1 Introduction

1.1 Salmonid Aquaculture and Related Stressors

Since 2010, the global output of aquaculture reached 160 millions of tonnes, twice the amount produced by fisheries (FAO/OMS 2015). Of this number, salmonids are second to carps as the most important group of farmed fish, with a total output of over 2 million tonnes. Atlantic salmon (Salmo salar), which makes for two thirds of this market, has been since 1982 almost exclusively mass-produced through fish farming (Fig. 1).

Fig. 1
figure 1

Global production output for Atlantic salmon from 1950 to 2014. (asterisk) The red line indicates the tipping point (1982) from which aquaculture output started to surpass fisheries. Data from FAO Fish Stats (accessed July 4, 2018)

Full size image

This implies a strong pressure on fish farmers to keep up with high demand for this source of animal protein. Intensive farming practices have been developed to maximize production but at the expense of exposing farmed fish to several simultaneous stressors including frequent handling procedures, overcrowding, and poor water quality (Madaro et al. 2015). Unlike wild fish, captive fish cannot escape from those stressors. Even though stress is necessary to survive a danger or challenge, prolonged inescapable stress factors related to intensive rearing reduce the capacity of fish to maintain homeostasis, putting energy allocation required for reproduction, growth, and persistence on hold (Schreck 1982).

The reallocation of energy that occurs during the stress response is triggered by increased levels of the glucocorticoid hormone cortisol (Barton 2002). One of the major functions to be downregulated by elevated glucocorticoid levels is the immune system (Pickering and Pottinger 1985). Through inhibition of key transcription factors, cortisol effectively suppresses humoral factors involved in the inflammatory response and immune cell trafficking (Fast et al. 2008). Excess plasma cortisol has been shown to increase the susceptibility of brown trout (Salmo trutta), Atlantic salmon (S. salar), rainbow trout (Oncorhynchus mykiss), and numerous other hosts to bacterial, fungal, and parasitic diseases (Pickering and Duston 1982; Maule et al. 1989; Wilk et al. 1989; Johnson and Albright 1992).

When fish are exposed to persistent and inescapable stressors over a prolonged period of time, cortisol levels tend to remain elevated, thereby hindering the ability of stressed fish to revert back to a resting state. This was observed in Atlantic salmon, where fish exposed to handling stress had 25–75 mg/mL cortisol levels for up to 23 days post-challenge compared to near-zero levels in control groups (Madaro et al. 2015). In another study, head kidney macrophages from stressed Atlantic salmon (15 s out of water daily for 30 days) showed decreased survival when exposed to Aeromonas salmonicida (Fast et al. 2008).

Sanitary (e.g., egg disinfection), prophylactic (e.g., vaccination), and curative (e.g., antibiotherapy) measures in an intensive farming environment is commonly used in order to compensate for the immune impairment that results from an over-elicited stress response. When an infection occurs, it can be laborious and time-consuming to identify the strain that causes the disease. Consequently, antimicrobial agents having a broad spectrum (i.e., targeting a wide range of bacterial species) are prioritized. Although the bacterial strain that caused the infection may be correctly targeted, a wide range of other bacteria (including beneficial symbionts) are also affected. This can disrupt global interactions between the host and its microbial flora (i.e., microbiota) that play a key role in maintaining fish health in the long term.

1.2 Host-Microbiota Interactions and Their Involvement in Health

All animals live in close association with trillions of microbial cells. Their abundance is so important that they outnumber host cells by a 2:1 ratio (Sender et al. 2016). Up to 1.5% of an individual’s biomass accounts for these microbes (Karlsson et al. 2013). Those constitute the host microbiota, i.e., the consortium of microbes residing on host surfaces (e.g., skin, intestines, etc.). In humans, the collective gene complement (i.e., the metagenome) of the microbiota may dwarf its host by a 150-fold factor in terms of unique functions (Qin et al. 2010). This vast gene repertoire assists the host by providing additional functions, such as metabolic pathways to digest otherwise indigestible compounds. As an example, humans cannot digest cellulose but gut bacteria in the large intestine can digest it into short-chain fatty acids (SCFAs), which humans can process with their own enzymatic toolbox (Cummings 1984). Host microbiota also contributes to immunity through (1) direct antagonism toward pathogenic microorganisms (Cherrington et al. 1991; Hammami et al. 2013), (2) signaling to the immune system (Swiatczak and Cohen 2015), and (3) reducing the carrying capacity of the host for exogenous pathogens (Kamada et al. 2013).

Benefits provided by host-microbiota interactions are highly dependent on environmental and physiological parameters. For example, acute stress responses typically shut down digestive (Mayer 2000) and immune functions (Morey et al. 2015) to react to a life-threatening danger. Alterations of these functions change both the availability of certain nutrients for microbial symbionts and reactivity of the immune system toward them. As a result, stress indirectly alters microbiota composition and, thereby, the interactions with its host.

The economic importance of salmonid fish calls for a better understanding of their host-microbiota interactions to develop therapeutic tools that are less damaging for the environment and human health (Llewellyn et al. 2014). This chapter aims, on the one hand, to present an overview of the current knowledge on the taxonomic composition (i.e., diversity and structure) of salmonid microbiota and processes governing its assembly (ontogenesis). On the other hand, the state of the art on microbial profiling and modulation will be discussed, as well as current research gaps and perspectives.

2 An Overview of Salmonid Microbiomes

A special attention has been given to the microbiota of salmonids that are significantly important in aquaculture.

2.1 Atlantic Salmon (Salmo salar)

Atlantic salmon has the most extensively characterized microbiota of all salmonids to this present day, with 17 dedicated studies published between 2007 and 2018.Footnote 1 Most of this research focused on the skin and gut microbiota, including assessments of its response to migration, nutrition, antibiotherapy, and captivity (Navarrete et al. 2008; Gajardo et al. 2017; Dehler et al. 2017; He et al. 2018).

2.1.1 Compositional Shifts During Freshwater-Seawater Migration

As other anadromous fish species, Atlantic salmon is exposed to two highly contrasted environments during its life cycle (i.e., freshwater and seawater), both of which differ greatly in terms of salinity, temperature, nutrient availability, and, potentially, environmental microbial exposure (Héry et al. 2014). Accordingly, major shifts in the abundance of dominant bacterial phyla were found in the skin and gut microbiota of Atlantic salmon before and after smoltification (Fig. 2). The most abundant phyla in skin and gut, respectively, Proteobacteria and Firmicutes, increase in abundance during freshwater to seawater transfers, while Actinobacteria decrease in both types of microbiota during this process (Lokesh and Kiron 2016; Rudi et al. 2018). However, the impact of this compositional shift on host physiology, and vice versa, remains unclear.

Fig. 2
figure 2

Relative abundance of the dominant bacterial phyla in the Atlantic salmon skin and gut microbiota. *3 weeks post-transfer in seawater. Data from Lokesh and Kiron (2016) and Rudi et al. (2018)

Full size image

2.1.2 Influence of the Diet and Protein Sources

Wild Atlantic salmon feed exclusively on animal protein; juveniles start with zooplankton and feed on larger fish as they grow (Harvey et al. 2016). Accordingly, farmed salmon should be specifically fed with fishmeal as a primary source of animal protein. However, there is a growing pressure on the aquaculture industry to reduce the fishmeal content of feeds for improved sustainability and reduced cost (Rimoldi et al. 2018). Plant-based protein sources are an increasingly popular replacement for fishmeal (Newaj-Fyzul and Austin 2015). However, marine carnivorous fish have not evolved mechanisms to efficiently digest carbohydrates and non-nutritious compounds present in plant-based meals (Naylor et al. 2000). Soybean meal, one of the most promising alternatives to fishmeal (Herman and Schmidt 2016; Park et al. 2017), contains compounds that trigger inflammation in the distal intestine of salmonids (Heikkinen et al. 2006). Even though those inflammatory compounds can be removed by alcohol extraction, the resulting soybean protein concentrate (SPC) still alters the intestinal microbiota (Table 1). Nevertheless, fish fed with a SPC-rich diet supplemented with mannan-oligosaccharide (MOS) had an alpha diversity index more similar to fish fed exclusively with fish and terrestrial animal meals (Table 1). In another study, Atlantic salmon fed with either soybean meal, SPC, or guar meal had higher levels of lactic acid bacteria (LAB), as well as higher expression levels of proliferating cell nuclear antigen (PCNA). The cause-effect relationship between legume-based diets, LAB, and PCNA levels is still elusive, however (Gajardo et al. 2017).

Table 1 Shannon alpha diversity index of the distal gut microbiota from Atlantic salmon fed with diets of varying protein sources
Full size table

2.1.3 Antibiotherapy

To our knowledge, few studies have addressed the impact of antibiotherapy on the Atlantic salmon microbiota. In 2017, a study investigated the impact of oxytetracycline (OTC), one of the most commonly used antibiotics against salmonid infectious diseases (Miranda and Zemelman 2002). OTC was administered daily in the form of medicated feed to salmon fingerlings. Microbiota composition was assessed by RFLP-PCR and sequencing of 16S rDNA amplicons. Whereas untreated microbiota was diverse and consisted mainly of Pseudomonas, Acinetobacter, Bacillus, Flavobacterium, Psychrobacter, and Brevundimonas spp., the microbiota of OTC-treated fish was dominated by Aeromonas sobria and A. salmonicida. Both species are known to harbor oxytetracycline resistance genes (Balassiano et al. 2007; Trudel et al. 2016). The latter is a well-known salmonid pathogen that causes furunculosis, a major opportunistic disease (Bullock et al. 1983). This study presented a textbook example of proliferation of opportunistic bacteria by collateral removal of competing microorganisms (see chapter “The Rise and Fall of Antibiotics in Aquaculture” for a detailed discussion on this topic). To our knowledge, no other study investigated specifically how antibiotics impact the Atlantic salmon microbiota. In this regard, there is a significant knowledge gap regarding other common antimicrobials such as florfenicol (Nordmo et al. 1998) and sulfamethoxazole-trimethoprim (Kadlec et al. 2011).

2.1.4 Captivity

Even though the bulk of salmon aquaculture is intended to produce food, government-led programs were also introduced to restore endangered salmon populations in rivers (Province of Quebec, Canada, is a notable example). Most involve stocking rivers with hatchery-reared juveniles (usually 0+ or 1+ parrs). Even though stocking parrs is preferred to stocking captive adults because of the latter’s low reproductive success, captive parrs do not survive as well in the wild as their wild-born counterparts. A 2018 study revealed a substantial mismatch between the microbiota of captive (meant for stocking) and wild parrs (Lavoie et al. 2018) sampled from two different rivers. Even though community composition from wild parrs was specific to the river, captive fish (born from wild breeders from either river) were not significantly differentiated despite their distinct genetic origin. Furthermore, their microbiota composition was highly distinct from their wild fish relatives. In addition, captive parrs’ microbiota was dominated by Firmicutes (Lactobacillaceae), whereas wild parr’s microbiota was enriched with Proteobacteria (Enterobacteriaceae). Those results were consistent with previous studies indicating that the microbiota composition is highly associated with the diet protein source (Desai et al. 2012; Gajardo et al. 2016). As such, captive parrs are fed with commercial pellets made from vegetable proteins, a great source of carbohydrates. The latter has been associated with an increase of lactic acid bacteria (LAB) such as Lactobacillaceae and contributes to the divergence between captive and wild parr’s microbiota composition.

As well as highlighting the substantial contribution to diet and environmental conditions on the microbiota composition, this study also confirmed that the bacterial species richness (alpha diversity) can be associated with the selective pressure of an environment. In comparison to captive parrs, wild juveniles showed a much lower diversity index and a higher homogeneity within the individual’s microbiota composition, suggesting that higher selective pressure translates into a more specialized microbiota composition (Derome et al. 2006).

Interestingly, some disparities were detected when studying the network interactions of taxa according to the parr’s origin. For instance, a higher proportion of negative interactions was found within captive parr’s microbiota (Fig. 3). Those results are of prime interest since it has been established that negative correlations can be associated with a higher dysbiosis index (Vázquez-Baeza et al. 2016). Overall, captivity and hatchery rearing highly contribute to the microbiota composition, even for parrs from the same genetic population. Studying microbial ecology in the aquaculture field is therefore totally pertinent for assessing the effect of captivity on physiology and microbiota, especially if reared fishes are meant to be released thereafter.

Fig. 3
figure 3

Network analysis (Spearman correlation coefficient) of the microbiota at the family level depending on parrs’ origin in two Canadian rivers. Node size is proportional to the relative abundance of each taxa, where low abundance taxa show more interactions than abundant taxa. Font size is proportional to the number of interactions for each taxa, illustrating keystone taxa for every environment. Negative correlations are illustrated by red edges. (a) Malbaie captive parrs; (b) Rimouski captive parrs; (c) Malbaie wild parrs; (d) Rimouski wild parrs. Data from Lavoie and Derome (2018, unpublished data)

Full size image

Since the diet of fishes appeared to be greatly associated with the host metabolism, stocking methods have been modified over time to mitigate the physiological mismatch of hatchery-reared fishes that are meant to be reintroduced in nature (Milot et al. 2013). For example, hatcheries are now raising Atlantic salmon juveniles until the alevin stage, which still have their unabsorbed yolk sac at the moment of stocking. However, hatchery rearing seems to have left a permanent imprint on the microbiota of stocked alevins, despite them being no longer exposed to an artificial environment and not being fed during the rearing (Lavoie and Derome 2018, unpublished data).

By analyzing the microbiota of wild and stocked juveniles that have been sampled in the river 4 months after stocking, differences between the microbiota of stocked and wild individuals are still highlighted. Overall, discrepancies are observed for the taxonomical composition of the microbiota as well as the diversity, suggesting a potential mismatch for metabolic functions. For instance, a higher diversity index is associated with stocked parrs’ microbiota, indicating that the rearing conditions have a permanent effect on the structure of the microbiota. Even though the exact contribution of this mismatch on host fitness is unclear and deserves further investigation, hatchery rearing is proven to drive microbial processes at various levels. Acquiring a better understanding on how the microbiota is affected by the environment after stocking will certainly lead to the optimization of the conservation methods of endangered fish species.

2.2 Rainbow Trout (Oncorhynchus mykiss)

Rainbow trout is second to the Atlantic salmon as the most produced fish in salmonid aquaculture with 812,000 tonnes produced in 2014, which represents 35% of global Atlantic salmon production (FAO Fisheries and Aquaculture 2018). A total of 14 studiesFootnote 2 on the rainbow trout microbiota were published from 2010 to 2018, most of which investigated the influence of diet on various physiological parameters. The remaining studies investigated the role of the microbiota in growth promotion and pathogen inhibition, as well as diet-immunity interactions and their impact on the microbiota. None of those studies addressed the impact of antibiotherapy on the microbiota composition and subsequent effects on fish health. Perhaps due to the relevance of this species in aquaculture, a special emphasis on the intestinal microbiota was found throughout most of the aforementioned studies.

2.2.1 Influence of Nutrition

Farmed rainbow trout was found to possess a core gut microbiota of 52 bacterial lineages (Wong et al. 2013). This core gut microbiota was remarkably resilient to interindividual variation, diet, and rearing density changes, with no significant change in the abundance of bacterial classes. Nevertheless, the gut microbiota as a whole responds to diet composition.

One of the hot topics in rainbow trout aquaculture is the use of alternative (plant-based) protein sources to improve the sustainability of fish farming. However, those are quite rich in carbohydrates compared to rainbow trout’s natural feed, which is very rich in protein (>40%) and poor in carbohydrates (<1%). A short hyperglucidic-hypoproteic stimulus (HHS) during early life stages was found to induce a long-term influence on the gut fungi (but not bacteria) profiles. Furthermore, it induced upregulation of glucose metabolism genes and downregulation of gluconeogenesis and amino acid catabolism genes in muscle tissue (Geurden et al. 2014). In the long term, HHS-treated fish did not differ in growth, feed intake, or efficiency of feed utilization. However, a significant effect on glucose homeostasis was observed. Up to 9 h after being fed the same commercial diet, HHS-treated fish had 1.5-fold higher glycemia than untreated fish. This hints to the possibility of nutritional programming as a way of optimizing the use of alternative plant-based feeds in fish farming.

The rainbow trout gut microbiota also responds to the inclusion of dietary additives. Supplementation with a plant essential oil mixture (MixOil) altered gut microbiota diversity indices and fillet quality metrics, but the link between those two remains unclear (Ceppa et al. 2018). The inclusion of organic acids in aquafeed was found to influence gut microbiota composition, but with unclear effects on host physiology (Jaafar et al. 2013).

2.2.2 Diet-Immunity Interactions

One of the main causes of mortality in rainbow trout aquaculture is Yersinia ruckeri, the causative agent of enteric redmouth disease (Tobback et al. 2007). The initial target organ for Y. ruckeri appears to be the gut (Méndez and Guijarro 2013). Accordingly, the administration of probiotic bacteria (via coated feed) enhanced resistance to this pathogen (Raida et al. 2003), but the mechanism of action was unclear.

In 2014, a Danish team investigated the missing link between microbiota, diet, and the immune response in fish challenged with Y. ruckeri (Ingerslev et al. 2014). Rainbow trout fry challenged by Y. ruckeri were split into two diet groups containing either (1) fishmeal + fish oil or (2) fishmeal/Pea meal (9:1) + rape seed oil. Microbiota composition was assessed by deep sequencing of the 16S rRNA gene, and immune gene expression was quantified by RT-qPCR. In summary, challenged fish fed with the marine-based diet had higher counts of Yersinia (as determined by either bacteriology or 16S profiling) and had also increased expression levels of interleukins 1-beta and 2. The plant-based diet may have had a prebiotic effect by favoring the presence of taxa that are protective against Y. ruckeri (Ingerslev et al. 2014). However, post-infection cumulative survival did not significantly differ between challenged fish fed either diet.

2.2.3 Pathogen Inhibition

The total cultivable microbiota of Chilean farmed rainbow trout harbors lactic acid bacteria (LAB) in high abundance (Araújo et al. 2015). Of those, 71% (mostly Lactococcus lactis isolates) possess inhibitory activity against one or more of the following pathogens: Lactococcus garvieae, Streptococcus iniae, Yersinia ruckeri, Aeromonas salmonicida, and Vibrio campbellii (Araújo et al. 2015). Whether those LAB isolates do possess inhibitory activity in vivo remains to be investigated.

2.3 Brook Charr (Salvelinus fontinalis)

The brook charr microbiota remains largely mischaracterized, except for the skin mucus (SM) microbiota, for which response to intensive rearing conditions and symbiont-pathogen interactions (including interindividual variations) were investigated (Boutin et al. 2012, 2013a, b, 2014). To our knowledge, no published studies have yet discussed the impact of diet or therapeutic tools on both microbiota structure and brook charr physiology.

2.3.1 Microbiota Structure and the Stress Response

The brook charr SM microbiota is dominated by the bacterial phyla Proteobacteria, Actinobacteria, and Bacteroidetes (Boutin et al. 2013b). Their relative abundance shifts abruptly when fish are exposed to hypoxia and high-density stress (Fig. 4). In addition to those abundance shifts, strong co-occurrence patterns were found. Some co-occurring genera associated with opportunistic diseases (Psychrobacter, Steroidobacter, Pseudomonas, Acinetobacter, Aeromonas) were specific to stressed and dead fish, whereas others (Sphingomonas, Methylobacterium, Propionibacterium, and Thiobacter) were abundant only in unstressed fish (Boutin et al. 2013b). Beneficial bacteria tended to decrease in a colinear manner following a stress event, thus resulting in an empty niche for opportunistic pathogens, which accordingly tended to increase as co-abundant groups. The role of SM microbiota in preventing infections in its host might be more important than previously thought. Indeed, several endogenous strains, in addition to those isolated from the gut microbiota, have shown inhibitory effects against common brook charr pathogens (Table 2).

Fig. 4
figure 4

Relative abundance shifts in the skin mucus microbiota of stressed versus unstressed brook charr. Stressed fish were exposed to high density (80 fishes in 10 L) until the oxygen concentration decreased to 3 mg/L (5 min). After stress exposure, fish were transferred in a new oxygenated tank to slowly recover. Data from Boutin et al. (2013b)

Full size image
Table 2 Endogenous isolates from brook charr microbiota known to inhibit salmonid pathogens
Full size table

2.3.2 Symbiont-Based Therapeutic Tools Against Opportunistic Pathogens

An indigenous brook charr isolate from skin mucus (Rhodococcus sp. CPM5) decreased mortality due to pathogens Flavobacterium psychrophilum and F. columnare by 47% without disturbing the natural microbiota of skin mucus (Boutin et al. 2013a). Unexpectedly, it was not by recolonizing the skin mucus microbiota that CPM5 conferred its protective effect, but rather by colonizing the filtering mass of the recirculation system, where it may have had a positive impact on water quality, as Flavobacterium spp. were observed to be virtually absent from the circulating water in treated groups (Boutin et al. 2013a). In addition, CPM5-treated fish’s water was dominated by Sphingomonas spp. unlike control tanks. Interestingly, Sphingomonas is the dominant bacterial genus in brook charr skin mucus. CPM5 may have indirectly improved resistance to flavobacteriosis by acting as a prebiotic for Sphingomonas spp., which in turn may have excluded Flavobacterium spp. from the surrounding water. This “symbiotic action at a distance” raised interesting questions regarding the nature of host-microbiota symbiotic relationships. Furthermore, it indicated that microbial symbionts may be recruited into novel ecological functions when readministered independently (Watson and Pollack 2001).

Other bacterial brook charr symbionts showed great promise as inhibitors of another major salmonid pathogen, Aeromonas salmonicida subsp. salmonicida (A. s. s.). One of those, Pseudomonas fluorescens ML11A, was also recovered from skin mucus (Gauthier et al. 2017a) and exhibited a strong antagonistic effect across a wide range of A. s. s. from different geographical origins (Gauthier 2016). In addition, gut isolates belonging to the Aeromonas sobria (sensu stricto) species, TM12 and TM18, showed tremendous inhibitive properties against A. s. s. through inhibitory compound diffusion on agar (Gauthier et al. 2017b). Those two probionts were shown to increase plasma lysozyme activity, respectively, by a 1.5- to 2-fold factor in in vivo preliminary experiments without A. s. s. challenge, suggesting a positive impact on host innate immunity (Gauthier et al. 2016, unpublished data). However, whether the inhibitive property of those probiont strains translates into a protective effect in vivo remains to be investigated.

3 An Overview of High-Throughput Methods and Their Contribution to Microbiota Studies

High-resolution study of the microbiota has been made possible with the advent of omics-based methods (genomics, transcriptomics, proteomics, etc.) following the emergence of high-throughput DNA sequencing during the late 2000s. This technological revolution made it possible to obtain several millions of nucleotide reads from tens to hundreds of samples at once, for costs that are orders of magnitude cheaper than Sanger sequencing for an equivalent amount of data (Vincent et al. 2017a). Knowing that a major fraction of microbial diversity is unculturable, high-throughput DNA sequencing technologies proved useful to the analysis of complex microbial assemblages. Several high-throughput sequencing methods have emerged, each offering distinct elements of information on the host-microbiota complex. For this reason, an integrative research strategy should ideally use a combination of these methods.

3.1 Whole-Genome Sequencing (for Specific Microbes)

Whole-genome shotgun (WGS) sequencing aims to obtain the whole DNA sequence of a specific organism. For individual microbes isolated from the microbiota, this method requires that the organism be in pure culture in order to avoid contaminating the sequence with exogenous DNA. Consequently, the DNA of individual nonculturable microbes cannot yet be sequenced independently (see paragraph below on metagenomics).

Briefly, a pure DNA extract of the organism is sheared into fragments of a few hundred base pairs long and is then sequenced using high-throughput technology, e.g., Illumina MiSeq (Tagini and Greub 2017). Typically, this process yields several hundred thousand DNA sequences (i.e., reads) that have to be assembled de novo (i.e., without prior knowledge on source DNA) to obtain the complete sequence of the organism’s genome. In most genome sequencing projects, the assembly does not reach completion and yields tens to hundreds of larger chunks (contigs) that span from a few kilobases to a few megabases in length. In fact, since 2014, there are more genomes published as incomplete drafts than complete ones (Gauthier et al. 2018) since it does not interfere much with downstream analyses.

When a draft assembly is complete, gene sequences are then annotated by similarity searches against databases of known genes. Some examples of web services offering this analysis are the NCBI Prokaryotic Genome Annotation Pipeline (Tatusova et al. 2016) and RAST (Overbeek et al. 2014). When gene annotation is complete, several downstream analyses can be applied: metabolic pathway reconstruction (Kanehisa et al. 2016), taxonomic assignment using average nucleotide identity (Konstantinidis and Tiedje 2005), phylogenomic inference (Delsuc et al. 2005), and virulence assessment (Chen et al. 2016) as well as antibiotic resistance gene prediction (Jia et al. 2017).

WGS sequencing was used, for example, to characterize the taxonomy and functions of individual bacterial symbionts in brook charr and rainbow trout which had potential as probiotic treatments against salmonid diseases (Boutin et al. 2012; Schubiger et al. 2015; Gauthier et al. 2017a, b), as well as pathogens (Reith et al. 2008; Rochat et al. 2017). However, WGS does not provide much insight on the higher organizational levels of the microbiota. The two following approaches (metabarcoding and metagenomics), respectively, address the following questions: (1) which microbes are involved and (2) how they contribute to the functional repertoire of the microbiome. Note that WGS data plays an important role in the annotation of functional data in metagenomics, as explained in the subsections below.

3.2 Metabarcoding (Who Is There?)

Metabarcoding is the massively parallel sequencing of a universal genetic marker to infer the taxonomic census of a community. It is currently the most common method in current microbiome studies, regardless of studied organisms (Garrido-Cardenas and Manzano-Agugliaro 2017; Mahato et al. 2017; Osman et al. 2018). The 16S ribosomal RNA gene is the most commonly used biomarker, due to its ubiquitous presence in bacterial, archaeal, and eukaryotic (organellar) genomes (Ju and Zhang 2015). Most, if not all, of the aforementioned salmonid-microbiota studies in Sect. 2 used 16S rRNA gene metabarcoding to assess microbiota composition.

First, DNA from the microbiota is extracted from host samples (e.g., skin mucus, gills, gut section). Then, a gene that is universally present across the widest range of organisms (a biomarker) is amplified by PCR. After this process, DNA “tags” are added to the amplified DNA products to allow “per sample” identification. PCR products are then sequenced simultaneously on a high-throughput apparatus (e.g., Illumina MiSeq or Ion Torrent PGM). There is a myriad of sample collection and DNA extraction methods that were developed in the last decade (Pollock et al. 2018), each with their own variability and biases for specific microbial groups. In a 2018 study, various combinations of gut sample collection, DNA extraction, and high-throughput sequencing were compared (Panek et al. 2018). However, optimization efforts (including the aforementioned study) have so far mostly been made for human gut microbiota studies. No methodological analysis on salmonid sample processing (skin mucus, gills, gut) has yet been published.

After sequencing, several files are produced, each containing several tens of thousands of sequence reads from a specific sample. Several computer programs are available to process the raw reads, each contributing to one or more of the steps involved in the analysis (Table 3). Typically, reads will be clustered in operational taxonomic units (OTU) under the assumption that similar sequences belong to a single microbial lineage. OTUs are then quantified by counting the reads that were involved in their construction. Then, reads are compared against a sequence database of known organisms, e.g., Greengenes (DeSantis et al. 2006), RDP (Cole et al. 2014), or ARB-SILVA (Yilmaz et al. 2014), to determine the consensus taxonomy of each OTU. Finally, an “OTU table” is obtained, on which a plethora of quantitative methods can be used to identify differences in microbiota structure between treatments, conditions, or samples (Table 3).

Table 3 Major steps in a typical 16S amplicon sequence analysis pipeline, including key software involved
Full size table

In summary, (16S) metabarcoding ultimately attempts to correlate the microbiota taxonomic composition to treatments dispensed to a host or several health metrics (i.e., growth rate, size, blood cell count, plasma lysozyme activity). However, it does so without providing much information on the function of microbes that are present (Zepeda Mendoza et al. 2015). Functional roles may be either (1) inferred grossly by reviewing what is known about specific families or genera in the literature or (2) inferred systematically by using “metagenome prediction” software to link taxa to whole-genome data (Langille et al. 2013; Aßhauer et al. 2015). However, most metabarcoding methods resolve taxonomy up to the genus level, and the functional repertoire of species within a single genus can be tremendously variable. For example, the genus Pseudomonas encompasses a wide array of mutualistic, commensal, and pathogenic species living across different habitats (soil, water, or in association with animals or plants) (Silby et al. 2011). Unsurprisingly, about only 1% of all known Pseudomonas genes is shared among all of its known genomes (Freschi et al. 2018). Therefore, using metagenome prediction from metabarcoding may overgeneralize the attributes of a single genus.

3.3 Metagenomics (What Are They Doing?)

Metagenomics, in principle, is not that different from WGS sequencing from a methodological point of view. However, input samples will be DNA extracts of complex communities (e.g., gut tissue sections or skin mucus swabs) instead of a single organism’s DNA, leading to the untargeted sequencing of all microorganism’s genomic sequences present in a given sample, in addition to the host organism’s genome (Quince et al. 2017). Due to the presence of DNA molecules from tens to thousands of different taxa in one sample, several hundred thousand fragments (contigs) per sample are typically obtained. Those contigs often correspond to individual gene sequences. Those are then annotated by homology search against known sequence databases, after which post-annotation analysis and visualization methods can be applied (Table 4). This methodological approach is therefore promising to improve the management beneficial microbial functions in aquaculture, as evidenced in other research fields (Culligan et al. 2014). However, the high sequencing depth required to get enough coverage for microbial sequences, the inherent complexity of analyzing metagenomic data (e.g., the lack of functional annotation of most nonhuman microbial transcripts), as well as the required computational power and storage make this approach highly challenging.

Table 4 Major steps in a typical meta-(genomics, transcriptomics) analysis pipeline
Full size table

Another challenge is that metagenomics gives insight on which gene functions are relevant in a biological system but does not predict their level of activity. Nevertheless, by sequencing host and bacterial messenger RNAs instead of total genomic DNA, one can indeed obtain a metatranscriptome, which can be annotated using similar methods; some assembly computer programs even allow the prediction of splicing variants resulting from the transcription of genes with intronic sequences (Haas et al. 2013). Moreover, the transcripts’ abundance can be quantified (as it is proportional to the number of mapped reads) making it possible to determine the level of expression of a given gene (Bashiardes et al. 2016). Therefore, by allowing to quantify simultaneously both microbial and host tissue gene expression, this methodological approach is thus suitable to shed light on active host microbiota (including pathogens) functional interactions. Accordingly, metatranscriptomics, combined with metagenomic analysis, has shown that, in the human gut, a substantial fraction of microbial transcripts are differentially regulated relatively to their microbial genomic abundances (REF). Though promising, this technique has inherent challenges such as high per-sample cost, depletion of both eukaryotic and bacterial ribosomal RNA transcripts (~90–95% of a total RNA sample), and the lack of standard bioinformatics methods ensuring repeatability across studies (Martin et al. 2018). To our knowledge, no metatranscriptome data of a salmonid has yet been published.

4 Future Perspectives for Microbiota Modulation

4.1 Host-Microbiota Interactions in Light of the One Health Perspective

As previously mentioned in chapter “The Rise and Fall of Antibiotics in Aquaculture,” an integrated view of biological systems is required to secure aquaculture production and ensure its sustainability. This is exemplified by the One Health perspective, which states that human, animal, and environmental health are co-dependent variables (Lebov et al. 2017). Therefore, acknowledging the contribution of the microbiota in animal health falls within this framework. Roles of the microbiota in salmonid health have been made increasingly clear throughout the last decade (see Sect. 2.2). Two relevant examples that would benefit from a holistic understanding are (1) the emergence of antibiotic resistance in salmonid pathogens, which is worsened by bacteria-to-bacteria gene transfer and the depletion of competing bacteria sensitive to antibiotics (Trudel et al. 2016), and (2) the impact of diet on the microbiota composition and its impact on immune function (see Sect. 2.2).

4.2 Fine-Grained Modulation Using Dietary Supplements

4.2.1 Probiotics

Probiotics are defined as a “live microbial culture added to feed or environment to increase viability of the host” (Gram and Ringø 2005). This positive effect on host physiology may originate from several mechanisms of action: (1) mechanisms targeting pathogens such as nutritional competition, diffusion of antimicrobial compounds, or competitive exclusion from epithelial surfaces (Bermudez-Brito et al. 2012; Kamada et al. 2013) and (2) mechanisms targeting the host itself, such as modulation of host immune signaling pathways (Kamada et al. 2013).

Although the specific mechanisms by which some probiotic strains exert their beneficial effects require further investigation, probiotic administration showed promising results on growth performance and general health of salmonid fish (Gatesoupe 2010). Some probiotic candidates showed great promise as prophylactic tools against opportunistic diseases (Boutin et al. 2012; Schubiger et al. 2015; Gauthier et al. 2017a, b). However, few of them, if any, have reached commercialization or even official approval for use in salmonid farming.

We may hope that, in the near future, probiotic administration will be guided by more and more thorough microbiota monitoring studies. For example, if a specific bacterial species is associated with increased immune function, then one or more isolates from this species could be administered to vulnerable fish as a prophylactic treatment. Perhaps microbial community assemblages that are reflective of a good health status will be engineered and administered as “microbiota transplants.” The microbiota transplant strategy, for instance, is the most effective therapies (90% efficacy rate) against human nosocomial Clostridium difficile infections (Liubakka and Vaughn 2016).

However, the bioengineering of host-associated microbial communities is a complex task, as most of microbial diversity is unculturable using basic microbiological methods (Tanaka et al. 2014). Nevertheless, recent progress in human microbiota studies (Zihler Berner et al. 2013; Auchtung et al. 2015; Dostal et al. 2015) could pave the way toward similar approaches in salmonid health management. For instance, gut bioreactors, a special class of continuous-flow fermenters, proved to be an excellent method to cultivate complex human gut microbiota systems in vitro in highly controlled simulated settings (Macfarlane and Macfarlane 2007).

4.2.2 Prebiotics and Synbiotics

In addition to probiotics, prebiotics are dietary additives that are fermented by the gut microbiota into short-chain fatty acids (SCFAs), which are the main energy source for colonic epithelial cells. The SCFAs also modulate lipid synthesis (Marcil et al. 2002), stimulate the immune system and increase host resistance against pathogens (Maslowski and Mackay 2010). Synbiotics are combinations of probiotics and prebiotics (Cerezuela et al. 2011). Synbiotics aim to simultaneously seed and maintain probiotic strains as dominant species in the gut. However, despite recent progress, there is limited information available on different aspects of synbiotics effects on fish (Cerezuela et al. 2011; Torrecillas et al. 2018), and their effect on the microbiota composition of salmonids is currently unknown.

4.2.3 Phage Therapy

Because phage particles are very specific to their bacterial hosts, they do not target both pathogens and the normal flora. Furthermore, phage particles replicate at the site of infection; thus curative doses can be fairly small. Moreover, although bacteria can become resistant to phages, these viral organisms can mutate and therefore evolve to counter phage-resistant bacteria (Matsuzaki et al. 2005), which synthetic antimicrobial treatments cannot do. The most important advantage of phages is that they might kill planktonic pathogens living in the surrounding water in addition to pathogens proliferating in carrier fish. Possible drawbacks of phage therapy include the possible transduction of virulence factors between bacteria; in addition, the vertebrate host may mount an immune response against the phage itself (see chapter “Would Bacteriophages Be a New Old Complement to Antibiotics in Aquaculture?” for a detailed discussion on phage therapy in salmonid aquaculture). Candidate phages that infect A. s. s. were isolated recently and opened the way to a broad-range treatment against multiple strains of this major salmonid pathogen (Vincent et al. 2017b). Other salmonid pathogens for which phage treatments are under development include Flavobacterium psychrophilum (Castillo et al. 2012; Madsen et al. 2013) and F. columnare (Prasad et al. 2011).

4.3 Coarse-Grained Modulation: The Case of K-Selection

In light of the One Health perspective, one may ponder over the efficacy of microbial management methods. Indeed, the strong demand for fish protein has resulted in a strong pressure on fish farmers to provide for an unprecedented increase of the human population (Duarte et al. 2009). Production strategies that may be efficient from an economical point of view (e.g., maximizing output at a minimal cost) may result in suboptimal rearing conditions such as overcrowding, hypoxia, and handling stress (Heikkinen et al. 2006). Disinfection methods and the aforementioned fine-grained microbiota management methods may help but may not address the root cause of certain diseases (i.e., poor rearing conditions). Current microbial management strategies aim to reduce the microbial load (e.g., surface disinfection of eggs and UV irradiation of incoming water) by assuming that fewer microbes translate into fewer risks of infectious disease. By doing so, those methods actually increase the water’s carrying capacity for generalist opportunistic microbes (r-strategists) whose life strategy is to proliferate as rapidly as possible, sometimes at the expense of their host (i.e., opportunistic pathogens) (Vadstein et al. 2018).

As a matter of fact, cod larvae (Gadus morhua) reared in recirculating aquaculture systems (RAS) had 72% higher survival rates than larvae reared in flow-through aquaculture (FTS) systems (Attramadal et al. 2014). RAS systems are specifically designed to promote the establishment of a complex microbial ecosystem that feeds of dissolved organic compounds and waste produced by the fish. Moreover, those systems are self-sustained (water is recirculating), resulting in a mature, specialist microbial community (K-strategists) and, therefore, low carrying capacity for opportunistic r-strategists. K-selection of microbial communities holds great promise as a microbiota management tool to promote fish health [for more details on K-selection, see chapter “Controlling Factors for Community Assembly in Developing Cod Larvae (Gadus morhua)”]. To our knowledge, research on K-selection of microbial communities in aquaculture has not yet been performed on salmonid fish.

4.4 Toward Real-Time Microbiota Monitoring

Novel DNA sequencing technologies have raised the bar in terms of throughput and scalability. For example, Oxford Nanopore has launched the MinION system, which allows amplicon and metagenome sequencing on a USB stick-side apparatus plugged on a laptop computer (Krehenwinkel et al. 2018). However, its high incorrect basecalling rate of 3%, which is equal to the well-adopted difference threshold for clustering OTUs, limits its usage in metabarcoding studies (Kerkhof et al. 2017). Because single-nucleotide differences are often critical to resolve distinct genera, it is imperative to distinguish biological sequence variation from amplicon sequencing errors (Callahan et al. 2016). With further improvements in sequencing accuracy and novel analysis pipelines, near-instant visualization and analysis of complex microbiome data from a laptop computer will perhaps be possible in the near future.

5 Conclusion

Since 2010, salmonids are second to carps as the most important group of farmed fish, with a total output of over 2 million tonnes. This results in a strong pressure on fish farmers to keep up with high demand for this source of animal protein. Intensive farming practices are commonly used in order to compensate for the immune impairment that results from an over-elicited stress response. This can disrupt global interactions between the host and its microbial symbionts (i.e., microbiota) that play a key role in maintaining fish health in the long term. This chapter presented an overview of the current knowledge on the taxonomic composition (i.e., diversity and structure) of salmonid microbiota and the state of the art on microbial profiling and modulation, as well as current research gaps and perspectives. A special attention has been given to the microbiota of salmonids that are significantly important in aquaculture.

  • Atlantic salmon has the most extensively characterized microbiota of all salmonids to this present day, with 17 dedicated studies published between 2007 and 2018. Most of this research focused on the skin and gut microbiota, including assessments of its response to migration, nutrition, antibiotherapy, and captivity.

  • Rainbow trout is the second most produced salmonid fish in salmonid aquaculture and is also the second most studied in terms of host-microbiota interactions. A total of 14 studies investigated the influence of diet, growth promotion, and pathogen inhibition, as well as diet-immunity interactions and their impact on the rainbow trout microbiota. However, none of those studies addressed the impact of antibiotherapy on the microbiota composition and subsequent effects on fish health.

  • The brook charr microbiota remains largely mischaracterized, except for the skin mucus microbiota, for which response to intensive rearing conditions and symbiont-pathogen interactions (including interindividual variations) were investigated. To our knowledge, no published studies have yet discussed the impact of diet or therapeutic tools on both microbiota structure and brook charr physiology, though ongoing studies are currently underway.

High-resolution study of the microbiota has been made possible with the advent of high-throughput DNA sequencing during the late 2000s. Several methods have emerged, each offering distinct elements of information on the host-microbiota complex.

  • Whole-genome shotgun (WGS) sequencing aims to obtain the whole DNA sequence of a specific organism.

  • Metabarcoding is the massively parallel sequencing of a universal genetic marker to infer the taxonomic census of a community. It is currently the most common method in modern microbiome studies, regardless of which organisms are studied.

  • Metagenomics is the whole-genome shotgun sequencing of total DNA extracts from complex communities instead of a single organism’s DNA, leading to the untargeted sequencing of all microorganisms’ genomic sequences present in a given sample, in addition to the host organism’s genome.

These methods allowed the study of salmonid microbiota from various levels of organization, i.e., from an individual microbe’s genome to the global functional interactions between host and the hundreds of microbial symbionts. Several health-promoting and therapeutic applications have benefitted from those omics-based methods. Whole-genome sequencing allows a thorough characterization of some bacterial probiotic candidates, as well as bacteriophages, that show great promise as prophylactic tools against opportunistic diseases. Metabarcoding and metagenomics allow modelling the microbiota structure as a function of health and treatment parameters. The latter are particularly useful in monitoring difference in microbiota composition across conditions that tremendously differ between each other (e.g., flow-through aquaculture vs recirculating aquaculture systems).

However, there are significant knowledge gaps in how salmonid-microbiota systems are affected as a whole by antibiotherapy, diet, and rearing conditions. The cause-effect relationships between treatments (or conditions), differential microbiota composition, and host physiology often remain unresolved in most studies. In addition, there are great inherent difficulties associated with the analysis of such a complex multivariate system as the microbiota, coupled with the lack of methodological consensus between studies. Perhaps the implementation of “good practices” and new technological advances should help resolving the complexity of host-microbiota systems to the benefit of salmonid aquaculture. Finally, there is growing awareness that fish and their rearing environments are complex ecosystems and, accordingly, that the well-being of both fish and microbiota must be considered when developing therapeutic tools or intensive rearing protocols.