Genetic and virulence characterization of Flavobacterium columnare from channel catfish (Ictalurus punctatus)


Mark L. Lawrence, College of Veterinary Medicine, Mississippi State University, PO Box 6100, Mississippi State, MS 39762-6100, USA. E-mail:


Aim:  To develop a method for conducting pulsed-field gel electrophoresis (PFGE) on Flavobacterium columnare, to use PFGE to characterize F. columnare channel catfish isolates, and to determine whether variation in pathogenic potential exists in F. columnare isolates from channel catfish.

Methods and Results:  On the basis of PFGE-derived profiles, similarity dendrograms constructed for more than 30 F. columnare isolates showed two major genetic groups with more than 60% similarity. Channel catfish fingerlings challenged with PFGE group A isolates by bath immersion had significantly higher average mortalities (>60%) than fish challenged with PFGE group B isolates (<9%). However, abrasion and skin mucus removal made channel catfish fingerlings susceptible to disease caused by group B isolates following immersion exposure.

Conclusion:  Our results suggest that two genetic divisions of F. columnare channel catfish isolates exist, and that isolates in PFGE group A isolates tend to be more pathogenic to immunocompetent channel catfish fingerlings than PFGE group B isolates.

Significance and Impact of the Study:  PFGE is a potentially useful tool for determining whether F. columnare isolates are more likely to be primary or secondary pathogens. Pathogenesis research for columnaris disease in catfish should focus on pathogenic isolates from PFGE group A.


Columnaris disease, caused by Flavobacterium columnare, has been described as a globally distributed acute to chronic bacterial infection of freshwater and brackish water fishes. Many commercially important species are affected by columnaris disease, including (but not limited to) channel catfish Ictalurus punctatus; common carp Cyprinus carpio; goldfish Carassius auratus; rainbow trout Oncorhyncus mykiss; Japanese eel Anguilla japonica and tilapia Oreochromis sp. Ictalurids are often severely affected (Plumb 1999; Shoemaker et al. 2003). Historically, columnaris disease has been considered the second most important bacterial infection in channel catfish. However, case reports from the Aquatic Diagnostic Laboratory of the Thad Cochran National Warmwater Aquaculture Center from 1997 to 2005 have shown F. columnare to be the most commonly diagnosed pathogen (Anonymous 2006).

Molecular analysis of the 16S rDNA gene or intergenic spacer regions is the most commonly used method for analysing the genetic diversity of F. columnare (Triyanto and Wakabayashi 1999; Arias et al. 2004; Darwish and Ismaiel 2005; Figueiredo et al. 2005; Schneck and Caslake 2006; Suomalainen et al. 2006a; Olivares-Fuster et al. 2007). Other methods used to analyse the intraspecific diversity of F. columnare have included rapid amplified polymorphic DNA-PCR (RAPD-PCR) (Thomas-Jinu and Goodwin 2004) and amplified fragment length polymorphism (Arias et al. 2004).

Because of its broad applicability, high discriminating power and tremendous epidemiological concordance, pulsed-field gel electrophoresis (PFGE) has been the gold standard of bacterial pathogen subtyping over the past 20 years (Barrett et al. 2006). In particular, PFGE has higher discriminatory power than methods based on PCR or restriction enzyme digestion with conventional electrophoresis (Gordillo et al. 1993; Olive and Bean 1999). However, only a few pathogenic bacteria of aquatic organisms (Pseudomonas anguilliseptica, Aeromonas salmonicida, Streptococcus iniae, Vibrio tapetis and Vibrio vulnificus) have been analysed by this technique (Castro et al. 1997; Weinstein et al. 1997; Umelo and Trust 1998; O’HIci et al. 2000; Blanco et al. 2002; Wong et al. 2004).

Despite the clinical and commercial significance of F. columnare infection, there is a paucity of information regarding the epidemiological relationships between the various bacterial isolates from the south-eastern United States, which is principally a warm water aquaculture region. The goals of this study were to develop a PFGE method to characterize F. columnare strains isolated from different locations in the south-eastern United States and to identify the possible correlation between F. columnare PFGE subgroups and virulence in channel catfish fingerlings.

Materials and methods

Bacterial strains

Our study included 30 clinical isolates of F. columnare obtained from channel catfish cultured in the south-eastern United States (MS, LA or GA) from 1989 to 2006 and the ATCC-49512 isolate from France. Isolates were identified by both morphology and by species-specific PCR. The 20 μl PCR reaction used primers described by Welker et al. (2005) (16S-14F and 23S-1R) and was composed of 0·2 μmol l−1 of each primer, 200 μmol l−1 dNTPs, 1·5 mmol l−1 MgCl2, 0·5 U of Taq DNA polymerase (Promega Corporation, Madison, WI, USA), 1× buffer B, and approx. 45 ng of template DNA. Cycling conditions consisted of an initial denaturation step of 5 min at 94°C, followed by 30 cycles of 30 s at 94°C, 45 s at 56°C, and 60 s at 72°C, and a final extension step of 10 min at 72°C on an Applied Biosystem’s 2720 Thermal Cycler (Applied Biosystems, Foster City, CA). The PCR product was subjected to electrophoresis on a 1·2% agarose gel and stained with Gelstar® nucleic acid stain (Cambrex, East Rutherford, NJ, USA).

In situ DNA isolation and PFGE

Flavobacterium columnare strains were grown on F. columnare growth medium (FCGM) agar plates (Farmer 2004) at 30°C for 2 days. Single colonies were used to inoculate 5 ml of FCGM broth, and cultures were incubated for 24 h at 30°C with rotary aeration to obtain final optical density at 600 nm (OD600) of 0·7. Genomic DNA was prepared following the CHEF Bacterial Genomic DNA Plug Kits Instruction Manual (Bio-Rad Laboratories, Hercules, CA, USA) with some modifications. Bacteria were harvested by centrifugation at 10 000 g for 2 min at 4°C, the supernatant was discarded, and bacteria were resuspended in 0·3 ml of cell suspension buffer. The cell suspension was mixed with an equal volume of 2% (w/v) low-melting-point agarose. Solidified plugs were incubated in lysis buffer containing 100 μg ml−1 lysozyme for 3 h at 37°C. The lysis buffer was replaced with fresh lysis buffer containing 100 μg ml−1 proteinase K, and the plugs were incubated at 56°C for 48 h. Plugs were washed four times with Bio-Rad Wash Buffer, with the last two washes containing 1 mmol l−1 PMSF.

Restriction endonuclease digestion was conducted using MluI and PmeI (New England BioLabs, Ipswich, MA, USA), and DNA fragments were resolved in 1% pulsed-field certified agarose using a CHEF Mapper System (Bio-Rad). Electrophoresis was performed under the following conditions: running time 36 h, temperature 14°C, voltage gradient 6 V, initial pulse time 1·17 s, final pulse time 10·30 s, included angle 120º. Gels were stained with 0·5 μg ml−1 of ethidium bromide, and DNA was visualized by ChemiImager 5500 imaging system (Alpha Innotech, San Leandro, CA, USA).

PFGE pattern analysis

Macrorestriction patterns (MRPs) were analysed both visually and by computer-aided methods. bionumerics v3 software (Applied Maths, Inc., Sint-Martens-Latem, Belgium) was used to normalize the DNA fragment migration distances relative to those of the lambda ladder and small DNA marker (New England BioLabs). Isolates with a DNA band pattern differing by ≥ 1 band were defined to be a distinct PFGE profile (Tenover et al. 1995). The similarities between MRPs were expressed by Pearson-coefficient correlation, and clustering by the unweighted pair group method using arithmetic averages (UPGMA) was used for dendrogram construction.

Experimental infection of channel catfish

The first experimental infection study was conducted to determine the virulence of 16 different F. columnare isolates that were classified by PFGE. The chosen isolates were clustered in PFGE group A (90-106, 1191-B, 94-060, 94-081, C03133K, C0084-4, Matt, S05-79) and PFGE group B (90-509, C91-20, 143-94, 92-002, ATTC-49512, Evans, ALG-92491, and C56-1). The challenge method used by Thomas-Jinu and Goodwin (2004) was followed with some variations. Briefly, F. columnare isolates were cultivated in 800 ml of FCGM broth in a shaking incubator for 24 h at 30°C, and bacterial concentrations were then adjusted to A600 of 0·7.

About 5- to 6-month-old channel catfish fingerlings were stocked in tanks (15 fish per tank) supplied with dechlorinated municipal water at approx. 25°C and allowed to acclimate for 1 week. For experimental infections, 100 ml of bacterial culture was added directly to 10 l of water in each tank, with four replicate tanks for each bacterial strain. Four control tanks were exposed to sterile FCGM broth. After a 5 h exposure with constant aeration, bacteria were gradually removed by restoring water flow. Feed was restricted throughout the experiment. Fish were held for 8 days and observed for clinical signs and mortalities. Posterior kidney samples of all moribund and dead fish were cultured on FCGM agar, and the identity of the bacteria isolated from at least three fish per isolate was evaluated and confirmed by PFGE.

A second experimental challenge was performed with 4-month-old SPF channel catfish to determine whether compromises in the fish skin/mucus barrier would increase the pathogenicity of F. columnare isolates. Four F. columnare isolates, two from PFGE group A (94-081 and Matt) and two from PFGE group B (143-94 and C56-1), were compared in three different skin treatments (abraded, skin mucus removal, and un-abraded fish). The bath immersion protocol used during the first experimental challenge was repeated with the following modifications. About 30 min before bacterial challenge, fish were sedated in a separate tank containing approx. 100 μg ml−1 tricaine methane sulpfonate (MS222) (Argent Chemical Labs, Redmond, WA, USA). After sedation, 180 fish were abraded with a sterile scalpel in the dorsolateral abdomen; abrasions were approx. 2 cm in length and penetrated through the epidermal and dermal layers. Skin mucus was removed from another 180 fish by dragging a paper towel twice, back and forth, along the lateral line and dorsal portions of the abdomen. A third group of 180 fish received no skin treatment and were considered the un-abraded controls. Following skin treatment, fish were returned to their respective treatment tank and allowed to recover from the anaesthetic prior to challenge. Three tanks containing 15 fish each served as controls and were sham exposed to uninoculated FCGM broth as previously described.

All fish experiments were conducted under the approval of the Mississippi State University Institutional Animal Care and Use Committee.

Statistical analysis

Per cent mortalities from immersion bath experiments were compared using the General Linear Model utility of the sas v. 9.1 software package (SAS, Cary, NC), and statistical differences between groups were determined by the least significant difference test at P < 0·01.


PFGE analysis

All 31 F. columnare isolates grew in the presence of neomycin and polymyxin B as yellow rhizoid colonies and were observed as gram-negative slender rods that exhibited wide differences in cell length among isolates. The F. columnare specific primers FCISRFL and FCISRR1 (Welker et al. 2005) confirmed that all 31 isolates were included in this study.

PFGE of F. columnare chromosomal DNA digested with PmeI yielded 11–17 fragments in the 32·08–243·07 kb range, while MluI digested DNA yielded patterns of 11–19 fragments in the 24·49–192·51 kb range. The PFGE patterns on the same isolate for each restriction endonuclease were found to be stable and reproducible on at least three separate gels. All isolates showed a high degree of genetic diversity, with unique PFGE patterns resulting from both MluI and PmeI digestion. PFGE DNA profiles produced by nine of the isolates using the restriction endonucleases MluI (Fig. 1a) and PmeI (Fig. 1b) are provided.

Figure 1.

 PFGE profiles of Flavobacterium columnare strains. (a) MluI digestion. Lanes: 1, 7 and 14, 8-48 kb PFGE molecular size marker (Bio-Rad); 2, 8 and 15, Lambda Ladder molecular size marker (Bio-Rad); Lane 3, S05-79; Lane 4, 90-106; Lane 5, 90-252; Lane 6, 143-94; Lane 9 L90-629; Lane 10, L90-659; Lane 11, C91-20; Lane 12, 90-497; Lane 13, 90-509. (b) PmeI digestion. Lanes: 1, 8 and 15, Lambda Ladder molecular size marker; Lane 2, 9 and 14, 8-48 kb PFGE molecular size marker; Lane 3, 90-509; Lane 4, 90-497; Lane 5, C91-20; Lane 6, L90-659; Lane 7, L90-629; Lane 10, 143-94; Lane 11, 90-252; Lane 12, 90-106; Lane 13, S05-79.

Clustering of the MluI restriction patterns divided the isolates into two groups (designated A and B), with 75·37% and 86·30% similarity, respectively, within the groups (Fig. 2). Between the two groups, there was 63·97% similarity. No geographical correlation was found between the clustered subgroups. Clustering of the strains using PmeI restriction patterns showed an overall 82·38% similarity index (data not shown). The isolates were clustered into two groups with 83·57% and 84·56% similarity. No geographical correlation was found with PmeI subgroups.

Figure 2.

 UPGMA dendrogram demonstrating PFGE-based genetic similarity of F. columnare strains digested with MluI.

Experimental infections

Lesions typical of columnaris disease (Bullock et al. 1986; Plumb 1999; Roberts 2001) were found in the challenged fish. Typical columnaris signs seen included ‘saddleback’ lesions, fin and skin necrosis and yellowish coloration. Final per cent mortalities from the first bath immersion experiment for the 16 F. columnare strains tested are shown in Fig. 3. Overall, the strains that were clustered into PFGE group A by MluI restriction pattern (90-106, 1191-B, 94-060, 94-081, C03133K, C0084-4, Matt and S05-79) were more virulent in channel catfish fingerlings, with an average per cent mortality of 60·8% after 8 days. By contrast, PFGE group B isolates (90-509, C91-20, 143-94, 92-002, ATTC-49512, Evans, ALG92-491 and C56-1) had an average per cent mortality of 8·5%. Group A isolates were significantly more virulent to healthy un-abraded channel catfish fingerlings than group B isolates (< 0·01). Flavobacterium columnare was reisolated from the kidneys of morbid fish, and a complete PFGE profile analysis was performed (Fig. 4). Results demonstrated that, in every case, the same F. columnare strains that were used for the experimental infections were isolated from the kidneys of dead fish. None of the control fish died.

Figure 3.

 Per cent mortalities caused by F. columnare isolates in channel catfish fingerlings following experimental infection by bath immersion. Black bars indicate group A isolates (based on PFGE analysis of MluI digest), and white bars indicate group B isolates.

Figure 4.

MluI PFGE profiles of F. columnare isolates before and after recovery from experimentally infected channel catfish fingerlings. Lanes 1 and 14, Lambda Ladder molecular size marker; Lane 2, original 94-081 isolate; Lane 3, original 143-94 isolate; Lane 4, original C56-1 isolate; Lanes 5, 6 and 7, recovered 94-081 isolate; Lanes 8, 9 and 10, recovered C56-1 isolate; Lanes 11, 12 and 13, recovered 143-94 isolate.

In the second experimental challenge, higher mortality rates were found in the abraded and skin mucus removal treatments than in the un-abraded treatment (Fig. 5). PFGE group A isolates (94-081 and Matt) were found to cause average per cent mortalities of 90%, 87% and 100% in the un-abraded, skin mucus removal and abraded fish treatments, respectively. PFGE group B isolates (143-94 and C56-1) caused average mortalities of 20%, 86% and 66%, in the un-abraded, skin mucus removal, and abraded fish treatments, respectively. Overall per cent mortalities from all four strains in the un-abraded treatment was 55%, which was significantly lower (< 0·01) than the overall per cent mortalities in the mucus removal (86·5%) and abraded treatments (83%).

Figure 5.

 Per cent mortalities of channel catfish fingerlings after bath immersion with four different isolates of F. columnare following three different skin treatments. (a) Un-abraded treatment (♦, C56-1; bsl00066, Matt; ▪, 94-081; •, 143-94), (b) abraded treatment (♦, Matt; ▪, C56-1; bsl00066, 94-081; •, 143-94), (c) skin mucus removal treatment (♦, Matt; ▪, C56-1; bsl00066, 94-081; •, 143-94).

Per cent mortalities in the un-abraded fish during the second challenge were higher for Matt, C56-1 and 143-94 than they were in the first challenge. However, per cent mortalities for 94-081 remained the same, and the same trend was present in both challenges (94-081 was most virulent, followed by Matt, C56-1 and 143-94). This difference between challenges may have been because fish in the second challenge were 1–2 months younger.


Due to F. columnare’s role as one of the most important pathogens in the channel catfish aquaculture industry, a thorough understanding of its genetic variation is necessary. Information is also needed on whether F. columnare isolates from channel catfish vary in their pathogenic potential.

PFGE is a valuable typing method that was used for epidemiological investigation of several human pathogenic bacteria, including Listeria monocytogenes, Escherichia coli and Mycobacterium (Izumiya et al. 1997; Autio et al. 1999; Hughes et al. 2001). In our study, PFGE proved to be a useful technique to distinguish intraspecific genetic variation among F. columnare isolates recovered from southeast United States catfish aquaculture facilities. Although both MluI and PmeI restriction patterns clustered F. columnare strains into two major genetic groups, MluI restriction patterns revealed a greater degree of variation between F. columnare strains than PmeI restriction patterns. Therefore, we used MluI as our primary method for F. columnare PFGE analysis. Using MluI restriction analysis, group A included 15 (48·4%) of the strains analysed, and group B included 16 (51·6%). Strains in group A can be further divided into two subgroups, one containing nine strains and the other six strains. Similarly, group B strains can be divided into two subgroups, one containing six strains and the other 10 strains.

A widespread distribution of F. columnare exists among channel catfish aquaculture facilities (Tucker et al. 2004). Isolates from LA, GA, Stoneville-MS and Starkville-MS were analysed, and no geographic correlation was observed between the two subgroups. One explanation for this could be the high exchange of channel catfish fingerlings between farms and research stations. Interestingly, all four of the Stoneville, MS isolates from 2003 (523-03, 521-03, S03-487 and S03-579) clustered together with more than 90% similarity, suggesting a common ancestor among these isolates.

Because the focus of the current study was on F. columnare isolates affecting channel catfish aquaculture, isolates from other fish species (except for strain ATCC 49512) were not included. Therefore, direct comparison of the genetic groups identified in the current study cannot be made to previously reported F. columnare genetic groups using 16S RNA (Triyanto and Wakabayashi 1999) or RAPD (Thomas-Jinu and Goodwin 2004). However, future studies using PFGE to analyse F. columnare from other fish species should allow a direct comparison.

Studies have shown that some F. columnare isolates are more pathogenic than others (Decostere et al. 1998; Thomas-Jinu and Goodwin 2004; Suomalainen et al. 2006b), but pathogenicity did not correlate strongly with genetic groups of F. columnare as determined by RAPD analysis (Thomas-Jinu and Goodwin 2004). For the first time, results from the current experimental challenge suggest a direct correlation between genetic groups of F. columnare and pathogenicity to channel catfish fingerlings. In our study, PFGE group A primarily contains strains that are pathogenic for channel catfish, and PFGE group B primarily contains strains that have low pathogenicity in catfish. However, there is variation in the virulence of individual strains within each group, with three strains in group A causing less than 40% mortality (one strain causing no mortalities) and one strain from group B causing more than 40% mortality. Therefore, genetic analysis of F. columnare by PFGE may not allow complete prediction of virulence for an individual strain, but it will allow classification of strains into either group A or B, which would allow classification as either a high or low risk for being a ‘primary’F. columnare pathogen.

The second experimental challenge strongly supported our conclusions from the first challenge. As in the first experimental challenge, isolates belonging to PFGE group A showed a higher degree of pathogenicity than isolates clustered in PFGE group B, especially when un-abraded fingerlings were challenged. On the other hand, once the skin/mucus barrier of fish was compromised, all of the isolates caused higher mortality rates. In particular, the mortality rates of the group B strains increased dramatically after skin/mucus compromise. These results support the conclusion that group B strains tend to be opportunistic pathogens, requiring host compromise to cause mortality, while group A strains tend to be primary pathogens capable of causing high levels of mortality in un-abraded fish.

Mucus removal was shown to increase susceptibility of channel catfish to F. columnare infection following immersion challenge (Bader et al. 2003). Skin damage also increased the susceptibility of zebra fish Danio rerio to F. columnare following bath immersion challenge (Moyer and Hunnicutt 2007), and mucus removal was used in a koi-Cyprinus carpio immersion infection model (Tripathi et al. 2005). Our study demonstrates that mucus removal was just as effective as skin abrasion in causing increased mortalities, indicating that mucus is an important barrier against F. columnare infection.

In conclusion, our data shows that PFGE is a useful, reliable and reproducible molecular technique for genetically fingerprinting F. columnare. PFGE also demonstrated heterogeneity of isolates from channel catfish. The experimental immersion challenges showed a correlation between the two genetic groups and virulence, indicating that PFGE is a potentially useful tool for determining whether F. columnare isolates are more likely to be a primary or secondary pathogen. This information may allow producers to determine whether treatment is warranted if F. columnare is detected in a catfish pond. In addition, future pathogenesis research for columnaris disease in catfish should focus on pathogenic isolates from PFGE group A.


This research was supported by the US Department of Agriculture, Agricultural Research Service (Cooperative Agreement No. 58-6402-2-0073, Catfish Health Initiative).

The authors thank Michelle Banes, Matt Griffin and Tim Streit for technical help and support in this study. The authors thank Dr Lora Petrie-Hanson, Claudia Hohn and MSU Laboratory Animal Resources and Animal Care for maintaining the MSU-CVM Specific Pathogen Free fish laboratory and providing support and care for the fish used in this study. The authors also thank Dr John Hawke and Dr Joseph Newton for providing some of the F. columnare isolates used in this study.