Limits of Detection Analysis of Advanced Technologies for Bacterial Detection in Food Samples: Review & Future Perspective

1. Introduction

Foodborne illnesses can be dangerously infectious, and they are predominantly caused by pathogens (e.g., bacteria, fungi, viruses, parasites, etc.) or toxins (e.g. dioxins, heavy metals, mycotoxins, etc.) entering the body through contaminated food [1]. Most of the pathogens that can cause foodborne diseases are bacteria [2]. Bacteria can cause acute poisoning, long-term diseases, serious disabilities and even deaths [3]. Among all, Salmonella species causes the most serious illnesses and deaths related to contaminated food [4-6]. Salmonella is commonly found in birds, eggs, vegetables and also in natural water. Its symptoms include fever, vomiting, pain and dehydration etc. Salmonella can be divided into over 2600 species. Among them, Salmonella enterica and Salmonella typhimurium are the most commonly found [7]. Listeria usually exists in processed products such as milk, meat, seafood and can grow in the refrigerators [8]. Listeria was shown to cause miscarriages in pregnant women or deaths of infants, although the chance is really low [9]. Around 20 species in Listeria can cause human diseases and Listeria monocytogenes is the type that causes the most harm to humans [10]. Most Campylobacter infections in humans are acquired by eating and touching contaminated poultry, seafood and meat [11-12]. More than 20 species of Campylobacter have been implicated in human disease, with Campylobacter jejuni and Campylobacter coli being the most well known ones [13]. The most common symptoms of Campylobacter infections are diarrhea, fever, vomiting, and stomach cramps [14-15]. S. aureus is normally found in birds, meat and milk [16]. S. aureus is one of the common bacteria that display antimicrobial resistance, to antibiotics like methicillin and vancomycin [17-18]. Common symptoms of S. aureus are shown on the skin, as painful red welts and sores [19-20]. Generally, E. coli can be found in contaminated meat, milk and vegetables [21]. E. coli can be divided into 3 main groups-Enteropathogenic, Enteroinvasive and Enterohemorrhagic. A subtype of Enterohemorrhagic E. coli is the most toxic variant that is also easily transferred [22-23]. Although E. coli does not cause any symptoms in most healthy humans, it can lead to diarrhea, vomiting, and fever [24-25].

As many bacterial species currently pose a major threat to humans, a quick, accurate and cheap method to detect bacteria in the environment is essential, especially for food samples [26]. The traditional method to detect bacteria is through culturing of bacteria, which includes isolating the bacteria and monitoring the growth of the colonies [27]. During the culture process, the bacterial colonies are fixed and stained on a glass slide and confirmed by microscopy observation in order to identify different types of bacteria. This process is usually very time-consuming and labor-intensive [28]. Other methods are more complex and can overcome some limitations of bacteria culture. Another common detection method is high-performance liquid chromatography (HPLC) which has high sensitivity [29]. When the concentration of bacterial colonies is very low but still cannot be ignored for human health, it will be a challenge for these methods [30]. Researchers have developed many alternative methods to overcome these problems [26]. One technology that has been widely used more recently is enzyme-linked immunosorbent assay (ELISA), which is available as a commercial test kit for the detection of bacteria. However, it has disadvantages, some of which include; low sensitivity and necessary cold chain that limits its application range [31]. As a result, it is really difficult to meet the demand for large-scale bacteria detection in food samples with current technologies.

LFIA is a relatively novel method in food safety analysis for the detection of bacteria in food. LFIA is cost-effective, simple to use, and can produce results rapidly with fewer samples [32-33]. It measures the concentration of bacteria by the darkness of color on the strip. Conjugated nanoparticles, dominate the porous membrane as an indicator [34-35]. Gold nanoparticles (GNPs) and iron oxide NP (IONP) are the most used NPs in the LFIA because of their low toxicity, and particles size and shape can be controlled by many factors [7,36]. Another relatively new method-PCR is a widely used technique for the detection of bacteria in food. It can make millions to billions of copies of a DNA sample rapidly so it has a high sensitivity and a relatively better detection limit than other common detection methods [37-38]. An alternative well-known technique is electrochemical analysis methods for the detection of bacteria in food. The electrochemical methods mainly measure the changes in the electronic properties caused by bacteria introduced to the solution [39-40].

In this review, PCR, LFIA and electroanalytical techniques and their efficiency in the detection of bacteria in food samples have been summarized. Recent developments (2013-2023) covering PCR, LFIA and electrochemical methods for detection of various bacterial species (Salmonella, Listeria, Campylobacter, S. aureus and E. coli) taking into account the different food types, analytical performances and the reported limit of detection (LOD) are discussed from 150 references. Current challenges and future avenues (including the role of nanomaterials /advanced materials) to further improve the analytical performance for bacterial detection are discussed.

2. Research Methods

Information was collected from Science Direct with keywords: bacteria, PCR, LFIA, electrochemical method, LOD. In Figure 1, 50 peer-reviewed articles by every detection method from 2013–2023 were compared to identify the limit of detection (LOD) for different bacteria: bacteria in the Pseudomonadota domain which includes Salmonella, E. coli; the Campylobacterota domain includes Campylobacter; and the Bacillota domain includes bacteria Listeria, S. aureus. In Figure 1, every type of bacteria by every detection method contains 10 articles and lowest LOD of it is listed. Please note that some papers included the LOD of more than one bacterium or more than one detection method, but only the detection of bacteria with the lowest LOD in that paper is discussed here. The data were collated by bacteria type, year of article, detection type: multiplex or not, food sample type, number of replicates, LOD (CFU/ml) in Table 1, Table 2 and Table 3.

3. Results

Table 1 provides a breakdown of the analysis of 50 peer-reviewed articles used in this review, based on the type of bacteria, multiplexing capability, the type of food sample, replicates and, reported LOD. The limits of detection are further shown in Figure 2, according to (a) multiplex detection capability and (b) food sample types, in an attempt to and highlight the analytical capabilities of the various techniques/methods considered. Since there are too many food samples in Table 1, they are divided into eight types: mammals (including beef, pork and sheep), birds (include chicken and duck), fish, egg, milk, plants (including lettuce, soybean, rice, cabbage and apple), natural water and bacterial solution for easier analysis.

Figure 2a indicates that the percentage of articles include multiplex detection of bacteria simultaneously by PCR, LFIA and electrochemical methods. It is a little more than half (52%) in PCR, but only around one fifth (22%) for LFIA, and one tenth (10%) for electrochemical methods. Figure 2b shows the percentage of articles according to food sample type with varying detection methods. Although milk was always included in most articles (13, 17, 22 in PCR, LFIA, and electrochemical methods respectively), the second is mammals in PCR and LFIA (12 and 11 respectively) but plants mainly feature for electrochemical methods (11). In addition, more common food samples in every detection method accounts for more than half of the articles together. On the other hand, articles about natural water and bacterial solution only appear in PCR, not LFIA or electrochemical methods.

Figure 3. Timeline of the annual number of articles collected and the annual average limit of detection in different years by different detection methods. (a) Number of articles. (b) Average limit of detection.

Figure 3. Timeline of the annual number of articles collected and the annual average limit of detection in different years by different detection methods. (a) Number of articles. (b) Average limit of detection.

Figure 3a presents the annual number of articles and average limit of detection over the years 2013 to 2023 by different detection methods. There was at least one article published every year for every detection method from 2013 to 2023. Articles published in 2019 to 2023 are always higher than articles published in 2013 to 2018 in every detection method indicating increased research interest. For PCR, the annual number of articles published was 5 in 2013, then it decreased to 2 in 2014. It increased gradually to 7 in 2018 and decreased to 3 in 2019. After 2019, fluctuations were observed every year. It reached the highest point of 10 in 2020 and decreased to the bottom at 2 in 2021. In the case of LFIA, it was only 1 in 2013. It increased gradually to 8 in 2018 and highly fluctuated every other year. It reached 8 again in 2019 and decreased to the lowest level at 4 in 2021. In the case of electrochemical analysis, it also started with 1 in 2013. It increased sharply to 3 in 2014 and decreased to 2 in 2016. Then it increased gradually to 9 in 2022, followed by a large decline to 5 in 2023.

In Figure 3b, the annual average LOD was usually the highest in LFIA, except for the electrochemical methods in 2018. Overall, the LOD was usually the lowest in PCR, except for the electrochemical methods in 2016 and 2022. In PCR, the average LOD was about 4 CFU/ml in 2013. Then it increased gradually to around 18 CFU/ml in 2016, and was followed by a large drop to roughly 3 CFU/ml in 2017. After 2017, fluctuations were observed every year and decreased in an overall trend to about 2 CFU/ml in 2021 (also the lowest for all detection methods in every year). It increased to about 6 CFU/ml in 2022. In LFIA, it was around 10 CFU/ml in 2013, it decreases to roughly 7 CFU/ml in 2014 and increased tremendously to about 75 CFU/ml in 2016 (also the highest for all detection methods in every year). Then it decreased sharply to about 17 CFU/ml in 2017 and was followed by an increase to nearly 40 CFU/ml in 2019 again. After that, it decreases gradually to approximately 8 CFU/ml in 2021 and increased again to roughly 25 CFU/ml in 2023. By the electrochemical method, it was about 4 CFU/ml in 2013, and it increased gradually to around 10 CFU/ml in 2015. After a decrease to about 7 CFU/ml in 2016, it increased again to around 35 CFU/ml in 2018. Then it decreases gradually to roughly 3 CFU/ml in 2022.

Figure 4. Detection of bacteria with LFIA with different nanoparticles (white: Gold, red: Iron; dark blue: Europium; green: Palladium; light blue: Silicon; pink: Carbon; yellow: Cobalt; brown: Manganese). (a) Number of articles with different nanoparticles. (b) Average LOD with different nanoparticles.

Figure 4. Detection of bacteria with LFIA with different nanoparticles (white: Gold, red: Iron; dark blue: Europium; green: Palladium; light blue: Silicon; pink: Carbon; yellow: Cobalt; brown: Manganese). (a) Number of articles with different nanoparticles. (b) Average LOD with different nanoparticles.

Figure 4a shows that gold is the most studied nanoparticles in articles with LFIA involving 33 articles, and it is followed by iron involving 8 articles. These two nanoparticles account for more than four fifths of the articles on LFIA. Only 2 articles involving europium, palladium, and silicon. Only 1 article involve carbon, cobalt, and manganese. Figure 4b illustrates that the average LOD is the highest for articles involving silicon (little over 50 CFU/ml), followed by carbon and palladium of roughly 40 CFU/ml. The average LOD for gold is about 26 CFU/ml, which is a little higher than the average LOD for all articles with LFIA (24 CFU/ml). It is followed by iron of around 12 CFU/ml, cobalt of 10 CFU/ml, Manganese of about 9 CFU/ml and Europium of only 4 CFU/ml.

Figure 5. Detection of bacteria by electrochemical method with different techniques (white: cyclic voltammetry (CV); red: differential pulse voltammetry (DPV); dark blue: square wave voltammetry (SWV); green: anodic stripping voltammetry (ASV); light blue: electrochemical impedance spectroscopy (EIS)). (a) Number of articles by different techniques. (b) Average LOD by different techniques.

Figure 5. Detection of bacteria by electrochemical method with different techniques (white: cyclic voltammetry (CV); red: differential pulse voltammetry (DPV); dark blue: square wave voltammetry (SWV); green: anodic stripping voltammetry (ASV); light blue: electrochemical impedance spectroscopy (EIS)). (a) Number of articles by different techniques. (b) Average LOD by different techniques.

Figure 5a shows that electrochemical impedance spectroscopy (EIS) accounts for the highest number of articles of 20, and it is followed by cyclic voltammetry (CV) of 14, differential pulse voltammetry (DPV) of 13, and these techniques occupy the majority (more than nine tenths) of articles in electrochemical methods. In addition, only 2 articles involve square wave voltammetry (SWV) and only 1 article involves anodic stripping voltammetry (ASV). Figure 5b illustrates that the average LODs for all articles, articles involving EIS and voltammetry have similar LOD at around 12 CFU/ml. The average LOD for CV is the highest at roughly 18 CFU/ml. It is followed by ASV of 15 CFU/ml, DPV of around 8 CFU/ml, and SWV of only 6 CFU/ml.

Figure 6. Average limits of detection of different bacteria and food samples by different detection methods. (a) Different bacteria. (b) Different types of food samples.

Figure 6. Average limits of detection of different bacteria and food samples by different detection methods. (a) Different bacteria. (b) Different types of food samples.

Figure 6 presents the average LOD of different (a) types of bacteria and (b) types of food/ water samples by different detection methods. Figure 6a shows that the overall average LOD is lowest for PCR, and highest for LFIA. The average LOD is higher for multiplex detection of bacteria simultaneous in PCR than single detection ones, but lower in LFIA and electrochemical methods. PCR has the lowest average LOD among all detection methods for Salmonella, Campylobacter and E. coli and these bacteria are all gram negative (-). In addition, electrochemical methods have the lowest average LOD among all detection methods for Listeria and S. aureus and these bacteria are all gram positive (+). On the other hand, LFIA always has the highest average LOD for every type of bacteria. The average LOD for bacteria in the Pseudomonadota domain are usually lower than bacteria in the Bacillota domain by PCR and LFIA, but similar to the latter by electrochemical method. For bacteria in the Pseudomonadota domain, the average LOD for E. coli is lower than it is for Salmonella by PCR and electrochemical methods, but higher than the latter by LFIA. For bacteria in the Bacillota domain, the average LOD for Listeria is lower than it for S. aureus by PCR and electrochemical method, but higher than the latter by LFIA. The average LOD of Campylobacter is usually the highest among all bacteria by every detection method, except that it is lower than S. aureus by PCR. In addition, only average LODs of Campylobacter LFIA and electrochemical methods are over 30 CFU/ml by all detection methods and among all bacteria.

Figure 6b shows that PCR has the lowest average LOD among birds, fish, milk, while the electrochemical methods show the lowest average LOD among mammals, egg and plants. LFIA always has the highest average LOD among all food samples, except that it is highest by PCR in egg. Among all food samples, egg has the lowest average LOD by LFIA and electrochemical methods among all food samples, while fish has the lowest average LOD by PCR (the lowest among all foods samples by all detection methods, also the only average LOD lower than 1 CFU/ml). In contrast, the data for birds demonstrated the highest average LOD by PCR, which is followed by mammals and milk (only these three have average LODs over 20 CFU/ml by all detection methods and among all bacteria). In addition, mammals exhibited the highest average LOD by PCR and milk the highest average LOD by PCR by electrochemical method among all food samples. Natural water and bacterial solution only show in detection by PCR, and their average LODs are lower than all other food samples by all detection methods except fish by PCR.

EU limits for Salmonella, Listeria, Campylobacter, S. aureus and E. coli in food for most people are 100, 100, 1000, 1000, 100 CFU/mL, respectively [21]. All the LODs in this review are far lower than the EU limits stipulate. Although some food intended for special groups such as infants, elderly people and patients with certain diseases require no presence of these bacteria at all, at least one article with LOD within 0.3 CFU/ml is included in every type of bacteria by PCR [21]

4. Discussion

The review of the LODs for PCR, LFIA and electrochemical methods has revealed trends in this research area that will inform food safety and public health experts. Figure 2a illustrates that the percentage of articles with multiplex is the highest in PCR, followed by LFIA and electrochemical method. That is the main reason that PCR is considered as a reliable standard detection method in the detection of bacteria under many circumstances. However, PCR does have disadvantages of high cost, time-consuming and complex procedure. As a result, PCR cannot replace LFIA and electrochemical method in the detection of bacteria completely. Milk is the most popular food sample for bacteria detection by every detection method. In addition, many articles involve mammals, plants and birds in every detection method (Figure 2b).

The results in Figure 3a show that more articles were published in 2019-2023 than in 2013-2018 by every detection method. In addition, the number of articles every year by every detection fluctuates highly. While the number of articles in LFIA and electrochemical method reached the peak in 2022 (8 and 9 articles respectively), articles in PCR reached the peak in 2020 of 10 articles. While the number of articles in LFIA and electrochemical method reached the lowest with only 1 article in 2013, articles in PCR reached the lowest in 2014 and 2015 in every year of 2 articles.

In Figure 3b, the annual average LOD was usually the highest by LFIA, and the lowest by PCR in every year. While the annual average LOD for PCR and LFIA reached the first peak in 2016 (about 75 CFU/ml and 18 CFU/ml respectively), the LOD for electrochemical method reached the first peak in 2015 of around 10 CFU/ml. Later, the annual average LOD for PCR and electrochemical method reached the second peak in 2018 (around 8 CFU/ml and 35 CFU/ml respectively), the LOD for LFIA reached the second peak in 2019 of around 40 CFU/ml. And the annual average LOD by PCR reached the lowest of only about 2 CFU/ml in 2019, and it was around 7 CFU/ml in 2014 by LFIA, and 3 CFU/ml in 2014 by detection method. The relationship between the annual average LOD and the year by every detection method is still unclear. Figure 3 shows that although the detection of bacteria has attracted more attention from researchers in recent years, LOD in PCR, LFIA and electrochemical method did not a decreasing trend. The main reason is that LODs in these articles are all lower than the EU limits, so the main purpose in many articles may not be lowering the detection limit.

Figure 4a shows that only 3 articles by LFIA involve nonmetal nanoparticles (silicon: 2, carbon: 1). The majority of articles with LFIA involve metal nanoparticles. Figure 4b illustrates that average LODs for nonmetal nanoparticles are usually higher than metal nanoparticles, except that the average LOD of palladium is little higher than carbon.

Figure 5a shows that EIS, CV, DPV account for the most articles by electrochemical method. Figure 5b illustrates that among these three main techniques in electrochemical method, average LOD is highest for articles with CV (18 CFU/ml), followed by EIS (12 CFU/ml) and DPV (8 CFU/ml). The main reason is that 2 articles with LODs of 100 CFU/ml for the detection of Campylobacter involve CV, and 1 article with a LOD of 100 CFU/ml for the detection of Campylobacter involve EIS. It also shows that few articles for the detection of Campylobacter of extremely LODs increase average LODs of different techniques by electrochemical methods.

In Figure 6a, the average LOD was usually the highest for LFIA in bacteria investigated. The average LOD was the lowest by PCR in gram (-) bacteria, and by electrochemical method in gram (+) bacteria. Campylobacter is gram (-), and its average LOD is usually the highest among all bacteria by every detection method, except it was lower than S. aureus by PCR. The recommended EU limits for Salmonella, Listeria, Campylobacter, S. aureus and E. coli in food for most people are 100, 100, 1000, 1000, 100 CFU/ml, respectively. All the LODs in this review are far lower than these EU limits. A possible reason for the relative higher average LOD of Campylobacter and S. aureus is that EU limits for them are relatively high. The average LOD for Salmonella and E. coli (both gram (-)) in the Pseudomonadota domain are usually lower than Listeria and S. aureus (both gram (+)) in the Bacillota domain by PCR and LFIA, but similar to the latter by electrochemical method. The difference between 2 bacteria in the same domain is much smaller than the difference between 2 different domains.

This review also shows that the average LOD for multiplex ones is higher than non-multiplex ones by PCR, but lower than them by LFIA and electrochemical methods. One of the main possible reasons is that PCR usually has lower LOD than LFIA and electrochemical method. It is difficult to keep both detection efficiency and sensitivity at the same time when LOD is already really low, but not that difficult when LOD is relatively high. This could be a promising focus for the development of bacteria detection in the future. This review also indicates that fish and egg have the lowest average LOD among all types of food samples. One of the main reasons is that bacteria in fish and egg can be distinguished FROM nearby animal cells. The complexity of food sample composition can lower the performance of detection and makes LOD higher. To address such limitations and challenges, sample enrichment and improvement in device properties of detection are needed. PCR, LFIA and electrochemical method have been used in detection of different bacteria, and many of them involve multiplex detection. It is often observed that different types of bacteria coexist in a single food sample. As a result, a multiplex detection that can fulfill the requirements of a low detection limit and high efficiency is necessary for food safety. These detection methods can also be combined with other technologies to obtain a better detection performance.

5. Conclusions

The development of detection technology for monitoring the quality and safety of foods has provided promising tools for improved quantitative performance. In order to improve the accuracy and precision of different detection methods (PCR, LFIA and electrochemical method), different parameters such as bacteria type, year of article, detection type: multiplex or not, food sample type, number of replicates have been considered as determinants of LOD. The results show that bacteria type and food sample type strongly contribute to predict the LOD. Average LOD is the highest for the detection by LFIA (24 CFU/ml), followed by electrochemical method (12 CFU/ml) and PCR (6 CFU/ml). Salmonella and Escherichia coli in the Pseudomonadota domain usually have lower LODs than other bacteria. LODs are usually lower for detections in fish and egg than in other food samples analyzed. Most articles by LFIA involve metal nanopaticles-especially gold and iron. The average LOD of articles involving gold (26 CFU/ml) is higher than it of iron (12 CFU/ml). EIS, CV and DPV are three major techniques among articles by electrochemical method. CV has the higher average LOD (18 CFU/ml) than EIS (12 CFU/ml) and DPV (8 CFU/ml). Sample enrichment and improvement in device properties of detection and the possibility of combination with other detection technologies are needed to lower LOD and improve performance of detection further. This review provides guidance for future developments of bacteria monitoring technologies, based on the enrichment of bacteria from samples, and the development of multiplex detection methods that can increase the detection efficiency but also keep the detection limit low. The integration and exploration of novel nanomaterials will help to further improve the detection limits (LODs) and advance the capabilities of bacterial detection technologies in the realm of food safety.