Edwardsiella ictaluri is a facultative intracellular bacterium that causes enteric septicemia of catfish (ESC). In this study, we aimed to develop bioluminescent E. ictaluri that can be monitored by noninvasive bioluminescence imaging (BLI). To accomplish this, the luxCDABE operon of Photorhabdus luminescens was cloned downstream of the lacZ promoter in the broad host range plasmid pBBR1MCS4. Edwardsiella ictaluri strain 93-146 transformed with the new plasmid, pAKlux1, was highly bioluminescent. pAKlux1 was stably maintained in E. ictaluri without any apparent effect on growth or native plasmid stability. To assess the usefulness of the bioluminescent strain in disease studies, catfish were infected with 93-146 pAKlux1 by intraperitoneal injection and by bath immersion, and in vivo bacterial dissemination was observed using BLI. This study demonstrated that bioluminescent E. ictaluri can be used for real-time monitoring of ESC in live fish, which should enable observation of pathogen attachment sites and tissue predilections.
Edwardsiella ictaluri, the etiological agent of enteric septicemia of catfish (ESC), is a facultative intracellular bacterium (Hawke, 1979; Hawkeet al,1981). Although it causes disease in wild channel catfish populations, it is primarily a problem in commercial channel catfish production, where it is the most prevalent disease (USDA, 2003). ESC occurs in acute and chronic forms, which are characterized by gastroenteric septicemia and meningoencephalitis, respectively (Newtonet al,1989; Baldwin & Newton, 1993; Morrison & Plumb, 1994). Currently, only two feed-additive antibiotics are approved for use in treating ESC in channel catfish in the United States, ormetoprim-sulfadimethoxine and florfenicol. To date, the only E. ictaluri virulence factors that have been identified and confirmed using isogenic mutants include the secreted enzyme chondroitinase (Cooperet al,1996) and O polysaccharide (Lawrenceet al,2001, 2003).
Conventional techniques have been used to investigate tissue distribution and persistence of E. ictaluri in catfish, including serial dilutions with plate counting (Lawrenceet al,1997) and radioisotopic labeling (Nusbaum & Morrison, 1996). However, sacrifice of animals is required to obtain data at each time point, which may limit the number of data collection points and cause more variability. In vivo bioluminescence imaging (BLI) is a technique that can be used for real-time quantification and tracking of bacterial infections in living hosts (Contaget al,1995; Contag & Bachmann, 2002). With this technique, the same host animals can be used for data collection over the entire experiment, reducing the number of animals required and reducing variability between time points (Contaget al,1995, 1997, 1998; Rocchettaet al,2001).
To enable BLI, bacteria are tagged with bacterial luciferase genes, which catalyzes the oxidation of a long-chain aldehyde and FMNH2 to cause emission of visible light (Frackmanet al,1990; Meighen, 1993; Hastings, 1996; Wilson & Hastings, 1998). The bacterial lux operon consists of five genes (luxCDABE) encoding luciferase enzyme and its aldehyde substrate. Addition of an exogenous substrate is not required for bioluminescence in bacteria expressing luxCDABE, and background luminescence from animal tissues is negligible (Troyet al,2004).
In this research, we aimed to develop bioluminescent E. ictaluri for in vivo BLI of the pathogen in living fish as a new noninvasive tool for use in fish disease investigations. The resulting broad host range vector we constructed should be useful for many gram-negative species, and the techniques we developed could be used for BLI in other host fish species.
Materials and methods
Bacterial strains and plasmids
Edwardsiella ictaluri strain 93-146 was originally isolated from an ESC outbreak in a commercial channel catfish pond in Louisiana. Escherichia coli XL1-Blue MRF' (Stratagene) and NovaBlue (Novagen) strains were used for cloning and blue-white screening. Escherichia coli SM10λpir was used as the donor in conjugations for transfer of plasmids into E. ictaluri. Escherichia coli was grown using Luria–Bertani broth and agar plates at 37°C, and E. ictaluri was grown using brain heart infusion (BHI) broth and agar at 28°C. Ampicillin (100μgmL−1), kanamycin (50μgmL−1), and colistin (12.5μgmL−1) were used for plasmid selection and conjugation. Bioluminescent Escherichia coli and E. ictaluri colonies were detected using a ChemiImager 5500 imaging system with AlphaEaseFC software (Alpha Innotech).
Construction of pAKlux plasmids and conjugation
An EcoRI cassette containing the luxCDABE operon from Photorhabdus luminescens was excised from pCGLS1 (Frackmanet al,1990) and ligated into broad host range plasmid pBBR1MCS4 conferring ampicillin resistance (Kovachet al,1995). Before ligation, shrimp alkaline phosphatase (USB) was used to remove 5′ phosphates from the EcoRI-digested vector. The resulting new plasmids were designated pAKlux1 (with luxCDABE oriented downstream of the lacZ promoter) and pAKlux1.1 (luxCDABE in the opposite orientation). Plasmid maps were prepared using pDRAW32 DNA analysis software freely available at http://www.acaclone.com/.
Plasmids pAKlux1 and pAKlux1.1 were transferred from Escherichia coli donor strain SM10 λpir into E. ictaluri recipient strain 93-146 by conjugation (Lawrenceet al,1997). Briefly, 750μL of donor and 1500μL of recipient culture were pelleted separately by centrifugation, washed three times, mixed, and transferred to sterile 0.45μM filter papers on the surface of BHI plates. Following incubation at 28°C for 12–18h, bacteria were washed from the filters and spread on BHI plates with ampicillin and colistin, resulting in bioluminescent E. ictaluri colonies.
Estimating the minimum detectable number of bioluminescent E. ictaluri
To determine the minimum number of E. ictaluri pAKlux1 detectable using an IVIS Imaging System (Xenogen), four separate dilution series were prepared in black 96-well microtiter plates with bacterial population densities ranging from c. 105 to 102CFU mL−1 (each row containing a 1/2 dilution). Photon emissions were collected for 1min, and each dilution series was then spread on BHI agar to determine viable bacterial population densities. The linear correlation between bacterial population densities and bioluminescence was determined.
Effect of pAKlux1 on native plasmids
To determine the effect of pAKlux1 on the two native plasmids of wild-type E. ictaluri (Newtonet al,1988; Fernandezet al,2001), plasmids were isolated from wild-type strain 93-146 and bioluminescent strain 93-146 pAKlux1 by QIAprep Spin Miniprep Kit (Qiagen), and unmodified circular plasmids were separated on 1% agarose gel. Native plasmid densities were calculated by AlphaEaseFC software (Alpha Innotech), and pEI2/pEI1 ratios were compared by single factor anova.
Growth comparison and plasmid stability
Growth kinetics of 93-146 and 93-146 pAKlux1 strains were analyzed by calculating their doubling times during exponential growth. Briefly, quadruplicate samples of 93-146 and 93-146 pAKlux1 strains were prepared by diluting overnight cultures 1000-fold in 12mL of BHI, and OD600nm readings were obtained over a 48h period using a THERMOmax microplate reader (Molecular Devices). Generation times were calculated using three time points (4, 8, and 12h) during exponential growth and compared by anova.
Plasmid stability was analyzed by subculturing quadruplicate cultures of 93-146 pAKlux1 in BHI with and without ampicillin for 15 days. New cultures were inoculated from 24h cultures at a 1000-fold dilution. OD600nm and bioluminescence were determined for each 24h culture immediately after starting new subcultures. Bioluminescence for each culture was measured in relative light units (RLU) using an Lmax microplate luminometer (Molecular Devices) and normalized by dividing RLU by OD600nm readings. The ratio of normalized RLUs between ampicillin-selected and nonselected 93-146 pAKlux1 groups was used to calculate percent stability.
Experimental fish and in vivo bioluminescence
Experimental models for infection of channel catfish with E. ictaluri are well established using both intraperitoneal (IP) injection and bath immersion (Lawrenceet al,1997, 2001). IP injection allows more controlled dosing of catfish, but the immersion route mimics the natural route of infection better because bacterial penetration of host mucosa is required.
Specific pathogen free (SPF) channel catfish (11.16±0.65cm) were obtained from the SPF channel catfish laboratory at the College of Veterinary Medicine, Mississippi State University and transferred into 20L flow-through tanks with dechlorinated municipal water. Fish were maintained in well-aerated tanks with a water temperature of c. 25°C throughout the experiments.
To determine whether luminescent E. ictaluri could be detected in the living host, we used IP injection to experimentally infect two nonalbino channel catfish and two albino channel catfish with 100μL 93-146 pAKlux1 (1 × 108CFU). A second experiment was carried out to determine the optimal experimental IP injection dose for bioluminescence detection. Seven nonalbino channel catfish were IP injected with serial dilutions of 93-146 pAKlux1 (2.5 × 108–2.5 × 102 CFU) in 100μL of phosphate-buffered saline (PBS). BLI was conducted at 0.25, 2, 18, 42, 54, 66, 90, and 114h.
A third experiment was then conducted to determine whether BLI could be used to monitor the progress of E. ictaluri infection following experimental immersion exposure, which requires bacterial penetration of host mucosal membranes. Five fish were infected by immersing them in water containing 1 × 107CFU mL−1 for 1h, and BLI was conducted at 12, 24, 36, 48, 60, and 72h.
Real-time bioluminescent signal detection and image analysis were accomplished using an IVIS Imaging System. Fish were anesthetized for 4–5min in 5L water containing 200mgL−1 tricaine methane sulfonate, and excess water was removed by drying fish with a paper towel. Then, fish were placed on black nylon plastic and transferred immediately to the dark collection chamber for image capture. Total photon emissions from the whole fish body (right and left lateral) were collected at exposure time of 1min. Following imaging, fish were returned to well-aerated water for recovery. The bioluminescent images were displayed as pseudocolor images in which red represented the most intense light emission. Bioluminescence was quantified from the fish images using Living Image software (Xenogen).
Detection of bioluminescent E. ictaluri from internal organs
To determine the ability of BLI to detect E. ictaluri in different organs, internal organs were surgically removed from terminally moribund fish in the experimental challenges described above, and bioluminescent signal was detected using the IVIS Imaging System. Then, tissues were homogenized in 500μL of PBS, serially diluted, and spread in duplicate onto BHI agar plates without antibiotics, and bioluminescence of the resulting colonies was confirmed.
Construction of the broad host range plasmids
Two broad host range vectors consisting of pBBR1MCS4 and the lux operon of P. luminescens were constructed and designated pAKlux1 and pAKlux1.1. In pAKlux1, the luxCDABE operon is expressed from the lacZ promoter (Fig. 1a). In pAKlux1.1, the luxCDABE operon is located in the opposite orientation (Fig. 1b).
Bioluminescent E. ictaluri strains
Edwardsiella ictaluri carrying pAKlux1 were highly bioluminescent, while E. ictaluri carrying the pAKlux1.1 exhibited faint bioluminescence (data not shown). The relationship between bacterial population density (determined by serial dilution and plate counts) and bioluminescence (determined by photon emissions) was linear (R2=0.99) for five orders of magnitude (Fig. 2). The minimum detectable number of E. ictaluri pAKlux1 in 96-well plate was less than 2500 CFU per mL.
The effect of pAKlux1 on the two native plasmids of E. ictaluri was assessed semi-quantitatively by growing 93-146 and 93-146 pAKlux1 strains under the same conditions. Results indicated that pAKlux1 can coexist together with the native plasmids, and the presence of pAKlux1 does not affect the copy numbers of native plasmids (P<0.05, Fig. 3).
The effect of pAKlux1 on E. ictaluri growth kinetics was also determined. Doubling times were calculated for three time points, at the beginning (4h), at the middle (8h), and at the end of the exponential growth (12h). No significant difference in the generation times was observed (P<0.05).
Stability of pAKlux1 was determined by subculturing 93-146 pAKlux1 broth cultures under selective (with ampicillin) and nonselective (no ampicillin) conditions for 15 days. For the first 10 days, the bioluminescent signal intensity between cultures grown with and without ampicillin was similar (the ratio of normalized RLUs ranged from 1.002 to 1.102). Bioluminescence of 93-146 pAKlux1 cultured under nonselective conditions started declining after 10 days, and at the conclusion of the experiment, the ratio of normalized RLUs was 0.7. Based on the data, it appears that the half-life of pAKlux1 in E. ictaluri is about 18 days under the described culture conditions.
Monitoring in vivo bioluminescence in living fish
Bioluminescence was successfully detected from albino and nonalbino catfish that were IP injected with 93-146 pAKlux1. As expected, bioluminescent signal intensity (mean of two fish) was almost 10-fold higher in albino catfish [>6.00E+08 photonss−1cm−2steradian−1 (sr)] than in nonalbino catfish (6.19E+07 photonss−1cm−2sr−1). BLI of albino catfish usually resulted in saturation of the images when injected with 108 CFU.
When nonalbino catfish were injected with serial dilutions of 93-146 pAKlux1, in vivo bioluminescence was immediately detectable from fish injected with 2.5 × 104CFU (Fig. 4). Bioluminescence was not detectable in fish injected with lower dilutions of 93-146 pAKlux1 immediately after injection, but bioluminescence became detectable in the following days as disease progressed (Fig. 4). Bioluminescence in fish injected with 2.5 × 103 and 2.5 × 102 CFU was detectable 2 days after injection (Fig. 4). In fish injected with higher doses, bioluminescent E. ictaluri had dispersed to the whole abdominal area at 15min after IP injection, but intense regions were noticeable in the approximate regions of the anterior kidney and the heart (Fig. 4).
Total photon emissions at each sampling point for the fish injected with serial dilutions of 93-146 pAKlux1 were calculated (Fig. 5). Interestingly, no matter what the initial dose, mortalities occurred when total emissions in fish reached c. 108–109 photonss−1cm−2sr−1 (Fig. 5). However, it was clear that as bacterial doses increased, disease progressed more rapidly and mortalities occurred sooner. For example, at high doses (2.5 × 107–2.5 × 108 CFU), fish died in about 2 days, but at low doses (2.5 × 102–2.5 × 103 CFU), fish died in about 6 days. There was also a notable lag period following the injections where bioluminescence did not increase, which was followed by a period of rapid increase in bioluminescence and death. The higher doses had shorter lag periods than the lower doses, with essentially no lag period detectable at the highest dose (2.5 × 108 CFU).
In the immersion-exposed fish, bioluminescence was detected around the eroded fin bases and fin epithelia at the earlier time points, but no signal was detected from the rest of the body (data not shown). As disease progressed, patches of bioluminescence became visible around the edge of the opercular opening and around the mouth area. Total photon emissions showed a similar trend to the photon emissions from fish that were IP injected with 2.5 × 105 CFU per fish (Fig 5). Bioluminescence increased after 30h (Fig 5), and by 48h postinfection, bacteria concentrated in the lateral area where posterior kidneys are located. Bioluminescence spread to the whole fish body at 60–72h postinfection, with concentration occurring on the gills and abdominal area, followed shortly by death.
Dissection of catfish to remove individual organs revealed that the anterior kidney, posterior kidney, and spleen were emitting intense bioluminescence, while localized intense signals were also visible from gills, stomach, intestines, and heart (Fig. 6a). 93-146 pAKlux1 were recovered from the moribund and dead fish tissues on BHI plates without ampicillin. All of the recovered E. ictaluri colonies were bioluminescent (Fig. 6b).
To express the luxCDABE operon in E. ictaluri, a suitable vector system containing necessary components for successful conjugal transfer, replication, and stability in E. ictaluri was required. pBBR1MCS4 is a highly stable broad host range plasmid carrying a mobilization locus (mob) and a lacZ promoter (Plac) located upstream of a multiple cloning site (MCS) (Elzeret al,1995; Kovachet al,1994, 1995). The lacZ promoter is suitable for constitutive expression of the luxCDABE operon in E. ictaluri because this species does not have a lacI repressor gene. Conjugation is the preferred method for plasmid transfer into E. ictaluri because of low electroporation efficiency (Maureret al,2001).
The luxCDABE operon was strongly expressed in E. ictaluri from the lacZ promoter on pAKlux1. The fainter bioluminescence resulting from 93-146 pAKlux1.1, which has the luxCDABE operon in opposite orientation from the lacZ promoter, is very similar to the findings of Frackman . (1990). The low level of luminescence from pAKlux1.1 is likely due to native promoter activity in the insert upstream of luxCDABE (Frackmanet al,1990). Although pAKlux1.1 is not as useful for BLI studies, it has potential for use as a promoter trap vector. Suitable restriction sites (SacI, SpeI, BamHI, SmaI, and PstI) are available for cloning upstream of luxCDABE (Fig. 1b).
Edwardsiella ictaluri contains two native plasmids that are consistently present in channel catfish disease isolates (Newtonet al,1988; Fernandezet al,2001). Our results showed that neither the native plasmids nor the growth of E. ictaluri are adversely affected by pAKlux1. It is likely that pAKlux1 is compatible with the native E. ictaluri plasmids because pEI1 has a ColE1-type replication system and pEI2 has a ColE2-type origin of replication (Fernandezet al,2001), and pBBR1MCS is compatible with ColE1 plasmids (Antoine & Locht, 1992; Kovachet al,1994). In addition, pAKlux1 was quite stable in E. ictaluri broth cultures without antibiotic selection and during experimental infections of catfish. These findings are similar to the stability of pBBR1MCS in Brucella species (Elzeret al,1995).
The current study demonstrates, for the first time, use of BLI for real-time monitoring of bacterial infection in living fish. Real-time monitoring of bacterial infection using BLI in a living host has been used mainly in mouse models (Contaget al,1995; Fordeet al,1998). In fish, Aeromonas salmonicida expressing luxAB was used to monitor colonization and transmission of the pathogen in Atlantic salmon (Salmo salar), but bioluminescence was not detected in live fish (Fergusonet al,1998). In addition, luxAB marked A. salmonicida required exogenous addition of aldehyde substrate for luminescence (Fergusonet al,1998). We have previously used luminescent E. ictaluri (luxCDABE carried on a ColE1 plasmid) for in vitro quantification of E. ictaluri serum resistance (Lawrenceet al,2003). However, this ColE1 plasmid was not stable enough in E. ictaluri to allow imaging of luminescent E. ictaluri in living fish.
Considerations for successful use of BLI to monitor bacterial infection include the requirement for oxygen in the tissues and signal loss because of absorption and scattering from host tissues. Our results indicate that a strong bioluminescent signal can be obtained through fish tissues, with the lower limit of detection being 104 CFU. Although pigmentation does reduce the total detectable emissions about 10-fold, it does not pose a problem owing to high sensitivity of signal detection and low background bioluminescence from fish tissues.
In this first study, BLI has already revealed new insights on the pathogenesis of ESC. First, using BLI, we were, for the first time, able to see bacterial attachment on the surface of fish. Novel attachment sites around the eroded fin arches and epithelia as well as the injured skin regions were detected, which would suggest the importance of mucus in prevention of initial bacterial attachment. Second, E. ictaluri infection appears to be characterized by an initial period of stable bacterial numbers followed by a period of rapid bacterial replication and dissemination. This period of rapid bacterial replication is followed shortly by death of the host. Although a quorum sensing system has not been described for E. ictaluri, this study fits the model that pathogenic bacteria must first reach a certain population density before causing disease and/or death (Donabedian, 2003), and it suggests that a quorum sensing system may regulate virulence gene expression in E. ictaluri. Third, it appears that death of catfish is imminent once a certain ‘lethal’ bacterial number in catfish tissues is reached, which in this study was c. 108–109 photonss−1cm−2sr−1.
The correlation between IP injection dose of E. ictaluri and the onset of catfish mortalities in this study is similar to previously reported findings (Lawrenceet al,1997). BLI showed that anterior and posterior kidneys and the spleen contain the highest concentrations of E. ictaluri during infection, which is also similar to previous reports (Lawrenceet al,1997; Thuneet al,1999). BLI also suggested that E. ictaluri has a predilection for gills, which has only been reported once previously (Nusbaum & Morrison, 1996), and heart, which has not been previously reported.
The present research demonstrates the development of a broad host range plasmid, pAKlux1, for luminescent labeling of gram-negative bacteria. It also reports the development of stable bioluminescent E. ictaluri and the application of BLI in fish. Future applications of BLI using luminescent E. ictaluri will enable new insights on the pathogenesis of ESC.
We thank Dr Scott Willard, Alicia Musselwhite, and Anna Chromiak at the Facility for Organismal and Cellular Imaging, Animal and Diary Sciences, Mississippi State University, for assistance with in vivo imaging. This project was supported by the National Research Initiative of the USDA Cooperative State Research, Education and Extension Service, grant number # 2004-35204-14211.