Flavobacterium psychrophilum, a member of the family Flavobacteriaceae within the phylum Bacteroidetes1, is currently one of the most devastating fish pathogens. Whereas most members of the family are free-living organisms, others are serious pathogens of humans and various animals1. F. psychrophilum is a gram-negative bacterium affecting various species of salmon and trout reared in freshwater2. The epithet psychrophilum refers to the water temperature at which outbreaks of the disease occur (3–15 °C). The optimum generation time of 2 h is achieved at 15 °C (ref. 3). Because some strains grow weakly up to 25 °C, F. psychropilum should be considered a psychrotrophic organism.

F. psychrophilum has occasionally been detected in samples of water, sediments and biofilms from rivers receiving outlet water from infected fish farms. In vitro studies have demonstrated that cells are able to survive for several months in stream water despite a rapid growth arrest4. The presence of F. psychrophilum within salmonid eggs and the resulting vertical transmission from broodfish to progeny suggest that this obligate fish pathogen has spread through the international trade of live fish and fish eggs5. Whereas in adult fish extensive necrotic lesions result in a clinical form called “cold-water disease,” young fish suffer severe mortality associated with hemorrhagic septicemia, a condition referred to as “rainbow trout fry syndrome.” Infections by F. psychrophilum are responsible for considerable economic losses in all major areas of salmonid aquaculture worldwide. Because a specific vaccine is currently unavailable, antibiotics represent the only recourse; however, because their use has been criticized6, satisfactory control methods are lacking. Moreover, the molecular pathogenesis of F. psychrophilum infections is not yet understood, and only scant data on virulence factors are available. To obtain insight into the disease mechanisms of F. psychrophilum, we determined and analyzed the complete genome sequence of the virulent strain JIP02/86 (ref. 7).


General genome features

The genome of F. psychrophilum JIP02/86 consists of a circular chromosome of 2,861,988 bp (Fig. 1 and Table 1) and the pCP1 cryptic plasmid8. The origin of replication has been identified (Supplementary Fig. 1 online). The relatively high number of rRNA and tRNA genes, six and 49 respectively, is in good agreement with the rather rapid growth of the bacterium9,10. The chromosome is predicted to contain 2,432 protein-coding genes. The rather small genome size of F. psychrophilum as compared with those of environmental members of the family is probably related to its restricted ecological niche.

Figure 1: Circular representation of the Flavobacterium psychrophilum genome.
figure 1

The outer scale is in kilobases. Circles 1 and 2 (from outside to inside), genes transcribed clockwise (in red) and counterclockwise (in blue). Circle 3, rRNA genes (in green) and tRNA genes (in purple). Circle 4, insertion sequence elements (in pink). Circle 5, GC skew (window size 5,000 bp).

Table 1 General features of the Flavobacterium psychrophilum JIP02/86 genome


Since early studies F. psychrophilum was shown to be actively proteolytic11, and proteases were suspected to be among its most important virulence factors12, leading to rapid and massive tissue destruction (Fig. 2). Accordingly, the F. psychrophilum genome encodes 13 putative secreted proteases probably involved in virulence and/or in the destruction of host tissues (Table 2). The two metalloproteases Fpp1 (ref. 13) and Fpp2 (ref. 14) hydrolyze a broad range of substrates, including basic elements of muscular tissues. Sequences of the first N-terminal amino acids of purified Fpp1 and Fpp2 (SSTGQLKTMRLAQS(C/W)NGQYA and TVYNIPV, respectively14) allowed the unambiguous identification of the corresponding genes. Pathogenic fungi secrete endopeptidases belonging to the fungalysin family of metalloproteases that cleave extracellular matrix proteins and are involved in the breakdown of proteinaceous structural barriers during host tissue colonization15. A similar mechanism may apply for F. psychrophilum, in that we identified two paralogous genes FP0280 and FP0281, tandemly arranged, predicted to encode metalloproteases of the fungalysin family.

Figure 2: Deep ulcerative lesion of the caudal peduncle provoked by F. psychrophilum in a rainbow trout (Oncorhynchus mykiss).
figure 2

Skin and muscle tissues have been destroyed, exposing the spinal cord. In the course of 'cold-water disease' of adult salmonid fish, such necrotic lesions typically occur around the adipose ('peduncle disease') or dorsal ('saddleback disease') fins, on the gills and on the flank. Scale is in centimeters.

Table 2 Putative secreted protease precursors of Flavobacterium psychrophilum JIP02/86 involved in the virulence and/or in the destruction of host tissues

Because many bacterial pathogens produce extracellular collagenolytic enzymes allowing the colonization of connective tissues of their vertebrate hosts, the secretion of extracellular collagenase was considered of importance to F. psychrophilum pathogenicity16. Surprisingly, the JIP02/86 genome contains a collagenase-encoding gene that is disrupted by an insertion sequence of the IS256 family. To further examine this unexpected finding, the integrity of the collagenase-encoding gene was assessed by polymerase chain reaction (PCR) and sequencing in 23 additional F. psychrophilum strains from different geographic areas and fish species. These experiments revealed that 10 of the 24 strains had an insertion sequence 256 (IS256)-disrupted gene (Supplementary Fig. 2 online). Strains containing the disrupted gene were all isolated from rainbow trout, indicating that the collagenase gene is not essential to virulence in this species and suggesting a clonal dissemination of strains with an IS256-disrupted collagenase-encoding gene in rainbow trout farms worldwide.

Bacterial hemolysins are cytolytic toxins that are important virulence factors. In another fish pathogen, Vibrio anguillarum, toxin VAH5 is able to lyse rainbow trout erythrocytes and a VAH5 mutant showed attenuated virulence17. In F. psychrophilum, FP0063 encodes a protein 53% similar to VAH5 that may have a similar role in pathogenicity and cooperate with secreted proteases for tissue destruction.

The thiol-activated cytolysin family of pore-forming toxins (TACYs) has been implicated in the pathogenicity of several gram-positive bacteria18. The cytolytic activity is conferred by the C-terminal region (domain 4) of the protein. TACYs are also able to trigger several signaling pathways in a variety of host cell types through domains 1–3 of the protein, as shown for listeriolysin O (LLO) and pneumolysin19. Signals triggered by LLO strongly influence the course of infection, and LLO lacking domain 4 is fully capable of inducing a cytokine response, acting as a pleiotropic pseudocytokine/chemokine. In F. psychrophilum, the FP0097 gene encodes a protein that is similar to domains 1–3 of TACYs. Hence, FP0097 may function as a bacterial modulin responsible for damage to host tissue by virtue of its cytokine-inducing ability.

Motility, adhesion and biofilm formation

Cells of F. psychrophilum move over surfaces by gliding motility, an active movement that does not involve pili or flagella. Indeed, no genes encoding flagella, pili or any motility organelles were found in the F. psychrophilum genome, but 13 of the 15 gld genes involved in the gliding motility of F. johnsoniae20 were identified. The missing gldE and gldO genes are the only two F. johnsoniae genes whose involvement in gliding motility was not unequivocally demonstrated, suggesting that GldE and GldO have an accessory function.

Adhesion mechanisms are greatly diversified among bacteria and are of particular importance in pathogenicity. Strong adhesive properties of F. psychrophilum cells to fish body surface, gills and eggs have been reported21. We identified 27 genes probably involved in bacterial adhesion (Supplementary Table 1 online). Fifteen of them are tandemly organized and encode proteins with leucine-rich repeats similar to BspA and LrrA of the periodontopathogenic bacteria Bacteroides forsythus and Treponema denticola, respectively. These immunogenic cell surface proteins bind strongly to extracellular matrix components and function in the attachment to human oral tissue22,23. Other proteins probably involved in adhesion are: (i) FP2413, which contains five fibronectin type III domains involved in cell surface binding and a CUB domain found in extracellular proteins of spermadhesin and in eukaryotic peptidase families; and (ii) FP1830, FP0016 and FP0595, which contain fibronectin type III domains.

Biofilm formation is of importance in several pathogenic bacteria, especially those living in water, conferring a selective advantage by increasing their ability to persist under adverse environmental conditions. Little is known regarding the biofilm formation capabilities of F. psychrophilum cells. A tlpB mutant was deficient in virulence and gliding motility but displayed enhanced ability to form biofilm, suggesting that gliding motility and biofilm formation are antagonistic properties24. Bacterial biofilms generally consist of cells entwined in a protective matrix of extracellular polysaccharides. The F. psychrophilum genome contains many determinants for the biosynthesis, export, modification and polymerization of exopolysaccharides, most of them located in a 70-kilobase (kb) region (1,425,425 bp to 1,496,190 bp). Four proteins similar to alginate O-acetyltransferases of Pseudomonas aeruginosa probably function in biofilm formation. In P. aeruginosa, acetylation of the extracellular polysaccharide alginate by alginate O-acetyltransferases is important for the architecture of biofilms and increases the ability of alginate to act as a defense barrier25.

Secretion systems

Secretion systems are of great significance for virulence by addressing toxins to the bacterial surface. ABC-type transport systems, the Sec-dependent transport system, the components of the main terminal branch of the general secretory pathway and the Sec-independent (TAT) transport system were all identified in the F. psychrophilum genome. F. psychrophilum is devoid of the type III and IV secretion systems usually used by gram-negative pathogens but encodes the PorT and PorR proteins. In the periodontopathogenic bacterium Porphyromonas gingivalis, the membrane-associated protein PorT is essential for the transport of major virulence factors, such as proteinases and adhesins, from the periplasm to the cell surface26, and PorR is involved in the biosynthesis of cell surface polysaccharides that probably act as anchors for these factors27. Hence, this mechanism may also apply to F. psychrophilum and other pathogenic members of the phylum.

Cell surface proteins

Cell surface proteins are important targets of the host immune system. We identified the host-recognized antigens P18 (ref. 28), P60 (ref. 29) and FspA (ref. 30), as well as two proteins similar to those that induce protective immunity against the poultry pathogen Ornithobacterium rhinotracheale31. We also identified 35 proteins containing a signal peptide and a conserved C-terminal region of 70 amino acids (Supplementary Table 2 online). Most of these proteins also belong to the aforementioned toxins and adhesins. Iterative searches of databases identified 170 proteins that contain this C-terminal conserved region exclusively in members of the phylum Bacteroidetes. Hence, members of this phylum contain the C-terminal motif [YPNPX21–23(N/D)X2GX18–27GXY] that probably addresses and/or anchors proteins to be exposed at the bacterial surface, some of which may be of importance for virulence.

Using purified fractionized cell extract, gel electrophoresis and mass spectrometry, we confirmed the location and validated the expression of several of these predicted outer membrane proteins (Supplementary Table 3 online). Among them, several are probably important for bacteria-host interactions.


Genome analysis allowed the first metabolic reconstruction of this bacterium (Fig. 3). An important aspect of the lifestyle of pathogens is the acquisition of nutrients from their host. F. psychrophilum is unable to use carbohydrates as sources of carbon and energy7. Indeed, sugar kinase and phosphotransferase systems usually used by bacteria for specific carbohydrate uptake are missing, but genes encoding the enzymes of the glycolytic pathway (except glucose kinase) and the pentose phosphate pathway were identified. Therefore, these two pathways are most likely involved in the generation of precursor metabolites, such as the nucleotide-sugar precursors for exopolysaccharides biosynthesis.

Figure 3: Overview of metabolism and transport in F. psychrophilum.
figure 3

Transporters are grouped by substrate specificity: ions (green), nutrients (red), drug/metabolite efflux (black), secretion systems (purple), polysaccharide export (blue) and energy production (yellow). Arrows indicate the direction of transport.

The breakdown of host proteins by the secreted proteases results in a mixture of amino acids and oligopeptides that probably constitute the main source of carbon, nitrogen and energy. Five importers probably involved in the uptake of host proteins degradation products were identified. The F. psychrophilum genome is rich in peptidases that probably degrade imported peptides to amino acids that are processed by extensive amino acid catabolic pathways. Their importance is highlighted by the presence of two genes, FP1110 and FP1111, that encode cyanophycinase and cyanophycin synthetase, respectively. Cyanophycin is a branched, non-ribosomally synthesized polypeptide originally described in cyanobacteria that occurs as insoluble inclusions in the cytoplasm and serves as a storage compound for carbon, nitrogen and energy32. One might predict that F. psychrophilum stores amino acids through this process during feast conditions that are mobilized for survival in nutrient-limited conditions.

Degradation of lipids to fatty acids is probably achieved by a putative phospholipase and by three enzymes of the esterase-lipase-thioesterase family. Beta-oxidation of fatty acids is performed by a long-chain fatty acid–CoA ligase, three fatty acid dehydrogenases, a crotonase and three thiolases.

Hence, most of the degradation products of host proteins and lipids by F. psychrophilum are available as tricarboxylic acid cycle precursors. Indeed, all enzymes of the tricarboxylic acid cycle are present in the F. psychrophilum genome, and at least two enzymes that interconvert C3 and C4 compounds were identified.

Fe acquisition, of utmost importance to bacterial pathogens, is probably performed by five TonB-dependent outer membrane Fe receptors. Moreover, FP0089 and FP0090 encode proteins highly similar to FeoA and FeoB that constitute a Fe2+ uptake system. In Campylobacter jejuni, feoB mutants were significantly affected in their ability to colonize the gastrointestinal tract during both commensal and infectious relationships. Similarly, P. gingivalis lost its virulence in a mouse model of infection when one of its two feoB paralogous genes was inactivated33. Therefore, this iron acquisition system may also function in the pathogenicity of F. psychrophilum.

In agreement with its strictly aerobic metabolism, the F. psychrophilum genome encodes an extensive aerobic respiratory chain. All described members of the family Flavobacteriaceae contain menaquinone 6 as their only or major respiratory quinone2, and the corresponding genes (menABCDEF and ubiE) have been identified. Among the 24 cytochrome oxidase family proteins identified, a probable cytochrome cbb3-type oxidase complex (cbb3) and the proteins necessary for its proper assembly were found. This complex with high O2 affinity is usually expressed in response to microaerobic conditions and allows human pathogens to colonize anoxic tissues34. Moreover, FP1152, located upstream of the cbb3 gene cluster, encodes a Fnr family transcriptional regulator probably involved in the direct regulation of cbb3 in response to O2 as reported in other bacteria35. In addition, a H-NOX family protein probably signaling shifts to low-O2 environments36 has been identified.

Psychrotrophic character and stress response

Because F. psychrophilum faces low temperature in its natural habitat, it has to successfully overcome the resulting physical constraints, including slower pace of biochemical reactions, increased viscosity and increased gas solubility37. The F. psychrophilum genome encodes proteins probably involved in the regulation of membrane fluidity, in the maintenance of protein synthesis and in the production of psychrophilic enzymes and proteins with antioxidant properties.

Modification of membrane lipid composition with temperature changes is a widely distributed process for regulating membrane fluidity in bacteria37. Incorporation of unsaturated fatty acids in lipid membranes is a consistent adaptation of many cold-experiencing prokaryotes38. In coherence with the description of several unsaturated fatty acids in the membrane of F. psychrophilum39, we identified FP0413 encoding a protein similar to the Δ6 acyl-lipid desaturase of Synechocystis sp.40 and two paralogous genes FP0051 and FP1155, encoding proteins related to Δ9 stearoyl-[acyl-carrier-protein] desaturases of higher plants41. In addition, a cluster of genes (crtIBZY) probably involved in the biosynthesis of carotenoids may also participate in the modulation of membrane fluidity in response to temperature shift as reported in other psychrotrophic bacteria42.

The importance of the control of RNA folding and degradation at low temperature is underlined by the presence of six ATP-dependent RNA helicases, three of them belonging to the DEAD/DEAH box RNA helicase family. In the Antarctic archaeon Methanococcoides burtonii, these enzymes are overexpressed at low temperature and are involved in the destabilization of secondary structures of nucleic acids43. Differential regulation of gene expression may function in the bacterial adaptation to temperature shifts. In F. psychrophilum, the cold shock nucleic acid–binding protein FP0276 and the chaperones FP0864, FP1509 and FP1985 might participate in these adaptive responses.

The most important selective pressure related to the exposure to low temperature is the exponential decrease in the rate of chemical reactions. Little is known about the psychrotrophic properties of F. psychophilum enzymes: only the metalloproteases Fpp1 and Fpp2 were biochemically characterized and reported to display a psychrophilic and thermolabile behavior13,14. The psychrotolerant Flavobacterium frigidimaris produces seven NAD(P)+-dependent dehydrogenases44; four of them were psychrophilic and thermolabile, whereas the other three were unexpectedly thermophilic and thermostable. Only the genes orthologous to the F. frigidimaris genes encoding psychrophilic and thermolabile dehydrogenases are present in the F. psychrophilum genome. Hence, some of the enzymes probably involved in the virulence and in essential metabolic pathways of F. psychrophilum seem to be optimally active at low temperature.

At low temperature, the solubility of gases, including the noxious and highly reactive molecule dioxygen, increases rapidly. The large repertoire of proteins having antioxidant capacities reported in Colwellia psychrerythraea and Desulfotalea psychrophila37 seemed essential for the survival in cold environments. Moreover, because one of the most efficient defense mechanisms against bacterial pathogens is the production of reactive oxygen species (ROS) by host macrophages45, the resistance of F. psychrophilum to ROS killing is necessary to establish infection46. Accordingly, a large panel of genes encoding enzymes probably counteracting the deleterious effects of ROS were identified (Supplementary Table 4 online): three superoxide dismutases, a bifunctional catalase-peroxidase, a thiol peroxidase and four proteins of the peroxy-redoxin family. In the strictly anaerobic bacterium Bacteroides fragilis, five genes (batA, batB, batC, batD, batE) are involved in the tolerance to O2 (ref. 47). Surprisingly, orthologous genes were identified in the genome of the strict aerobe F. psychrophilum, suggesting their involvement in the protection of bacterial cells against ROS. Among the OxyR regulons present in Escherichia coli, the suf operon is specially devoted to the synthesis of [Fe-S] clusters when Fe or S metabolism is disrupted by Fe starvation or oxidative stress48. Moreover, this system is necessary for the virulence of the plant pathogen Erwinia chrysanthemi49. F. psychrophilum probably uses similar strategies to counteract the deleterious effects of oxidative stress and Fe limitation, in that proteins highly similar to SufABCDSE of E. coli have been identified. Therefore, sophisticated stress response mechanisms and the associated regulatory repertoire reflect the ability of F. psychrophilum to cope with various stressful conditions, in particular the oxidative stress resulting from the synergy of ROS production by hosts and their enhanced diffusion in cold environment.


The complete genome sequence of a fish pathogen has not been published previously. Sequence analysis of F. psychrophilum has revealed a combination of strategies that probably confer upon it the ability to colonize and degrade fish tissues and to exploit proteinaceous compounds for growth. Versatile respiratory and stress response systems may allow F. psychrophilum to deal with microaerobic conditions and oxidative stress. Moreover, the ability to form biofilms and to store cyanophicin could explain the long survival of F. psychrophilum outside the host. These features, together with its strong adhesive properties, probably contribute to its dissemination in salmonid farms worldwide. The availability of different genetic tools8 and transposon mutant libraries24 will allow large-scale functional analysis to assess the precise role of the predicted proteins identified. Thus, the F. psychrophilum genome sequence should facilitate the understanding of flavobacterial virulence mechanisms in fish and provide a basis for the development of better control strategies. Ongoing genome sequencing of other flavobacteria will provide new insights into pathogenicity and niche adaptation through comparative genomics.


Genome sequencing strategy.

The genomic DNA of F. psychrophilum JIP02/86 was sequenced by Qiagen GmbH using a conventional whole-shotgun strategy. Briefly, mechanically sheared 1.5-kb, 3-kb and 5-kb DNA fragments were isolated, inserted into pUC19 and cloned. Double-ended plasmid sequencing reactions were carried out using PE BigDye Terminator chemistry, and sequencing ladders were resolved on a PE 3700 automated DNA sequencer. The 33,900 useful readings were assembled into 56 contigs, providing a 9.3 X genome coverage. Gaps were closed by primer walking on gap-spanning clones, and remaining physical gaps were closed by multiplex PCR. Assessment of final assembly quality was done by long-range PCR and by ensuring that the physical map deduced from the genome sequence was identical to that obtained experimentally by pulsed-field gel electrophoresis (data not shown).

ORF prediction and annotation.

The prediction of coding sequences was generated using the self-training gene detection software SHOW based on Hidden Markov Models ( The ribosome-binding sites and transcriptional terminators were detected using the SHOW and Petrin softwares, respectively, whereas tRNA and rRNA were detected using the tRNA-scan and rRNA-scan softwares, respectively. Genome annotation was performed using the AGMIAL annotation platform50. Because of a lack of a reference set of annotation for genomes in the family Flavobacteriaceae, web-based softwares and databases were used to manually curate all predicted genes. ISs were classified using the IS finder database (, ABC transporters using the ABCISSE database ( and proteases using the MEROPS database ( Genome comparisons were performed using the web interface MaGe, which allows graphic visualization enhanced by a synchronized representation of synteny groups ( Orthologous genes between two genomes were defined as gene couples satisfying the bidirectional best-hit criterion or a BLASTP alignment threshold—that is, a minimum of 35% sequence identity on 80% of the length of the smallest protein. These orthologous genes were subsequently used to search for conserved gene clusters, for example, synteny groups among several bacterial genomes. All possible kinds of chromosomal rearrangements were allowed (inversion, insertion/deletion). A gap parameter, representing the maximum number of consecutive genes that were not involved in a synteny group, was set to five genes.

Nucleotide sequence accession number.

The genomic sequence reported in this article has been deposited in the EMBL database under the accession number AM398681. The annotated genome and the relevant AGMIAL database (genome navigator, Blast server and metabolic comparisons) are available at (

Cell surface protein identification.

Isolation of the outer membrane fraction, sodium dodecyl sulfate–polyacrylamide gel electrophoresis, protein elution and protein identification by liquid chromatography–tandem mass spectrometry were performed following published procedures28 detailed in Supplementary Table 3.

Note: Supplementary information is available on the Nature Biotechnology website.