An assessment of adult mosquito collection techniques for studying species abundance and diversity in Maferinyah, Guinea

Background Guinea is a West African country with a high prevalence of vector-borne diseases where few entomological studies have been undertaken. Although several mosquito collection methods are routinely used for surveillance in vector control programmes, they target different behaviours causing bias in species diversity and abundance. Given the paucity of mosquito trap data in West Africa, we compared the performance of five trap-lure combinations and Human Landing Catches (HLCs) in Guinea. Methods Five mosquito traps were compared in a 5×5 Latin Square design for 15 days in three villages in Guinea between June and July 2018. CDC light traps, BG sentinel 2 traps (with BG and MB5 lures), gravid traps and Stealth traps were deployed for 24-hour intervals with mosquitoes collected every 12 hours (day and night collections). HLCs were also performed for 15 nights. A Generalised Linear Mixed Model was applied to compare the effect of the traps, sites and collection times on the mosquito abundance. Species identification was confirmed using PCR-based analysis and Sanger sequencing. Results In total, 10,610 mosquitoes were captured across all five traps. Significantly more mosquitoes (P<0.005) were collected by Stealth traps (7,096) compared to the rest of the traps. Stealth traps and BG sentinel 2 traps were the best at capturing An. gambiae and Ae. aegypti mosquitoes respectively. HLCs captured predominantly An. coluzzii (41%) and hybrids of An. gambiae s.s. / An. coluzzii (36%) in contrast to the five adult traps, which captured predominantly An. melas (83%). Senguelen (rural) presented the highest abundance of mosquitoes and overall diversity in comparison with Fandie (semi-rural) and Maferinyah Centre One (semi-urban). To our knowledge, four species are reported for the first time in Guinea. Conclusions Stealth traps presented the best performance overall, suggesting that this trap may play an important role for mosquito surveillance in Guinea and similar sites in West Africa. We recommend the incorporation of molecular tools in entomological studies since it has helped to reveal, together with morphological identification, the presence of 25 mosquito species in this area.

from each sample were manually checked, edited, and trimmed as required, followed by 182 alignment with ClustalW and checking to produce consensus sequences. Consensus 183 sequences were used to perform nucleotide BLAST (NCBI) database queries (37,38). Full 184 consensus sequences were submitted to Genbank and assigned accession numbers XXX-185 YYY. Confirmation of species was considered complete for sequences with an identity to a 186 particular species given by BLAST of greater or equal to 98%, and where no other species 187 also gave identities at this level. 188

Comparison of five adult mosquito traps 202
A total of 10,610 mosquitoes were trapped by the five adult mosquito traps across the 30 203 collection intervals (15 days and 15 nights) of the study. In terms of abundance, the ST 204 captured the highest percentage of the total number of mosquitoes collected (67%), followed 205 by the LT (24%), the BG2-MB5 lure (4%), the GT (3%) and the BG2-BG lure (2%) ( Table 1). 206 The diversity of species was measured using the Simpson's diversity index. Results showed 207 that the BG2-BG captured the most diverse range of mosquito species (Simpson's diversity 208 index = 0.157), followed by the GT (0.241), BG2-MB5 (0.24), LT (0.415) and ST (0.484) 209 (Table 1). 210 The majority of the mosquitoes collected across this study belonged to the main genera: 211 Anopheles, Aedes and Culex. However, the ST and LT captured one and two Uranotaenia 212 mosquitoes respectively, the BG2-MB5 captured two Mansonia and the BG2-BG captured 213 one Eretmapodites. Further information on species captured by each trap is shown in Table  214 2A. Regarding the sex of collected mosquitoes, 38% of specimens captured by the LT and 215 ST were males, whereas for the other traps, males were less than 22%. GT caught the 216 highest proportion of gravid females, whereas unfed females represented the highest 217 proportion of the catch in other traps. Bloodfed females made up the smallest group, with the 218 BG2-MB5 lure trapping the highest relative proportion. The total numbers of bloodfed 219 females were too low for comparative bloodmeal analysis (Table 1). 'Damage state' of the 220 specimens was also annotated and assessed. No specimens were damaged by the gravid 221 trap, less than 10% of the specimens were damaged in both BG2 and 10% of specimens 222 were damaged in the LT (data not shown). However, the ST resulted in the highest 223 proportion of damaged mosquitoes at approximately 20%, of which nearly one quarter could 224 not be morphologically identified (Table 1). Although the ST captured the largest number of 225 mosquitoes, this trap also collected a large number of non-target Diptera and ants, making 226 sorting of the specimens time-consuming ( Figure 3). 227

Generalised Linear Mixed Model for mosquito abundance 228
A negative binomial GLMM was used to determine statistical differences between the 229 abundance of mosquitoes captured by each trap. The results indicated that the following 230 parameters influenced the number of mosquitoes collected: Site (Maferinyah Centre One, 231 Senguelen and Fandie), Time Period (evening and morning), Trap (BG2-BG, BG2-MB5, GT, 232 LT, ST) and Sampling Point (random factor). Rainfall, temperature and humidity did not 233 significantly influence the data, however, humidity was included as a random factor. The 234 final, best-fit model was: Abundance ~ Site + (1|Point) + (1|Humidity) + Time + Trap. 235 According to this model, there were no significant differences between the abundance of 236 mosquitoes captured by the GT, the BG2-MB5 and the BG2-BG (Table S3). There were no 237 differences either between the abundance of mosquitoes captured by the GT and the LT. 238 However, there were significant differences between the abundance of mosquitoes captured 239 by LT and BG2-MB5 (p=0.057) and LT and BG2-BG (P<0.005). Finally, significant 240 differences were found between the abundance of mosquitoes captured by the ST and all  (Table S3). Regarding sites and collection intervals, 243 more mosquitoes were captured in Senguelen than in Maferinyah Centre One and Fandie 244 (P<0.001) and significantly more mosquitoes were captured during the night than during the 245 day (P<0.001). 246 The above model was used to assess the effectiveness of the different traps at capturing 247 Aedes, Anopheles and Culex mosquitoes in general, and An. gambiae s.l. and Ae. aegypti 248 species in particular, since they are the main vectors of disease. The results showed that 249 while no differences are shown between the abundance of Aedes mosquitoes captured by 250 the traps, both BG2 are significantly better at capturing Ae. aegypti mosquitoes (Table 3). 251 The ST resulted to be the best at capturing the Anopheles genus and An. gambiae s.l. in 252 particular, although it presented significant differences only when compared with the GT. 253 Finally, the ST was significantly better at capturing Culex mosquitoes than any other trap, 254 followed by the LT (Table 3). 255 were selected for molecular identification and comparison of species composition (Figure 4). 262

Comparison of
Results showed that An. melas was the predominant species (85%) caught by adult 263 mosquito traps, whereas it was collected the least (10%) using HLCs. Anopheles coluzzii 264 and An. gambiae s.s. / An. coluzzii hybrids were the most abundant species collected using 265 HLCs (40% and 35% respectively), whereas these were 12% and 2% of the collections 266 respectively using adult traps. Anopheles gambiae s.s. represented 15% of the individuals 267 collected using HLCs whereas this species was only 1% of the individuals collected using 268 adult traps. 269

Species composition in the Maferinyah subprefecture 270
Senguelen was the site with the highest number of mosquitoes (5,784) followed by Fandie 271 (4,094) and Maferinyah Centre One (732) ( Table 4). The diversity of the species from the 272 day collection (07:00 to 19:00) was similar to the night collection (19:00 to 07:00) in 273 Senguelen and Maferinyah Centre One, presenting a Simpson's diversity index of around 274 0.2 and 0.3 respectively. However, Fandie showed a high diversity in the day collection 275 (0.142) and a low diversity in the night collection (0.48) (Table 4). Further information on 276 species captured in each site and during each collection period are shown in Tables 2B and  277 2C. A total of 25 species were found across the three sites (using a combination of 278 morphological and/or molecular identification), belonging to the Aedes, Anopheles, Culex, 279 Eretmapodites, Mansonia and Uranotaenia genera. One Toxorhynchites (Tx.) brevipalpis 280 was also captured during a morning collection in Fandie by the BG2-BG lure combination. 281 However, the power failed to one of the traps during this round, and therefore the collection 282 could not be included in the analysis. 283 A subsample of 370 specimens were selected for molecular identification. This subsample 284 included 249 Anopheles, 24 Aedes, 96 Culex and 1 Eretmapodites individual. These 285 numbers represented 47.2%, 2.7%, 1.1% and 100% respectively of the total number of 286 collected mosquitoes within each genera (Table S4A). The 370 specimens selected for 287 molecular identification were chosen in order to confirm the species identity of mosquitoes 288 collected using all traps across the three sites, representing 1.4%, 8.5% and 4.4% of the 289 total collections from Fandie, Maferinyah Centre One and Senguelen respectively ( Table  290 S4B). In total, 20 species were confirmed by Sanger sequencing (Table S5) This study provides the first entomological survey in Guinea that compares the mosquito 319 species abundance and diversity using a range of different adult mosquito traps. Other  (Table 1), so they could be utilised in 337 studies looking at male behaviour. In general across the traps, sites and collection intervals, 338 all the study collections presented a greater number of females than males. However, 339 interestingly this composition was reverted in two collections, and a greater number of males 340 was capturedin sampling points C and E in Fandie. The fact that these two sampling 341 points may have been located next to a swarm could be a potential explanation (44). 342 Previous studies suggest that the LT are optimal for catching Anopheles (45), however, the 343 main genus captured by the LT was Culex. In contrast, the ST was the best at capturing the 344 largest number of Anopheles mosquitoes in general and An. gambiae s.l. in particular. 345 According to Costa-Neta et al. (46), the higher the intensity of the light source, the higher the 346 number of Anopheles captured. This may be one reason why the ST captured the highest 347 number of Anopheles (Table 1 and 3). 348 Previous studies suggest that the GT are good at catching Culex (47), and this was indeed 349 the main genus captured by this trap. However, the ST was significantly better at capturing 350 large numbers of Culex mosquitoes (Table 1 and 3). As expected, this trap also captured the 351 highest proportion of gravid females. Additionally, all of the specimens were un-damaged, 352 since the design of the trap allows the collection of specimens without passing through a fan, 353 so its use could be beneficial to capture mosquitoes with the objective of establishing a 354 colony or screening for arbovirus transmission. 355 Due to the small sample size, no conclusions can be made regarding the best collection 356 method for Eretmapodites, Mansonia and Uranotaenia mosquitoes. Although the ST showed 357 the best performance in terms of abundance of mosquitoes captured, this trap also caused 358 significant damage to specimens, making morphological identification time-consuming and 359 inaccurate. One reason for this damage could be the high density of collected specimens 360 ( Figure 3A), which remained in the trap for up to 12 hours during trapping intervals, 361 depending on trap entry time. In addition to this, the presence of ants and big Diptera could 362 have also contributed to this damage ( Figure 3A and 3B). Another reason could be the low 363 protection that this trap confers to the collected specimens from rainfall, due to the small 364 surface area of the cover / rain shield, resulting in wet and clumping specimens ( Figure 3C). 365 Therefore, the performance of the ST could potentially be improved by using it for shorter 366 periods of time or by swapping collection bags more often, to reduce the high densities of 367 mosquitoes within the same collection bag. Also, by choosing locations offering greater 368 protection from rainfall, which could help reduce damage to the specimens. 369 The BG lure is designed to attract mainly Aedes whereas the MB5 lure was specifically 370 designed for Anopheles (48,49). Although BG2 with BG lure have been used in Burkina 371 Faso (50), to our knowledge no traps have been used in West Africa with the MB5 lure so 372 far, so both lures were tested in the two BG2 in this study. Previous studies suggest that the 373 BG2 in general are effective for catching Aedes mosquitoes (51), and that the addition of the 374 BG lure improves this (51). In this study no significant differences were seen in the number 375 of Aedes mosquitoes (at genus level) captured by the five different traps, although the high 376 proportion of Aedes specimens captured by the BG2-BG (Table 1)  Additionally, both BG2 presented the best performance at capturing Ae. aegypti mosquitoes 381 in comparison with the rest of the traps, with no differences between the two lures (Table 3) s.l. However, no significant differences were seen between both (Table 3), indicating that the 390 MB5 lure needs further improvement in order to obtain more effective collections of this 391 genus. Since the ST showed the best performance for Anopheles (and An. gambiae s.l..) 392 and no significant differences were shown between the ST and the two BG2, its use could 393 be recommended for studies specifically looking at these genera, although an increased 394 number of trapping intervals would be required to increase the number of mosquitoes 395

captured. 396
Diversity takes into account richness (number of different species) and evenness 397 (comparison of population size of each species). Although the number of species captured 398 by the LT and ST was higher than the other traps (higher richness), the difference in the 399 number of specimens from each species was higher than the other traps (low evenness). 400 Therefore, the diversity of the mosquito populations captured by LT and ST was the least 401 diverse. The BG2-BG presented the most diverse collection of mosquitoes, followed by the 402 GT and the BG2-MB5. Therefore, these three traps would be recommended for studies 403 looking at species diversity, as opposed to LT and ST, which would be recommended for 404 studies requiring a large number of mosquitoes of a particular species, with exception of 405 some species (see Table 2A). 406 Human landing catches are the gold standard method for measuring exposure of humans to 407 mosquito bites (53). However, this method is labour-intensive and faces ethical 408 considerations (54), as operators are potentially exposed to pathogens during collections. 409 Since adult mosquito traps are an affordable and easy to use alternative which provides 410 reliable entomological data about malaria transmission (55), we compared both methods 411 specifically targeting the major malaria vectors in the An. gambiae complex. Human landing 412 catches captured predominantly An. coluzzii, An. gambiae s.s. and hybrids, but they only 413 captured a small percentage of An. melas. Alternatively, more than three quarters of the trap 414 collections were An. melas and only a small percentage was An. coluzzii, followed by a 415 smaller proportion of An. gambiae s.s. and hybrids. Anopheles gambiae s.s. and An. coluzzii 416 are highly anthropophilic, whereas An. melas is considered opportunistic, feeding on 417 humans when available and on other mammals otherwise (56). Although different cues such 418 as lights and lures that mimic human odours are used in mosquito traps to try to attract host-419 seeking females, HLCs are more effective at attracting anthropophilic Anopheles species. 420 Therefore, this method would still be recommended for targeting species with this behaviour. 421 These results also suggest that an improvement in lures or trap design is needed to better 422 mimic human cues and increase the number of anthropophilic species captured. Some 423 studies have tried this in the past by modifying BG sentinel traps to increase the captures of 424 An. darlingi (57) and An. arabiensis (58) mosquitoes and use them as an alternative for 425 HLCs. However, others have also shown that HLCs are still more effective at capturing 426 Anopheles species in comparison with adult traps, whose main collections comprise 427 Culicines (59), as seen in the present study. Since in our study both methods -HLCs and 428 mosquito trapswere undertaken outdoors, no conclusions can be made about which 429 method would work more effectively for targeting different feeding and resting behaviours. 430 Senguelen, a rural site, presented the highest relative abundance of mosquitoes, whereas 431 Maferinyah Centre One, a semi-urban site, presented the lowest relative abundance. In 432 terms of mosquito species diversity, the former was also more diverse than the latter. The 433 fact that the rural site was surrounded by dense vegetation and breeding sites, as opposed 434 to the semi-urban environment, could explain these differences. Both Senguelen and 435 Maferinyah Centre One presented similar diversities between day and night collections. Traditionally, identification of mosquitoes has been carried out using morphology. However, 447 morphological identification can be time-consuming and inaccurate sometimes, especially 448 when the specimens do not present obvious and exclusive features or when they are 449 damaged, as seen in the mosquitoes collected by the ST in this study. Molecular tools such 450 as PCR and sequencing can improve entomological studies by overcoming these limitations. 451 As an example, one of the female mosquitoes collected using HLCs presented long palps -452 typical from the Anopheles genusbut it was white in colour and did not present the 453 common wing and leg patterns of many species of the Anopheles genus ( Figure 5). This 454 individual female could not be identified by experienced entomologists using Anopheles keys 455 so DNA was extracted from this individual and PCR with Sanger sequencing revealed this 456 species to be An. coluzzii. Random mutagenesis could be a potential explanation for this 457 phenotype. Since molecular tools can complement and improve morphological identification 458 of mosquitoes, it would be recommended to combine both for further entomological 459

investigations. 460
Among the species whose presence was confirmed in Guinea using molecular methods, we 461

AVAILABILITY OF DATA AND MATERIALS 546
The dataset supporting the conclusions of this article is included within the article and its 547 additional files. 548

AUTHORS' CONTRIBUTIONS 549
CCU designed the study, conducted and analysed the field and laboratory work and wrote 550 the first draft of the manuscript. CLJ designed the study, analysed the sequencing data and 551 co-wrote the manuscript. MK undertook fieldwork and labororatory analysis and contributed 552 to the practical design and logistics of the study. VAB designed the study and provided R 553 codes for Latin Square analysis. PH undertook fieldwork and labororatory analysis. MS and 554 HB helped to obtain local authorisation. MS and GC contributed to setting and collecting 555 traps and mosquito identification. LAM contributed to field data analysis and supervision. TW 556 designed the study, analysed the laboratory data, wrote the manuscript and provided overall 557 supervision. TW and LAM obtained funding for the study. All authors read and approved the 558 final version of the manuscript. 559