Anal. Chem. 2004, 76, 433-440
Detection of Campylobacter and Shigella Species in Food Samples Using an Array Biosensor Kim E. Sapsford,† Avraham Rasooly,‡ Chris R. Taitt,§ and Frances S. Ligler*,§
George Mason University, 10910 University Boulevard, MS 4E3, Manassas, Virginia 20110, United States Food and Drug Administration, 5100 Paint Branch Parkway, College Park, Maryland 20740, and Center for Bio/Molecular Science & Engineering, Naval Research Laboratory, Washington, DC 20375
Campylobacter and Shigella bacteria are common causes of food- and water-borne illness worldwide. There is a current need in food, medical, environmental, and military markets for a rapid and user-friendly method of detecting such pathogens. The array biosensor developed at the NRL encompasses these qualities. In this study, 25-min, sandwich immunoassays were developed for the detection of Campylobacter and Shigella species in both buffer and a variety of food and beverage samples. The limit of detection for Shigella dysenteriae in buffer and chicken carcass wash was 4.9 × 104 cfu mL-1, whereas Campylobacter jejuni could be measured at concentrations as low as 9.7 × 102 cfu mL-1. The limits of detection and dynamic range were found to vary depending on the sample matrix, but could be improved by running the sample over the waveguide surface for longer periods of time. Samples were run with no preconcentration or enrichment steps and little-to-no sample pretreatment prior to analysis. Food- and water-borne infections are a worldwide problem, affecting both developed and developing countries alike.1,2 Campylobacter is the most common cause of intestinal and diarrheal disease in the U.S., with an estimated 2.4 million cases per year.3,4 Campylobacter jejuni and Campylobacter coli together are responsible for ∼95% of these cases. The bacteria are found to inhabit the intestinal tracts of a variety of healthy mammals and birds, leading to common causes of the infection arising from the consumption of unpasteurized milk and milk products and undercooked poultry. The minimum infective dose in human studies was found to be 500 bacteria for C. jejuni.4 Symptoms of infection include diarrhea, fever, abdominal and muscle pain, nausea, and headache, but typically only last 7-10 days, because the infection is self-limiting. In a very small number of cases, reactive arthritis and Guillain-Barre´ syndrome may develop * Corresponding author. Phone: 202-404-6002. Fax: 202-404-8897. E-mail:
[email protected]. † George Mason University. ‡ U.S. Food and Drug Administration. § Naval Research Laboratory. (1) Alocilja, E. C.; Radke, S. M. Biosens. Bioelectron. 2003, 18, 841-846. (2) Sharma, S.; Sachdeva, P.; Virdi, J. S. Appl. Microbiol. Biotechnol. 2003, 61, 424-428. (3) Mao, Y.; Zhu, C.; Boedeker, E. C. Curr. Opin. Gastroenterol. 2003, 19, 1122. (4) Kothary, M. H.; Babu, U. S. J. Food Safety 2001, 21, 49-73. 10.1021/ac035122z CCC: $27.50 Published on Web 12/09/2003
© 2004 American Chemical Society
following infection. As with most of these bacterial infections, young children, the elderly, and immune-compromised individuals are most at risk and suffer the severest symptoms. Shigella species account for 0.1); however, responses between different batches of bacterial cells were found to be variable (P < 0.005), as shown in Figure 2A. The limit of detection for batch no. 2 was 5 × 104 cfu mL-1, a factor of 10 increase relative to batch no. 1. One of the main problems with using colony-forming units per milliliter as a measure of the concentration of bacterial cells when coupled with antibody capture is that antibodies will bind both live and dead cells. In both batches, the colony-forming units per milliliter, which is a measure of only the live cells in the batch, was determined before the cells were killed and shipped to the NRL. To independently quantify the amount of bacterial cells in the sample, a protein quantification kit from Molecular Probes was used and the concentration of each batch was expressed in micrograms per liter. The results of the S. dysenteriae samples used and those of the Campylobacter bacterial cells used later in this study are shown in Table 1. Figure 2B depicts the dose-response curves in Figure 2A replotted with the concentration expressed in micrograms per liter. The dose-response curves in the linear region are now highly reproducible, not only between slides using the same batch but also batch-to-batch (P > 0.25). Therefore, a greater number of dead cells must have been present in batch no. 1 prior to the live cell count to account for the apparent batch-to-batch variation originally observed (Figure 2A). This result was important under the conditions of this study to account for the batch-to-batch variability observed. However, in real-world food samples, only the level of live bacterial cells is usually of interest, since dead cells, although a useful indication of food quality, are not considered a health risk.
Figure 2. Comparison of slide-to-slide and batch-to-batch variation of the dose-response curve for S. dysenteriae in PBSTB; slide 0102batch no. 1 (b), slide 0119-batch no. 1 (O), slide 0103-batch no. 2 (2) and slide 0108-batch no. 2 (4). The dose-response curves are plotted as net intensity verses (A) S. dysenteriae concentration in colony-forming units per milliliter and (B) S. dysenteriae concentration in micrograms per milliliter
Table 1. Concentration of Bacterial Cells at the Highest Concentration concn sample S. dysenteriae batch no. 1 S. dysenteriae batch no. 2 C. jejuni ATCC43486 C. jejuni ATCC35918 C. jejuni ATCC43480 C. jejuni ATCC43478
cfu
mL-1
4.7 × 107 4.3 × 107 2.3 × 108 3.2 × 107 1.2 × 108 4.3 × 109
µg mL-1
µg cfu-1
380 ( 30 1600 ( 200 45 000 ( 2000 53 000 ( 4000 90 000 ( 10 000 49 000 ( 2000
8.1 × 10-6 3.7 × 10-5 2.0 × 10-4 1.7 × 10-3 7.5 × 10-3 1.1 × 10-5
Dose-response curves for S. dysenteriae (batch no. 2) spiked into several different food types were determined to assess how the array biosensor system performed with real-world samples. Food samples were spiked with S. dysenteriae ∼2 h prior to the assay to allow time for the bacterial cells to interact with the sample matrix. Positive controls of 2.5 × 107 cfu mL-1 S. dysenteriae in buffer (no food matrix) were included on all the food sample slides, and all data were normalized to these controls for the dose-response curves in different food samples. Figure 3A shows the normalized S. dysenteriae dose-response curves determined in PBSTB, ground turkey meat (GT), chicken carcass wash (CW), buffered milk (BM), and lettuce leaf wash (LL). The limits of detection, linear range, and regression fits are summarized in Table 2. In most food matrixes studied, there was found to be an increase in the value of the limit of detection and significant differences in the slopes of the linear regression fits (P > 0.005). Ground turkey and buffered milk were found to cause the greatest effects on the dose-response curve relative to the buffer control. This suggests that there is some interference from the matrix which affects binding in the sandwich assay and highlights the need to run a dose-response in the matrix under study when quantification of samples, containing unknown amounts of S. dysenteriae, is required. Figure 3B demonstrates that S. dysenteriae in different food matrixes can be detected in the presence of 10-fold excess Salmonella with little loss in the net
intensity measured, highlighting the specificity of the capture and trace antibodies used in the assay. S. dysenteriae was also measured in the presence of 20-fold excess E. coli with a 20% increase in the net intensity of the data squares measured (data not shown). E. coli cells alone gave a low response in the Shigella sandwich immunoassay, ∼10% of the Shigella net intensity. Therefore, E. coli cross-reacts with the polyclonal Shigella capture and tracer antibodies and can be detected when present at high concentrations. This result is not altogether surprising, because the two bacteria are very similar organisms and share among other things similar virulence-related genes.5,9,13,15,2828 Campylobacter Immunoassays. The Campylobacter sandwich immunoassay format was tested in PBSTB for its ability to measure two different species of Campylobacter bacteria as well as two different strains of each species. Figure 4A shows the dose-response curves obtained for C. jejuni, strains ATCC43486 and ATCC35918, and C. coli, strains ATCC43480 and ATCC43478. The Campylobacter sandwich immunoassay format was found to respond to all the bacterial samples tested. The curves for each of the Campylobacter species could be fitted with a four-parameter asymmetric sigmoidal, giving Y ) -143 + (9.78 × 103)/{1 + [(x/ 2.04 × 105)-0.71]}, Y ) 78.0 + (1.44 × 104)/{1 + [(x/1.87 × 104)-0.99]}, Y ) 282 + (4.01 × 103)/{1 + [(x/1.95 × 106)-1.37]}, and Y ) 176 + (3.84 × 103)/{1 + [(x/8.60 × 105)-1.40]} for C. jejuni, strains ATCC43486 and ATCC35918, and C. coli, strains ATCC43480 and ATCC43478, respectively. Limits of detection of 4.8 × 103 and 9.8 × 102 cfu mL-1 for C. jejuni strains ATCC43486 and ATCC35918, respectively, and 7.8 × 105 and 3.1 × 105 cfu mL-1 for C. coli strains ATCC43480 and ATCC43478, respectively, were determined. The lower limits of detection and greater affinity of the Campylobacter sandwich immunoassay for the C. jejuni strains is not surprising, because the antibodies were raised against this species of the bacteria. These limits of detection are similar to those obtained by Che et al. (2 × 104 cfu mL-1)14 and Endtz et al. (3 × 104 cfu mL-1)11 for ∼2-h assays using immu(28) Pupo, G. M.; Lan, R.; Reeves, P. R. PNAS 2000, 97, 10567-10572.
Analytical Chemistry, Vol. 76, No. 2, January 15, 2004
437
Figure 3. (A) Dose-response curves for S. dysenteriae in different food matrixes: PBSTB (b), ground turkey (GT, O), chicken carcass wash (CW, 2), buffered milk (BM, 4), and lettuce leaf wash (LL, 9). (B) Detection of S. dysenteriae (2.5 × 107 cfu mL-1) in the presence of excess Salmonella (2.5 × 108 cfu mL-1) in different food and beverage matrixes. Solution (1) PBSTB, (2) Salmonella in PBSTB, (3) Shigella in PBSTB, (4) Shigella/Salmonella in PBSTB, (5) Shigella in CW, (6) Shigella/Salmonella in CW, (7) Shigella in GT, (8) Shigella/Salmonella in GT, (9) Shigella in BM, (10) Shigella/Salmonella in BM, (11) Shigella in LL, and (12) Shigella/ Salmonella in LL.
Table 2. S. dysenteriae Detection in Food and Beverage Matrixes
a
matrix
limit of detection cfu mL-1
linear range cfu mL-1
eq of linear regression
R2
PBSTB ground turkeya carcass wash buffered milk lettuce leaf
4.9 × 104 7.8 × 105 4.9 × 104 7.8 × 105 2.0 × 105
4.9 × 104 to 7.8 × 105 7.8 × 105 to 1.3 × 107 4.9 × 104 to 7.8 × 105 7.8 × 105 to 1.3 × 107 2.0 × 105 to 7.8 × 105
Y ) 3 × 10-7x + 0.005 Y ) 3 × 10-8x + 0.006 Y ) 2 × 10-7x + 0.003 Y ) 2 × 10-8x + 0.004 Y ) 1 × 10-7x + 0.010
0.997 0.997 0.987 0.985 0.983
cfu g-1.
Figure 4. (A) Dose-response curves of different Campylobacter species and strains in PBSTB: C. jejuni ATCC43486 (b), C. jejuni ATCC35918 (O), C. coli ATCC43478 (2) and C. coli ATCC43480 (4). (B) Detection of C. jejuni (ATCC35918) (1 × 105 cfu mL-1) and C. coli (ATCC43480) (1.25 × 107 cfu mL-1) in the presence of excess Salmonella (2.5 × 108 cfu mL-1), E. coli (2.5 × 108 cfu mL-1), or both in PBSTB. Solution (1) C. jejuni, (2) C. jejuni/Salmonella, (3) C. jejuni/E. coli, (4) C. jejuni/Salmonella/E. coli, (5) PBSTB, (6) C. coli, (7) C. coli/Salmonella, (8) C. coli/E. coli, (9) C. coli/Salmonella/E.coli, (10) PBSTB.
noassay-based detection. Much lower limits of detection, 0.008120 cfu mL-1, have been obtained using PCR-ELISA when used 438 Analytical Chemistry, Vol. 76, No. 2, January 15, 2004
in conjunction with a 48-h enrichment step prior to analysis.16-18 The Campylobacter sandwich immunoassay was also able to detect
Table 3. Campylobacter jejuni (ATCC35918) and coli (ATCC43480) Detection in Food and Environmental Matrixes
a
matrix
limit of detection cfu mL-1
linear range cfu mL-1
eq of linear regression
R2
PBSTB river water ground turkeya carcass wash
9.7 × 102 3.1 × 103 1.6 × 103 3.1 × 103
C. jejuni 9.7 × 102 to 7.8 × 103 3.1 × 103 to 1.3 × 104 1.6 × 103 to 1.3 × 104 3.1 × 103 to 1.3 × 104
Y ) 4 × 10-5x + 0.025 Y ) 4 × 10-5x + 0.034 Y ) 4 × 10-5x + 0.017 Y ) 4 × 10-5x + 0.018
0.970 0.986 0.998 0.971
PBSTB river water ground turkey* carcass wash
7.8 × 105 1.6 × 106 7.8 × 105 7.8 × 105
C. coli 7.8 × 105 to 3.1 × 106 1.6 × 106 to 3.1 × 106 7.8 × 105 to 3.1 × 106 7.8 × 105 to 3.1 × 106
Y ) 2 × 10-7x + 0.097 Y ) 1 × 10-7x + 0.031 Y ) 3 × 10-7x + 0.069 Y ) 7 × 10-8x + 0.019
0.979 0.984 0.938 0.938
cfu g-1
Figure 5. The CCD image of a single slide exposed to multiple analytes. The slides were patterned with two rows of capture antibodies directed against Salmonella (Sal, 20 µg mL-1 rabbit anti-Salmonella), Shigella (Shig, 20 µg mL-1 rabbit anti-Shigella), and C. jejuni (Camp, 20 µg mL-1 rabbit anti-C. jejuni), incubated overnight, rinsed, blocked, and assembled with the assay PDMS flow cell. Samples containing C. jejuni (ATCC35918) (1 × 105 cfu mL-1), S. dysenteriae (2 × 106 cfu mL-1), and S. typhimurium (5 × 105 cfu mL-1) in chicken carcass were flowed over the slide as indicated, followed by a tracer cocktail containing 10 µg mL-1 each of AlexaFluor-rabbit anti-C. jejuni, AlexaFluor-rabbit antiShigella, and Cy5-monoclonal anti-S. typhimurium. The net intensity values (divided by a factor of 10) reached for each data square are given in the grid table; regions in which a positive response is expected (because of the patterning of the capture antibodies) are highlighted in white.
C. jejuni (1 × 105 cfu mL-1) and C. coli (1.3 × 107 cfu mL-1) in the presence of either Salmonella (2.5 × 108 cfu mL-1) or E. coli (2.5 × 108 cfu mL-1) or both, with no effect on the net intensity measured from the data squares, P > 0.25 and P > 0.1 for C. jejuni and C. coli, respectively (Figure 4B). C. jejuni (strain ATCC35918) and C. coli (strain ATCC43480) were chosen for the food studies. As with the Shigella experiments, ground turkey and chicken carcass wash were investigated. River water was also included as a sample matrix because of its frequent contamination with Campylobacter from the excreta of infected birds. A PBSTB positive control was run as part of the assay and, as with the Shigella data, the net intensities were normalized using this positive control. The dose-response curves were calculated from the data squares, and the limits of detection, linear range, and regression fits are summarized in Table 3 for C. jejuni (ATCC35918) and C. coli (ATCC43480). As found in the Shigella studies, the limit of detection increased in value in the different samples tested for C. jejuni, although the slope of the normalized data remained unchanged (P > 0.1, relative to the buffer). The limit of detection for C. coli, however, only increased in the river water sample, with 7.8 × 105 cfu mL-1 being detected in both the ground turkey and the carcass wash, whereas the slopes of the dose-response curves were found to be sampledependent (P < 0.005 for RW and CW; P > 0.25 for GT, relative to the buffer). Again, this highlights the need, in situations in which the array biosensor is being used to test real-world samples,
for the development of a dose-response curve in the matrix of study if quantification as well as identification of the bacteria is required. Multianalyte Immunoassays. One of the main advantages of the PDMS patterning manifolds used in our assays is the ability to immobilize multiple capture antibodies for different species onto a single slide. Such an arrangement not only allows multiple samples to be analyzed simultaneously for multiple pathogens, but a series of standards and positive and negative controls can also be analyzed for quantification and standardization of the data on each slide. Figure 5 shows the final CCD image of a slide functionalized with three different capture antibodies and demonstrates the detection of S. typhimurium, S. dysenteriae, and C. jejuni (ATCC35918) spiked into chicken carcass wash. The high specificity of the antibodies is also demonstrated with positive responses relative to the blank only observed on the slide where expected independent of whether the sample contained one, two, or more of these different species. CONCLUSIONS This study has demonstrated the ability of a sandwich immunoassay to detect Campylobacter and Shigella bacteria in buffer and a number of complex food matrixes. The sandwich immunoassay was rapid, taking only 25 min, and required little-to-no sample pretreatment or preconcentration steps prior to analysis. The limit of detection for S. dysenteriae in buffer was 4.9 × 104 Analytical Chemistry, Vol. 76, No. 2, January 15, 2004
439
cfu mL-1, whereas C. jejuni could be measured as low as 9.7 × 102 cfu mL-1. The limits of detection and dynamic range were found to vary depending on the sample matrix for both bacteria. Therefore, it is recommended that if quantification of the bacteria in a particular food sample is required, a standard curve be generated in that matrix. Although the limits of detection for these pathogens are comparable to those reported previously for immuno-based detection, the fact remains that as little as 500 bacteria for C. jejuni and 10 bacteria for S. dysenteriae can cause human illness. Consequently, for the array biosensor to be utilized effectively by the food industry for the detection of C. jejuni or S. dysenteriae, the technique would need to be coupled with a sample preconcentration or pre-enrichment step prior to analysis. Such steps would therefore achieve levels of bacterial cells detectable by the array biosensor. The use of immunomagnetic beads, for example, may be used to preconcentrate the bacterial cells and would probably increase the analysis time by only a couple of hours.19,20 On the other hand, pre-enrichment of the sample, using a commercially available specific enrichment broth, typically requires longer time periods and could increase the analysis time to as much as 2 days. Despite the potential requirement for preconcentration in order to reach limits of detection equivalent to the infectious dose, the
440
Analytical Chemistry, Vol. 76, No. 2, January 15, 2004
array biosensor still has advantages over other assay methods. One of the main advantages of the array biosensor over current existing technology is its ability to detect multiple analytes in multiple samples on a single slide. This was clearly demonstrated in a 25-min assay for S. typhimurium, S. dysenteriae, and C. jejuni (ATCC35918) spiked into chicken carcass wash. Another advantage is the simplicity of the capture immunoassay, as compared to ELISA and PCR assays in terms of number of steps and manipulations involved. ACKNOWLEDGMENT The authors thank Lisa Shriver-Lake for her helpful discussions regarding the food spiking and preparation. This work was supported by funding from the Department of Defense Joint Science and Technology Panel for Chemical and Biological Defense, the United States Department of Agriculture (USDA) and the Food and Drug Administration (FDA). The views expressed are those of the authors and do not represent those of the U.S. Navy, U.S. Department of Defense, or the U.S. Government. Received for review September 24, 2003. Accepted November 4, 2003. AC035122Z
