Multiplex vs. singleplex assay for the simultaneous identification of the three components of avian malaria vector-borne disease by DNA metabarcoding

Introduction

Accurate detection and identification of vector-host-parasite interactions is crucial for understanding the complex evolutionary dynamics of these systems and for managing prevention campaigns (Almeida et al., 2024; Casman & Dowlatabadi, 2010; Martinez et al., 2021). Different methodological approaches have been used to analyse this tripartite interaction, traditionally focusing on arthropod blood meals, using serological techniques such as enzyme-linked immunosorbent assay or precipitin tests (e.g., Beier et al., 1988; Tempelis, 1975; Washino & Tempelis, 1983). However, in these studies taxonomic resolution was low (Lardeux et al., 2007) and the detection was limited to pre-selected taxa. Over the past twenty years, blood meals (or digestive parts) of various blood-feeding invertebrate taxa have been analysed with molecular techniques (i.e., polymerase chain reaction, PCR) to better characterise these interactions (reviewed in Borland & Kading, 2021). This has been specifically conducted for the identification of either parasites (e.g., Ciloglu et al., 2019; Szentiványi et al., 2020), hosts (e.g., fed on by leeches (Abrams et al., 2019; Schnell et al., 2018; Weiskopf et al., 2018)), triatomines (Balasubramanian et al., 2022), midges (Cutajar & Rowley, 2020), sandflies (e.g., Kocher et al., 2017) or mosquitoes (e.g., Estrada-Franco et al., 2020), hosts and parasites (e.g., Dumonteil et al., 2018) or vectors and parasites (Makunin et al., 2022). Often based on Sanger-sequencing, these approaches fail to distinguish mixed blood meals or parasite co-infections, and require a phasing step (or a nested PCR protocol). Furthermore, most of these assays target relatively large-size DNA barcodes (>250 bp; e.g., Alcaide et al., 2009), preventing the successful amplification of degraded samples and/or with very low amount or degraded DNA of the target species.

Until recently, singleplex PCR assays (the use of a single primer pair per PCR reaction) were the only way to detect the components of a vector-borne system, i.e., by performing separate PCR reactions for the host, vector and parasite. Some studies have already used singleplex to identify the host-vector-parasite triad, e.g., using blackflies (Chakarov et al., 2020) or triatomines (Dumonteil et al., 2018).

Nowadays, advances in primer design, next-generation sequencing and the use of DNA metabarcoding approaches have opened the possibility for identifying vector-borne interactions in a single PCR reaction combining multiple primer pairs, i.e., multiplex PCR, which greatly reduces the laboratory costs and time. Currently, the application of multiplexing is however limited, as there is a general lack of multiplexing routine among the community and of its potential for downscaling costs. However, the use of in silico PCR (i.e., computational tools to simulate PCR amplification) has simplified the application of multiplex, facilitating the design of new primer pairs to be both taxonomically informative and compatible in terms of PCR reaction conditions. A priori, singleplex PCR is easier to design and is expected to result in higher detection scores of the target species, because only one target is amplified per reaction without potential primer competition during the PCR reaction, and conditions are optimised for only one primer pair. When performing multiplex PCRs, the initial concentration of DNA from each target taxonomic group may compromise the successful amplification with primer pairs aimed at less abundant taxa. This could occur either due to the consumption of PCR reagents by one or a few primers, or because disproportionate amplification by one primer pair saturates the sequencing clusters.

In this proof-of-concept study, we used short DNA metabarcodes for the amplification and high-throughput sequencing of partially digested mosquito blood meals in an avian malaria framework as a case study system. This vector-borne disease is caused by apicomplexan parasites belonging to the Plasmodium genus, which are transmitted in birds by mosquitoes. For the first time, mosquito blood meals were used here to simultaneously identify, using multiplex DNA metabarcoding, the host-vector-parasite triad in a tripartite vector-borne system.

In order to improve the identification methodology, we compared the detection probabilities by singleplex and multiplex DNA metabarcoding of the three components of this system and the differences in their taxonomic assignments. Studies quantifying the detection differences between singleplex and multiplex reactions are still rare (Eastwood et al., 2024). As such, expanding our knowledge on which factors determine the successful detection of the components of vector-host-parasite interactions comparing singleplex and multiplex approaches will help to develop more cost-efficient laboratory procedures and improve medical diagnosis of vector-borne diseases, whose transmission and dispersion into non-endemic areas are exacerbated by current climate change (Thomson & Stanberry, 2022).

Materials & Methods

Results

After quality filtering, we retained 2,930,324 reads from the singleplex approach and 1,740,807 reads from the multiplex approach. We visualised the number of samples (out of 51) with a successful detection of the host-vector-parasite components in Fig. 1A, distinguishing between singleplex and multiplex assays (see Data S1 for the full dataset). We found a higher detection rate of Plasmo01 in singleplex (43 out of 51) than multiplex (23 out of 51). For Aves02_mix and Culi01, the detection rates were similar between methods. Five out of 51 of the Aves02_mix amplifications were unsuccessful because the host was non-avian, which was verified using a generalist vertebrate primer, Vert01 (Riaz et al., 2011); the putative maximum detection of Aves02_mix is thus 46 (Table S4).

The same results were put in perspective using the taxonomic assignment. We visualised the detections per primer pair and sequence retained (Fig. 1B). If a taxon was detected in the same sample, using the same primer pair and in both singleplex and multiplex, but the sequence detected was different, we labeled it accordingly and visualised the results in Fig. 1B. Finally, we also visualised each sample individually to compare their performance for each primer pair in Fig. 1C.

Multiple hosts and co-infections

The DNA metabarcoding approach revealed multiple hosts, which originated from the vector feeding on various bird species, and coinfections with several Plasmodium lineages within a single mosquito abdomen (Fig. 2). We also detected clear differences in co-infection ratios between the two approaches, with singleplex being the approach with the highest number of detected co-infections (Fig. 2).

Taxonomic detection

Based on the DNA extracted from the mosquito samples, we detected all the vectors combining singleplex and multiplex (Fig. 1). These were mainly Culex sp. (C. pipiens or C. torrentium, the resolution of the Culi01 barcode being limited due to its short size), but we also detected one Ochlerotatus sp. mosquito. As for the hosts, we identified 12 different bird taxa, with Parus sp. and Passer sp. being the two dominant genera among the sampled mosquitoes. Together with Passer domesticus, Serinus canaria was used as positive control host. The sequences of the latter could only be assigned to family-level: Fringillidae (Fig. 3). Regarding the Plasmodium parasites, we detected four different taxa, mainly P. relictum and P. vaughani.

In order to compare the taxonomic assignment between singleplex and multiplex, we visualised the taxonomic assignment of the sequences produced with the three primer pairs (Fig. 3). For Plasmo01 there was no discrepancy between any resulting sequence pair. There was greater variability in the sequences retrieved for Aves02_mix than for Culi01. For example, in seven samples, the sequences obtained through singleplex were assigned to Passer domesticus but the sequences obtained through multiplex could only be assigned to Passer sp.

Metabarcoding performance

We also explored the performance differences between the two approaches, comparing the difference in RRA for each primer pair (Fig. 4). We found the singleplex approach yielded higher RRA for Aves02_mix. At a lower scale, this was also the case for Plasmo01, although for Plasmo01 we could not detect any taxon in many of the samples. As for Culi01, the relative read abundance was overall greater for multiplex than singleplex. To further visualise these differences in the performance, we compared the following variables within the two approaches: (i) the total amount of reads, (ii) RRA, (iii) RRA correcting by the primer relative concentration within multiplex (Fig. S1).

Discussion

In this study, we assessed the performance of both singleplex and multiplex approaches to study the host-vector-parasite components using the avian malaria system as a case study. The goal of this investigation was twofold: (i) to provide a proof-of-concept of the feasibility of the identification of all the components of vector-host-parasite interactions in a single PCR reaction and (ii) to highlight the methodological differences between the conventional singleplex and the multiplex approaches.

The use of DNA metabarcoding for detecting parasites is not new (Diaz-Real et al., 2015; Ondrejicka et al., 2014; Šigut et al., 2017), but this approach had until recently been limited by the separate detection of each component of a vector-borne system. Despite the great accuracy of a singleplex approach, a better comprehension of the interactions between the components of the host-vector-parasite triad might enable a better understanding of their evolutionary dynamics and improve prevention campaigns of vector-borne diseases. A multiplex approach using the vector as a template (e.g., blood meals) could highlight which host species are mostly infected or which are the dominant associations between certain parasite, vector and host species. These are just a few examples of potential applications for elucidating the interactions between the components of a vector-borne system using multiplex assays (see also Campana et al. (2016) for an in-solution capture approach, and Orantes et al. (2018) for a RADseq-based pipeline).

In terms of detection of each component of the tripartite association, our results reveal clear differences when comparing singleplex and multiplex. The results were similar for both host and vector, but the detection of the parasite was higher using the singleplex approach (Fig. 1A). We suggest this is due to the low initial DNA concentration of Plasmodium in the multiplex mix. The parasite template DNA competes with the DNAs from host and vector to be amplified in the multiplex, and its low initial concentration (in relation to very low infection intensities in the blood of hosts—chronic stage of infection, Pigeault et al., 2018) translates into lower amplification, and thus lower detection. Host and vector DNA is more abundant in the blood-meal DNA extracts, and given their shared resources for amplification, have a greater likelihood to be amplified. This is also the reasoning behind the standard use of a nested PCR protocol developed more than 20 years ago to detect and identify haemosporidian infections (Bensch et al., 2000; Waldenström et al., 2004). In this case, two PCRs are necessary: the first one amplifies a longer fragment of the mtDNA cytochrome b gene (approx. 580 bp including primers) which contains the barcode of the target parasite; the second one amplifies a smaller fragment (approx. 520 bp) using the first amplicon as template. This method allows for an increased sensitivity and specificity because two primer sets successively bind to the same target sequence, although nested PCRs increase the risk of PCR mutations and cross-contaminations. However, amplification of DNA fragments of this length can be difficult with potentially degraded samples such as blood meals in the midgut of the vector (Jo et al., 2017). When comparing the sequences retrieved using both approaches, we observed clear differences among primer pairs. Plasmo01 and Culi01 yielded the same DNA sequence in both approaches, whereas half of the samples amplified with Aves02_mix produced different sequences (Fig. 1B). This was expected given that Aves02_mix is a mix of three primer pairs. We suggest that the competition between these three primer pairs with the Culi01 and Plasmo01 in multiplex could have bound unevenly to the template DNA producing slightly dissimilar sequences, but we cannot rule out potential heteroplasmy. However, the taxonomic assignment of these sequences revealed these were equally resolutive (Fig. 3), except for one case. This confirms the equal performance of both approaches to retrieve consistent taxonomic information, despite the small differences within each sequence.

Regarding co-infections, they were more frequently detected when using singleplex. These results highlight the higher resolution of the data when a single PCR reaction is dedicated to each primer pair, despite the higher cost and greater time investment when analysing a large number of samples compared with multiplexing. When designing the laboratory procedure for an experiment, using one approach or another has to be assessed having this trade-off in mind.

To further understand the performance differences, we compared singleplex and multiplex in terms of the RRA for each approach (i.e., singleplex or multiplex), primer and sample, calculated dividing by the total amount of reads of each approach as reference, instead of the sum of reads per sample (Fig. 4). We did this to account for the amplification and sequencing biases across primer pairs. The mix of primers for the multiplex approach was not equimolar; we purposely increased the concentration of the Plasmo01 and Aves02_mix pairs. This was done in order to reduce the potential over-amplification of Culi01, given the blood-meals were extracted from mosquito abdomens, and we expected their DNA to be more abundant and of better quality. We used the singleplex results as unbiased reference of primer performance for each sample, which highlighted the differences between primer pair performance when used together in the multiplex. The results were distinct for each primer pair: RRA was greater in singleplex for Aves02_mix and Plasmo01 (although by a smaller margin); and greater in multiplex for Culi01. We interpret these results as follows: the reduction of Culi01 concentration did not compensate for the abundance and quality of mosquito DNA, which still represented a higher proportion of reads. This can be observed when comparing Fig. S1: correcting for the primer concentration in multiplex only reduced Culi01 amplification, but not Aves02_mix or Plasmo01.

Nevertheless, this did not affect the detection of avian hosts (only three samples that were amplified in singleplex were not amplified in multiplex). However, it had a great impact on the Plasmo01 detection, as half of the Plasmodium detections occurred only through singleplex. This suggests the low abundance of Plasmodium DNA in the blood-meal was not compensated by the greater concentration of Plasmo01 and undermined the detection potential of the multiplex approach. However, in terms of Plasmo01 RRA, we did not find clear differences between the two approaches (Fig. 4).

Put in perspective, the use of multiplex offers major advantages in terms of cost and time over the use of singleplex. Their detection and taxonomic performance are comparable for both host and vector. However, the parasite, and its low initial DNA concentration, poses greater challenges. We suggest the relative concentration of parasite primers should be greatly increased compared to the host and vector ones in order to increase the reliability of the detection results when using a multiplex approach. More generally, several relative primer concentrations within a multiplex could be tested, in order to optimise detection of each of the targeted taxa involved in the final assay. This can be for instance achieved by using a fragment analyzer, which quantifies DNA size and relative abundance, provided that the barcode lengths have non-overlapping ranges between the three components. Doing so, the relative primer concentration above which the parasite (or least abundant component) is revealed in the multiplex can be adjusted to maximize its detection. New generation Taq polymerases optimized for co-cycling of different PCR targets (i.e., multiplexing) can also be used in the assay (e.g., Platinum II Taq Hot-Start DNA polymerase, Thermo Fisher), improving the likelihood to reveal the minor component in a DNA mixture after the amplification step. Furthermore, other combinations of DNA primers can be developed in the multiplex assay, either by targeting other taxonomic groups or by achieving different taxonomic coverage and resolution, in the case of the malaria system but also in any vector-borne interaction involving hematophagous arthropods.

Overall, we conclude from this proof-of-concept study that the applications of multiplex are promising, but further research is needed to better understand multiplex detection probabilities and to refine the technique for large-scale applications in vector-borne disease prevention management.

Supplemental Information