Between 2020 and 2023, we collected 260 adult apple snails (Pomacea spp.) from 15 cities across 11 provinces in mainland China (Suppl. material 1: table SS1; Fig. 1). All specimens were preserved in 100% ethanol or frozen at –80 °C immediately after collection. Genomic DNA was extracted from foot muscle tissue using the 5 fz Genome DNA AutoExtraction Kit- DeepWell Plate (Magnetic Beads) by Nuoto™ AutoExtracter-32 Nucleic Acid Extractor (KANGMA-HEALTHCODE (SHANGHAI) BIOTECHCO, China) according to the manufacturer’s instructions. DNA concentration and integrity were assessed using a Qubit 3.0 Fluorometer (Life Technologies, USA) and 1% agarose gel electrophoresis. Paired-end libraries (~300 bp insert size) were constructed using the MGIEasy DNA Library Prep Kit and sequenced on the DNBSEQ-T1 sequencer (MGI Tech, Shenzhen, China) at BGI-Shenzhen, yielding 100 bp paired-end reads. In addition, publicly available whole-genome sequencing data of 168 individuals of the genus Pomacea (originally labeled as P. canaliculata) were collected from NCBI (BioProject PRJNA951872) (Lu et al. 2024) (Fig. 1).

Raw reads were initially processed to remove adapters and low-quality bases using TRIMMOMATIC (v0.39) (Bolger et al. 2014) software with the following parameters: ILLUMINACLIP:adapter.fa:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:35. To enrich for mitochondrial DNA, clean reads were mapped using BWA (v0.7.17-r1188) (Li and Durbin 2009) with the mem algorithm against a combined reference consisting of the P. canaliculata nuclear genome (GCF_003073045.1; Liu et al. 2018) and representative mitogenomes of P. canaliculata, P. maculata, and P. occulta. Subsequently, properly paired reads mapping specifically to the mitochondrial sequences were extracted and converted to paired-end FASTQ format using SAMTOOLS (v1.17) (Danecek et al. 2021).

The mitogenome for each individual was assembled by NOVOPLASTY (v4.3.1) (Dierckxsens et al. 2017), MITOZ (v2.4) (Meng et al. 2019), and GETORGANELLE (v1.7.7) (Jin et al. 2020), and the assembly with the best support was selected for subsequent analysis based on gene completeness, circularization status, and read-mapping support. To assess the assembly quality, raw reads were mapped back to these assemblies using BWA mem algorithm (v0.7.17-r1188), and the alignments were visualized in IGV (v2.16.0) (Robinson et al. 2011). Visual inspection primarily targeted the protein-coding regions to ensure the absence of assembly errors. The final mitogenomes were annotated using the annotate module of MITOZ (v2.4) (Meng et al. 2019). Gene boundaries, specifically start and stop codons, were manually validated and curated by comparison with the reference sequences. The 13 protein-coding genes (PCGs) were aligned using MAFFT (v7.526) with the --auto strategy (Katoh and Standley 2013). To establish a comprehensive global dataset, we additionally retrieved 2,704 publicly available COI sequences from GenBank. This public dataset was rigorously filtered to exclude sequences containing ambiguous residues, exhibiting insufficient overlap with the target alignment region, or lacking precise geographical metadata. Accession numbers, geographic information, and corresponding references for public COI records are provided in Suppl. material 1: table S2 and Suppl. material 2: table S14. Finally, these public sequences were combined with our newly sequenced data and aligned using CLUSTAL OMEGA (v1.2.4) (Sievers et al. 2011).

Phylogenetic analyses were conducted based on both the COI sequences and the concatenated 13 mitochondrial PCGs. For the COI dataset, four congeneric species were used as outgroups: P. diffusa (MF462137.1), P. lineata (FJ710311.1), P. paludosa (EF514960.1), and P. scalaris (EU528585.1). For the mitogenome dataset, the tree was rooted using P. aulanieri (NC_084053.1), P. diffusa (NC_041142.1), and P. reevei (NC_084052.1) (Suppl. material 2: table S14). Public sequence labels were checked against their phylogenetic placement relative to reference sequences. Bayesian phylogenetic inference was performed using MRBAYES (v3.2.7) (Ronquist et al. 2012), with substitution model uncertainty accommodated by setting lset nst = mixed rates = invgamma. Two independent runs, each comprising one cold and three heated Markov chains, were run for 10,000,000 generations, sampling every 1000 generations. Convergence was confirmed when the average standard deviation of split frequencies fell below 0.01. The first 25% of samples were discarded as burn-in, and a 50% majority-rule consensus tree was constructed from the post-burn-in trees. Maximum-likelihood phylogenetic analysis was performed using IQ-TREE (v2.3.6) (Minh et al. 2020), with the best-fit substitution models selected by MODELFINDER (Kalyaanamoorthy et al. 2017). Branch support was assessed using 1,000 replicates of ultrafast bootstrap approximation (Hoang et al. 2018) and SH-like approximate likelihood ratio test (aLRT).

Nucleotide composition, AT/GC skew, and codon usage bias for the 13 PCGs were calculated using MEGA11 (Tamura et al. 2021). Haplotype statistics, including number of haplotypes, haplotype diversity (Hd), nucleotide diversity (π), average number of nucleotide differences (k), number of variable sites, singleton sites, and parsimony-informative sites, were estimated using DNASP v6 (Rozas et al. 2017). Neutrality tests (Tajima’s D, Fu’s Fs, Fu & Li’s D* and F*) and mismatch distribution analyses were also performed by DNASP v6. Median-joining haplotype networks for COI and concatenated PCGs were constructed using POPART (v1.7) (Leigh et al. 2015) with the TCS algorithm. Pairwise uncorrected p-distances were calculated from the 13-PCG haplotype alignment using MEGA11 (Tamura et al. 2021). Gaps were handled using the partial deletion option, retaining only alignment positions present in at least 95% of sequences.

Approximate Bayesian computation with random-forest model choice (ABC-RF) was used to evaluate alternative invasion scenarios for the Kenyan P. canaliculata population (Pudlo et al. 2016; Collin et al. 2021). Because mitogenomic data were unavailable for Kenya, this analysis was based on the 556-bp COI fragment. Sequences were grouped into four populations: Kenya, Asia, the native range of P. canaliculata in South America, and other introduced regions. All Kenyan sequences were retained, and ten replicate datasets were generated by randomly sampling 20 sequences from each of the other groups. Five scenarios were tested using DIYABC-RF (v1.1.54), including direct native-range introduction, Asia-mediated introduction, introduction from another invaded region, and two Asia-involving admixture scenarios. For each replicate, 100,000 datasets were simulated under a K2P model with broad demographic priors, and model choice was performed in ABCRANGER (v1.16.69) using 1,000 trees. Because COI provides limited resolution for detailed route inference, scenarios were also grouped into non-Asia alternatives (S1 and S3) and Asia-related alternatives (S2, S4, and S5).

Raw sequencing reads newly generated in this study have been deposited in NCBI Sequence Read Archive under BioProject PRJNA1456661 (SRA run accessions SRR38230055–SRR38230314). The corresponding mitochondrial genome sequences have been submitted to GenBank under accession numbers PX899771–PX900030. These newly generated mitogenome assemblies and their complete annotations have also been deposited in GenBase of the National Genomics Data Center (NGDC), Beijing Institute of Genomics, Chinese Academy of Sciences/China National Center for Bioinformation (https://ngdc.cncb.ac.cn/genbase/) under accession numbers C_AA137052–C_AA137311 (e.g., the sequence C_AA137052 can be found at https://ngdc.cncb.ac.cn/genbase/search/gb/C_AA137052). Public genome sequencing data used for additional mitogenome assembly were obtained from Lu et al. (2024) under BioProject PRJNA951872, with SRA run accessions listed in Suppl. material 2: table S15. The mitogenome assemblies generated from these public data are available in GenBase under accession numbers C_AA136884–C_AA137051. All mitogenome assemblies are also available in the China National Sequence Archive (CNSA) under accession numbers CNA0749791–CNA0750218. Other supporting data are available from Figshare (Fan et al. 2026).

We obtained and analyzed 428 mitogenomes from three Pomacea species, including P. canaliculata (n = 317), P. maculata (n = 21), and P. occulta (n = 90), representing more than 40 localities in China and several Southeast Asian countries (Fig. 1). Among them, 260 mitogenomes were newly generated in this study and 168 were derived from the public data. Of the 428 mitogenomes, 322 were classified as confidently circularized. All Pomacea mitogenomes retained in the final dataset contained the complete set of 13 protein-coding genes. Mitogenome lengths across the three Pomacea species exhibited slight intraspecific variation: the total length of the mitogenome ranged from 15,453 bp to 15,916 bp in P. canaliculata, 15,561 bp to 15,690 bp in P. maculata, and 15,495 bp to 15,696 bp in P. occulta (Suppl. material 1: tables S3–S5).

To reconstruct global invasion pathways, we retrieved 2,704 COI sequences from GenBank (Suppl. material 1: table S2). These sequences covered the native range in South America as well as invasive populations across Asia, North America, Europe, and the Pacific islands. When integrated with 428 COI fragments extracted from our newly sequenced mitogenomes, these sequences yielded a comprehensive dataset of 3,132 COI sequences (556 bp) (Suppl. material 1: table S6). Notably, all sampled P. occulta sequences were derived exclusively from Chinese populations, whereas samples for other species represented a broader global distribution.

We identified a total of 131 unique COI haplotypes across the three Pomacea species. We first corrected the species information associated with sequences in public databases through phylogenetic relationship analysis (Suppl. material 1: table S2). The haplotype networks based on this curated dataset revealed distinct genealogical structures for each species (Fig. 2). Among the 3,132 sequences, P. canaliculata exhibited the highest number of haplotypes (n = 75), followed by P. maculata (n = 49) and P. occulta (n = 7) (Suppl. material 1: tables S6, S8). In addition, the COI-based phylogenies inferred by maximum-likelihood (Suppl. material 2: fig. S1a) and Bayesian inference (Suppl. material 2: fig. S1b) both showed well-supported monophyletic clades for each species but consistently placed P. canaliculata and P. occulta as sister taxa, rather than the pairing of P. maculata and P. occulta reported in previous studies.

For P. canaliculata, the network was dominated by two massive star-like clusters centered on the PC2 and PC3 haplotypes, each representing more than 1,000 individuals (Fig. 2A). The cluster centered on the PC2 exhibited a widespread, trans-oceanic distribution, encompassing native lineages from Argentina and Chile, as well as introduced populations in Papua New Guinea, Hawaii, and widespread Asia. In contrast, the PC3 cluster appeared to be endemic to East Asia, with samples from mainland China, Korea, Japan and Hong Kong sharing this haplotype. Notably, the sharing of the central haplotype PC2 between native and invasive ranges was exclusively observed with Argentine populations, including those sharing PC17 and PC9. In stark contrast, other native populations, such as those from Chile and Uruguay, did not share any haplotypes with the introduced regions. Specifically, sequences from Chile formed distinct tips radiating from the PC2 haplotype, while Uruguayan haplotypes were located elsewhere in the network, entirely separate from the main invasive clusters. Within mainland China, the central haplotype PC2 was present in almost all sampling sites, with the sole exception of the Beijing population (Suppl. material 2: fig. S2a). The network also featured several secondary high-frequency haplotypes, such as PC1, PC4, PC5, and PC6, which consisted primarily of samples from mainland China and other Asian regions. Of particular note, PC1 provided a unique genetic link between the extensive Asian population and the invasion in Kenya, Africa. For P. maculata (Fig. 2C), the network revealed two distinct lineages. The PM78 cluster, shared with native populations from Brazil, encompassed the vast majority of samples from mainland China. Specifically, the genetic diversity within mainland China was restricted to only three haplotypes: the dominant central PM78 and two minor derived haplotypes, PM80 and PM124. In contrast, the second lineage was anchored by haplotype PM79, which was shared by native populations in Argentina and Uruguay and invasive populations in China-Taiwan and South Korea; notably, samples from China-Taiwan were detected in both PM78 and PM79 lineages. This PM79 node is centrally positioned and directly links to two major derived haplotypes, PM77 and PM82, which are dominated by samples from the United States, whereas mainland China was completely absent from this entire lineage, indicating a distinct genetic composition from that of the US. The haplotype network for P. occulta displayed a single, star-like phylogeny centered on the dominant haplotype PO125 (Fig. 2B). Although no sequences from outside mainland China were available for this species in our dataset, P. occulta harbored seven haplotypes within mainland China, surpassing the three identified in P. maculata.

In this study, we found that the 13 concatenated PCGs provided much higher resolution in delineating phylogenetic relationships than the single-locus COI fragment. The globally dominant COI haplotypes were resolved into multiple mitogenomic haplotypes. Specifically, the COI haplotype PC2 was subdivided into nine distinct mitogenomic haplotypes which formed Clade a in the mitogenome network. The PC_Hap9 (n = 54) was broadly distributed across Fujian, Guangxi, Hunan, Jiangxi, and Sichuan, whereas PC_Hap14 (n = 25) was detected in Guangdong, Sichuan, Yunnan, and Vietnam. Most PC2-derived mitogenomic haplotypes showed regional clustering, but the 13-PCG data also revealed local heterogeneity within the same COI haplotype; for example, PC2 samples from Shanghai were separated into PC_Hap1 and PC_Hap24, and samples from one Chengdu site into PC_Hap9 and PC_Hap14. Similarly, the PC3-derived haplotypes accounted for 10 of the 17 mitogenomic haplotypes in Clade c, which was dominated by PC_Hap3 (n = 119). Furthermore, the PC1, PC6, PC9, and PO125 were also divided into several mitogenomic haplotypes. In contrast, PM78 corresponded to a single mitogenomic haplotype, PM_Hap4. A same-sample comparison between the 556-bp COI fragment and the 13-PCG dataset further confirmed the higher haplotypic resolution of mitogenomes (Suppl. material 2: fig. S3; Suppl. material 1: tables S10, S11). This higher resolution of mitogenomes in phylogeny analysis enabled us to partition the global P. canaliculata into three major distinct lineages (Clades a, b, and c) for the subsequent analyses.

The mitogenome haplotypes separated three Pomacea species into distinct lineages, with the mean interspecific p-distance of 10.17% between P. canaliculata and P. maculata, 10.47% between P. canaliculata and P. occulta, and 7.44% between P. maculata and P. occulta (Fig. 3A; Suppl. material 1: table S9). Within P. canaliculata, the network displayed three distinct lineages, with mean p-distances ranging from 4.20% to 4.55% among Clades a, b, and c. The Clade a and c exhibited extensive geographic overlap, with both clades encompassing populations from Putian, Xiamen, Hangzhou, Changsha, Nanning, and Vietnam. However, the population from Beijing was restricted to the Clade c, while the Clade b was detected only in Shenzhen and Yuxi. The P. maculata population was identified as a single mitochondrial lineage, Clade d. The single haplotype PM_Hap4 was broadly distributed across Jiangling, Chengdu, Hangzhou, Nanchang, and Vietnam. The Clade e consisted of the P. occulta population, and this clade included samples from Chongqing, Yunnan, Hunan, Hubei, Guizhou, Jiangxi, and Shenzhen, with two minor haplotypes specific to Chongqing and Nanchang.

Genetic diversity and neutrality tests revealed distinct patterns in the three Pomacea species and in the major P. canaliculata clades (Table 1). The P. canaliculata showed high haplotype diversity (Hd = 0.813) and the highest nucleotide diversity among the three Pomacea species (π = 0.02498). However, this high overall diversity contrasts sharply with the low nucleotide diversity observed within each clade: Clade a (π = 0.00082), Clade b (π = 0.00002), and Clade c (π = 0.00432). This pattern indicates that the high overall nucleotide diversity of P. canaliculata mainly resulted from the divergence among clades. Furthermore, neutrality tests for the separated clades revealed divergent patterns. Clade a presented a significant positive Tajima’s D (2.329), and Clade c showed a non-significant positive value (0.396). In contrast, Clade b exhibited negative values for all indices (Tajima’s D = –1.11; Fu’s Fs = –0.34), although these values did not reach statistical significance. For the other invasive species, the genetic variation was extremely limited. The P. maculata population exhibited zero nucleotide diversity as it was represented by only one haplotype, which precluded neutrality testing. Similarly, the P. occulta population displayed low diversity, with an Hd of 0.107 and π of 0.00001, alongside non-significant negative neutrality indices.

Given the deep genetic divergence revealed in our network analysis, we performed lineage-specific analyses using the COI dataset to characterize demographic patterns while minimizing the confounding effects of population substructure. For the invasive P. occulta, we detected strong and consistent signals of expansion, evidenced by significant negative values for Tajima’s D (–2.0, P < 0.05) and Fu’s Fs (–13.80) (Suppl. material 1: table S12), complemented by a strict unimodal L-shaped pattern of the mismatch distribution (Suppl. material 2: fig. S4). In the case of P. canaliculata, a sharp demographic contrast was observed between the invasive and native populations. Within the invasive populations, the Clade a exhibited the strongest signal of population expansion, characterized by significantly negative neutrality indices (Fu’s Fs: –68.38; Tajima’s D: –2.43, P < 0.001) and a strict L-shaped mismatch distribution (Suppl. material 2: fig. S5). The dominant Clade c displayed a non-significant Tajima’s D; however, its Fu and Li’s D* (–7.33) and F* (–5.27) values were significantly negative (P < 0.02). In contrast, the minor Clade b presented a different profile, exhibiting positive neutrality indices, such as a Tajima’s D of 1.42, which likely reflects a recent bottleneck. Native-range samples generally showed non-significant or positive neutrality indices and multimodal or ragged mismatch distributions (Suppl. material 2: fig. S5), reflecting long-term demographic stability in their home ranges. For the P. maculata, demographic patterns diverged significantly between the two major invasive groups. The group of mainland China displayed significant negative values in neutrality tests (Tajima’s D: –1.97, P < 0.05; Fu’s Fs: –4.51) and a steep L-shaped mismatch distribution (Suppl. material 2: fig. S6), indicating a recent rapid expansion. In contrast, the USA-dominated clade exhibited positive values for all neutrality indices, including a Fu’s Fs of 0.703, and a broad mismatch peak spanning 2–4 differences, reflecting a distinctly different demographic history compared to the rapidly expanding Chinese population.

The recently reported occurrence of P. canaliculata in Kenya provided an opportunity to examine the origin of this new African invasion. In the COI network, Kenyan samples shared the haplotype PC1 with Asian invasive populations, suggesting a possible maternal connection between Kenya and Asia (Fig. 2). To evaluate whether the Kenyan population was directly introduced from the native range or secondarily introduced through invaded regions, we performed COI-based ABC-RF scenario testing. Five detailed scenarios were compared, including direct introduction from the native range, introduction through Asia, introduction through other invaded regions, and Asia-involving admixture routes. The five-scenario comparison did not identify a single well-supported route, as selected scenarios varied among replicate analyses and posterior probabilities were moderate, ranging from 0.400 to 0.517 (Fig. 4A, B). We therefore grouped the scenarios into broader Asia-related and non-Asia alternatives. In this grouped comparison, Asia-related scenarios were selected in 9 of 10 replicate analyses, with posterior probabilities of 0.507–0.652 (Fig. 4C; Suppl. material 1: table S13). These results provide tentative support for an Asia-related origin of the Kenyan population, although the precise invasion pathway remains unresolved.

Our study demonstrates that full mitogenomes provide higher resolution in phylogenetic analysis than the classical DNA barcoding marker. Previous studies based on the COI fragment have established a broad framework for Pomacea invasions in China, including multiple maternal lineages in P. canaliculata population and a more restricted mitochondrial composition in P. maculata (Lv et al. 2013; Yang et al. 2018b; Liu et al. 2019b). Our results are consistent with these previous studies, but show that the mitogenomic data can further refine the population genetic structure in apple snails. Our single-locus COI phylogeny showed that the P. canaliculata and the P. occulta are sister taxa (Suppl. material 2: fig. S1), which was also reported in previous COI-based analysis (Lv et al. 2013). However, the 13-PCG dataset showed that the P. maculata and the P. occulta are sister lineages (Suppl. material 2: fig. S7), consistent with most previous mitogenomic studies (Yang et al. 2018b; Yang and Yu 2019; Yang et al. 2022). We attributed this discrepancy to a topological artifact that may have resulted from mutational saturation in short fragments (Xia et al. 2003; Chan et al. 2022). Furthermore, the mitogenomic data unveiled hidden genetic diversity; for instance, the globally dominant PC2 and PC3 haplotypes were resolved into 9 and 10 distinct mitogenomic haplotypes, respectively, several of which showed distinguishable geographic distributions. The detailed genetic information provided by mitogenomes allowed us to distinguish lineages that were difficult to be identified based on short DNA fragments. Such higher resolution can facilitate the tracing of secondary spread, identifying potential source regions of newly invasive populations, and improving monitoring of ongoing population expansion, which complemented the study based on genome-wide variations (Lu et al. 2024). Our analyses added a maternal-lineage perspective, while recognizing that populations sharing the same maternal haplotype may still differ at the genome-wide level because of recombination, admixture, drift, or local adaptation (Galtier et al. 2009; Toews and Brelsford 2012). Nonetheless, this study provides essential genetic data for future global monitoring of invasive apple snails.

Our results demonstrated that Pomacea species in China have established populations with distinct maternal-lineage patterns. For the P. canaliculata, high genetic diversity has been maintained through three genetically distinct clades. The genetic differences between the three clades likely resulted from pre-existing population divergence rather than accumulated mutations in each population after introduction because of the short invasion history of apple snails in Asia and mainland China (Mochida 1991; Yang et al. 2010). Notably, these distinct lineages co-occur in many different regions, from coastal Fujian and Zhejiang to inland Hunan and Guangxi. This geographic overlap supports human-mediated dispersal rather than natural spread. Anthropogenic vectors, such as aquaculture (Naylor 1996) and the aquarium trade (Patoka et al. 2018), have likely transported these ancestral lineages across vast distances, resulting in anthropogenic sympatry among clades that were historically isolated in their native range. This admixture counteracts the founder effect, potentially enhancing the invasion success of P. canaliculata. In contrast, the P. maculata populations in this study showed extreme genetic uniformity. All sampled individuals shared a single haplotype (PM_Hap4), which was broadly distributed across several localities including Jiangling, Chengdu, Hangzhou, Nanchang, and Vietnam, suggesting that the regional establishment and spread can also occur with a restricted maternal composition (Schrieber and Lachmuth 2017). This implies that the initial invasion success of the P. maculata may have been driven by the proliferation of one well-adapted lineage. However, broader sampling and genome-wide analyses are still needed to verify this pattern and clarify the invasion mechanism of P. maculata. It is worth noting that the invasive threat from P. occulta may have been underestimated, because P. occulta has a wide distribution and high diversity in China. This suggested that the P. occulta may have a stronger dispersal potential or a more complex introduction history than the P. maculata. Given its resemblance to P. canaliculata, the P. occulta has likely been misidentified in the past. Molecular identification is therefore needed in future management to monitor this widespread but more cryptic invader.

Integrating our mitogenomic data with global COI haplotypes allowed us to better resolve the invasion corridors of Pomacea species. For the P. maculata, our analysis suggested that introduced populations in Asia involve at least two maternal lineages. The first one was represented by the PM78 lineage (Fig. 2), which was inferred to have originated from Brazil based on the haplotype sharing analysis. This lineage is ubiquitous across China, including China-Taiwan and Hong Kong, as well as Japan and Southeast Asia, supporting a Brazil-associated P. maculata lineage in East Asia (Hayes et al. 2008; Yang et al. 2018b; Banerjee et al. 2024). The second was represented by the PM79 lineage, which was inferred to be derived from Argentina or Uruguay (Fig. 2). This lineage has successfully established populations in China-Taiwan and South Korea but was completely absent from mainland China. In addition, this lineage also connected to the PM77 and PM82 haplotypes, which were dominated by samples from the United States. These results suggest that invasion histories inferred from mainland China should not be generalized to China-Taiwan, which has been recognized as an early introduction area and a potential source region for apple snail spread in East Asia (Hayes et al. 2008; Yang et al. 2010).

Regarding the P. canaliculata, our data supported that the primary invasion source for this species is Argentina, with close connections between haplotypes of the two countries (Yang et al. 2018b). The coexistence of several high-frequency haplotypes in the Chinese P. canaliculata population supported multiple introductions and the following population expansions. Interestingly, we provided key insights into the invasion pathway of the African population. Our results showed that the PC1 was shared with the invasive populations in Asia and Kenya (Fig. 2), suggesting a possible maternal connection between Kenya and Asian invasive lineages. The ABC-RF analyses provided tentative support for an Asia-related origin and a possible bridgehead-like invasive pathway for the Kenyan population. This genetic connection between Kenya and Asia populations aligns with trade and agricultural relationships between the two places (Liu et al. 2019a). The invasion of the P. canaliculata in Kenya was first reported in 2021, a timeline that coincides with the intensification of trade and agricultural partnerships between Asia and East Africa (Liu et al. 2019a; Buddie et al. 2021). Furthermore, Asia currently serves as a global hub for the aquarium trade and aquaculture, acting as a major redistribution center for aquatic species (Padilla and Williams 2004). Therefore, while a direct introduction from an unsampled South American population couldn’t be strictly ruled out, the convergence of genetic data and trade evidence still supported the hypothesis that established Asian populations served as the bridgehead for the recent spread of P. canaliculata into Africa.

In P. canaliculata, we observed the coexistence of deeply divergent maternal lineages within China. When these samples were analyzed as a single population, statistical tests such as Tajima’s D and mismatch distribution analyses generated misleading signals of population stability. However, this stability is an artifact that resulted from mixed sampling of two genetically distinct groups, a phenomenon consistent with findings in the brown anole lizard, Anolis sagrei, where the admixture of divergent lineages inflated genetic diversity (Kolbe et al. 2004). Previous COI-based studies also used neutrality tests and mismatch distributions to infer demographic signals in invasive Pomacea populations (Liu et al. 2019b; Yang et al. 2022; Yi et al. 2024), but most analyses did not explicitly separate the divergent maternal lineages resolved in this study. By separating the COI data into specific clades, we were able to reveal that both Clade a and Clade c exhibit clear signs of recent expansion. However, the 13-PCG dataset showed a different pattern. Considering the smaller sample size of the 13-PCG dataset and the fact that all nine mitogenomic haplotypes in Clade a corresponded to the single COI haplotype PC2, the broader COI dataset is likely more informative for detecting lineage-level demographic signals. However, the 13 PCG dataset performs better at revealing hidden genetic structures that are difficult to detect using COI haplotype data. This suggests that lineage-specific analyses, when supported by sufficient data, are important for accurately interpreting biological invasion processes, since treating the species as a whole can mask distinct demographic signals among divergent maternal lineages (Simonsen et al. 1995).