen, Hilden, Germany), brie y, 1-2mL fermentation broth was centrifuged at 20,000×g for 1 min at 4℃ to collect the precipitate (approximately 200-250mg) for the RNA extraction, and the RNA puri cation following steps was referenced the protocol of manufacturer.
The DNA was used to quantify the relative abundance of Odoribacter splanchnicus and Bacteroides fragilis NCTC 9343. Brie y, the primers of these two bacteria was designed by using the software Primer 3, the sequences of the two bacteria was referenced the 16S rRNA sequencing in the web of NCBI, the details of the primers were showed in Table S5, the primer of bacterial16S rRNA was referenced previous study [45]. q-PCR was used to con rm the relative abundance of the two bacteria, the q-PCR reaction steps were followed the protocol of SYBR ® Green PCR Kit (SYBR, Japan) ( Table S6). The relative abundance of the two bacteria was calculated as 2 △Ct , where △Ct represents the difference in the Ct value for the 16S rRNA gene minus that for the genes [46].
Extracted RNA was reverse transcribed into cDNA by using the PrimeScript TM RT reagent kit (TaKaRa, Kusatsu, Japan). The cDNA was used to quantify the expression of Odosp_3416 and BF9343_2953. The primers were designed using the NCBI website with the total bacterial 16S rRNA gene as the reference gene (Table S5). q-PCR was used the con rm the relative expression level of the two genes, the q-PCR reaction steps were followed the protocol of SYBR ® Green PCR Kit (SYBR, Japan) with a little modi cation (Table S7). The relative expression level of the two genes was calculated as 2 △Ct , where △Ct represents the difference in the Ct value for the 16S rRNA gene minus that for the genes [46].

In vitro bacterial growth measurements
The anaerobic bacterium Bacteroides fragilis NCTC9343 were cultured at 37ºC by inoculating 40 mL aliquots of anaerobic basal medium (Becton Dickinson and Company, Lincoln Park, USA) and then grown anaerobically in an anaerobic chamber (Mitsubishi Gas Chemical Company, Inc. Tokyo, Japan). gga-miR-222a and the control mimic were supplied in the culture at a concentration of 2 µM. (RiboBio, Guangzhou, China). Growth was monitored as absorbance at 600 nm once per hour for up to 24 h with a spectrophotometer. The cultured bacterial cells were collected at 10h and used for BF9343_2953 gene expression measurement with the Bacteroides fragilis 16S rRNA gene as the reference gene. The concentrations of methionine in culture medium at 10h were tested as above mentioned.
In situ hybridization detection of the uptake of gga-miR-222a The bacterial cells of Bacteroides fragilis NCTC9343 were centrifuged at 12,000×g and washed twice with ice cold PBS. Then, the cells were xed in 4% PFA/0.25% glutaraldehyde. A 5'-DIG and 3'-DIG dual labeled probe for gga-miR-222a was used for in situ hybridization. The detection of the uptake of gga-miR-222a by bacteria was imaged using a Thermo Fisher Talos L120C transmission electron microscope Thermo (Fisher Scienti c, MA, US).

Statistical analysis
The data of the comparison of fermentation incubation indexes, the comparison of the relative abundance of miRNA, bacteria and genes, gas production, H 2 S production and growth curve were examined by analysis of variance (ANOVA) with Statistical Package for the Social Sciences (SPSS) software, version 22.0. Signi cant differences between the means were determined by Tukey's test.
Differences were considered signi cant at P 0.05.
The metatranscriptomic results of each sample was analyzed by HTSeq software, and the model used was union, the number of genes in different expression levels and the expression level of individual genes were statistically analysed. In general, the value of FPKM is 0.1 or 1 as the threshold for determining whether genes are expressed. The software DESeq was used for normalization of the read counts from analysis of genes expression levels [47].
The expression level of miRNA was calculated by the TPM formula (normalization read counts= (readCount*1,000,000)/libsize)), libsize was the sum of the read count of all miRNAs.