To characterize the spawning activities of the three M. cephalus cryptic species that reside around Taiwan, three sampling schedules that covered the different life stages of fish from the spawning ground, feeding ground, and nursery ground were applied (Fig. 1). Spawning adults were sampled using purse seine nets during three consecutive years (2005–2007) [8] off the southwest coast of Taiwan near Kaohsiung, an M. cephalus spawning ground exploited by fishermen since the seventeenth century [15,23]. Adults of various sizes present at a feeding area were collected from the Kaoping River Estuary, the broadest tropical drainage in Taiwan (Fig. 1). Sampling was conducted monthly between August 2008 and March 2009, a period consisting of the months preceding and following the known spawning period of the fish. Juveniles from the Gonshitian Creek Estuary in Northern Taiwan and the Baoli Creek Estuary in Southern Taiwan (Fig. 1) were collected monthly from December 2011 to March 2012 by using a set net with the mesh size of 1 mm in areas of shallow water that enabled easily sampling. The recruitment of M. cephalus juveniles in estuaries along the western coast of Taiwan that are known nursery areas of the fish was observed mainly in these four months [15,18]. In this study, feeding and nursery areas were distinguished in consideration of the size of the inhabited system of fish; a feeding area is also a nursery area because of the simultaneous presence of juveniles and adults, whereas only juveniles are observed in nursery areas.

The fork length, total weight (TW), and gonad weight (GW) of adult fish were measured. The total length of the juveniles was measured because the caudal fin in such fish was undeveloped. The GSI was calculated using the following equation: 100 GW/(TW – GW) [25]. The differences in mean values were tested using either the t-test or an analysis of variance (ANOVA). For each fish, muscle tissues were removed from the epi-axial region under the second dorsal-fin base and preserved in 95% ethanol.

Genomic DNA was extracted from muscle tissue by using a DNA Purification Kit (Bioman, Taipei, Taiwan), preserved in a Tris EDTA (TE) buffer, and then quantified and diluted to 1 ng/μL for use in polymerase chain reaction (PCR).

A rapid screening method that involved detecting the three cryptic species of M. cephalus in the NW Pacific by using a multiplex Cytochrome Oxidase I (COI) haplotype-specific PCR (MHS-PCR) was developed in a previous study [8] according to the recommendations in related studies [26,27]. Table 1 lists the primers used for rapid screening. PCR was performed using a Biometra TGradient Thermocycler with a 15-μL reaction volume containing 0.2 μm of dNTP, 1.5 μL of a 10 × PCR buffer (Bioman), 0.5 μm of forward and reverse primers, 0.2 U of Taq DNA polymerase (Bioman), and 1.0 μL of template DNA. Mitochondrial DNA amplification was conducted using the following PCR program: 35 cycles of denaturation at 94 °C for 15 s, annealing at 55 °C for 15 s, and extension at 72 °C for 30 s after heating at 94 °C for 5 min. The PCR products were electrophoresed in a 2.0% agarose gel (Bioman) and stained with ethidium bromide to characterize the size of bands by using ultraviolet transillumination. The PCR products of NWP1, NWP2, and NWP3 were 362 bp, 283 bp, and 549 bp, respectively, according to the COI sequences (Fig. 2).

Among the 151 spawning adults sampled from the spawning ground off the southwest coast of Taiwan in December 2005, 2006, and 2007, only one species (NWP1) was represented (Table 2). By contrast, NWP1 accounted for only 1.8% of the 400 adults collected at the feeding ground of the Kaoping River Estuary, where the two other cryptic species (NWP2 and NWP3) were dominant, representing 46.0% and 52.8% of the species composition, respectively. Nevertheless, despite the low composition, NWP1 was observed in most monthly samples. From August to March, NWP3 was common, whereas the composition of NWP2 decreased in November and December (8.3% and 2.6%, respectively).

In the Gonshitian Creek Estuary in Northern Taiwan, no fish belonging to NWP3 were observed among juveniles from December to March. Although NWP2 was the most abundant species among juveniles at this location (75% of the samples), in March, a clear shift occurred in NWP1, which accounted for all juveniles collected that month.

In the Baoli Creek Estuary in Southern Taiwan from December to March, no fish belonging to NWP1 were observed among the juveniles, whereas NWP2 (51.9–95.8%) and NWP3 (4.2–48.2%) were abundant. The composition of these two species exhibited an inverse trend compared with that of the samples collected from the Kaoping River. NWP2 was abundant in all four months, whereas the composition of NWP3 fluctuated widely.

The mean length of NWP1 on spawning and feeding grounds ranged from 526.0 ± 38.1 mm (in 2007) to 475.7 ± 34.6 mm (in 2005), and significantly exceeded the mean length of NWP2 (412.4 ± 39.3 mm) and NWP3 (392.2 ± 44.6 mm) on feeding grounds (ANOVA, P < 0.05; Table 3). The GSI of NWP1 on the spawning ground ranged from 17.0 to 24.6, exceeding the GSI of all NWP2 and most NWP1 and NWP3 on the feeding ground (>0.1–20.1).

Species in the Kaoping River Estuary exhibited various reproductive characteristics (Fig. 3). Among seven large NWP1 fish, only two, measuring more than 480 mm, were mature (GSI > 17) (Fig. 3a). By contrast, although sexual maturation began at shorter lengths in NWP2 than in NWP1, no NWP2 collected from August to March exhibited a GSI exceeding 10 (Fig. 3b). Full sexual maturity (GSI > 15) was observed in NWP3 at lengths as short as 350 mm (Fig. 3c).

The spawning activity inferred according to temporal changes in GSIs revealed different patterns among the three species (Fig. 4). According to the limited quantity of data, the GSIs of NWP1 peaked at 20.1 in December (Fig. 4a). The GSIs of NWP3 increased gradually from October (7.2), peaked in November (17.9) and December (19.4), and remained relatively high until March (11.9–14.8; Fig. 4c). By contrast, the GSIs of NWP2 increased earlier, to approximately 10 in September, remained at that level between October and November, and then decreased to 3.9 in December (Fig. 4b).

The lengths of juveniles sampled from the Gonshitian Creek Estuary (north) and the Baoli Creek Estuary (south) of Taiwan from December 2011 to March 2012 ranged from 22.0 to 63.0 mm (Fig. 5). Recruitment of NWP2 juveniles occurred in both Northern and Southern Taiwan, whereas that of NWP1 and NWP3 juveniles was in Northern and Southern Taiwan, respectively.

In the north, juveniles collected from December to March exhibited lengths ranging from 22.0 to 33.5 mm. Recruitment of small NWP2 fish with a mean length of 26.8 ± 2.1 mm occurred during December to February, whereas NWP1 juveniles (30.9 ± 1.9 mm) were observed only in March. No modal progression in the length frequency distribution was observed in either species in Northern Taiwan (Fig. 5a).

In the south, recruitment of small NWP2 and NWP3 fish occurred in December–February. Slight modal progression was observed in the length frequency distributions, except for a modal shift that occurred between February and March (Fig. 5b). The NWP2 juveniles (37.1 ± 11.4 mm) were significantly larger than the NWP3 juveniles (27.2 ± 4.8 mm) from December to March (t-test, P < 0.01).

Several studies have investigated the life history traits of M. cephalus in Taiwanese waters; however, detailed information on the three cryptic species is lacking. Although previous studies did not recognize the cryptic species, they nonetheless suggested the coexistence, in Taiwanese waters, of fish in different stages of the life cycle that live in various aquatic environments [15,28–30]. According to recruitment and temporal patterns of fish landings, some authors suggested the existence of migratory and resident populations [23]. Some fish may remain in Taiwanese estuaries all year, whereas other fish may migrate offshore over a wide geographic area, from the south of Taiwan to the Japan and Yellow Seas. According to the trace elemental composition of the otoliths of M. cephalus in Taiwanese estuaries, some M. cephalus were found to move between estuaries and offshore waters, whereas other fish moved between estuaries and freshwater habitats [28–30]. If the three migratory behaviors (one offshore and two coastal and estuarine migrations) correspond to species characteristics, then these behaviors indicate the existence of distinct ecological niches that limit hybridization. According to the species composition observed in the various aquatic environments sampled in our study and in previous studies [8,12], NWP1 is clearly the dominant species in marine environments, whereas NWP2 (as Species 1 [12]) and NWP3 are the dominant species in estuaries. No freshwater location has been sampled, thus limiting inference on the ecological niches of NWP2 and NWP3. Therefore, whether the freshwater migration inferred on the basis of otoliths corresponds to a specific ecological niche (for NWP3) or species plasticity (of NWP2 and NWP3) remains undetermined. Adults can migrate to freshwater environments to avoid competition, which suggests the existence of specific ecological niches in mugilids [30]. However, on the basis of the level of species admixture in some key habitats such as nursery and feeding grounds, whether the existence of a specific ecological niche is a prezygotic barrier sufficiently powerful to prevent hybridization is uncertain.

Our previous study [8] and the present study confirmed the large sympatry of the three cryptic species belonging to the M. cephalus species complex in Taiwanese waters, even though NWP1 appears to be more prominent in marine environments than are NWP2 and NWP3, which are dominant in estuarine environments. An absence of spatial isolation or incomplete ecological niche segregation does not necessarily imply that hybridization is possible. Species isolation may be only temporary, and species isolation that occurs during the reproductive period at locations such as spawning grounds represents a strong prezygotic barrier. Philopatry in fishes occurs in numerous species, including marine species [31–33]. During the reproductive season, 90% of the Atlantic herring Clupea harengus present in Newfoundland waters return to one of the three major spawning grounds in the Gulf of Maine [34]. If 10% of herring adults are vagrants and maintain gene flow, then an absence of genetic misassignment between mitochondrial and nuclear genomes in NW Pacific M. cephalus species suggests an absence of hybridization [8]. The yearly return of mature NWP1 fish to the offshore spawning ground in the Taiwan Strait supports the existence of specific spawning. The absence of other species also attests to the strong natal philopatric behavior. The problem that remains unresolved is determining the location of the spawning grounds for NWP2 and NWP3. The composition of NWP2 adults in the estuaries decreased as the GSI increased, suggesting that mature adults leave the estuaries; conversely, the composition of NWP3 adults remained stable in the estuaries, and their GSIs increased to values similar to those of NWP1 at their spawning grounds. This indicated that the NWP3 spawning grounds might be in or near estuaries, a phenomenon that is inconsistent with the typical life cycle of M. cephalus worldwide, which necessitates migration between estuarine and marine environments (for a review, see [14]). This phenomenon can explain why the NWP3 juveniles in the estuaries were the shortest; the juveniles of other species may have hatched at sea and spent more time recruiting in estuaries.

Asynchrony of spawning periods is generally considered to explain the genetic isolation of species with overlapping distributions [35]. Such asynchrony has been observed in various marine organisms, including hermit crabs [36], bivalves [37,38], sponges [39], polychaetes [40], gastropods [41], and fishes [42–44]. Knowlton [35] listed 26 examples of spawning asynchrony in cryptic and sibling marine species. Regarding M. cephalus, asynchrony can also be considered a cause of species isolation because spawning periods are inconsistent between NWP1 and NWP2 and only partially overlapped between NWP2 and NWP3. Although the spawning period of temperate NWP1 overlapped with that of tropical NWP3, the reproductive behaviors of NWP1 differ completely. Therefore, the differences in the temporal evolution of the GSI and the juvenile recruitment in estuaries between species pairs clearly suggest the importance of temporal isolation as a prezygotic barrier. Water temperature (21 °C) and short photoperiods (6 h of light and 18 h of dark) are likely crucial factors determining vitellogenesis in M. cephalus [45] and imply that spawning periods vary among the species of the M. cephalus complex located at different latitudes [14]. Thus, the spawning nonsynchrony of NW Pacific cryptic species is expected because these species are partially isolated over different latitudes. Similarly, some specific reproductive traits may result from the environmental conditions in which these species live outside the contact zone. Thus, the large-scale reproductive migration of the NWP1 species is likely directly caused by the large temperature decrease in the Japan and Yellow Seas during the northeast monsoon season. The onset of the northeast monsoon in winter leads to the emergence of the CCC, which blocks the northward incursion of the warm Kuroshio branch water, causing the temperature to decrease to 16 °C [46]. This temperature decrease drives the NWP1 species to migrate southward to find sea-surface temperatures conducive to spawning; in December, these conditions are present in the Taiwan Strait, where the respective warm and cold waters of the Kuroshio branch and the CCC mix [47].

Differences in the duration of spawning activity among species (the NWP1 breeding season is the shortest) can also be attributed to water temperature because high-latitude species have shorter spawning seasons than do their tropical relatives [48]. A previous study [8] emphasized that NW Pacific M. cephalus species have distribution ranges that overlap only partially; NWP1 prefer more temperate environments that do other species, whereas NWP3 are present mostly in tropical areas and NWP2 range throughout the warm Kuroshio Current in the NW Pacific. Thus, the extended spawning and recruitment seasons for the NWP2 and NWP3 species could be attributed to the minor seasonality of the environments in which their juveniles recruit.

In contrast to the notion that the M. cephalus historically fished on the spawning ground located in the Taiwan Strait belong to only one species, a previous study [12] and our studies have indicated that juveniles harvested in estuaries consist of a mixture of three species. A management policy that does not consider the presence of multiple species in a stock may have severe consequences on the species sustainability. The demographic productivity of the species may differ such that each species may not support equal levels of exploitation [34]. To ensure that these species are not overfished, improving the knowledge of the demographic characteristics of these species by monitoring each species individually is crucial [49].

In addition, because aquaculture and artificial breeding programs seem to be successful in slowing the decline in populations of M. cephalus worldwide [50–52], more comprehensively characterizing the physiological traits of these species is vital for improving aquacultural practices. The slight differences in ecological niches observed in our study indicate that the existence of specific physiological requirements must be considered in aquacultural practice. For example, NWP1 may adapt more easily to cold water than do the other two species but adapt less easily to aquaculture ponds in Southern Taiwan where water temperatures are high. Differences in thermal tolerance among the Taiwanese species have been suggested. The migratory type (assumed here to be NWP1) of fish has an allele at the glucose-6-phosphate isomerase locus that differs from the most frequent allele observed in the resident type (probably NWP2) of fish, and the enzymatic activity of fish with these alleles varied under different thermal regimes [21]. Similarly, the freshwater affinity of the NWP3 species may substantially affect the ability of these species to grow in brackish aquaculture ponds. By contrast, NWP2 could not mature fully in estuaries because NWP2 fish with high GSIs were observed in the ocean [8]. Finally, understanding the physiological requirements of these cryptic species may improve aquacultural practices by reducing juvenile mortality.