New Insight into the Demography History, Evolution, and Phylogeography of Horseshoe Crabs with Special Emphasis on American Species

Abstract

Xiphosurids (Merostomata, Xiphosura) are a group of chelicerates with a rich and complex evolutionary history that is constantly being updated through new discoveries. In this study, we re-estimated the divergence time of the extant horseshoe crab species with new fossil calibration points and addressed the inter- and intraspecific relationships of the American horseshoe crab through a phylogeographic perspective. In order to achieve our objectives, three datasets were compiled from fragments of different lengths of the COI gene that include sequences from 154 individuals, representing the Mexican populations. In addition to these, the datasets also included previously published sequences corresponding to individuals from different US populations and Asian horseshoe crab species. Firstly, we estimated the divergence times of extant horseshoe crab species by Bayesian methods using multiple fossil calibration points. Subsequently, we investigated the phylogeographic relationships and demographic history of Limulus polyphemus in the Americas utilizing various datasets. The time of divergence of the two Asian species clades was estimated to be approximately 127 million years ago (Ma). Phylogeographic relationships between the Asian and American species are linked through a minimum of 86 mutational steps. In America, phylogeographic relationships reflect differentiation between US and Mexican populations of L. polyphemus. We detect signs of demographic expansion for the Mexican population during the last 75,000 years, as well as an absence of phylogeographic structuring. The evolutionary history of horseshoe crabs is older than previously believed; however, the current distribution and demographic changes have probably been influenced by environmental events of the recent past, such as the glacial–interglacial periods that occurred during the Pleistocene.

1. Introduction

Xiphosurids (Merostomata, Xiphosura), also known as horseshoe crabs, are one of the two extant marine chelicerates groups with a fossil record dating back as far as 480 Ma during the Ordovician [1]. Due to their apparent ecological and morphological conservation, they have been considered as “living fossils” [2], even if this term is subject to numerous debates, with various definitions and alternatives proposed (see synthesis in [3]). Despite the prevailing notion that these organisms have remained relatively unchanged over time, the fossil record has unveiled a complex and dynamic evolutionary history, with 76 species exhibiting morphological and ecological variation [4,5]. Xiphosurids have experienced at least five invasion events, with shifts in niche occupation from marine generalists to specialists inhabiting brackish/freshwater environments throughout their evolutionary history, resulting in the radiation of two of the largest Xiphosurid clades (i.e., Bellinurines and Austrolimulids) during the Paleozoic and Mesozoic [6]. These two clades became extinct, respectively, during the Permian and Cretaceous mass extinction events, with only the Limulid clade remaining as a representative of the group [3].

Despite having survived all five Phanerozoic mass extinctions [7], at present, the order Xiphosura is composed of four extant species within the family Limulidae (Merostomata, Xiphosura), representing merely a fraction of the group’s historical diversity, both ecologically and morphologically [6]. All four species are restricted to brackish-marine environments and are considered generalists [8]. The American horseshoe crab, Limulus polyphemus Fabricius, 1973, has a disjunct distribution along the Atlantic coast of North America, from the Gulf of Maine to the coastal islands of eastern Louisiana in the United States (US), and from Laguna de Términos (Campeche State) to Bahía de la Ascensión (Quintana Roo State) in the Yucatán Peninsula, Mexico (Figure 1a) [9,10]. The Asian horseshoe crab species, coastal horseshoe crab Tachypleus gigas Müller, 1785, tri-spine horseshoe crab Tachypleus tridentatus Leach, 1819, and mangrove horseshoe crab Carcinoscorpius rotundicauda Latraeille 1802, inhabit the shallows waters of the Indo-Pacific in Southeast Asia (Figure 1b) [11]. The current distribution of horseshoe crabs is determined by four large-scale environmental parameters, which influence their dispersal between estuaries: (1) continental geomorphology (e.g., presence of estuarine environments, continental shelves as dispersal pathways, or ocean depths as physical barriers), (2) temperature, (3) tidal types (determinant for the choice of potential spawning sites), and (4) benthic currents [8,12]. In North America, low temperatures limit the northern distribution of L. polyphemus, while amplitude, tidal types, and the depth of the seafloor (>50 m) impede the establishment of spawning populations in the western section of the Gulf of Mexico (resulting in its disjunct distribution). The limitation to their distribution further south may be due to the lack of suitable habitats for spawning and juvenile development in conjunction with different conditions that may create barriers (e.g., narrow continental shelf or current circulation patterns) [8,12].

Phylogenetic analyses currently indicate that the order Xiphosura is positioned within the class Arachnida, as a sister group to Arachnopulmonata (Tetrapulmonata and Scorpiones) [15]. These findings suggest that Arachnida is paraphyletic with respect to the horseshoe crabs [16]. At the family level, the phylogenetic relationships of the four extant species of horseshoe crabs reveal the sister-group relationship of the subfamilies Limulinae (L. polyphemus) and Tachypleinae (Asian horseshoe crabs) [17]. The divergence between the extant Asian species and the American species has been linked to the opening of the Atlantic Ocean, approximately 150–130 Ma ago [18]. The fossil species Mesolimulus walchi (Demarest 1822), which inhabited Europe during the Cretaceous, a geological time close to the Atlantic opening, has been used to calculate divergence times, as it is considered the representative of the stem group of extant horseshoe crabs [17]. However, the fossil record of this group extends back to the Triassic, with the oldest species recorded as Tachypleus syriacus Woodward, 1879 (=Mesolimulus syriacus). This suggests an underestimation of the divergence times of this group by at least 90 Ma [19]. The presence of unresolved relationships or uncertainties among fossil taxa has implications for the estimation of the divergence time of extant species. Consequently, the use of multiple fossil calibrations, considering the most current hypotheses of species fossil relationships, may improve the estimation of divergence times [20]. Understanding the evolutionary history of species through their relationships and diverging times is imperative when interpreting how extant species may respond to current or future changes [4]. This is of particular importance in the context of developing conservation strategies for species that are experiencing diverse challenges across their respective geographical ranges (see [21,22]).

The American horseshoe crab (L. polyphemus), also known as Cacerolita de Mar or Me’ex (from the Mayan language) in Mexico [10], has been the focus of significant research due to its economic importance in the fisheries and pharmaceutical sectors. Most research has been conducted on the distribution of the species in the United States and has focused mainly on ecological (e.g., [12,23]), physiological (e.g., [24]), demographic (e.g., [25]), and genetic aspects [26,27,28]. Among the main findings of the genetic and demographic studies are the detection of discontinuities or phylogeographic breaks [26,29], the structuring into five genetic regions or populations along the east coast of the United States [27,28,30], and an analysis of demographic shifts within the species’ core range in the United States over the past 150 years that alludes to anthropogenic activities such as overharvesting and habitat modification as probable contributing factors [25]. In the southernmost distribution of the species, the Yucatán Peninsula, there are at least two genetically distinct populations: one located in the southern Gulf of Mexico and the other covering the northern coast of Yucatan and the Caribbean coast [31]. Recent studies in these populations suggest the absence of bottlenecks [31,32]. However, the demographic history of these populations remains to be fully elucidated. Given the incipient knowledge concerning Mexican populations, as well as the intrinsic habits of the species (e.g., low larval vagility and female philopatry), it is important to address the genetic structuring from a phylogeographic perspective throughout its Mexican distribution in order to ascertain the current regional status.

The International Union for Conservation of Nature (IUCN) has identified L. polyphemus as a species at risk of extinction in certain regions. This is attributed to the decline of small and vulnerable populations due to overexploitation, habitat loss caused by shoreline modification, and climate change (e.g., sea level rise and abnormal weather events) [21,33]. Globally, this species is categorized as ‘vulnerable’ according to the IUCN Red List; however, in Mexico, it is listed as ‘in danger of extinction’ (NOM-059-SEMARNAT-2010). Three spatial units have been defined in this region, characterized by their ecological conditions, genetic structure, and the particular risk they face [21]. The major risks faced by Mexican populations are of anthropogenic origin, such as habitat loss caused by coastal development, contamination of water bodies, and illegal fishing, particularly relevant on the Yucatán north coast, where adult horseshoe crabs are used as bait for octopus fishing [10,34]. The effects of climate change on Mexican populations have not yet been addressed but it is anticipated that they will be influenced by disturbances to nesting sites resulting from sea level rise, fluctuations in water temperature, and variations in salinity [21,22,33]. A comprehensive understanding of current population status, their interrelationships, and their historical performance could facilitate the establishment of guidelines for predicting potential responses to both present and future events. This would enable strategic conservation efforts to be focused on those populations at most risk of extirpation.

This work proposes a phylogenetic and phylogeographic perspective that considers all extant species of horseshoe crabs, with the objective of improving the inter- and intraspecific understanding of current and historical relationships, with special emphasis on the American horseshoe crab. In order to achieve this objective, the following three aims have been established: (1) re-estimate the divergence times among horseshoe crab extant species using multiple fossil calibration points, (2) assess the phylogeographic relationships and demographic history of L. polyphemus in different subpopulations at the scale of its entire distribution range, and (3) assess the phylogeographic structure and demographic history of L. polyphemus at the regional scale of its Mexican distribution range. For the first time, the Mexican population data were integrated with available information on extant species of horseshoe crabs to facilitate a phylogeographic and demographic comparison.

2. Materials and Methods

2.1. Datasets Configuration

Sequences from L. polyphemus samples from the Mexican population were obtained (see the following section for details), and in addition, sequences from two different regions of the COI gene were downloaded in order to compile databases addressing the different objectives of the study. For intraspecific analyses, we downloaded 645 bp sequences from L. polyphemus specimens from the US through the GenBank and Barcode of Life Data System (BOLD) databases, as well as those previously reported by Pierce et al. [26] from two Mid-Atlantic populations (Chesapeake and Delaware Bays). For interspecific analyses, we used 477 to 645 bp sequences available in GenBank for the three Asian horseshoe crab species and the tick Rhipicephalus microplus Canestrini, 1888, the latter as an outgroup (Table S1).

We compiled three datasets to meet our objectives (Table S1): (1) the first, hereafter referred to as the Long Fragment (LF) dataset, consists of an 1190 bp fragment of the COI gene sequenced for individuals of L. polyphemus from the Mexican populations (details below); (2) the second, designated as the Short Fragment 1 (SF1) dataset, corresponds to the first 645 bp of the LF, and it includes individuals from the Mexican region (this study), sequences obtained from GenBank and BOLD databases for three different locations of the US region (Florida, Maryland, and Connecticut), sequences for the three Asian species, and the sequence of the outgroup; and (3) the third, designated as the Short Fragment 2 (SF2) dataset, corresponds to the last 496 bp of the LF, which were trimmed in the Mexican individuals in order to align and compare with available sequences from two Mid-Atlantic populations, Chesapeake and Maryland Bays (aligning with 496 bp of the originally 515 bp published by Pierce et al. [26]).

2.2. Study Area and Sample Collection in the Mexican Distribution Range

We sampled at eight locations representing the southernmost distribution area of L. polyphemus in the Yucatán Peninsula, Mexico (Figure 1a): three localities in the Gulf of Mexico, namely Laguna de Términos (N = 31), Champotón (N = 28), and Ría Celestún (N = 25); four localities on the northern coast of Yucatán at Chelem (N = 7), Chuburná (N = 20), Ría Lagartos (N = 6), and Yum Balam (N = 24); and one locality on the Caribbean coast at Sian Ka’an (N = 13). To collect the specimens, we carried out transects of approximately 100 m along the coastline, collecting on nights during a new moon or full moon, when individuals’ activity peaks in the intertidal zone. We used nonlethal sampling, which consisted of removing a fragment of the terminus of one locomotive appendage (approximately 2 cm), after cleaning and using sterile material [35]. Individuals were kept captive in plastic containers with sand and water for less than 30 min and released at the site of collection. All tissue samples were preserved in 96% ethanol and stored at 4 °C until DNA extraction.

2.3. Laboratory Procedures

Total genomic DNA was extracted from the muscle tissue using a high salt extraction protocol [36] and stored at −20 °C. DNA concentration was determined using a Qubit^® 2.0 fluorometer (Invitrogen, Carlsbad, CA, USA) and DNA quality was assessed by electrophoresis in agarose gel (1.0% with 1 × TBE buffer, Promega, Madison, WI, USA) using GelRed^® (Biotium, Fremont, CA, USA) as post-staining. We amplified an 1190 bp of the cytochrome oxidase subunit I (COI) gene using the primers LCO1490 [37] and Lim 582 [26]. PCR amplifications were performed in 25 μL reaction volume containing ~20 ng of DNA template (2 μL), 5 μL of 5x Green Buffer (Promega, Madison, WI, USA), 0.4 μL of dNTP Mix (Promega, Madison, WI, USA), 2 μL of MgCl2 (Promega, Madison, WI, USA), 1.3 μL of each primer and 1 U of Taq Polymerase (0.2 μL) (GoTaq Flexi, Promega, Madison, WI, USA), and the total volume was completed with ultrapure water. All amplifications were carried out in a gradient thermocycler (T100 Bio-Rad, Hercules, CA, USA). The cycling conditions were as follows: initial denaturation at 94 °C for 4 min, 30 cycles of denaturation step at 95 °C for 1 min, primer annealing temperature at 45 °C for 1 min, and extension at 72 °C for 1.5 min, followed by a final extension at 72 °C for 5 min. The amplification products were sequenced in forward and reverse directions by the Genomic Services Laboratory of CINVESTAV. Sequences were edited, assembled, and aligned using the default Geneious Alignment parameters in Geneious Prime v2023.0.4 (https://www.geneious.com, accessed 7 April 2025, Boston, MA, USA).

2.4. Analyses

2.4.1. Phylogeny and Phylogeography of Horseshoe Crabs

The phylogenetic reconstruction was performed using sequences of the mitochondrial COI gene of L. polyphemus from Mexico (N = 1, representing the most frequent haplotype) and the US (N = 5), of Asian horseshoe crabs (N = 7; [17,38]), and an outgroup (SF1 dataset). Firstly, we obtained the best-fit partitioning schemes and models of molecular evolution using PartitionFinder2 v2.1.1 [39], supported by the corrected Akaike Information Criterion (AICc) and the greedy algorithm [40]. The best partitioning scheme/model of molecular evolution was as follows: 1st codon position/TIMEF + I, 2nd codon position/TVM, and 3rd codon position/HKY + G. The phylogenetic analysis was generated by a Bayesian inference (BI) model using MrBayes v3.2.6 [41,42]. Two independent runs were carried out for the partitioned dataset, each using a random starting tree, one cold, and three heated chains, for 40,000,000 generations, sampling one tree every 4000 generations. The first 25% of the trees were discarded as burn-in each run. Posterior probabilities for supported clades were determined by a 50% majority rule consensus of the trees retained after burn-in. The BI analyses were run on the CIPRES Science Gateway (https://www.phylo.org/, accessed 14 Decembre 2023, San Diego, CA, USA; [43]).

To recognize the phylogeographic relationships of all horseshoe crab species, we generated a haplotype network by statistical parsimony using TCS v1.2.1 [44] under the 95% statistical parsimony criterion. For these, we considered the SF1 dataset (excluding the T. gigas specimen from the Bay of Bengal because it generated ambiguity and collapse of the haplotype network) plus all Mexican specimens truncated at the adequate COI section. A connection limit of 100 mutational steps was considered to link the divergent networks of the four species.

2.4.2. Divergence Time Estimation

To estimate the divergence time between the horseshoe crab clades, we considered a sample section of the SF1 dataset which consisted of the following: (1) for the Asian horseshoe crabs, one individual of T. tridentatus from the East China Sea, two individuals of T. gigas from the Bay of Bengal and the South China Sea, corresponding to geographically distant populations with a genetic gap [38], as well as two individuals of C. rotundicauda from the Andaman Sea and South China Sea, reported as possible cryptic species [17] (Figure 1b); (2) for the American horseshoe crab, we included one specimen from Laguna de Términos (Mexico) and one from Connecticut, northeastern US, consistent with the two extremes of its distribution.

The divergence time analysis was based on a calibrated species tree under a Bayesian approach carried out in BEAST v1.8 [45]. The evolutionary model GTR + I was selected based on AICc (jModelTest 2.1.10; [46]). The analyses were run under an uncorrelated lognormal relaxed clock model using the Yule speciation process prior to the modeling of the tree. We used two calibration points, both treated as minimum age constraints. To calibrate the root, we constrained Limulidae with the fossil of Tachypleus gadeai Vía Boada and De Villalta, 1966 (=Heterolimulus gadeai; Figure 1c) from the Ladinian stage (237 ± 1 Ma) [47] according to the Fossil Calibration Database (https://fossilcalibrations.org/, accessed 29 November 2023) [48]. The second calibration point was placed on the divergence between T. tridentatus and T. gigas with the fossil T. syriacus (Figure 1d) from the Cenomanian (93.9 Ma; [49]) [19]. For this analysis, we performed a single run of 50,000,000 generations with a random starting tree, sampling every 5000 generations. To evaluate convergence and effective sample size (ESS) for all parameters, we used Tracer v1.6 [50]. TreeAnnotator v1.8.2 [51] was used to obtain the tree with maximum credibility, burning in the first 10% of the trees and limiting the posterior probability of the clades to 0.5. The final tree was built with FigTree v1.4.2 [52].

2.4.3. Phylogeography and Demographic History in America

For our three datasets (SF1 excluding Asian species, SF2, and LF), we generated haplotype networks by statistical parsimony using TCS v1.2.1 [44] and considering a probability connection limit of 95%. Additionally, we obtained summary statistics such as the number of haplotypes (NHap), haplotype and nucleotide diversity (h and π, respectively), and the number of segregating sites (S) with DNAsp v5.10 [53].

To address historical demographic changes of L. polyphemus in America, we analyzed populations from the US Mid-Atlantic region (Chesapeake and Delaware Bay) and populations from the Yucatán Peninsula, employing the SF2 dataset. For a finer regional scale analysis, we addressed Mexican populations through the LF dataset. For both datasets, we computed Tajima’s D [54] and Fu’s F_s [55] indices, and analyzed the mismatch distribution under the sudden population expansion model with the raggedness index (1000 permutations) [56]. These analyses were conducted in DNAsp v5.10. To visualize the inference of changes in effective population size (N_e) over time, we constructed a Bayesian skyline plot for both the US (Chesapeake and Delaware Bay specimens from the SF2 dataset) and Mexico (from the LF dataset) populations using BEAST v1.8. We used a rate of 0.001 substitutions/site/lineage/million years, obtained from our previous estimation of divergence times, to configure the time axis in the skyline plot. We used the substitution model HKY for both SF2 and LF datasets, based on AICc (ran using jModelTest v2.1.10).

3. Results

3.1. Phylogeny and Phylogeography of Horseshoe Crabs

The BI analysis from the partial COI sequence (SF1 dataset) resulted in a tree with a resolved and well-supported relationship among the Asian and American horseshoe crabs (Figure 2a). The interspecific nodes had posterior probability (PP) values of 0.9–1.0. Intraspecific relationships were also well supported within populations of the Asian species C. rotundicauda and T. gigas. Within the L. polyphemus clade, intraspecific relationships resulted in two major clades with low support, one grouping individuals from Florida (PP = 0.64) and the other including Mexico, Maryland, and Connecticut (PP = 0.54). A clade with good support (PP = 0.91) is derived from the latter, including individuals from Maryland and Connecticut.

Horseshoe crabs’ phylogeographic relationships through haplotype networks linked the L. polyphemus clade to Asian species via the haplotype SF1_H12 (corresponding to Florida) through 68 mutational steps to the first divergent point (Figure 2b). The closest linkage is with C. rotundicauda (Gulf of Thailand), through 86 mutational steps, while the furthest linkage is with T. tridentatus (South China), through 148 mutational steps.

3.2. Divergence Time Estimation

The estimated divergence times (Figure 3) between both clades of the Asian horseshoe crabs, C. rotundicauda and Tachypleus spp., corresponded to the Lower Cretaceous, approximately 127 Ma (172.72–95.08, 95% HPD). The divergence time between C. rotundicauda populations occurred ca. 34 Ma (67.80–10.89, 95% HPD), during the Paleogene, whereas divergence between T. gigas populations was found at 20 Ma (38.26–5.71, 95% HPD), during the Neogene. Finally, the extreme points of the distribution of the American horseshoe crab (Mexico and Connecticut) diverged ca.14 Ma (28.51–3.58, 95% HPD), during the Neogene.

3.3. Phylogeography and Demographic History in America

We identified 16 haplotypes (Table S2) for L. polyphemus within its Mexican distribution through the LF dataset. Haplotypes LF_H1 and LF_H6 were more common (73.38% and 12.34%, respectively), and together represented more than 85% of individuals. From these haplotypes, 12 were unique among localities while 4 were present in at least two localities (Figure 4a). Considering the SF1 dataset (without Asian species), we obtained 15 haplotypes, 10 of which were exclusive to Mexico, with SF1_H1 being the most represented (89.9% of Mexican individuals). The remaining 5 haplotypes were specific to the US populations. The US haplotypes were linked to the most common Mexican haplotype (SF1_H1) by two (Florida haplotypes) and eight (Mid-Atlantic haplotypes) mutational steps in two different branches (Figure 4b). Using the SF2 dataset, we determined a total of 14 haplotypes, with only 5 Mexican haplotypes (SF2_H1 the most common at 93.5%) linked to 9 US haplotypes through five mutational steps via Florida haplotype (SF2_Lim B) (Figure 4c).

The genetic diversity of L. polyphemus varies between populations (Mexico and the US) and the dataset used (

Table 1. Summary statistics for the SF1, SF2, and LF datasets presented by locality and region (Mexico and the US) for Limulus polyphemus. N = number of sequences, N Hap = number of haplotypes, S = segregation sites, h = haplotype diversity, SD = standard deviation, π = nucleotide diversity, F_s = Fu’s Fs index, D = Tajima’s D index, significant level of raggedness index * p < 0.05, NA not applicable.

Locality	Code	N	N Hap.	S	h	π	Fs	D
Short Fragment 1 (SF1)
Laguna de Términos	LT	31	2	2	0.280 (SD 0.090)	0.0009	0.741	0.237
Champotón	CH	28	3	2	0.140 (SD 0.087)	0.0002	−2.352	−1.510
Ría Celestún	RCL	25	4	3	0.230 (SD 0.110)	0.0004	−2.352	−1.511
Chuburná	CHU	20	2	1	0.100 (SD 0.088)	0.0002	−1.648	−1.164
Chelem	CHE	7	2	1	0.286 (SD 0.196)	0.0004	−1.101	−1.006
Ría Lagartos	RL	6	1	0	NA	NA	NA	NA
Yum Balam	YB	24	5	4	0.377 (SD 0.122)	0.0006	−2.067	−1.689
Sian Ka’an	SK	13	1	0	NA	NA	NA	NA
All MX		154	10	10	0.208 (SD 0.044)	0.0004	−3.285 *	−2.029 *
Florida	FL	2	2	3	1.000 (SD 0.500)	0.0046	NA	NA
Maryland	ML	2	2	2	1.000 (SD 0.500)	0.0031	NA	NA
Connecticut	CN	1	1	0	NA	NA	NA	NA
All US		5	5	15	1.000 (SD 0.126)	0.0127	1.082	1.015
Short Fragment 2 (SF2)
Laguna de Términos	LT	31	3	2	0.127 (SD 0.080)	0.0003	−2.396	−1.505
Champotón	CH	28	2	1	0.071 (SD 0.065)	0.0001	−1.747	−1.151
Ría Celestún	RCL	25	1	0	NA	NA	NA	NA
Chuburná	CHU	20	2	1	0.100 (SD 0.088)	0.0002	−1.647	−1.164
Chelem	CHE	7	2	1	0.286 (SD 0.196)	0.0006	−1.101	−1.006
Ría Lagartos	RL	6	1	0	NA	NA	NA	NA
Yum Balam	YB	24	2	1	0.344 (SD 0.099)	0.0007	0.623	0.480
Sian Ka’an	SK	13	1	0	NA	NA	NA	NA
All MX		154	5	4	0.124 (SD 0.036)	0.0003	−2.875*	−1.521
Chesapeake bay	CHB	14	6	6	0.846 (SD 0.061)	0.004	0.658	0.362
Delaware bay	DWB	41	3	3	0.262 (SD 0.083)	0.001	−0.519	−0.659
All US		55	7	7	0.486 (SD 0.079)	0.0023	−0.540	−0.688
Long Fragment (LF)
Laguna de Términos	LT	31	5	5	0.652 (SD 0.063)	0.001	−0.548	−0.111
Champotón	CH	28	4	3	0.206 (SD 0.100)	0.0001	−2.754 *	−1.733
Ría Celestún	RCL	25	5	4	0.420 (SD 0.117)	0.0004	−2.039	−1.529
Chuburná	CHU	20	2	2	0.100 (SD 0.088)	0.0002	−2.188	−1.513
Chelem	CHE	7	2	2	0.286 (SD 0.196)	0.0005	−1.374	−1.237
Ría Lagartos	RL	6	1	0	NA	NA	NA	NA
Yum Balam	YB	24	7	5	0.554 (SD 0.116)	0.0006	−1.526	−1.267
Sian Ka’an	SK	13	1	0	NA	NA	NA	NA
All MX		154	16	15	0.447 (SD 0.048)	0.0005	−3.664 *	−1.999 *

). In Mexico, the first section of the partial COI region (i.e., the SF1 dataset) showed higher values of genetic diversity (h = 0.208, π = 0.00004, S = 10) than the last and shorter section (i.e., the SF2 dataset), which had considerably lower values (h = 0.124, π = 0.00003, S = 4). Considering the longer fragment (i.e., the LF dataset), the genetic diversity parameters increased (h = 0.447, π = 0.00005, S = 15).

Tajima’s D and Fu’s F_s demographic indices (Table 1) were significantly negative in Mexico for both LF and SF2 datasets, suggesting population expansion. This is reinforced by the mismatch distributions, which showed an unimodal distribution, typical of a population that has undergone expansion, and a non-significant raggedness index value (LF dataset: r = 0.332, p = 0.121; SF2 dataset: r = 0.532, p = 0.519) (Figure 5(a1)). The Bayesian skyline plot for this population demonstrated a sharp increase over the last 75,000 years (for the LF dataset; Figure 5(a2)).

US Chesapeake and Delaware populations, based on the SF2 dataset, presented a genetic diversity four-fold greater (h = 0.486, π = 0.0023, S = 7) than the Mexican population (compared to its analog dataset) despite a smaller sample size (Chesapeake, N = 14; Delaware, N = 41).

Demographic analyses of the US populations (Chesapeake and Delaware: SF2 dataset) showed non-significant negative values of Tajima’s D and Fu’s F_s indices (perhaps due to a low sample number), indicating an excess of rare mutations in these populations. The mismatch distributions present a multimodal distribution and a non-significant raggedness index value (r = 0.280, p = 0.743), indicating a diminishing population size or structured size (Figure 5(b1)). Lastly, the Bayesian skyline plot presented a slight increase during the last 150,000 years (Figure 5(b2)).

4. Discussion

Our research was the first to integrate a large sample of Mexican individuals of the extant horseshoe crab species representing its complete distribution range into a phylogenetic and phylogeographic study employing multiple COI segment lengths. This study corroborated the findings of a preceding investigation [17] regarding the phylogenetic relationships between Asian and American horseshoe crabs, and the well-supported interspecific relationships within the Tachypleinae subfamily. However, by expanding the number of representatives from several localities (Mexico, Florida, Maryland, and Connecticut) we were able to increase the resolution of intraspecific relationships for L. polyphemus, which had weak support in the previous study [17]. The individuals from Florida are the most basal in these relationships, while those from Mexico are more closely related to individuals from Maryland and Connecticut (US Mid-Atlantic), despite greater geographical distance.

An understanding of the relationships among American populations can be gained by examining the phylogeographic relationships of L. polyphemus with respect to Asian species, which are linked through the Florida population. This phylogeographic pattern can be related to changes in the past distribution of L. polyphemus, particularly during the last two million years, when marked glacial cycles during the Pleistocene period generated dynamics of latitudinal changes in the distribution of most species [57]. The cold phases of the Pleistocene resulted in the expansion of continental glaciers, the reduction of sea level leading to the exposure of new continental shelves, and the reduction of water temperature in coastal regions near ice masses [57]. Given that temperature is one of the limiting factors in the distribution of L. polyphemus, glacial events would have been pivotal in shaping the population dynamics of this species. Sekiguchi and Shuster [8] pointed out that during the Last Glacial Maximum (ca. 23–19 ka) the distribution of L. polyphemus did not extend northwards beyond Florida, therefore this northernmost population possibly may have served as a Pleistocene refugium. During this glacial period, the sea level lowered, exposing large areas of the seafloor and modifying the circulation patterns of ocean currents, creating connections between regions that are now separated [58], which probably facilitated the dispersal/colonization of L. polyphemus to the Yucatán Peninsula, Mexico. In contrast, the aforementioned authors suggested that during interglacial periods, L. polyphemus would attain a more northerly distribution than it does today, which would explain why the most derived or recent haplotypes are found in the northern region of its distribution range. This northern paleodistribution was recently confirmed by detecting the presence of L. polyphemus in a Greenland ecosystem with a dating of approximately two million years, suggesting the presence of warmer surface water conditions that allowed the colonization by this species [59].

The evolutionary history of extant horseshoe crab species has been the subject of various research [6,17,60] largely due to the interest in their ancient origin and the fact that they have succeeded in surviving and adapting to major climatic and geologic changes. In this context, the divergence of extant species had been linked to fossil species such as Tachypleus decheni Zincken, 1862 and Limulus coffini Reeside and Harris, 1952, hypothesizing a minimum age of 75 Ma for the apparently most recent common ancestor [18]. More recently, the age of divergence of the subfamilies Limulinae and Tachypleinae was estimated by considering the fossil specimen M. walchi, recognized as the crown group of extant species, in conjunction with the rifting process of the Atlantic Ocean opening, which occurred around the Lower Cretaceous (i.e., ~140 Ma) [17].

Currently, several studies have used different calibration points to obtain divergence times for this group. For example, Shingate et al. [61], through genome-wide protein data, estimated divergence times by calibrating their tree by means of a soft maximum constraint for all Xiphosurids between 236 and 636.1 Ma ago. In this calibration method, the authors considered the fossil T. syriacus (236 Ma) as the minimum age, as is also the case in this study, and the maximum age was defined by the presence of the last common ancestor of the Ecdysozoa (see [47]). This wide time range has influenced their results since they obtained an age of 436 Ma (Silurian period) for the divergence between the American and Asian horseshoe crabs. However, the order Xiphosurida does not appear until the Devonian period, and the family Limulidae, until the Permian (299–251 Ma) [6], suggesting that the origin of extant horseshoe crab species must be more recent.

In another study, Zhou et al. [62] used mitochondrial genomes from seven different species of chelicerates, including horseshoe crabs, in addition to information obtained from a fossil specimen considered the ancestor of all chelicerates, to calibrate their tree root at 530 Ma. Their results were closer to those obtained by Obst et al. [17], estimating the divergence time of the American and Asian species at 130 Ma.

In this study, we endeavor to present a comprehensive overview of existing knowledge concerning the relationships between fossil species and extant species with the aim of assessing divergence time. Using the fossil species T. gadeai we update the minimum divergence time between the Asian and American horseshoe crab clades to the Upper Triassic, a time when the global climate was predominantly warm and dry, and when the Pangea supercontinent was beginning to separate, forming smaller oceans and potentially driving the diversification of this group [3,63]. According to the hypothesis proposed by Obst et al. [17], L. polyphemus would have originated in a western region of the Neo-Tethys Ocean, in its portion of the Laurasia paleocontinent, and then moved westward through the rifting process that led to the opening of the Atlantic Ocean in a period between the Upper Triassic and Lower Jurassic [64]. The divergence between C. rotundicauda and Tachypleus spp. occurred during the Lower Cretaceous (mean = 127 Ma; 172.72–95.08, 95% HPD); previously, this divergence had been estimated at approximately 45 Ma [17], implying an underestimation by at least 80 Ma. During this period, the fragmentation of the supercontinent Pangea led to the formation of new marine habitats and changes in connectivity between them [65]. The expansion of the oceans allowed marine species to disperse and colonize different regions [66]. This geographical fragmentation may have promoted the speciation of Asian horseshoe crabs by limiting gene flow between populations. At present, C. rotundicauda shows a notable differentiation between the populations of the Andaman Sea and those of the Gulf of Thailand and South China Sea, as observed both in the global haplotype network and in the phylogeny, which could represent an incipient process of speciation, or as mentioned by Obst et al. [17], the potential presence of cryptic species. A comparable situation is observed in the case of T. gigas, where the Bay of Bengal population is clearly differentiated from those of the South China and the Andaman Seas, evidently attributable to the presence of geographical barriers [38]. This internal split between T. gigas populations occurred approximately 20 Ma ago, during the Miocene, an epoch during which this region was characterized by extensive mangroves and estuaries [67].

In relation to the American horseshoe crab, L. polyphemus, our findings indicate, for the first time, that genetic differentiation between the extremes of the species distribution, namely Mexico and the northeastern United States, commenced during the Neogene period (mean = 14 Ma; 28.51–3.58, 95% HPD), and is probably indicative of historical biogeographical events (e.g., changes in distribution, colonization, and isolation) and environmental changes (e.g., changes in surface water temperatures) that influenced population divergence within L. polyphemus. As a result of the update in the knowledge of fossil species of horseshoe crabs and their relationships with extant species, this study has been able to use new calibration points and refine the estimation of divergence times within this group. Although the use of partial mitochondrial genome sequences, as compared to the use of complete mitochondrial genomes or complete nuclear genome sequences, presents limitations with respect to resolution and phylogenetic robustness, the accuracy of divergence time estimates is highly dependent on adequate fossil calibration [20].

Focusing on the American horseshoe crab, L. polyphemus, the present study provides valuable insights into the genetic diversity and population structure of this species across its distribution range by analyzing three datasets (SF1, SF2, and LF). The analyses revealed distinct haplotypes and highlighted notable differences in genetic composition between Mexican and US populations. Notably, the SF1 and SF2 datasets revealed different relationships between haplotypes, specifically with respect to the linkage between the Mexican and US populations. For example, by means of the SF1 dataset, we showed that the connection between Mexican and US populations occurs through two different lines, one linking Mexico to the US through Florida and the other connecting them through the Mid-Atlantic populations, with the linkage to Florida being the closest as it presents a lower number of mutational steps. A possible cause of this configuration could be due to the lack of sample representation between Florida and the Mid-Atlantic, where haplotypes linking both populations could exist for this dataset. Based on this dataset and the global haplotype network, Florida represents the least derived population of all. On the other hand, the relationships from the SF2 dataset show a unique connection between Mexico and the US through Florida and continuing to the North Atlantic, showing a latitudinal gradient relationship. These different relationships may be due to the position of the COI sequence fragment analyzed. For example, at least for Mexico, the first section (corresponding to the SF1 dataset) proved to be more informative (i.e., higher number of segregating sites, number of haplotypes, and genetic diversity) compared to the second section (corresponding to the SF2 dataset), which appears to be more conserved. In addition, it is important to underline the origin of the Florida specimens: for the SF1 dataset, at least one individual came from the Florida east coast (i.e., SF1_H12), while for the SF2 dataset, the specimen was attributed to the Florida west coast, in the Gulf of Mexico. In the northeastern Florida region, Saunders et al. [29] identified a major genetic discontinuity that differentiated northern and southern US populations, evidencing a phylogeographic break at Cape Canaveral, Florida, which would explain the different linkages visualized in the haplotype networks of the SF1 and SF2 datasets.

Despite the fact that a significant proportion of the US haplotypes corresponded to single individuals (all of whom were contained within the SF1 dataset), we were able to successfully recover multiple individuals previously analyzed from two localities, Chesapeake and Delaware Bays, and include them in the SF2 dataset. As Pierce et al. [26] noted, Chesapeake Bay has more haplotypes than Delaware Bay (N hap = 6 vs. 3, respectively) despite a smaller sample size (N = 14 vs. 41, respectively). Yet, as mentioned above, they share the most common haplotype (SF2_Lim C). They mention that the differences between the two Bays are due to restricted gene flow. This is a possibility given the matrilineal inheritance of mtDNA and the dispersal habits of the species, which is highly philopatric [27]. However, previous genetic studies, approached from nuclear genetic markers, indicated that the two Bays are part of a large population in the US Mid-Atlantic region [27,28]. The lower haplotype diversity in Delaware Bay, which harbors the largest population [68], could be more associated with overexploitation (by which it is also historically the most impacted; [69]) than limited gene flow. These populations likely increased modestly around 150,000 years ago (Figure 5(b2)) because of changes in their distribution, since both bays have a relatively recent origin in the middle of the Holocene [70]. In contrast to our results, Faurby et al. [25] detected population declines in the core of the distribution (including these populations), indicating that they may have occurred at the end of the last Ice Age. However, demographic signals in the current gene pool may not accurately reflect the historical past of these populations, given recent anthropogenic influences that may have eroded genetic diversity.

At a regional scale in the Mexican distribution of the species, the haplotype networks of the three datasets (i.e., LF, SF1, and SF2) are arranged in a star-like formation, centered on the most common haplotype. The dominance of one or few haplotypes may be due to different causes, such as a possible founder effect, population bottleneck, or selective pressures influencing their prevalence [71]. Nevertheless, the topology of the haplotype network (i.e., star-like) and the pattern of genetic diversity (i.e., high-to-moderate haplotypic diversity and low nucleotide diversity) are indicative of population expansion [72]. The negative Tajima’s D and Fu’s F_s indices and the unimodal mismatch distribution are also typical of a population that has undergone expansion [56]. This expansion of the Mexican population, according to the Bayesian Skylineplot, has been occurring over the last 75,000 years. A possible explanation for this relatively recent population expansion lies in the dispersal/colonization of a population subgroup (possible founder effect suggested by the dominance of a common haplotype) towards the Yucatán peninsula during the glacial period, when the distribution of the species attained its southernmost descent. It can be hypothesized that, during the transition from the glacial to the interglacial period, with the resultant rise in sea level, this population will have become isolated from the remaining populations (currently, this population shows differentiation and restricted gene flow with its US counterpart; [27,28]), establishing itself and increasing its effective population size over time. Similar biogeographic and demographic histories are reflected in Asian horseshoe crab species, such as the case of T. tridentatus, in which Yang et al. [73] detected signs of population expansion (star-like haplotype network and unimodal mismatch distribution) in three populations in Southeastern China in addition to the isolation of a population in the bays of Japan. These authors attribute these processes to Pleistocene glacial sea-level fluctuations, where the LGM is associated with a 120–140 m decrease in the south and east China seas. They propose the existence of an ancestral population distributed along the coasts connecting China and Japan, which fragmented after the late Pleistocene sea-level rise during the LGM, leaving the Japanese populations isolated. In relation to its sister species, T. gigas, which demonstrates sympatry with T. tridentatus in some regions of its current distribution, populations that have also undergone expansion have been detected, particularly in populations in the Bay of Bengal, India [38] and in Indonesia [74]. This occurrence appears to originate from similar causes attributable to historical dispersal and isolation events during the Pleistocene. This hypothesis is logical, despite the particularities of each region inhabited by extant horseshoe crab species, given that their distribution obeys the same four large-scale environmental factors (i.e., continental geomorphology, temperature, tidal types, and benthic currents). Therefore, past changes could have similarly influenced these species.

Despite the philopatric habits of L. polyphemus females and larvae [23,27] and at least two Mexican populations previously identified through SSR nuclear markers [31], our results point to the absence of a clear structure across Mexican populations. As posited above, this phenomenon could be attributed to a recent colonization event which resulted in a reduced gene pool through a founder effect. Furthermore, insufficient time has elapsed for the migration–drift equilibrium to be reached, which would establish a phylogeographic structure [75]. An alternative explanation is overexploitation. Indeed, in some localities characterized by low haplotypic diversity (≤2 haplotypes), such as Chuburná, Chelem, and Ría Lagartos, this species is exploited for its use as bait for octopus fishing (Octopus maya Voss and Solis, 1966), despite its endangered status [10]. Overexploitation could erode haplotypic diversity in these localities [76]. Finally, we suggest other explanations for the case of Sian Ka’an, which also presents low haplotype diversity, notwithstanding its location within a natural protected area (Sian Ka’an Biosphere Reserve); it is located in the southernmost portion of the species distribution in the Caribbean Sea, and typically edge-of-range populations of a species with limited dispersal capabilities tend to be smaller and demonstrate lower genetic diversity when compared to core populations [77]. In addition, this region is influenced by ocean currents that circulate northwards, favoring migration from Sian Ka’an but not vice versa. This could explain the absence of haplotypic diversity in this locality, as previously shown through SSR nuclear markers [31].

The difference in haplotypic composition between Mexican and American populations, both of which have at least one dominant haplotype, testifies to the genetic differentiation and diversity that exists in this species throughout its distribution, from the southernmost to the northernmost part of its range. In order to obtain a more comprehensive understanding of the phylogeographical relationships and evolutionary history of L. polyphemus throughout its American range, improved representation of American populations is imperative. Given that single-gene datasets have limited capabilities for evidencing robust phylogenetic relationships, particularly in closely related species, the use of next-generation sequencing technologies should be included in the future, in conjunction with the appropriate use of calibration points, to refine the understanding of interspecific and intraspecific horseshoe crab evolution and divergence times. Despite this limitation, our results concerning the demographic history of Mexican populations suggest an evolutionary history influenced by environmental changes in the recent past. Reconstructing the evolutionary history of the American horseshoe crab, as well as its populations, allows us to understand how it has responded to changes in the past and how it might cope with possible changes in the future.