Molecular Methods for the Detection of Toxoplasma gondii Oocysts in Fresh Produce: An Extensive Review

Human infection with the important zoonotic foodborne pathogen Toxoplasma gondii has been associated with unwashed raw fresh produce consumption. The lack of a standardised detection method limits the estimation of fresh produce as an infection source. To support method development and standardisation, an extensive literature review and a multi-attribute assessment were performed to analyse the key aspects of published methods for the detection of T. gondii oocyst contamination in fresh produce. Seventy-seven published studies were included, with 14 focusing on fresh produce. Information gathered from expert laboratories via an online questionnaire were also included. Our findings show that procedures for oocyst recovery from fresh produce mostly involved sample washing and pelleting of the washing eluate by centrifugation, although washing procedures and buffers varied. DNA extraction procedures including mechanical or thermal shocks were identified as necessary steps to break the robust oocyst wall. The most suitable DNA detection protocols rely on qPCR, mostly targeting the B1 gene or the 529 bp repetitive element. When reported, validation data for the different detection methods were not comparable and none of the methods were supported by an interlaboratory comparative study. The results of this review will pave the way for an ongoing development of a widely applicable standard operating procedure.


Introduction
Toxoplasmosis is a zoonotic parasitic disease caused by the protozoan Toxoplasma gondii (T. gondii; [1]). The clinical manifestations of toxoplasmosis in humans, including congenital, cerebral and ocular toxoplasmosis, cause a substantial disease burden worldwide [2,3] Moreover, T. gondii can also cause clinical disease in its animal hosts, resulting in major losses in livestock industry and lower welfare for the affected animals [1].
It has been estimated that 42-61% of acquired toxoplasmosis cases are foodborne [4]. The food-and waterborne transmission routes of T. gondii are numerous, including ingestion of infective tissue-dwelling stages of the parasite in raw or undercooked meat of infected animals and ingestion of oocysts, shed by infected felines and sporulated in the environment, in contaminated water or food, such as fresh produce (fruits, vegetables, and juice) [5].

Materials and Methods
We searched two online databases, PubMed and Scopus, for all potentially relevant records on molecular methods applied to detect T. gondii oocysts, irrespective of the matrix, following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines where applicable [10] (Figure 1). The search terms were grouped into 14 combinations (Supplementary File S1). The databases were searched for records in English that were published up to 12 February 2020. Publications were initially screened by three independent reviewers for eligibility based on title and abstract. Then, records were excluded if they were: (i) letters, editorials, notes, comments, and reviews; (ii) studies describing methods not applicable to T. gondii or (iii) methods using reagents that are not widely available (e.g., antibodies). Full texts of the records were screened by six independent reviewers. Records were excluded at this stage if (i) full text was unavailable or if (ii) the study did not describe methods applicable to molecular detection of T. gondii oocysts.
For data extraction, we focused on three key steps of the detection: oocyst recovery, DNA extraction and DNA detection. From each eligible record, data were extracted in predefined tables (Supplementary File S1). For the oocyst recovery step, only data extracted from studies using fresh produce as matrix were included in the final analysis and assessment of the data. Experimental contamination (spiking) studies were considered to provide information of great relevance for the oocyst recovery step, as they are performed under controlled experimental conditions. Data on the three main steps were analysed independently.
The extracted data were complemented with the output from a survey conducted in February 2020 among 24 expert laboratories with experience in T. gondii detection in food and non-food matrices [11]. The questionnaire collected information about current practices, details and facilities for molecular testing for T. gondii in different matrices, as well as expert opinions on methods for molecular detection of T. gondii oocysts.
In the included spiking studies, the matrices used were berries (strawberries, raspberries, blackberries, blueberries and cranberries), leafy greens (basil, lettuce, spinach, cilantro, dill, mint and parsley) and other vegetables (radish, thyme and green onions) ( Table 1). The sample amount used ranged from 10 g to 60 g (Table 1). For spiking, six studies used a dripping method (pipetting the oocyst suspension onto the matrix surfaces, mimicking vegetable contamination by irrigation), while two studies used an immersion method (i.e., the material was immersed in water containing a known amount of oocysts) ( Table 1). A post-spiking incubation to allow oocyst adherence after applying a dripping method ranged from 30 min to overnight, and temperatures used were room temperature and +4 • C. Oocysts of T. gondii are highly resistant to environmental conditions but do not multiply in the environment; consequently, the first step for parasite detection requires parasite enrichment by parasite recovery and concentration. All procedures for oocyst recovery from leafy greens involved washing and pelleting of the washing eluate by centrifugation. Additional steps reported prior to centrifugation included overnight flocculation using CaCO 3 solution, filtration through cellulose ester membrane and flotation using Sheather's sucrose solution (Table 1). Limited information on the impact of these additional steps on recovery rate could be extracted. Despite the fact that it was not considered in our literature review effort, it is worth mentioning that even the use of a immunomagnetic separation (IMS) step with non-commercially available in-house anti-T. gondii oocyst monoclonal antibodies did not result in any improvement of the recovery rate (as quantified by qPCR) [13]. Filtration, flocculation and flotation might partially replace centrifugation, which can be a limiting factor when using large volumes of wash buffer or a large number of samples [16][17][18][19]. Although flocculation and flotation might also reduce soil particles and other contaminants that could potentially inhibit DNA amplification [16][17][18][19], one study underlined that if flotation is used, residual Sheather's solution could inhibit downstream PCR reactions [17]. Washing of vegetables was performed either manually, using an automatic horizontal orbital shaker from 15-30 sec to 60 min, or by stomaching ( Table 1). The most commonly used washing buffers were aqueous solutions of 1 M glycine pH 5.5 (four studies) or 0.1-1% Tween 80 (four studies) in the range of 4-6 mL of washing buffer per gram of sample (Table 1). Two studies compared different washing methods: stomaching of leafy herbs contaminated with E. papillata oocysts with 1 M glycine pH 5.5 buffer provided a higher recovery percentage than horizontal orbital shaking did, whereas for spinach spiked with heat-inactivated T. gondii oocysts, manual shaking with 0.1% Tween 80 was more effective than stomaching [15,17]. One of the studies highlighted that washing buffers containing a surfactant or detergent are not recommended for stomaching, since the bubbles produced during the homogenisation seemed to interfere with oocyst recovery [17]. An evaluation of washing buffers was reported in two studies, both using leafy greens but different washing protocols (manual washing vs automatic shaking or stomaching): Both studies reported that 1 M glycine pH 5.5 performed better than PBS [12,17]. Among the questionnaire survey participants who reported testing fresh produce, stomaching, manual washing and pelleting by centrifugation were the most common sample processing techniques [11]. The reported washing buffers were similar to those reported in the published literature. In the six prospective surveys included, a larger variety of leafy greens were tested, including mixed salads (Table 2). One study [19] used a large amount of the tested sample, up to 1 kg, as well as a large volume of washing buffer (≥2 L) followed by flocculation. In four studies, the amount of tested samples ranged from 35 g to 100 g, with an average volume of washing buffer of 2 mL/g of sample [18,21,22,24]. The washing buffers used in these four studies contained either Tween-80 (0.1-1%) or glycine. Three studies reported washing using an automatic shaker for 15-20 min or up to 2 h, followed by centrifugation [17,19,22]. Additionally, in the largest survey with over 1000 samples, 35 g of sample was tested, washing was done with 200 mL of 1 M glycine pH 5.5 using an orbital shaker or stomacher, and oocysts were recovered by centrifugation and flotation with Sheather's sucrose solution [21]. Prior to DNA extraction, a step to break the wall of the oocysts was included in all spiking studies (Table 1). Bead-beating (BB) using a commercial mix of beads in combi-nation with a high-speed mechanical homogenizer, with single or double cycles at speeds in the 4-6.5 m/s range for 30 s to 2 min, were used in four studies [12,14,19,20]. The freeze and thaw (FT) method, with 1 to 10 cycles, temperature ranges from −196 • C to 100 • C, and incubation times of 1-5 min, was used in four studies [13,15,16,18]. FT was as effective as to ultrasound (US), when compared to no pre-treatment [16]. In two studies, US or incubation with proteinase K at 56 • C was used as an additional step after FT cycles [13,17]. Three spiking studies on non-vegetable matrices evaluated the performance of different DNA extraction procedures for T. gondii oocysts including BB and/or FT [25][26][27] (Table 3). Table 3. Studies comparing the performance of freeze and thaw cycles vs bead-beating as pre-treatment procedure to DNA extraction from Toxoplasma gondii oocysts.  Use of commercial kits, including sample homogenisation by BB, performed better than the procedure using sedimentation/flotation in combination with FT followed by in-house phenol-chloroform extraction and the DNA extraction kit without BB, even when an additional step using glass beads was included [25]. In one study, the combination of BB, FT and proteinase K treatment together with a commercial DNA extraction kit showed higher sensitivity than vortexing and BB followed by DNA extraction using another commercial kit [26]. One study suggested that increasing the number of FT cycles did not enhance oocyst DNA detection and may have resulted in decreased sensitivity due to DNA degradation [27]. According to the questionnaire results, BB was used in the majority of the participating laboratories for testing of fresh produce [11]. When all the included studies where considered, 15 reported on the use of BB for DNA extraction from T. gondii oocysts [14,19,20,[24][25][26][28][29][30][31][32][33][34][35] (Supplementary File S1) with two commercial kits most frequently used (Supplementary File S1). FT associated with silica spin-column kits was reported in 58 studies (Supplementary File S1). Concerning molecular detection, conventional PCR (cPCR) was used in 17 of the 77 reviewed studies (22%), nested or seminested PCR in 20 studies (26%) and two papers reported using loop-mediated isothermal amplification (LAMP) (Supplementary File S1). For the cPCR assays, most studies targeted the B1 gene or the 529 bp repetitive element (529RE) ( Table 4). Six studies compared the sensitivity of cPCR targeting B1 vs. 529RE, expressed as the limit of oocysts providing a positive amplification (Supplementary File S1). For fresh produce, B1-cPCR was shown to be 10 times more sensitive than 529RE-cPCR [16], with a limit of detection (LoD) of 10 and 100 oocysts/heads of lettuce, respectively. For soil and faeces, the results were the opposite [36][37][38]. The sensitivity of the 529RE-cPCR is also affected by the efficiency of the DNA extraction method [25] and reducing amplicon size was beneficial [38]. Sensitivity appeared generally higher in water or DNA-poor matrices than in complex matrices [e.g., 26]. NR faeces [40] 1 oocyst in water and mussel hemolymph mussels [26] 100 oocysts in mussel gills and dig. Glands mussels [26] NR oysters [ NR soil [36] Assays relying on qPCR accounted for almost 50% of studies (38 studies) and were mainly qualitative, with Taqman assays targeting the 529RE, which was the most often applied (Table 5 and Supplementary File S1).

Discussion
Detection of T. gondii in vegetables is challenging due to the low sensitivity of existing detection methods. This also holds true for other foodborne parasites (e.g., Cryptosporidium spp. and Giardia duodenalis) [65]. As oocysts of T. gondii are highly resistant to environmental conditions and do not multiply in the environment, oocyst recovery from fresh produce is the first and key step to enable successful detection. Molecular detection must then rely on efficient DNA extraction from the robust oocysts, together with a reduction of possible contaminants that could inhibit the DNA amplification. Finally, amplification must be specific and sensitive to detect DNA from low numbers of oocysts, ideally a single oocyst, avoiding any cross-amplification with closely related species.
As shown in this review, many different methods have been described for each step of the molecular detection of T. gondii oocysts and different combinations of them have been used to analyse fresh produce as well as other matrices. This variability, which was also evident in the results of the questionnaire survey [11], prevents a direct comparison of the studies to identify the most promising method for a sensitive and reliable detection of T. gondii oocysts in fresh produce (as well as in other matrices). Although specific characteristics of different vegetable matrices can interfere with oocyst recovery due to e.g., trapping and adhesion force and, later on, with molecular detection (i.e., different concentrations of PCR inhibitors), the overall molecular detection procedure should be harmonised and standardised. The oocyst recovery step from fresh produce is particularly important but challenging to standardise due to a large variability in the reported methods (e.g., washing procedure, washing buffers and oocyst concentration). For instance, stomaching with an appropriate setting of homogenisation power and speed to account for brittleness of the vegetable samples, would be a fast procedure to apply for large scale analysis and easy to standardize. Due to the presence of high amounts of natural detergents in some types of fresh produce (e.g., saponins in spinach), the use of washing buffers with detergent (i.e., Tween-80) might not be recommended as they could exacerbate foaming and potentially trap oocysts in the foam, thus lowering the recovery rate. The 1 M glycine solution is potentially the buffer of choice, as it is inexpensive and did not generate an excess of debris during stomaching of lettuce as the sample matrix [66], which could eventually interfere with downstream oocyst concentration and DNA extraction. Although oocysts concentration by centrifugation might be time consuming and require a centrifuge, other procedures might be more complicated or less efficient. For instance, flocculation of water samples with Fe 2 (SO 4 ) 3 resulted in PCR inhibition [65]. The risk of oocyst loss following NaNO 3 flotation was highlighted in one study on soil samples [34], suggesting that NaNO 3 flotation is suitable when oocyst contamination is ≥103/40 g soil. One paper discussed that while flocculation is simple and inexpensive, filtration is more robust for processing turbid wastewater (and possibly the washing suspensions of vegetables), and PCR inhibitors appeared to be eliminated by using 1-µm pore-sized polyethersulfonate membrane filters [43]. Additionally, filtration would be preferable when large volumes (litre) of a sample need to be processed.
The reported DNA extraction protocols substantially differ in their approach to break the robust oocyst wall (FT, US and BB), whereas further DNA purification and clean up from inhibitors are mostly performed using silica-column-based DNA extraction kits. Although FT does not require expensive equipment, in contrast to the use of a bead beater, the choice of the most promising and efficient FT procedure is difficult due to the large variability of settings applied in different studies (e.g., length and number of reported freeze and thaw cycles were quite different). Moreover, the requirement of several cycles of FT is time consuming especially when a large panel of samples is tested. According to most of the manufacturer's protocols, kits using pre-packed silica spin columns allow the use of only a fraction of the supernatant obtained from the initial sample lysis per single extraction. This might lead to a considerable loss of material and reduction of the final assay sensitivity, as either only a portion of the original sample is used for the DNA extraction step, or might require multiple DNA extraction from the same sample with consequent increase in assay time and costs. Commercial kits including a mechanical disruption step (e.g., BB) have already been successful in detecting Hammondia spp. and T. gondii oocysts by PCR with a high sensitivity [25]. Furthermore, they have the advantage of using larger sample volumes without substantial adaptation of the kit that are loaded with a silica matrix onto empty columns and could, therefore, favour a higher assay sensitivity. Whether the performance of different commercial kits based on BB is comparable or not was not specifically assessed in any of the papers included in this study, but might be presumed by the comparability of two kits tested in Herrmann et al., 2011 (specifically NucleoSpin Soil from Marcherey-Nagel vs ZymoResearch fecal DNA Kit from Zymo) [25]. However, since available kit formulations and producers might change over time and in different countries, kit performance should always be evaluated prior to a study, in order to select the most suitable kit.
For the purpose of this review, we did not further consider nested-PCR and LAMP (loop-mediated isothermal AMPlification) assays as suitable for routine testing of fresh produce. Despite their higher sensitivity and specificity compared to conventional PCR, both techniques suffer from a high risk of background and cross-contamination, and nested PCR requires two consecutive rounds of amplification. Concerning the reported molecular assays, qPCR targeting either the B1 gene and/or the 529RE both provide a very high sensitivity, due to multiple copies of both targets in the T. gondii genome. Although double-strand DNA-intercalating fluorescence dyes (e.g., SYBR Green) combined with melting curve analysis (MCA) are relatively cost beneficial and easy to use, duallabeled TaqMan probes have the advantage of combining detection with confirmation of the amplification products without the need for further amplicon sequencing. It should be noted that the specificity of the amplification product can be of concern, especially when targeting 529RE, due to potential cross amplification with parasites closely related to T. gondii (i.e., Hammondia hammondi, Sarcocystis spp. Neospora caninum) [38,60].
PCR inhibitors are important confounders that must be addressed in any PCR-based detection effort. Molecular detection of pathogens in food can be challenging due to a large variety of PCR inhibitors that can be co-extracted with DNA. Especially, DNA extracts from pelleted washing suspensions of plant-based food may contain diverse PCRinhibiting substances from debris of the plants themselves (e.g., phenols, polyphenols, polysaccharides), but also from residual soil or irrigation water components (e.g., humic and fulminic acids) [67]. Depending on the food matrix and type and mechanism of inhibitory substances, different strategies can be evaluated to decrease their concentration in the sample or to reduce their inhibitory effect by e.g., using less-sensitive polymerases or specific PCR additives (e.g., BSA, DMSO) [67]. It should be noted that, for the detection of PCR inhibitors and to exclude false-negative results, the use of an internal amplification control (IAC) is mandatory for diagnostic PCR detection of foodborne pathogens according to CEN/ISO 22174 [68]. Competitive IACs are synthetic oligonucleotides that are amplified with the same set of primers as the target gene. Although they are amplified under the same conditions and thus mimic the amplification of the target gene, they also have a stronger potential to reduce the assay sensitivity and may require more optimization work [69]. As low sensitivity and inhibition is already an issue when analysing fresh produce for T. gondii, we rather propose the application of non-competitive IACs, ideally as a synthetic sequence with no homology with either the target parasitic DNA or with the matrix, as for example used in the US FDA-BAM 19b for "Molecular Detection of Cyclospora cayetanensis in Fresh Produce Using Real-Time PCR" [9]. As these non-competitive IACs are amplified with a different set of primers, they can universally be applied in different PCR detection systems and have the advantage of generally not competing with the target amplification, when used in low concentrations and with limiting primer concentrations.
We would like to stress that for any published qPCR assay, it is important to report the performance characteristics according to the MIQE guidelines [70][71][72]. This includes: (i) use of an IAC to check for PCR inhibition; (ii) preparation of a standard curve (10-fold serial dilution of at least five template concentrations) with background matrix (e.g., pelleted washing suspensions from uncontaminated food matrix); (iii) evaluation of amplification efficiency and linearity with a R 2 value (ideally ≥0.98); (iv) determination of the LoD95%, supported by spiking studies.
A standardized procedure to be applied for the detection in fresh produce would not only be desirable for T. gondii but also for other foodborne protozoan parasites. Of course, implementation of slightly different methods might be necessary to reliably detect the target parasite. The problems associated with the availability of a large number of laboratory methods for pathogen detection are manifold. If prevalence data are not comparable from different regions or countries, they might result in inaccurate risk assessment conclusions. For instance, if quantification is required, it is necessary to define the quantified target (DNA amount, number of oocysts, target copy number) as well as a standardized and harmonized procedure to convert this to equivalent numbers of oocysts (indeed oocysts load is the data that food stakeholders might expect). This is exacerbated by the fact that the large majority of published methods are not or insufficiently validated. Although ISO standards for the validation of parasitic methods are currently not available, method validations can be based on a number of available documents [70,[73][74][75][76][77][78][79].
Noteworthy, none of the articles reviewed reported on any attempt to evaluate the applied methodology through an inter-laboratory comparison. Results of ring trials are an important indicator for the inter-assay precision (reproducibility) of a method and an essential step for the better understanding of the method characteristics. As already described for the validation of microbiological methods in the food chain, inter-laboratory comparisons should also be performed as part of the validation of parasitological methods. In light of the increasing internationalisation of food supply chains, the need for conclusive data to better understand food-borne transmission or to provide a solid basis for risk assessments, has increased. For this, robust, validated and standardized laboratory methods for the detection of contamination of food sources with a high level of confidence are essential. This is especially important for T. gondii, a highly prioritized zoonotic foodborne pathogen, where laboratories are currently using a multitude of different diagnostic approaches.