healthmedicinet

Differential susceptibility of Atlantic salmon families to P. salmonis infection

In this study, groups of approximately 20 fish from forty full-sibling Atlantic salmon
families were used in controlled experimental challenges. Tagged fish were infected
by intra-peritoneal injection (IP) with P. salmonis (1?×?10
⁴
PFU/ml) and randomly distributed among ten tanks in order to reduce the possibility
of bias during fish culturing and handling. Even though an IP challenge is not a natural
form of infection, it is an effective method for presenting a naïve animal with a
known and controlled amount of bacteria 11], 12]. Cumulative fish mortality was used as a measure of susceptibility to P. salmonis as survival reflects the cumulative effects of all host-pathogen interactions during
infection and is therefore the best criteria to determine the level of susceptibility
18]. The cumulative mortalities of fish families injected with P. salmonis ranged from 0 to 64.3 % at 40 dpi (Fig. 1), with most families presenting cumulative mortalities between 5 and 16 % (?=?28). In families with some degree of mortality, fish showed several clinical signs
characteristic of infection: lethargy, anorexia, darkening of the skin, respiratory
distress, and/or surface swimming 5]. To confirm the presence of P. salmonis in the challenged fish, the 16S rRNA gene of P. salmonis was quantified using real-time PCR (qPCR) in at least three fish from each family.
Results indicated that the bacterium was present in the head kidney and spleen of
all challenged fish, whereas unchallenged control fish presented negative results
(data not shown). Based on cumulative mortality data, two groups of families were
defined (Fig. 1). The low susceptibility (LS) group was comprised of families with cumulative mortalities
of 0 %, and the high susceptibility (HS) group included families with a cumulative
mortality greater than 30 %. In both groups, mortality caused by P. salmonis began 15 to 16 dpi (data not shown), but variances in cumulative mortality were evidenced
by the end of the first challenge. The results of the first challenge indicated a
differential distribution of cumulative mortality among infected families, suggesting
that the susceptibility of fish families to P. salmonis infection is a result of differences in the genetic backgrounds.

Fig. 1. Cumulative mortality following P. salmonis infection. Forty full-sibling Atlantic salmon families were challenged for 40 days
with an intra-peritoneal dose of P. salmonis (isolate PS889). Families were enumerated according to their cumulative percentages
of mortality. The families with the highest mortality levels (31.3 – 64.3 %, grey
column) were named families of high susceptibility (HS), while the families with the
lowest mortality levels (0 %, green column) were named families of low susceptibility
(LS)

The influence of genetic factors on fish susceptibility to P. salmonis13], 14] and other bacterial infections has been previously reported in other fish species
15], 17], 19], 20]. For example, Camp et al. 20] challenged fifteen full-sibling families of juvenile channel catfish (Ictalurus punctatus) with the bacterium Edwardsiella ictaluri, the causative agent of enteric septicemia of catfish. Cumulative mortalities among
families ranged from 5 to 35 %. Using the most resistant and sensitive families, they
also detected differences between families in macrophage aggregations and in the amount
of lymphocytes (B and T) in peripheral blood. In a different work, the expression
of transcripts encoding for complement component 3 (C3) and lysozyme C II were induced
to a greater extent in a resistant rainbow trout (Oncorhynchus mykiss) line compared to a susceptible line in response to Flavobacterium psychrophilum infection 15]. Moreover, changes in the expression of transcripts encoding several toll-like receptors
and innate immune components were observed in genetic groups of catfish (Ictalurus punctatus) and Japanese flounder (Paralichthys olivaceus) with different susceptibilities to Edwardsiella spp infection 16], 17]. Thus, the results of these and other studies indicate an association between infection
susceptibility levels in fish and changes in the relative abundances of transcripts
involved in immune and stress responses, among other processes.

Gene expression analysis of LS and HS families

For gene expression analysis, a second group of naïve fish belonging to the six most
resistant and six most susceptible families (LS and HS, respectively) was IP injected
with P. salmonis. Fish from each family were divided into control (non-infected fish, ?=?10) and treated (infected fish, ?=?10) groups, and head kidney samples from each fish were collected at 14 dpi, before
the onset of mortality, and analysed using microarrays. No differences in fish weight
between LS (487.5?±?104.5 g) and HS (485.5?±?122.6 g) or between control and infected
groups were observed. Head kidney samples were also examined using qPCR assays to
confirm the presence of P. salmonis in all IP-injected fish and its absence in control fish.

Three LS and three HS families were selected for microarray assays. For each assay,
head kidney RNA from groups of five non-infected and five infected individuals were
pooled, reverse transcribed, and hybridized to microarrays. In order to produce a
general description of the transcriptional response to P. salmonis in fish families with different susceptibilities to the infection, each of the three
LS families and each of the three HS families were regarded as biological replicates
of the high and low susceptibility groups, respectively (experimental design in Additional
file 1).

Results indicated that 2491 and 2602 probes were differentially expressed (false discovery
rate (FDR) adjusted, p??0.05) between infected and non-infected fish of the LS and HS groups, respectively
(Fig. 2a). Fold change (FC) values of differentially expressed probes were well correlated
(Pearson correlation?=?0.92) between the two groups of families (Fig. 2b), indicating that the response to P. salmonis infection was similar in terms of these indicators. A complete list of probes differentially
expressed following bacterial infection is shown in Additional file 2. Some of these probes contained different regions of the same genes or paralogs.
A total of 1430 probe sets (735 up-regulated and 695 down-regulated) were differentially
expressed between infected and non-infected fish of the LS group, of which 1288 probe
sets corresponded to genes with predicted functions. For the HS group, statistically
significant gene expression differences between infected and non-infected fish were
observed for a total of 1300 probe sets (625 up-regulated and 675 down-regulated),
of which 1154 corresponded to genes with predicted functions.

Fig. 2. Global transcriptome response of fish head kidney to P. salmonis infection. a Venn diagram of differentially expressed probes between infected and control fish
from LS and HS families. b Representation of Fold Change (log
₂
) of all common probes significantly modulated by P. salmonis infection in LS (X axis) and HS (Y axis) families. Parameters of the linear regression are indicated

To validate results from the microarrays, the relative transcript abundance of a random
set of 40 genes was examined by qPCR. A microarray result was considered validated
when the RNA expression profile of a gene is statistically differential in response
to infection and followed the same trend when tested by both microarray and qPCR in
the six families. Among the 40 genes that were subjected to validation, 33 (83 %)
displayed the same trend observed in the microarray analysis of the six families (Additional
file 3). Overall, a strong positive correlation of 0.82 (Pearson correlation) was determined
between microarray and qPCR analyses for the combined data set (p??0.00001) (Fig. 3). Thus, although qPCR showed a broader dynamic range than microarrays, these two
platforms correlated well with each other.

Fig. 3. qPCR validation of microarray results. Mean log
₂
ratios (infected/control) of gene expression (N?=?33, Additional file 3) calculated from microarrays were plotted against the mean log
₂
ratios derived from qPCR assays. Each circle represents the mean of five technical
replicate (N?=?990 assays). Correlation between microarrays and qPCR was calculated by Pearson
product moment correlation and a p??0.01 was considered statistically significant

Common transcriptional response of LS and HS families to P. salmonis infection

We determined the number of common (shared) probe sets that were significantly up-
(?=?54) or down-regulated (?=?155) in the two groups of families following P. salmonis infection; a list of representative shared genes is shown in Table 1 (see the complete list of probes in Additional file 4). Of the genes that increased their expression in response to infection, we detected
functional categories that were associated with the antibacterial response, such as
the immune response, energy metabolism, and cytoskeleton rearrangement, among others
21]. In particular, among transcripts encoding proteins with predicted roles in the innate
immune response we found lysozyme C II (lyz), which has a hydrolytic activity against Gram-positive and Gram-negative bacteria
in tissues and body fluids 22], 23]. Its increased transcript abundance and enzyme activity has been widely described
in fish infected with different pathogens 16], 24], 25], indicating a conserved and relevant function in antimicrobial defence. Additionally,
among the transcripts with functions involved in the adaptive immune response, one
was found that encoded for a component of the major histocompatibility complex (MHC)
class I (hla-UBA). Moreover, the expression of some components linked to the organization and regulation
of the actin cytoskeleton, such as cytoplasmic actin, thymosin, tropomyosin, and myosin
light chains, were also up-regulated. In this regard, dynamic rearrangements and the
organization of the actin cytoskeleton are critical for lymphocyte migration, as well
as for the formation and stabilization of the immunological synapse at the interface
between antigen-presenting cells and T cells 26].

Table 1. Representative probe sets differentially expressed between infected and non-infected
fish

On the other hand, genes that had a significantly lower expression in response to
infection were associated with different cellular processes and provided some insights
on how this pathogen modulates the host response. Thus, transcripts that showed decreased
relative abundance following bacterial infection were mainly involved in the processes
of protein synthesis (ribosomal proteins), transport of oxygen and selenium, and homeostasis
of metals. Twenty-four probe sets encoding 40S ribosomal subunit proteins and 38 encoding
60S ribosomal subunit proteins were down-regulated in response to infection, suggesting
that the transcriptional repression of translation machinery might be an antibacterial
response or part of a general reduction in host metabolic activity. It has been suggested
that the shut-down of translation machinery is a bacterial and viral strategy to control
the translation of pathogenic proteins 27]–30] and to suppress innate host defences by inhibiting the capacity of infected cells
to synthesize immune system proteins 31], 32]. This strategy may be used by P. salmonis to control the host response in order to survive and replicate inside infected cells
8]. Since large numbers of probe sets represent the same protein or processes, these
results suggest a coordinated gene expression response to P. salmonis infection.

The relative abundance of transcripts encoding for hemoglobin subunits and selenoprotein
P was also significantly decreased in response to infection, suggesting that P. salmonis might impact the plasma transport of oxygen/iron and selenium. The down-regulation
of hemoglobin subunits may be part of a host defence mechanism to limit the availability
of hemic-iron, an important source of iron for intracellular bacteria 33]–35]. Moreover, decreased relative abundance of transcripts encoding for proteins involved
in intracellular non-hemic iron binding (ferritin middle and heavy subunits) and in
hemic binding (hemopexin) suggest that P. salmonis infection induces changes in iron metabolism in Atlantic salmon. These changes may
affect the expression of genes directly involved in the synthesis of hemoglobin, as
has been reported for other pathogens 36]. Finally, Selenoprotein P, an extracellular protein that transports most plasma selenium
37], was consistently down-regulated following infection. Selenoprotein P has been associated
with oxidative 38] and immune defence 39] mechanisms, and it has been proposed as a viable candidate molecular marker for responses
to P. salmonis11] and anaemia virus (ISAv) 40] infections. Thus, the down-regulation of Selenoprotein P transcripts seems to be
a conserved response to different types of pathogens infecting Atlantic salmon.

Since knowing the processes in which differentially expressed genes are involved helps
to understand the host-pathogen interaction, a functional analysis was used to identify
biological processes (Gene Ontology and Reactome) and metabolic pathways (KEGG) in
all probe sets, including those that were mutually or distinctly up- or down-regulated
in the family groups (Table 2). Functional annotation of genes with increased expression highlighted the central
place of immune processes, such as toll-like receptor signalling, bacterial infection
(Salmonella), and phagosome/lysosome pathways, which are crucial for innate immune responses involved
in the recognition, phagocytosis, and degradation of pathogens 41], 42]. Genes with decreased expression in both LS and HS families were mainly annotated
to functional categories associated with protein complex assembly and translation.
This result was consistent with that obtained with the functional analysis of shared
down-regulated genes, and it strengthens the idea that P. salmonis hijacks the translation machinery of the host cell. In addition to this, hemopoiesis
was another common functional term among down-regulated probes, and this might reflect
the fish response to anaemia induced by P. salmonis infection (Table 2).

Table 2. Functional annotation of common probes significantly up- or down-regulated in LS and
HS groups of families

Gene expression differences in LS and HS families following the P. salmonis infection

We hierarchically clustered the complete list of probes differentially expressed between
infected and non-infected fish from the LS and HS families (Additional file 2 and Additional file 4), and a correlation analysis was applied to measure the degree of association among
the gene expression patterns of the six families. A Pearson correlation (Additional
file 5: Figure S1A) and Euclidian distance (Additional file 5: Figure S1B) were used as metrics, and an average linkage clustering described the
data. Both clustering analyses clearly separated the three LS families from the three
HS families in two distinct branches and showed a close proximity among families with
similar levels of susceptibility to infection. These results suggest that salmon families
with different levels of susceptibility to the infection differentially modulate transcript
abundance in response to the pathogen.

To further examine the varied responses to P. salmonis infection between LS and HS families, and to identify potential mechanisms of natural
resistance, we analysed the biological processes and metabolic pathways in two groups
of probe sets differentially expressed between infected and non-infected fish (Table 3). The first group included 1138 probe sets that were significantly modulated (up-
or down-regulated) in the LS but not the HS group. The second group included 127 probe
sets that were up-regulated in LS families but down-regulated in HS families. Analysis
of these underlined the central place occupied by both innate and adaptive immune
systems, as represented by transcripts encoding for complement proteins (C3, C4, factor
B), Myeloperoxidase (MPO), CXC and CC chemokines and receptors, interleukin 18b, and
immunoglobulins (Additional file 6). Interestingly, whereas the alpha polypeptide of MHC class I was up-regulated in
both LS and HS families, ?2-microglobulin, another polypeptide of MHC class I, was
consistently down-regulated only in HS families.

Table 3. Functional annotation of genes differentially expressed between LS and HS in response
to P. salmonis infection

It is worth mentioning that the characterization of host transcriptional changes at
a late stage of P. salmonis infection (14 dpi) included genes directly involved in fighting infection, as well
as genes involved in general physiological processes. Among these were genes with
predicted functions in histone modification, protein folding, and carbohydrate and
fatty acid metabolism (Table 3). This observation suggests that potential infection biomarkers could be involved
in more general cellular processes and may not be limited to genes directly involved
in the immune response.

Differential activation of the iron-depletion system in LS and HS families after P. salmonis infection

The functional classification of genes differentially expressed between infected and
non-infected fish revealed that part of the core response to P. salmonis infection included the down-regulation of several probes representing transcripts
that encode for heme-proteins (hemoglobin and cytochrome), heme-binding proteins (hemopexin),
and non-hemic iron binding proteins (ferritin, middle and heavy subunits). Moreover,
transferrin, the iron-binding glycoprotein that transports iron in the plasma 43], was up-regulated in LS but down-regulated in HS families. However, hepcidin, the
principal regulator of iron efflux in vertebrates, which controls access of iron into
circulation 44], 45], was significantly up-regulated in HS families but not in LS families (Additional
file 2). These results suggest that the regulation of iron homeostasis could be crucial
for the natural resistance of Atlantic salmon to P. salmonis infection.

Based on the central role that iron plays in both pathogen virulence and host anti-microbial
resistance 46], we further examined whether differences in iron content and metabolism could be
detected between LS and HS families in response to P. salmonis infection. To do this, the total iron content in non-infected and infected fish head
kidneys was measured. Head kidney iron content ranged between 1.5 and 2.0 ?g Fe/mg
dry weights in non-infected fish from all families regardless of susceptibility (Fig. 4). However, at 14 dpi with P. salmonis, a significantly lower level of iron content was detected in infected fish from all
LS families and in only one of the HS families (HS3) (Fig. 4a). Since there is evidence for a metabolic interaction between trace metals such as
Fe and Zn 47], 48], Zn content was also assessed in infected fish families. The Zn content was similar
in control fish of the LS and HS families, and levels were unaffected by infection
(Fig. 4b). In view of the importance that iron availability has for bacterial proliferation
34], 49], 50], bacterial load was measured in infected tissues (Table 4). The results showed that the bacterial load was significantly lower in LS compared
to HS families, indicating a positive correlation between the bacterial load and the
iron content in kidneys after infection. This is a relevant aspect because it has
been demonstrated that iron depletion limits intracellular bacterial growth in murine
macrophage models 50].

Fig. 4. Cellular iron and zinc content in infected and non-infected fish from LS and HS families.
a Cellular content of iron (?g) in dry weight (DW) head kidneys (mg) from non-infected
(white bars) and infected tissues (black bars). b Cellular content of zinc (?g) in dry weight (DW) head kidneys (mg) from non-infected
(white bars) and infected tissues (black bars). In all case, bars represent the mean of five biological replicate determinations
(± SEM); *, p??0.05 (Student’s?t test)

Table 4. Bacterial load in infected fish

As iron content between LS and HS families was similar without infection, the correlation
between decreased metal content, reduced bacterial load, and pathogen resistance might
be explained by an ability of LS families to reduce iron content in the head kidney
in response to infection, and not by the iron status preceding infection. This suggests
that decreased cellular iron content is a physiological response to infection.

To begin to understand the molecular mechanisms by which fish are able to reduce iron
content in response to the infection, we first measured the relative abundance of
transcripts with potential functions in iron transport and metabolism in both infected
and non-infected fish kidneys from every LS and HS family. The data revealed that
following infection, the LS3 family decreased the abundance of transcripts encoding
iron uptake transporters (dmt1, divalent metal transporter 1; and trfr, transferrin receptor), whereas families LS1 and LS2 increased the abundance of transcripts
encoding for the efflux transporter (ireg1, ferroportin) (Fig. 5). This might account for the reduced cellular iron measured after infection, as has
been reported for Salmonella infection in murine macrophages 51].

Fig. 5. Expression analysis of iron metabolism genes in head kidneys from infected and non-infected
fish. Relative changes in the expression of genes were determined using qPCR in non-infected
(white bars) and infected fish (black bars) from LS and HS families. For each gene
the relative abundance of mRNA was normalized towards the elongation factor 1 alpha
(EF1A) mRNA. Bars represent the mean of five replicate determinations (± SD); *, p??0.05 (Student’s?t test)

Although HS families were also able to modify the abundance of transcripts encoding
iron transporters after infection, in some cases, modulation was opposite to that
observed in LS families. For example, the HS1 family (that most susceptible to infection,
with 64.3 % cumulative mortality) showed a significant increase in the abundance of
dmt1 transcripts and a decrease in the abundance of transcripts encoding for Ireg1. The
HS3 family, the only HS family that showed a significant decrease in iron content,
displayed increased Ireg1 while no changes were detected for uptake transporters.
On the other hand, the HS2 family had a decreased abundance of uptake transporters
without alterations in the abundance of Ireg1. Since the HS2 family was not able to
reduce cellular iron content in response to infection, this suggests that Ireg1activity
may be pivotal to reduce iron availability in infected fish. Interestingly, the transcript
encoding for the peptide hepcidin, which causes ferroportin internalization and degradation
44], was up-regulated only in those families that did not reduce iron abundance following
infection, thus supporting the idea that the regulation of iron efflux is fundamental
for reducing the intracellular content of this metal.

Finally, the effect of P. salmonis infection on the abundance of transferrin (trf) and ferritin light chain (ferl) transcripts was evaluated. These encode for ubiquitous proteins that bind and store
extracellular and intracellular iron, respectively 52]. Results showed a significant increase in the abundance of trf and ferl in all LS families and a reduction of trf in two HS families, suggesting that infected fish were able to trigger strategies
to limit the access of P. salmonis to cellular iron with different efficiencies between LS and HS families. In view
of this, our results support that the iron-deprivation mechanism of nutritional immunity
53] could be an important defence mechanism against P. salmonis infection.

Genome sequence of P. salmonis reveals the presence of iron-acquisition genes

To cope with iron-deprivation mechanisms, pathogens have evolved mechanisms for iron
acquisition that are tightly controlled by the availability of iron in the environment
46]. Herein, we sought to identify P. salmonis iron acquisition genes that could act as potential virulence factors. To do this,
we sequenced and annotated the genome of P. salmonis and identified a set of orthologous genes with reported roles in the synthesis and
uptake of siderophores and heme, ferric iron active transport (energy system), ferrous
iron acquisition, and transcriptional regulation (Fur, ferric uptake repressor). Some
features of these genes are shown in Table 5.

Table 5. Iron acquisition systems of P. salmonis

From the genome sequence, three orthologous genes were predicted encoding for synthetases
of vibrioferrin, an unusual marine carboxylate siderophore (pvsA, pvsB, pvsD); its membrane-spanning exporter (pvsC); and its TonB-dependent siderophore receptor (PvuA) 54]. All classic components of the ExbB???ExbD???TonB energy system were found, supporting
the TonB-dependent active transport of iron siderophores across the outer bacterial
membrane 55]. Furthermore, components of both fhu and feo operons, required for the acquisition of hydroxamate siderophores (or heme) 56] and ferrous iron 57] respectively, were predicted from the P. salmonis genome. Finally, an orthologue of Fur was also identified. Fur controls the intracellular
concentration of iron in bacteria, thus in the presence of intracellular iron, Fur
binds DNA and represses the transcription of genes involved in siderophore biosynthesis
and iron acquisition 58].

A schematic representation of P. salmonis iron acquisition genes in a genetic context (Fig. 6a) shows that except for fur and fhuA/hemeR, genes were distributed into three separate iron gene clusters. The presence of these
components suggests that P. salmonis can acquire iron through multiples mechanisms, including those for the transport
of ferric and ferrous iron, heme iron, and both endogenous and exogenous siderophores.
This allows us to presume that iron is a crucial element for the survival and virulence
of P. salmonis. The capacity of pathogenic bacteria to acquire iron in an animal host is important
for establishing infection 59]. Since animal hosts have essentially no free iron but do have different heme sources,
it is probable that mechanisms for heme capture are relevant during infection. Moreover,
orthologous genes that encode for hemolysin and its related secretion components (hlyb1 and hlyb2; hlyd, tolc1 and tolc2) were also identified 60]. Hemolysins are cytolytic toxins to erythrocytes and other cell types that are produced
by some heme-acquiring bacteria and are considered to be virulence factors 61]. The presence of these genes in the P. salmonis genome is consistent with the haemorrhagic and anaemic response previously described
in fish during P. salmonis infection 5], 10].

Fig. 6. Putative siderophore biosynthesis and iron transport gene clusters of P. salmonis and Fur binding-site prediction. a Siderophore biosynthetic genes are indicated by black arrows. The congate siderophore/heme
outer membrane receptor/exporter genes are depicted in white and the component of
TonB-dependent active transport across the bacterial outer membrane system is in grey.
Green arrows denote components involved in the ferrous uptake system and blue arrows
indicate ferric hydroxamate/heme uptake genes. A single homolog of the ferric uptake
regulator (fur) was predicted in P. salmonis genome. The blue box represents Fur-binding sites. b Sequence logo for predicted Fur-binding sites in P. salmonis

To explore whether the predicted iron gene clusters were regulated by Fur, we sought
putative Fur-binding sites in the whole genome of P. salmonis. For this purpose, the Fur-binding site motif of ?-proteobacteria was reconstructed
from 656 manually curated intergenic sequences, and this information was used to search
for this motif in the entire P. salmonis genome. The sequence logo for the predicted Fur-binding sites in P. salmonis is shown in Fig. 6b.

The results indicated that two hundred open-reading frames (ORFs) had at least one
putative Fur-binding site in the intergenic region upstream of the respective start
codon (data not shown). More importantly, a putative Fur-binding site was found upstream
of the three iron gene clusters described above (Fig. 6a), suggesting that these gene clusters might be regulated by Fur and become activated
during iron deficiency.

To address this, the bacterium was cultured for twelve days in a free-blood standard
liquid media supplemented by 0.1 mM of ferric iron (Fe-NTA) (reference condition),
which permitted optimum bacteria growth 62]. Two other experimental conditions were also used; one without ferric iron (0 mM)
and another with 1.0 mM of ferric iron supplementation, representing deficit and excess
iron availability, respectively. Data showed that carrying capacity (K) was significantly
lower in both experimental conditions than in the reference condition, however no
significant differences in growth rates (exponential phase, at 5 d) were observed
under the different supplementation conditions (Fig. 7a). On the other hand, the intracellular bacterial contents of iron, measured by Atomic
Absorption Spectroscopy (AAS) during the exponential growth phase, showed that in
the absence of iron supplementation there was a significant decrease in the intracellular
concentration of iron when compared to the reference condition. However, no significant
differences in iron contents were detected among bacteria grown in iron supplemented
conditions (Fig. 7b). These results indicate that P. salmonis senses iron fluctuations in the environment and adaptively responds to both iron
deficiency or excess, and since both conditions are deleterious for the bacteria,
expression of acquisition and detoxification systems are finely regulated 63], 64].

Fig. 7. Effect of iron availability on P. salmonis growth and expression of iron-acquisition genes. a Growth curves of P. salmonis treated with three concentrations of iron supplementation (0 mM, 0.1 mM and 1 mM
of Fe-NTA) during twelve days. Each point represents the mean of nine determinations
(± SEM); *, p??0.05 (Student’s?t test). b Iron content of P. salmonis treated with three concentrations of iron supplementation (0 mM, 0.1 mM and 1 mM
of Fe-NTA) during five days. Columns represent the mean of five determinations (±
SEM); *, p??0.05 (Student’s?t test). Iron content was normalized towards mg of total proteins. c Relative expression of iron acquisition genes was determined using qPCR. These genes
were significantly modulated (Student’s?t test p??0.05) in response to iron deficit in comparison with the iron supplemented conditions.
The relative abundance of each mRNA was normalized towards the recombinase A (Rec
A) mRNA of P. salmonis. Bars and numbers represent the mean value of five determinations (± SEM). Different
colors represent different iron cluster genes

The expression of predicted iron acquisition genes under supplemented and non-supplemented
iron conditions was also measured. A significant increase in the relative abundance
of all predicted transcripts occurred in response to iron deficiency (0 mM Fe-NTA)
as compared to both supplemented conditions (Fig. 7c). This behaviour was supported through a putative Fur-binding site found present
upstream of the three iron gene clusters described.

Although fhuA/hemeR was not predicted as part of an iron gene cluster, and a Fur box was not identified
in its upstream region as has been described in other bacterial species 56], 65], the transcriptional behaviour of fhuA/hemeR indicates that it might be part of a still un-annotated polycistronic operon regulated
by Fur. However, future assays are necessary to confirm this hypothesis.

Interestingly, the abundance of transcripts that encoded a putative Fur transcription
factor was significantly reduced in the iron-restricted condition. The absence of
a predicted Fur-binding site, and the role of Furas a repressor of iron acquisition
genes, support its transcriptional behaviour, which has also been reported in other
bacterial species 66], 67]. Finally, from previously published information and the present results, we formulated
a model for the possible spatial organization of the predicted proteins involved in
iron acquisition in P. salmonis (Additional file 7). In light of the severity of P. salmonis infection and the risk of acquiring antibiotic-resistant bacterial strains, fully
describing iron acquisition systems and the roles of these in P. salmonis pathogenesis is a crucial step towards developing therapeutic agents.

To the best of our knowledge, this is the first study that compares the transcriptional
response to P. salmonis infection among Atlantic salmon families with different levels of susceptibility.
Nevertheless, this study has some limitations, for instance, fish were infected by
intra-peritoneal injection, a method that does not represent the natural form of P. salmonis infection. Moreover, for gene expression analysis, a single tissue (head kidney)
was sampled and only at one time-point, so we could have missed other differentially
expressed transcripts that appear in other tissues affected by the infection and before
and after 14 dpi. Such issues can be addressed by using different infection protocols,
multiple tissues and time-point measurements, and in future studies, by using RNA
Sequencing in order to identify potential splicing variants and polymorphisms among
salmon families, information that could be relevant to understand fish resistance
to infection. Another limitation of our study is that the complexity of P. salmonis genome only permitted us the assembly of a draft genome (198 scaffolds). Thus, in
order to identify the complete set of genes involved in iron metabolism and acquisition
and their genetic contexts, further studies are necessary to assembly the complete
P. salmonis genome.