Discussion
In this paper, we describe a method of mosquito species identification based on a traditional endpoint PCR and a precise amplicon size detection, avoiding the need for sequencing. We found that ITS2 sequence size differences between species are detectable by capillary electrophoresis. This enabled us to assign species-specific profiles for nine Anopheles species (seven species from theNyssorhynchus subgenus and two from the Anophelessubgenus) present in French Guiana. Identification is based on ITS2 amplicon size information combined with a simple morphological observation, the color of the fifth hind tarsus. This additional piece of data is quick and easy to acquire, by eye or under a stereomicroscope, when collecting mosquitoes in the field and does not require any advanced knowledge in taxonomy.
This species identification method has the advantage of being simple and relatively quick to set up, and can be used in routine in a laboratory avoiding the requirement for sequencing. During the three steps of method development, we were not able to sequence our samples on-site and we had to send samples in 96-well plates between French Guiana and France for sequencing. For 7.8, 9.2 and 1.4 % of samples, at steps 1, 2 and 3 of method development respectively, it was not possible to obtain definitive results as forward and reverse sequencing outcomes were different. Moreover, we observed that 3.0 and 3.7 % of sample sequencing results, at steps 1 and 2 of method development respectively, were erroneous. We were able to detect these errors because of contradictory results between sequencing and our method, whether because of fragment size not matching the putative species or because of conflicting leg color. Without the implementation of the method, these errors would have gone unnoticed. In case of doubt after sequencing, we amplified the ITS2 region again and sent the sample for sequencing to check the initial result. Hence, sequencing can be a source of error, especially when samples need to be shipped, as cross-contaminations may occur during plate processing, shipment and sequencing. In our case, plates were sealed and packaged carefully but we noticed that plate caps that have undergone temperature changes during PCR and storage in the freezer are slightly easier to open. Moreover, we used adhesive PCR plate seals for one shipment of two plates (one of step 1 and one of step 2), which resulted in more misidentifications and uncertain results. During subsequent shipments (steps 2 and 3), we used plate caps that were replaced with new ones just before shipping and observed a relatively low error rate, yet did not investigate this further.
In our method, missing or erroneous leg color is the major source of problem. It rarely led to incorrect identification (1.2 % samples species at step 2), but more often led to uncertainties (10, 12 and 0.27 % at steps 2, 3 and Routine, respectively), which represents the majority of method uncertainties (12, 16 and 0.54 % samples, at steps 2, 3 and routine, respectively). Wrong leg color specifically led to 4.3 and 8.2 % indeterminations at steps 2 and 3 respectively, indicating that efforts to ensure correct annotation of this information at the time of capture can significantly lead to improvement in determination rates. A simple double check by a colleague may be the key, as these data are usually collected in a repetitive way, sometimes at night, in a non-usual environment. Nonetheless, this problem cannot be completely solved, as some mosquito samples lack both hind legs. One could decide to exclude them from the study, yet we were able to identify the species of 54/94 mosquitoes (steps 1, 2 and 3) with missing fifth hind tarsus color information, and of 16/17 in routine. The fifth hind tarsus color element is therefore a piece of data that should not be neglected, but even if it is missing, identification remains possible in many cases.
The second key point for the reliability of our method is, obviously, the proper use of the capillary electrophoresis device and its components. The channels of the capillary electrophoresis cartridge are prone to clogging if they are used incorrectly and the results can be erroneous in these cases. The only recommendation on this point is to follow the supplier’s instructions carefully and check results individually, in order to detect any aberrant results. In case of aberrant results, we either excluded the problematic channel or replaced the cartridge.
While the current method has been developed on Anopheles species from French Guiana, this approach may be extended to species identification of individual fieldwork specimens in any taxon. Widening the scale besides Anopheles (Nyssorhynchus ), our method clearly discriminated our Anopheles samples from four mosquitoes of the Culex and Psorophora genera during step 1, and we expect development steps to require much less specimens if the species of interest are more distant from each other. More testing would be needed to determine whether all mosquito genera have a significantly different profile from one another and whether, within these genera, species identification is possible. Our tool may then facilitate identification of a broad range of mosquitoes, and be particularly advantageous during epidemics in order to target species that are potential vectors of arboviruses or parasites. It could be compatible with sampling by non-experts combined with photographs during collaborative work or citizen science projects. Finally, when presence/absence information are sufficient, pools of different samples sharing the same morphological trait (in our case, the color of Ta-III5) could be processed.
To specify the interest of our method with currently available methods, several aspects can be considered: the possibility to go back and check results, the level of local diversity and the cost-effectiveness. Our method allows to easily check results as DNA extracts can be stored for a long time. They can be used for a second similar analysis by PCR and capillary electrophoresis as well as for sequencing of ITS2 or any other sequence, which may notably apply to population genetics studies. The other methods of molecular biology have similar advantages, while morphological observations and audio recordings may not, unless samples have been properly stored.
In terms of diversity, we detected nine different Anophelesspecies, including four dominant ones. With a higher local diversity, too many overlaps between intervals would reduce the efficiency of the method with the current experimental conditions. The latter can be improved, for instance by running capillary electrophoresis straight after PCR, which could allow to deal with a slightly higher diversity. Sequencing would remain the best option in case of high species diversity. With a lower diversity of only two or three species, multiplex PCR with species-specific primers would remain the simplest and cheapest method. Hence, our species identification method is a reliable alternative for locally moderate species diversity.
Considering costs, capillary electrophoresis requires an initial equipment investment, yet the processing cost of each sample is much lower with the proposed method than with sequencing. When considering the initial cost of the device and of capillary electrophoresis reagents on one hand, shipping and sequencing cost of samples on the other hand (and ignoring the cost of labour and our interest in the other applications of the device), our investment on the capillary electrophoresis apparatus would be repaid after 4,000 to 5,000 samples. Beyond that, more than 5 \euro per sample can be saved. Hence, episodic needs may rather be addressed with sequencing, while our approach seems reasonable for a laboratory requiring steady species identification. Alternatively, even with very low consumable cost per sample [42] , the MALDI-TOF approach requires a 10-fold higher investment and maintenance budget than capillary electrophoresis. It would thus require a much higher number of samples to be more cost-effective than our method.
In sum, we introduce PCR and capillary electrophoresis combined with simple morphological observation as a convenient method to discriminate species of different field-collected samples. We specifically deployed it to identify nine Anopheles species found in French Guiana, mainly from the Nyssorhynchus subgenus. After deployment, this method allows to save time and money and to keep control over experimental schedule.