The rainbow trout reference miRNAome
In order to evaluate the complexity of the rainbow trout circulating miRNAome, we first established a comprehensive rainbow trout miRNAome that could be used as a reference. We identified and annotated 354 mature rainbow trout miRNAs corresponding to at least 280 miRNA genes by adapting a strategy previously used in different fish species [30, 32] and by using a total of 52 sequencing libraries composed of 14 new libraries (blood plasma and ovarian fluid) and 38 libraries that we had previously been generated in a wide variety of tissues, cellular populations, organs, and whole embryos [28]. The rainbow trout miRNAome was previously incompletely characterized with only 123 mature miRNAs described based on sequence homology with other species [28]. In a recent study, the number of known miRNAs identified in rainbow trout mucus, blood plasma, and surrounding water based on sequence homology with other species ranged from 94 to 192 depending on the samples [25]. With 354 annotated mature miRNAs, the present study therefore corresponds to a major increase in our knowledge of the rainbow trout miRNAome. In addition, existing studies in rainbow trout often rely on miRNA annotations available in other species (e.g., Atlantic salmon) [25] as no rainbow trout miRNA annotation is available in miRBase [35] and other databases [36]. Finally, the rainbow trout miRNAome annotation reported here is consistent with the recently described evolution of miRNA genes in teleost fish [29]. While slightly lower, the number of mature rainbow trout miRNAs reported here is in agreement with previous reports in other ray-finned fish species using similar genome-wide annotation strategies that led to the annotation of 362, 495, 396, and 408 individual mature miRNAs in gar, zebrafish, stickleback, and icefish, respectively [30,31,32]. In summary, the present report provides a comprehensive and evolutionarily supported rainbow trout miRNA repertoire annotation that was previously incompletely characterized in this species.
The high complexity of circulating miRNAome
Among the 354 annotated rainbow trout miRNAs, 331 were detected on average above a threshold of 10 reads per million reads (RPM), either in one of the two biological fluids studied or in one of the 21 other sample types analyzed (brain, pituitary, gills, heart, muscle, myoblasts, myotubes, stomach, intestine, liver, spleen, head-kidney, leucocytes, trunk-kidney, skin, gonad, testis, spermatogonia, ovary, eggs, whole embryos). In biological fluids, 211 miRNAs were identified and corresponded to 64% of the overall expressed miRNAome diversity (Fig. 2A). Among these 211 miRNAs, 172 (82%) were detected in both blood plasma and ovarian fluid, while 24 (11%) were detected only in the blood plasma and 15 (7%) were detected only in the ovarian fluid (Fig. 2A). Notably, two miRNAs (miR-365-2-5p, miR-23b-2-5p) were detected above the 10 RPM threshold only in the blood plasma and not in any other studied sample, while one miRNA (miR-726-5p) was found only in the ovarian fluid (Fig. 2A). A comprehensive analysis of miRNA expression levels revealed that the overall distribution patterns of miRNA read counts were similar in biological fluids and in other samples (Fig. 2B). In each analyzed library, including ovarian fluid and blood plasma libraries, a few miRNAs accounted for most of the reads per million. Together, these data illustrate the relatively large complexity of c-miRNAomes of the two biological fluids studied here. Our results are consistent with existing data in the human, chicken, and cow blood plasma in which 349, 649, and 468 miRNAs were reported, respectively [4, 21, 24]. This result in trout, however, is to our knowledge the first comprehensive characterization of blood plasma and ovarian fluid miRNAomes in fish.
Origin and specificity of c-miRNAs in the blood plasma and ovarian fluid
To investigate the possible organs of origin of miRNAs present in the blood plasma and ovarian fluid, we categorized c-miRNAs based on the organ in which they exhibited the highest expression, under the hypothesis that an organ strongly expressing a miRNA is likely an organ secreting the miRNA into the fluid, or at least one of the major contributors. This analysis investigated a subset of 13 different organs from females (brain, pituitary, gills, heart, muscle, stomach, intestine, liver, spleen, head-kidney, trunk-kidney, skin, ovary) and excluded male samples, complex libraries (e.g., whole embryos), and individual cell types (e.g., myoblasts). We observed that miRNAs present in both ovarian fluid and blood plasma had maximum expression (i.e., were detected at the highest level) in a wide diversity of organs (Fig. 3A). For both analyzed fluids, the brain, gills, pituitary, ovary, and liver were the organs in which most miRNAs were the most highly expressed and no major differences in potential organs of origin could be identified between these two fluids (Fig. 3A). These data suggest that many organs might contribute to the complexity of c-miRNAomes in both blood plasma and ovarian fluid. It is however noteworthy that the different organs used in the analysis likely contained some blood at the time of sampling. This could have led to an overestimation of the possible organs of origin of specific c-miRNAs, especially for the organs exhibiting low expression of these c-miRNAs. In addition, a recent study in rainbow trout showed that the abundance of specific miRNAs following a stressful event exhibited an inverse relationship between tissues and blood plasma extracellular vesicles that could indicate that the liver and head kidney secreted these miRNAs [37]. Further analyses monitoring miRNA abundance in fluids and putative tissues and organs of origin over time are needed to further understand the origin of c-miRNAs.
When miRNA expression in fluids was analyzed together with expression data in the panel of 13 female organs from which they could originate, we observed that both blood plasma and ovarian fluid exhibited specific expression patterns as distinct as in the other analyzed organs, if not more. The heatmap presented in Fig. 3B clearly shows that, while sharing common strongly expressed miRNAs, each fluid was nonetheless characterized by the overabundance of several miRNAs (the yellow ones towards the top right of the panel) that were not overabundant in any organs analyzed in the present study (Additional file 1). The PCA analysis carried out using all available samples clearly illustrated that biological fluid miRNAomes, while distinct, were also clearly different from all other “solid” tissue and organ miRNAomes (Fig. 4A). When analyzing the presence of miRNAs in the different libraries, we observed that most miRNAs present in the blood plasma and ovarian fluid were also detected in most organs (Fig. 4B).
When analyzing the expression of the miRNAs exhibiting strong expression in the ovarian fluid and blood plasma, we observed that among the 10 most abundant miRNAs in the ovarian fluid and plasma, seven (miR-451-5p, let-7a-5p, miR-21-5p, miR-16b-5p, miR-26a-5p, let-7e-5p, miR-30d-5p) were common to both fluids. In the blood plasma, miR-92a-3p, miR-150-5p, and miR-128-3p were the three other most abundant miRNAs. In the ovarian fluid, miR-202-5p, miR-22a-1-3p, and miR-146a-5p were the three other most abundant miRNAs. For miR-451-5p, miR-16b-5p, miR-26a-5p, and miR-92a-3p, a clear over abundance was observed in both fluids in comparison to organs (Fig. 5). In contrast, the other most abundant miRNAs in fluids were also highly abundant in at least one other organ (Fig. 5).
Together, these data indicate that c-miRNA repertoires in rainbow trout are complex. Similar to what was observed for miRNAs in different organs, c-miRNAomes in the blood plasma and ovarian fluid each exhibited specific expression profiles, with fluid-specific combinations of highly expressed miRNAs, major differences in miRNA abundances, and fluid-type-specific miRNAs in comparison to the other fluid and organs analyzed. These data suggest that the complexity of c-miRNAomes in the blood plasma and ovarian fluid results, at least in part, from the accumulation of miRNAs originating from a wide diversity of organs. The presence in body fluids of miRNAs that cannot be detected in other organs, or detected at much lower levels, suggests that these miRNAs originated from other sources that were not investigated here. These specific miRNAs could also have originated from miRNA-expressing cells present in these biological fluids or in their vicinity. For example, miR-451-5p could have originated from erythrocytes that greatly express this miRNA during the late stage of red-blood cell maturation [38,39,40,41] and miR-21-5p could have been expressed by the endothelial cells forming the vasculature [42]. Finally, it is also possible that the level of these miRNAs resulted from their progressive accumulation in these fluids over time permitted by their high stability in nuclease-rich fluids [43].
Differences and similarities of ovarian fluid and blood plasma miRNA repertoires
As indicated above, the blood plasma and ovarian fluid had many miRNAs in common. Marked differences between blood plasma and ovarian fluid miRNAomes, however, existed in terms of both miRNA profiles and expression of fluid-specific miRNAs. The PCA analysis of fluid samples only (Fig. 6A) clearly showed that the overall c-miRNA profiles differed between the blood plasma and ovarian fluid samples. Accordingly, 138 miRNAs were significantly differentially abundant between the ovarian fluid and blood plasma (Additional file 2). Among these 138 miRNAs, 67 were over-abundant in ovarian fluid and 71 in the blood plasma.
Our results are consistent with previous studies in humans showing distinct miRNA compositions in different body fluid types [4]. While the authors suggested a common origin for miRNAs present in the different body fluids, they also reported fluid-specific enrichment of several miRNAs, including in the blood plasma. In the present study, two miRNAs (miR-202-5p and miR-194b-5p) exhibited a dramatic, over 100-fold, enrichment in ovarian fluid in comparison to the blood plasma (Additional file 2). In contrast, two miRNAs (miR-460-5p and miR-365-2-5p) exhibited the opposite pattern. While the function of extracellular miRNAs remains unclear [43], it has been hypothesized that fluid-specific miRNAs could have regulatory functions in surrounding tissues [4]. In rainbow trout, as in many vertebrates, miR-202-5p is predominantly expressed in gonads [28, 44,45,46,47,48]. The strong abundance of miR-202-5p in ovarian fluid therefore agrees with its strong ovarian expression because the ovarian fluid, in which ovulated oocytes (i.e., unfertilized eggs) are held in the body cavity until spawning, is, at least in part, from ovarian origin [49]. Conversely, the blood perfuses all organs and transports molecules throughout the body, including in the ovaries [4] and many ovarian fluid components such as proteins are also known to be brought in from the blood [50, 51]. Ovarian fluid c-miRNAs may thus also, in part, be brought in the ovarian fluid via the blood, which would be consistent with the presence of a high proportion of common miRNAs in both fluids. In medaka, miR-202-5p plays a major role in female reproduction, specifically in the control of egg production and egg ability to be fertilized [44]. The overabundance of miR-202-5p in ovarian fluid, compared to the blood plasma, would be consistent with a physiological role of miR-202-5p in the ovarian fluid before, at, or after ovulation, a period associated with major events, including the final maturation of oocytes and the onset of the next reproductive cycle.
Varying c-miRNA abundance in response to reproductive and metabolic states
In the present study, we also aimed at identifying blood plasma c-miRNAs that change in expression level during the reproductive cycle or in response to different metabolic states resulting from different feeding levels. The PCA analysis (Fig. 6B) revealed clear differences in blood plasma c-miRNAomes during the reproductive cycle. Differences were especially noticeable between samples taken at the beginning of the reproductive cycle (i.e., at previtellogenic stage) compared to samples taken later during the reproductive cycle. The statistical analysis led to the identification of 107 differentially abundant miRNAs during the reproductive cycle (Additional file 3). The heatmap of these miRNAs presented in Fig. 7A revealed four different clusters of miRNA expression profiles during the reproductive cycle. Most changes in expression occurred between previtellogenesis (PV) and early-vitellogenesis (EV). The differential expression analysis resulted in the identification of 48 downregulated (cluster 1) and 50 upregulated (cluster 2) miRNAs in PV compared to EV (Fig. 7B). In addition, we identified five miRNAs upregulated at ovulation (OV, cluster 3) and four miRNAs exhibiting the opposite pattern (cluster 4). Together, these results indicate that major changes occurred in the blood plasma c-miRNAome during the reproductive cycle and that a significant proportion of the blood plasma c-miRNAome (107 c-miRNAs, 55% of the overall blood plasma c-miRNAome) exhibited a differential abundance between at least two stages of the reproductive cycle. These results show that the blood plasma miRNAome exhibits marked stage-specific signatures during the reproductive cycle.
In contrast, we were not able to detect any significant differences in c-miRNA abundances in the blood plasma in response to metabolic levels (i.e., feeding level) (“al” for ad libitum and “r” for restricted diet in Fig. 7). This result could, however, originate from the low number of replicates (two sample pools per diet) that composed our RNA-seq dataset.
Circulating blood plasma miRNAs as non-invasive biomarkers of metabolic and reproductive states
To further evaluate the potential of blood plasma c-miRNAs to respond to differences in metabolic and reproductive states, we selected the most promising miRNAs (i.e., exhibiting the highest fold-changes between different metabolic levels or reproductive stages in our small RNA-seq data) and conducted an extended analysis of their expression by quantitative PCR (QPCR) using five individual replicates and an additional time point during the reproductive cycle (Late vitellogenesis, LV) (Fig. 8). The potential origin of these candidate c-miRNAs was also analyzed by QPCR in a panel of organs to shed light on their possible origin of expression (Fig. 9). Quantitative PCR demonstrated that selected biomarker c-miRNA candidates exhibited highly significant changes in their blood plasma abundance throughout the reproductive cycle. In most cases, these changes occurred in a feeding level-dependent manner, indicating that circulating miRNA levels in the blood plasma can be deeply influenced by metabolism. Among analyzed candidate biomarkers, miR-1-1/2-3p, miR-133a-1/2-3p, and miR-206-3p exhibited a similar pattern throughout the reproductive cycle with a dramatic increase in blood plasma abundance during vitellogenesis (i.e., the reproductive phase characterized by major yolk protein uptake from the blood stream by the oocyte) when fish were fed ad libitum but not when the food was restricted (Fig. 8A–C). When investigating the organs expressing these three miRNAs, we observed a predominant expression in skeletal muscle (Fig. 9A–C). These miRNAs are known to be muscle-specific miRNAs, often referred to as “myomiRs” [52], and have been associated with myogenesis and with various biological processes in the skeletal muscle and heart [53,54,55,56,57,58,59,60,61,62]. In Nile tilapia, an increase in the expression of these three myomiRs was observed in muscles throughout the fish life [63]. The association between these miRNAs and active myogenesis thus appears to be evolutionarily conserved in vertebrates. A higher level of blood plasma myomiRs in well-fed animals compared to animals under a restricted diet would be consistent with the significant increase in growth rate observed in the present individuals when fed ad libitum [27]. Together, these observations suggest that blood plasma levels of miR-1-1/2-3p, miR-133a-1/2-3p, and miR-206-3p have the potential to identify episodes of active myogenesis. Under the hypothesis that these potential biomarker myomiRs reflect muscle growth rate, which requires experimental validation using additional samples and individuals held in a variety of experimental conditions, this result could offer a wide range of possible applications. For wild population management, these biomarker candidates could for instance offer the possibility to assess the quality of an ecosystem through the ability to monitor fish growth throughout the year. In aquaculture, it could allow fine phenotyping of muscle growth in response to specific diets or rearing conditions. More importantly, these biomarker candidates could allow to specifically question muscle growth in comparison to global body growth that can be influenced by the development of other tissues such as fat deposits, an information that is currently not easily accessible without sacrificing the fish.
Among the miRNAs that we investigated by QPCR to assess their potential use in non-invasive phenotyping, four c-miRNAs exhibited significant changes in their blood plasma abundance in response to feeding rate, either globally in the case of miR-375-3p (Fig. 8D) or in interaction with the reproductive stage in the case of miR-214a-3p, miR-30c-3p, and miR-221-3p (Fig. 8E–G). The latter, miR-214-a-3p, miR-30c-3p, and miR-221-3p, was also expressed in a wide variety of organs (Fig. 9E–G), making it hazardous to speculate on their organ of origin and the biological processes in which they may be involved. The expression profiles of these c-miRNAs, however, indicate that they could be used, most likely in combination with other c-miRNAs, to estimate the metabolic state of the fish at a given reproductive stage. Interestingly, these three c-miRNAs have been used as blood plasma biomarkers for several human pathologies such as cancer (liver, prostate, ovarian, and pancreatic cancers) and cardiovascular and renal diseases [64,65,66,67,68,69]. In contrast, the highly predominant expression of miR-375-3p in the pituitary (Fig. 9D), which is consistent with existing data in other vertebrate species [70, 71] can tentatively be associated with the neuroendrocrine control of biological processes such as nutrition and reproduction. This c-miRNA exhibited highly significant differences in blood plasma abundance in response to feeding rate both globally and in a reproductive-stage-dependent manner (Fig. 8). The difference in miR-375-3p levels in response to feeding rate was especially marked during late vitellogenesis and ovulation period. In other animal species, miR-375-3p has been implicated in the regulation of insulin [71] but a role in reproduction metabolism has also been suggested [70, 72]. Even though the role of miR-375-3p in animal reproduction remains unclear, this miRNA appeared to be abundant in the blood plasma and highly responsive to changes in reproductive and metabolic states in female rainbow trout. For these reasons, miR-375-3p is a highly promising candidate biomarker for non-invasive phenotyping of neuroendocrine response in rainbow trout and possibly other animal species.
Among the c-miRNAs that we monitored by QPCR throughout the reproductive cycle, miR-202-5p had the most striking profile (Fig. 8H). Independent of the feeding regime, miR-202-5p exhibited a dramatic increase in its blood plasma abundance at ovulation and was also among the most highly abundant c-miRNAs in ovarian fluid according to the small RNA-seq data. In fish, miR-202-5p plays a major role in reproduction and female medaka lacking expression of miR-202-5p produced fewer eggs and of lesser quality [44]. It is therefore possible that miR-202-5p in the blood plasma and in ovarian fluid plays an important biological role around the time of ovulation that would require further investigations. As already described in rainbow trout, teleost fishes and other vertebrates, miR-202-5p is predominantly expressed in the ovary and was also detected in unfertilized eggs (Fig. 9H) [28, 44, 45, 73, 74]. In the rainbow trout ovary, miR-202-5p was differentially expressed during oogenesis with a peak of expression during vitellogenesis followed by a progressive decrease during final oocyte maturation [75]. The profile of miR-202-5p in the blood plasma reported here with a peak of expression at ovulation thus differs from its ovarian expression. It is possible that this discrepancy results from the delay between expression in the ovary during vitellogenesis and accumulation in the blood plasma during periovulatory period. This is, however, unlikely given the 2–3-month periods between mid-vitellogenesis and ovulation. The sharp increase in blood plasma miR-202-5p levels at ovulation (Fig. 8D), in contrast, suggests a release during the periovulatory period, either from the ovary or from the eggs. It is thus possible that a dynamic accumulation of miR-202-5p in the blood plasma occurs either immediately prior to or following ovulation. Under this hypothesis, circulating miR-202-5p levels could serve as a biomarker to predict approaching ovulation, if the accumulation in the blood plasma occurs prior to ovulation, or to estimate post-ovulatory egg ageing, if the accumulation in the blood plasma occurs at or after ovulation. In both cases, this c-miRNA would be of major interest as a non-invasive phenotyping biomarker enabling, in aquaculture or wild resource management settings, the selection of females that are close to or at ovulation to prevent the occurrence of post-ovulatory ageing of the eggs, a phenomenon associated with a dramatic decrease in egg quality [76].
Together, both small RNA-seq and QPCR data revealed that the levels of selected circulating miRNAs exhibited major differences during the female rainbow trout reproductive cycle and, for some of them, also in response to changes in metabolic state. Some of these c-miRNAs therefore appear to be highly relevant candidate biomarkers that could serve for non-invasive phenotyping of sexual maturation (i.e., progress into the reproductive cycle) and episodes of muscle growth. These results are consistent with recent observations in rainbow trout showing that specific c-miRNAs were differentially abundant in the blood plasma, mucus, and surrounding water very rapidly after a stressful event [25]. Together, these observations highlight the strong potential of c-miRNAs to serve as biomarkers and non-invasive indicators of stress, reproductive and metabolic states, and myogenic activity. Further investigations are however needed to explore their potential in other physiological and pathological contexts in various fish species and to validate them as biomarkers. The identification of other biomarker c-miRNAs, such as markers of viral and bacterial infections, would allow the collection of a panel of relevant complementary information from a single blood sample and thus offer tremendous phenotyping possibilities.