There are different kinds of bacteria which can contaminate the work areas in food processing plants that are a danger to human health. They can spread to food items which are shipped out to consumers in the supermarket. The three main bacteria which are the subject of this technote are E-coli, Listeria and P-Fragi. One way to identify bacteria is through a standard dye test, which would classify the bacteria as gram-positive or gram-negative. If the dye test produces a red colour, the bacteria is classified as gram negative, and if it produces a blue colour, the bacteria is classified as gram positive. Of the three types of bacteria considered, E-Coli and P-Fragi are gram-negative and Listeria is gram-positive. Gram-positive bacteria generally have thicker walls that are rigid and do not have an outer membrane. Gram-negative bacteria on the other hand have thinner walls, are more elastic and have an outer membrane. There are other differences as well (Reference 1).
Figure below shows Gram-Negative and Gram-Positive bacteria.
Figure 1: Gram-Positive and Gram-Negative bacteria dye test (Reference 1)
Rapid detection of the distribution, concentration and types of these bacteria is important to prevent the disease spread and if patients are infected, it can help the patients to be cured in a short time (Reference 2).
In this article we will discuss two methods of identification and quantification of the three aforementioned bacteria. The pros and cons of each method are also discussed. These two methods are 2D camera methods which visually detect the fluorescence of the bacteria under Deep UV illumination (265 nm excitation) and spectrometer method which measures the fluorescence spectra using the same deep UV excitation. The spectrometer method uses chemometrics modeling such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to separate different types of bacteria.
Method 1: Camera Method
In this method a camera with the following specs were used:
Model: MaxCam2020-UV-TE from EHD imaging,
Format: 2048x2028 pixels,
Spectral range: 200-1100 nm
Active cooling 35° below ambient
The Deep UV excitation source was a 265 nm LED with 320 mW of power from Violumas. A 300 nm long pass filter was placed in front of the camera to block any 265 nm reflection from the source. Several samples of E-coli, Listeria and P-Fragi at different concentrations (e^6, e^5 , e^4 and e^3 cfu/mm^2) were provided by a collaborating company. These samples at different concentrations were placed on Stainless Steel plates. Figure 2 shows the experimental set-up.
Figure 2: The imaging method experimental setup
As depicted in figure 2, the incident angle is 30° and the reflectance angle is 60° to avoid any specular reflection to enter the camera. Background subtraction was done to obtain the images without background illumination. Figure 3 shows a series of images of E-Coli at different concentrations. Colour coding was used to observe the bacteria images more clearly. The concentration was in cfu/mm^2.
Figure 3: Images of e-coli at several concentrations in cfu/mm^2
Shape differences between different concentrations were observed as shown in Figure 3. The highest concentration sample with 1e6 cfu/mm^2 looked like a solid disk but as the concentration was decreased to 1e5 cfu/mm^2, 1e4 cfu/mm^2 and 1e3 cfu/mm^2 the shape became more ring-like with hollow interior. There was no observable difference between the shape of different kinds of bacteria at the same concentration, so this method is not suitable for distinguishing between different types of bacteria, but it is suitable to distinguish between different concentration levels.
Method 2: Spectrometer Method
A quick search in the literature showed a variety of Optical biosensors already available in the market. Biosensors are available to detect Listeria in lettuce and milk, Salmonella in apple juice and e-coli in tap water (Reference 3). The spectrometer method measures the fluorescence spectrum of the bacteria using the deep UV excitation at 265 nm. There are two distinct peaks for the bacteria fluorescence. One peak is around 335-350 nm which is due to tryptophan which is an essential amino-acid. The other sub-fluorescence peak at 515.9 nm is contributed by Flavia (Reference 2). There are subtle differences between the spectra of three types of bacteria which may not be visible visually. However, a Principal Component Analysis (PCA) model followed by a Linear Discriminant Analysis (LDA) model showed clusters formed for each bacteria. LDA is one of the most popular supervised methods for food analyses such as authentication, characterization and adulteration (Reference 4). For the spectrometer method, the excitation part was similar to the camera method and used a 265 nm, 300 mW LED shining at 30° on the sample. Instead of the camera, an Ocean Optics USB2000+ spectrometer with a spectral range of 200-850 nm and a ball lens attached to its input aperture were used. A 300 nm long pass filter was in front of the spectrometer to block any incoming deep UV light. The spectrometer was at 60° with respect to the normal to the sample plate to make sure it only detects the fluorescence. The ambient light was subtracted since in a real food processing environment where this instrument may be used, the ambient lights are always on and need to be subtracted. Figure 4: shows the fluorescence spectrum measured from E-Coli.
Figure 4: Fluorescence spectrum from E-coli sample.
In Figure 4, the part of the spectrum below 300 nm has been filtered out by the 300 nm long pass filter. Measurements from samples of three different types of bacteria at different concentrations were made. From each sample 3 average measurements were made at integration time of 1000 milliseconds so it took 3 seconds for each measurement. Only the part of the spectrum from 300-500 nm was used for the analysis. For all samples, first a PCA analysis with 8 components were used. The output of PCA models, which were called PC1 to PC8 were used as the input for the LDA analysis. The application of PCA to the data would remove the covariance of the data and the LDA would be able to separate the clusters better. Since there were 5 measurements from each sample plate, to create more data, combinations of 3 out of 5 [(1,23,),(1,2,4),….(3,4,5)] were averaged. So the total number of spectra from each sample plate increased from 5 to 10. A total of 90 spectra from 3 types of bacteria were generated and fed to the PCA and LDA models (3 concentrations for each bacteria and 5 measurements from each sample).
Figure 5 shows the result of this LDA analysis in 3D.
Figure 5: LDA analysis of samples showed a clear separation of 3 types of bacteria
A small overlap is observed between Listeria and P-Fragi but E-Coli is well separated from the other two.
The spectra recorded using the spectrometer did not show a good correlation with intensity and was very angle sensitive. It seems that the combination of the two methods is needed to separate the bacteria (Spectrometer method) and get an estimate of their intensity (camera method).
The instrument was finally packaged in a case made of acrylic and testing was done in a pork- processing factory in Quebec with positive results which could be the subject of a future technote.
References:
- Gram-Positive vs Gram-Negative Bacteria- 31 Differences with Examples, Sagar Aryal, Microbe notes, Jan 9,202
- Rapid Detection of Three Common Bacteria Based on Fluorescence Spectroscopy, R.Du et.al, Sensors, 22, 2022
- Rapid detection of Listeria monocytogenes, Salmonella, Campylobacter spp., and Escherichia coli in food using biosensors, A. Cossetini et.al, Food control, 137, 2022.
- Use of spectroscopic methods in combination with linear discriminant analysis for authentication of food products, M. Esteki et.al, Food control 91, 2018.
Author: Rez Mani (PhD)