Transposable elements are constantly exchanged by horizontal transfer reshaping mosquito genomes

Transposable elements (TEs) are a set of mobile elements within a genome. Due to their complexity, an in-depth TE characterization is only available for a handful of model organisms. In the present study, we performed a de novo and homology-based characterization of TEs in the genomes of 24 mosquito species and investigated their mode of inheritance. More than 40% of the genome of Aedes aegypti, Aedes albopictus, and Culex quinquefasciatus is composed of TEs, varying substantially among Anopheles species (0.13%–19.55%). Class I TEs are the most abundant among mosquitoes and at least 24 TE superfamilies were found. Interestingly, TEs have been continuously exchanged by horizontal transfer (212 TE families of 18 different superfamilies) among mosquitoes since 30 million years ago, representing around 6% of the genome in Aedes genomes and a small fraction in Anopheles genomes. Most of these horizontally transferred TEs are from the three ubiquitous LTR superfamilies: Gypsy, Bel-Pao and Copia. Searching more 32,000 genomes, we also uncover transfers between mosquitoes and two different Phyla—Cnidaria and Nematoda—and two subphyla—Chelicerata and Crustacea, identifying a vector, the worm Wuchereria bancrofti, that enabled the horizontal spread of a Tc1-mariner element of irritans subfamily among various Anopheles species. These data also allowed us to reconstruct the horizontal transfer network of this TE involving more than 40 species. In summary, our results suggest that TEs are constantly exchanged by common phenomena of horizontal transfers among mosquitoes, influencing genome variation and contributing to genome size expansion. Author Summary Most eukaryotes have DNA fragments inside their genome that can multiply by inserting themselves in other regions of the genome, generating variability. These fragments are called Transposable Elements (TEs). Since they are a constituent part of the eukaryote genomes, these pieces of DNA are usually inherited vertically by the offspring. To avoid damage to the genome caused by the replication and insertion of TEs, organisms usually control them, leading to their inactivation. However, TEs sometimes get out of control and invade other species through a horizontal transfer mechanism. This dynamic is not known in mosquitoes, a group of organisms that acts as vectors of many human diseases. We collected mosquito genomes available in public databases and characterized the whole content of TEs. Using a statistic supported method, we investigate TE relations among mosquitoes and discover that horizontal transfers of transposons are common and occurred in the last 30 million years among these species. Although not as common as transfers among closely related species, transposon transfer to distant species also occur. We also identify a parasite, a filarial worm, that may have facilitated the transfer of TE to many mosquitoes. Together, horizontally transferred TEs contribute to increasing mosquito genome size and variation.

genomes available in public databases and characterized the whole content of TEs. Using a statistic 48 transposition mechanism: subclass I, are small elements (2 to 7 kb) normally coding for a single protein 75 comprising elements that transpose by cleaving both strands of DNA, and subclass II is composed of large 76 (Mavericks reaching 40 Kb) and medium-sized TEs (Helitrons reaching 15 Kb) that transpose with cleavage 77 of only one DNA strand [8,11,12]. 78 TEs are inherited by vertical transfer, the transfer of genetic from ancestral to descendant species, 79 to all host offsprings [13]. However, there is growing evidence that TEs can also move horizontally 80 between independent species by a phenomenon known as horizontal transfer (HT) [14]. Recent large-scale 81 studies on Insects and Vertebrates have revealed thousands of Horizontal Transposon Transfer (HTT) 82 events and general patterns, such as the occurrence of HTT more frequently between closely related host 83 species which share a spatiotemporal overlap; the substantial contribution of horizontally transferred TEs 84 to the genome size of the host species; and that DNA transposons are transferred horizontally much more 85 frequently than Retrotransposons [15]. However, many open questions need further attention to understand 86 HTT at different host taxa, its functional impact on host genomes, and by which mechanisms HTT occurs. 87 In a simplistic view, HTT might occur by direct transfer of a TE between species or mediated by a 88 vector/intermediate species that transport TEs between hosts [16]. Currently, there is evidence scattered on 89 different host/TE systems that these elements can 'hitch a ride' on other parasite genomes/particles, such 90 as those of parasitic nematodes, macroscopic blood-sucking insects, and viruses that interact intimately 91 with different host species [17][18][19]. However, the exact mechanisms by which HTT occurs has remained 92 as a large mystery since it is difficult to reproduce such phenomenon in the lab. Moreover, reconstruct TEs 93 horizontal steps from sequence data is full of pitfalls and missing information mainly due to incomplete 94 sampling of host-intermediate species [16,20]. 95 TEs make up a significant portion of the genome of some species: 80% of the model plant organism 96 Zea mays [21], and almost half of the genome of the mosquito Aedes aegypti [22]. Mosquito belongs to 97 Culicidae family, which comprises more than 3500 known species, dispersed trough all continents except 98 Antarctica. It is divided into two large subfamilies, Culicinae and Anophelinae, the former is composed of 99 11 tribes with several genera, while the latter has only 3 genera, having the Anopheles genus as the most 100 species-rich group [23]. Many mosquito species of the Culicidae family are important human pathogen 101 8 and Anopheles mosquitoes (Fig 2). However, it is clear that Cu. quinquefasciatus has the most distinct TE 184 superfamily landscape of the Culicidae family, with the greatest abundance of Zator,Sola, like elements compared to all other mosquito species studied. Another distinctive feature of the Cu. 186 quinquefasciatus mobilome is the low genome proportion of ubiquitous superfamilies among mosquitoes 187 such as Tc1-Mariner, RTE, and I (Fig 2)

202
We found 212 TE families with significant HT signals among the species studied. However, it is 203 important to note that this is certainly an underestimate of the true number of HTT events since most TE 204 families have more than one significant pairwise HTT signal ( Fig 3A). We, therefore, decided to describe 205 it in more general terms, retaining at least one HT event for each TE family, in view of the complexity of 206 the signals found and the lack of currently available algorithms for determining the most likely HTT 207 scenario and estimating the minimum number of events required to explain the signal. 208 9 More than three-quarters of transposons involved in HT events belong to Class I TEs, covering 209 eight different superfamilies ( Fig 3B). The superfamily with the largest number of families undergoing HT 210 is Gypsy, in which HT accounts for more than 35% of all events, followed by Tc1-Mariner elements. 211 Although Class I TEs are more abundant in mosquito genomes and cover most TE families involved in 212 horizontal transfer events, Class II TEs were found to contain more superfamilies (9) undergoing HT. In 213 this group, the Tc1-Mariner superfamily deserves special mention, since it has 11 TEs involved in HT 214 events involving four or more mosquito species. 215 Almost 25% of the HT signals detected (47 families) occurred only between Ae. aegypti and Ae. 216 albopictus. One hundred and sixty-one TE families involved in HT events were detected among species of 217 the Anopheles genus, accounting for 18 of the 21 species. Only tree mosquito species, An. darlingi, An. 218 punctulatus and An. koliensis, were not involved in HTT events. One Anopheles species, An. merus-an 219 early divergent species of the A. gambiae complex-was involved in the largest number of HTTs (Fig 3A). 220 The majority of species from the Anopheles genus are involved in HTTs from both Class I and Class II TEs 221 (Table 1). In three species we detected only HTTs of Class I TEs: non-LTR elements in An. albimanus and 222 An. farauti, and LTRs elements in An. christyi. Concerning HTT between different mosquito genera, six 223 events were found. Five out of six of those belong to the Tc1-Mariner superfamily, while only one R4-a 224 non-LTR retrotransposon -was transferred from Ae. albopictus to the ancestor of An. gambiae complex 225 mosquitoes, as already described in a previous study [32]. No horizontal transfer events involving the Cu. 226 quinquefasciatus species were found. It is important to note that all HTT events showed very low p-values 227 (S4 Fig) associated with several additional HTT evidence as patchy distribution and phylogenetic 228 incongruences between TE and host phylogeny, as further discussed below (2.8 -Horizontal transfer of 229

TEs involving distantly related eukaryotic species). 230
We estimated the HTT rate for Class II, LTR, and non-LTR elements based on the number of TE 231 families with HT signal on the total number of families tested with VHICA. Class II TEs showed the highest 232 rate (24.1%), followed by non-LTRs (20.2%) and LTRs (19.4%). Around 14% of the HTT networks involve 233 at least four species. This proportion is small, particularly among LTR families, where four or more 234 mosquito species were detected in less than 10% of the total number of LTR families involved in HTT. Of 235 all the superfamilies studied, Tc1-Mariner and RTE are the ones whose elements are most often horizontally 236 transferred across many species, 17 and 13 respectively. Details about the participation of species in each 237 TE family horizontal transfer network can be seen in Supplementary File 2, which lists all VHICA output 238 images of positive HTT cases. The relative age of each TE family undergoing HT per mosquito genome is 239 given in S5 Fig  TEs horizontally transferred between species represent a significant fraction of some mosquito 244 genomes ( Fig 4A). We estimated that about 6% of Ae. aegypti and Ae. albopictus genomes are covered by 245 horizontally transferred TEs representing more than 10% of the total TE content of these species. In the 246 Anopheles genus, the impact of horizontal transfer on genome size is much smaller. Only An. gambiae and 247 An. coluzzii showed more than 1% of their genome covered with copies derived from horizontally 248 HTT among LTRs and TIR elements intensified more recently ( Fig 4C). In general, horizontal transfers 260 occurring between species of the Anopheles genus are more recent than those taking place within the Aedes 261 genus or between these two genera. Most of the HTTs between Ae. aegypti and Ae. albopictus identified 262 are relatively ancient transfers, most having occurred more than 15 million years ago ( Fig 4D). The species involved in HTT events belong to three metazoan phyla. The first phylum was 278 Arthropoda, which covers the majority of species. It is represented by the class Arachnida, by the common 279 house spider Parasteatoda tepidariorum, by Hexanauplia, represented by the copepod Lepeophtheirus 280 salmonis, and by the class Insecta, which includes most of the mosquito species involved in HTT. The 281 second phylum was Cnidaria, with only one representative, Hydra vulgaris. The third was Nematoda, where 282 we found a horizontal transfer case involving the parasitic worm Wuchereria bancrofti and other Anopheles 283 species that transmit this nematode to humans. Besides the high identity between mosquito TEs and those 284 of distantly related species, sometimes separated by more than 700 million years, we also observed a patchy 285 distribution of these elements, providing further support for the horizontal transfer events. 286 Two of these TE lineages (HTa and HTe) belong to the ITmD37E group, which includes TEs that 287 have the DD37E motif on their transposase sequences (Fig 6A). One lineage (HTd) clustered with 288 previously described elements from the ITmD37D family, also known as the mat family. The other two 289 lineaged (HTb and HTf) are from the DD34D or Mariner family. Elements from the HTf network belongs 290 to the mauritiana subfamily, and the HTb network elements, which are involved in most HTT events with 291 species from outside of the Culicidae family, belong to the irritans subfamily. To obtain a more in-depth 292 understanding of the HTb network, we performed a second BLAST search using as queries all sequences 293 recovered from the search against non-mosquito genomes. We found HTb related elements to be present in 294 13 42 species of divergent invertebrates, including W. bancrofti (Fig 6B). Intragenomic dating of the 295 horizontally transferred TEs provides more specific data and enables a more accurate interpretation of the 296 HT events, including the source and sink species. The relative intragenomic age of TEs showed that the 297 most ancient elements belong to the Lepidoptera species Adoxophyes honmai. Most of the flies' elements 298 are ancient and have similar ages. The elements of Asian mosquitoes (An. maculatus, An. epiroticus, An. 299 dirus, An. sinensis) are more ancient than those of African ones (An. gambiae, An. coluzzii, An. arabiensis, 300 An. funestus). Interestingly, the age of the W. bancrofti element is very similar to that of those of the Asian 301 mosquitoes, suggesting that W. bancrofti acquired the element from the Asian Anopheles species and 302 donated it to African species of Anopheles more recently (Fig 6C). Further evidence that this element is 303 prone to transpose horizontally is provided by the multiple pairwise HTT significant signal among 304 mosquitoes ( Fig 6D). 305

306
Most evolutionary biologists are still puzzled by the abundance and variability of the mobilome, 307 even in closely related species. These rather simple sections of DNA are able to override the rules of 308 Mendelian inheritance that govern the rest of the genome by two main ways: hijacking the host molecular 309 machinery to generate more copies of themselves increasing in copy number while the majority of host 310 genes remains stable and through horizontal transfer, a phenomenon that allows TEs to invade the genomes 311 of other species despite being inherited vertically to all original host descendants [3,34]. More than 312 hijacking the host molecular machinery, TEs also can be detrimental to their host due to transposition and 313 recombination [4,35]. Nonetheless, host genomes possess a full arsenal of molecular mechanisms that can 314 silence these elements [36][37][38]. Therefore, the abundance and diversity of the mobilome in the genome of 315 each species is thus the result of an endless arms race between TEs and host [6,39,40]. Although much 316 knowledge has been discovered about the TE "life cycle" since the genomic revolution there are still several 317 open questions about a specific role of horizontal transfer in reshuffling TEs between different eukaryotic 318 species and their impact on receptor species. In this study, we performed an in-depth mobilome 319 characterization in all available mosquito genomes and detected a very large and variable TE content that 320 is constantly exchanged by HTs between mosquitoes and distantly related species. Moreover, we found 321 strong evidence of an intermediate worm species that mediated HT between mosquitoes and showed that 322 horizontally transferred TEs significantly impacted the mosquito genome size. 323 The 24 mosquito genomes available have been scarcely studied from the mobilome perspective, 324 except for the two model species An. gambiae and Ae. aegypti [22,41]. However, even though some TE showed that 29% of the Cu. quinquefasciatus genome was composed of TEs. A further study using a de 338 novo method based on a structural search found other TEs not described at the time in this genome [44]. 339 Here, we describe entire TE superfamilies that had not yet been described in some mosquito species and 340 further expand the TE annotation for Cu. quinquefasciatus reaching a TE content of 43.55%. We also report 341 for the first time the occurrence of Penelope elements in the genome of Cu. Quinquefasciatus and Crypton-342 like elements in the Ae. albopictus genome. Thus, a consistent characterization of the TE content of a given 343 species requires that both de novo and homology-based approaches (with different reference TE database) 344 be used for TE identification and classification. 345 Our in-depth mobilome characterization allowed us to investigate the influence of TEs on the size 346 of the mosquito genomes. We observed that the mosquito genome fraction covered by TE correlates 347 significantly with genome size although it is much more pronounced in species from the Aedes genus. Our 348 15 results corroborate findings in arthropods [45] and vertebrates [46], However, it is interesting that the An. 349 gambiae species complex showed a higher genome size to TEs ratio than the other Anopheles species. This 350 suggests that other sequences are also responsible for genomic expansion in these species such as 351 microsatellites and segmental duplications as shown for other species [47,48]. Another host intrinsic feature that may impact HTT rate is the phylogenetic relatedness of the host 371 species. A number of studies have reported that horizontal transfer occurs more frequently in more closely 372 related species [15,50]. However, long-range HTTs, between distantly related taxa, have also been observed 373 [51,52]. These occur frequently with specific TE groups, such as Class II elements from the Tc1-Mariner 374 [53] and hAT superfamilies [54]. Our data corroborate these findings, showing that most HTTs of 375 Retrotransposons occurred between closely related species of the same genus, and long-range transfers 376 among species of Aedes and Anopheles genus occurred mostly with DNA transposons TEs. Five TEs that 377 underwent transfer to even more distant metazoan species were also Class II TEs from the Tc1-mariner 378 superfamily. A similar pattern was observed in a study that investigated HTT in 195 insect species [15]. In 379 general, HTTs occur more frequently between closely related species, but Class II TEs are more prone to 380 long-range transfers probably due to host factor transposition independency and promoters that make them 381 able to transpose in a wide set of divergent host species [55]. 382 An interesting aspect of the large number of HTT events found among mosquitoes is that it is 383 heterogeneous distributed on the mosquito evolutionary tree, showing a much higher prevalence of HT 384 between species from the Aedes genus ( Figure 2). Although there is a clear bias of more genomes available 385 from Anopheles genus, we still found much more HTTs between Aedes species, the largest mosquito 386 genomes sequenced so far and not a single HT involving Culex quinquefasciatus, a species with 387 intermediate genome size (600Mb) but with a large TE content (around 40%). More unbiased studies 388 including a diverse set of mosquito species should be performed to test these patterns, but our findings 389 highlight that some taxa are more prone to exchange TEs horizontally than others, which is corroborated 390 by other large-scale studies [56]. 391 One of the most debated topics concerning HTT is how exactly do TEs move from one species to 392 another and which vector species bridge the gap between species and facilitate HTT. Several long-range 393 HTT events across phyla involve a Tc1-Mariner element of the irritans subfamily ( Fig 6B). Interestingly, 394 we found a number of copies of these elements in four different genome assemblies of W. bancrofti-a 395 parasitic nematode transmitted by mosquitoes that causes lymphatic filariasis in humans [57]. At least one 396 of these assemblies was reconstructed using only DNA sequences from adult worms isolated from the blood 397 of an infected human patient (GCA_001555675.1), excluding any possibility of mosquito DNA 398 contamination. The detection of copies of the element in independently sequenced W. bancrofti strains 399 confirms that these elements are true components of the genome of this species. The W. bancrofti element 400 has a similar age to that of those found in Anopheles species from Europe and Asia, although this element 401 is younger in African Anopheles species. This suggests that W. bancrofti acquired this element from 402 Measuring the direct influence of horizontally transferred TEs is a difficult task since precise gene 411 annotation of all genomes is required to associate TE insertions with the impact on surrounding genes. 412 However, it is possible to estimate the proportion of the host genome that is derived from horizontally 413 transferred TEs. This can be used as a proxy for estimating the likely impact/burden on the host genome. 414 We found that horizontally transferred TEs contribute to genome size in varying degrees, depending on the 415 mosquito genome. For both species of the Aedes genus, HT derived TEs represent a substantial portion of 416 the genome (5.77-6.23%), that is, around 110Mb in the Ae. albopictus genome. This is similar to the 417 Drosophila melanogaster genome size (143.7Mb) [60] and larger than that of the two-spotted mite 418 Tetranychus urticae (91Mb) [61]. These results show that horizontally transferred TEs reach a high copy 419 number since they replicate unchecked by the host genome after their invasion and can contribute 420 substantially to the expansion and structure of the host genome. In line with that, genome size is also 421 correlated with the proportion of the ge nome covered by horizontally transferred TEs, confirming that 422 HTT contributed substantially to the mosquito genome size ( Figure 2C). 423 Our study shows that complementary methodologies should be used for the precise characterization 424 of the mobilome of a host species. Besides, our main findings are: mosquito mobilome varies in several 425 orders of magnitude and is highly diverse; more than 212 TE families underwent horizontal transfer but we 426 found no association between mosquito species spatiotemporal overlap and HTT rate which differ from 427 other large-scale studies on Insects; there is a significant influence of the mobilome and horizontally 428 annotation. Although each approach has its strengths, the use of either one of these methodologies in 453 isolation almost always leads to an underestimation of the true TE content and adoption of a complementary 454 strategy is therefore advised [42]. 455 19 4.3. Identification of transposable elements using a de novo approach 456 A de novo approach was employed to identify mosquito TEs, using the TEdenovo pipeline from 457 the REPET package [75]. See Supplementary Methods for details of pipeline execution. To complement 458 our initial identification, we also ran RepeatScout, another software that uses a de novo approach [72], on 459 the genome of each species using the default parameters. The repeated elements identified were then passed 460 to the TEdenovo pipeline. 461 Given the virtual absence of elements characterized by TEdenovo or RepeatScout in some genomes, 462 we examined the raw reads of these species further using Tedna [68], to ascertain whether TEs could have 463 been removed before or during the genome assembly step. 464

Identification of transposable elements using a homology-based approach 465
At this stage, the TEs were identified based on homology with previously described TEs in the 466 RepBase and TEfam databases. Initially, we executed a blastn and a tblastn search of the two databases, 467 individually, against the genome of each species. Only HSPs (high-scoring segment pairs) with a bit-score 468 above 200 were retained, to exclude random hits. Thereafter, blastn and tblastn-derived hits were merged 469 into a single file to reconstruct the copies derived from HSPs that matched elements from the same family 470 spaced no more than 1000 nucleotides apart. 471 All TE copies in a given genome were clustered by the CD-HIT-Est [76] algorithm with 80% 472 identity and coverage using a global alignment strategy. Copies of each cluster were then extracted and 473 aligned using the MAP algorithm, which reconstructed the representative consensus of each structural 474 We performed an analysis of the correlation between the fraction of transposons and the genome 488 size of each species studied. This was achieved using the cor() function (method = "spearman") in the R 489 software. The correlation coefficient was used to test the strength of the correlation. Resulting graphics 490 were created using the ggpubr package [81]. 491

492
To determine the inheritance mode of the TEs characterized, either vertical or horizontal, we used 493 the R software's VHICA package [82]. This package compares the relation between dS ratio and codon 494 usage bias (CUB) linear regression of vertically-inherited single-copy orthologous host genes with 495 transposable elements. A vertical transfer is the most likely scenario if the dS-CUB of a TE is not 496 significantly different from that of the host genes. By contrast, a significant deviation in the host genes' dS-497

CUB values indicates horizontal transfer. 498
Although the species that make up the An. gambiae complex are morphologically indistinguishable, 499 we decided to investigate each one separately, in view of the many natural pre-mating barriers that restrict 500 species hybridization. Furthermore, when mating does occur, there is evidence that male progeny are non-501 viable or sterile [83,84]. 502 Fifty randomly selected single-copy orthologous genes from mosquito genomes were obtained 503 from OrthoDB [85]. The ID of each of these genes was used to retrieve the nucleotide coding sequence for 504 each gene from VectorBase (S4 File). The sequences for each ortholog gene set were then codon-aligned 505 using the MACSE software [86]. 506 The TE sequence consensuses of all species recovered using TEdenovo, excluding chimeric 507 elements, were submitted to clustering by the CD-HIT-est algorithm, with 80% identity and 80% coverage 508 using the refinement parameter (-g 1) and global alignment. As a validation method, we extracted the copy 509 that was most similar to each consensus (based on the best bit-score hit) by conducting a BLAST search 510 for each consensus against the genome of its respective species. This created a second set of clusters with 511 the same structure as the consensus clusters, using the TE copies instead of the TE consensus. Clusters of 512 representative TE copies remaining in the analysis were as follows: i) those having sequences in at least 513 two species; ii) those in which all sequences had at least 600 nucleotides; iii) those having at least one 514 sequence with ORF codifying a polypeptide greater than 200 amino acids in size. We used MACSE 515 software to perform codon alignment for each cluster/family (as defined by Wicker et al.), taking the 516 nucleotide sequence of the largest ORF found among the clustered sequences as a reference (-seq 517 parameter) and the remaining sequences as a FASTA file in the -seq_lr parameter. The flanking regions of 518 sequences, based on the beginning and end of the reference sequence after alignment, were trimmed, as 519 only the coding region was of interest for the dS-CUB analysis. 520 The alignment of the orthologous genes and TEs were then passed as input to the VHICA package 521 to ascertain whether vertical/horizontal transfers occurred. Those TE clusters whose p-value was less than 522 0.01 in a one-tailed statistical test were considered horizontal transfer events. Additionally, the percentage 523 of the genome involved in horizontal transfer events, as well as the K2P parameter for each TE family in 524 each species, was calculated using the RepeatMasker software and its auxiliary scripts. 525 To investigate whether any of the TE families that were involved in horizontal transfer events 526 among the 24 studied mosquito species could also be involved in horizontal transfers to other species, we 527 performed a blastn (dc-megablast) of the sequences of these TEs against: the NCBI nt database; all genomes 528 of protostomes, plants, fungi, protists, flatworms, viruses, Echinodermata, Hemichordata, and Chordata 529 organisms present in NCBI as of January 2019. Those matches that had a high degree of identity at the 530 nucleotide level, with more than 80% of the mosquito's TE sequence coverage, and those that had a copy 531 number greater than five were considered to be probable HTTs. When a species had more than one assembly 532 from different samples, we dispensed with the need for a copy number if the matches were present in the 533 majority of the different genome assemblies of this species. The dating of HTT events was performed by applying the formula T = k/2r [87], where T is time, 536 k is the synonym substitution rate (dS) between TE copies from two species, and r is the evolutionary rate 537 of the species groups. We obtained dS estimates per mosquito taxon from the 50 single-copy ortholog genes 538 as follows: 17.567 x 10 -3 mutations per million years for transfers among mosquitoes from the Anophelinae 539 subfamily, 9.205 x 10 -3 for transfers within Culicinae subfamily of mosquitoes, and 10.006 x 10 -3 for 540 transfers between the Anophelinae and Culicinae subfamilies (S1 File). It should be noted that we found 541 very few orthologous TE copies and these were restricted to species of the An. gambiae complex. As the 542 speciation time is not well defined among these species, we performed HTT dating using host gene 543 estimates. The distribution region of each mosquito species of the genus Anopheles was taken from the 546 distributions presented by  and the distribution predicted using the Malaria Atlas Project 547 [92]. These regions were considered the ancestral habitats of these mosquitoes since most Anopheles 548 mosquitoes are not invasive species and thus disperse very little. We consider sub-Saharan Africa to be the 549 ancestral habitat of Ae. aegypti [93] and the east and southeast Asian region extending to India to be the 550 native habitat of Ae. albopictus [94]. To evaluate whether the overlap and/or proximity of the mosquitoes 551 species distribution has any impact on the likelihood of the horizontal transfer, a point-biserial correlation 552 analysis was performed using the cor.test () function of the R software, considering the number of transfers 553 that occurred between two species of mosquitoes and their geographical overlap or non-overlapping 554 distribution. 555