Foraging ecology of the amphibious mudskipper Periophthalmus chrysospilos (Gobiiformes: Gobiidae)

Sampling sites

Specimens were collected from April 2020 to March 2021 at four locations along the Hau river estuary in Vietnam (Fig. 1): Duyen Hai-Tra Vinh (DHTV; 9°41′18.6″N 106°30′35.8″E), Tran De-Soc Trang (TDST; 9°29′26.8″N 106°11′58.5″E), Dong Hai-Bac Lieu (DHBL; 9°06′03.2″N 105°29′49.1″E); and Dam Doi-Ca Mau (DDCM; 8°58′17.5″N 105°22′51.8″E). There is little fluctuation in temperature at the sites between wet (June–December) and dry (January–May) seasons; the temperature remains at approximately 27 °C. Conversely, precipitation measures varied significantly—a monthly average of 20 mm in the dry season and 400 mm in the wet season (Le et al., 2006). The pH was 7.6–7.9 and varied with sites but not seasons, whereas the salinity was 12.3–23.5% and varied with season but not site (Dinh et al., 2021b).

Vegetation typical at these sites include Acanthus ebracteatus Vahl., Avicennia marina (Forssk.) Vierh., Bruguiera gymnorrhiza (L.), Nypa fruticans Wurmb., Savigny Sonneratia caseolaris (L.) A. Engl., and Rhizophora apiculata Blume. The dominant plant species at DHTV, TDST and DHBL were Sonneratia caseolaris, Avicennia marina and Bruguiera gymnorrhiza (L.) Savigny., respectively. At DDCM, Avicennia marina and Bruguiera gymnorrhiza (L.) Savigny. were equally dominant (Dinh et al., 2021b).

Sample collection

Fishes were randomly collected ∼4 hours during the ebb tide by hand in an area of 120 square metres (6 m × 15 m) on the mudflat at night for five consecutive days per month. Fishes were identified following Murdy & Jaafar (2017) and sexed. The genital papillae of females are bulbous, pinkish, and equally broad at the base and tip, whereas those for males are slender and whitish, broad at the base and tapers towards the tip (Dinh et al., 2020a). Fishes captured were immediately euthanised in a solution of tricaine methanesulfonate (MS222) before being transferred into a solution of 5% buffered formalin and transported to the laboratory. In the laboratory, specimens were soaked in water for 2 hours before analyses. The total length (TL, nearest to 0.1 cm) and weight (W, nearest to 0.01 g) of each specimen was recorded before dissection and analyses of content of the digestive tract. The length at first maturity (L_m) was used to divide the fishes into two groups: immature group if TL < L_m and mature group if TL ≥ L_m. Individual of females and males from DHTV with total lengths of 7.3 cm and 6.9 cm or below were considered immature. L_m values of females and males from TDST were 6.4 cm and 6.6 cm; 7.0 cm and 6.2 cm from DHBL; and 6.8 cm and 8.6 cm from DDCM (Q. M. Dinh, 2021, Unpublished data).

The use of fishes in the present study was assessed and approved by The Council for Science and Education, School of Education, Can Tho University (Animal Welfare Assessment number: BQ2020-03/KSP). Altogether, 1,031 specimens were used for RGL and 546 specimens for diet composition analyses.

Relative gut length

The digestive tract of each specimen was carefully removed and measured (nearest 0.1 cm) to obtain the relative gut length (RGL = total length of the gut/the total length of fish). The association of the length of digestive tract to the length of the fish has been used as an indicator for feeding guild (see Al-Hussaini (1947), Kapoor, Smit & Verighina (1976), Drewe et al. (2004)). Typically, RGL value greater than 3 signifies that the species is herbivorous, RGL value between 1 and 3 signifies omnivorous fishes, and RGL value below 1 signifies carnivorous fishes.

Diet composition

Contents of each digestive tract were analysed under the stereomicroscope (Motic DM-143-FBGG) and identified to the lowest possible taxonomic level following Dill (2002) and Nguyen et al. (2013). The occurrence of each food item was calculated by the following equation: %O_i = 100 × O_i/N (Hynes, 1950), where O_i is the number of fishes consuming prey i and N is the total number of fishes examined. The weight of each food item was calculated by the following equation: %W_i = 100 × W_i/W_total (Hyslop, 1980), where, W_i is weight of prey i, W_total is total weight of all prey individuals. Biovolume of prey was calculated by the following equation: %V_i = (100 × O_i × Wi)/Σ(O_i × W_i), where V_i, O_i and W_i are the percentage of biovolume, occurrence and weight of prey i respectively. The diet composition was analysed against sex, size, site, and season to understand the interactions of these parameters to food acquisition (Natarajan & Jhingran, 1961; Hyslop, 1980; Clayton, 2017; Dinh et al., 2020b).

The modified graphical method of Costello (1990) was used to visualise the diet composition of Ps. chrysospilos (Amundsen, Gabler & Staldvik, 1996). Diet components that are highly abundant and essential appear at the upper right quadrant of the graph, while less abundant but essential components appear at the lower right quadrant of the graph; less essential components that are abundant or scarce appear on the left upper and lower quadrants respectively (Adámek, Andreji & Gallardo, 2007).

Data analyses

PRIMER v.6.1.11 (Clarke & Gorley, 2006) with PERMANOVA+ v.1.0.1 add-on package (Anderson, Gorley & Clarke, 2008) was used to test if variation in diet can be attributed to the sex and size of the fish, or to site and season of catch (Baeck, Yoon & Park, 2013). If variations were detected, nonparametric tests would be applied to identify the driver for the differences (Dinh et al., 2017). The Mann-Whitney U test was used to test for changes in diet composition between factor pairs: sex (male and female individuals); size (immature and mature were determined basing on the L_m value as mentioned in the section on fish collection); and season (dry and wet). If diet composition varied for more than two factors, the Kruskal-Wallis H test was applied to identify the component driver.

T-test was applied to test the variations of RGL between fish sex, sizes, and season. One-way ANOVA was used to test the variation of RGL between sites. The effect of interactions between factors such as: gender × size; gender × season; gender × site; size × season; size × site; season × site; gender × size × season; gender × size × site; gender × season × site; size × season × site; and gender × size × season × site on RGL were quantified using General Linear Model. The significant level was p < 0.05 in all tests. To decrease the likelihood of Type I error of all test, the Benjamini–Hochberg procedure was applied (Benjamini & Hochberg, 1995; McDonald, 2014).

Relative Gut Length (RGL)

The RGL values were significantly different between individuals of different sizes (t-test, t = −5.39, n = 1.031, df = 1.029, p < 0.001, CI_95% [−0.07 to −0.03]); mature individuals exhibited higher RGL values when compared to immature individuals (0.56 ± 0.006 and 0.51 ± 0.007), respectively, (see Table 2). However, the RGL values were not significantly different between males and females (t = 0.45, n = 1.031, df = 1.029, p = 0.15, CI_95% [−0.01 to 0.02]; see Table 2). When testing for temporal variations, RGL values were significantly different between seasons (t = −12.90, n = 1.031, df = 1.029, p < 0.001, CI_95% [−0.13 to −0.10]). The RGL value was higher in the wet season than the dry season (0.59 ± 0.01 and 0.47 ± 0.01, respectively, see Table 2). During the study duration, RGL values were the highest from July to November 2020 (around 0.61 ± 0.01–0.65 ± 0.016), while the lowest value was observed in January 2021 (0.44 ± 0.01) (One-way ANOVA, F_2,11 = 28.22, p < 0.001, Tukey Post Hoc comparison analysis) (Table 2). When testing for spatial variations, RGL values were found to be significantly different between sites. The RGL value at Dam Doi, Ca Mau was the highest (0.57 ± 0.009, F_2,3 = 3.52, p = 0.015), and the lowest was found in Duyen Hai, Tra Vinh and Tran De, Soc Trang (0.53 ± 0.01).

Although the RGL of Ps. chrysospilos varied ontogenetically, all values were RGL < 1, indicating that this species is primarily carnivorous (Al-Hussaini, 1947). The interaction between factors, as assessed using GLM, such as gender × size (F_2,1 = 1.48, p = 0.23); gender × season (F_2,1 = 0.78, p = 0.38); gender x site (F_2,1 = 0.92, p = 0.43); size × season (F_2,1 = 0.24, p = 0.63); size × site (F_2,1 = 1.39, p = 0.25); gender × size × season (F_2,2 = 1.94, p = 0.16); gen × size × site (F_2,2 = 2.31, p = 0.08); gender × season × site (F_2,2 = 0.48, p = 0.70); size × season × site (F_2,2 = 0.89, p = 0.45); gender × size × season × site (F_2,3 = 0.33, p = 0.81), did not exhibit significant impact to the RGL with the exception of interaction season × site (F_2,1 = 3.93, p = 0.008).

Diet composition

Analyses of the digestive tracts of 546 individuals of Ps. chrysospilos revealed clear patterns of food preferences (see Table 3 and Fig. 2). The three most-dominant food items from the pooled data of all individuals were, in descending order of percentage biovolume—Acetes spp. (small shrimps, 26.8%), Uca spp. (fiddler crabs, 21.6%) and Dolichoderus sp. (ant, 19.7%). Other food items present, by percentage biovolume, were Mollusca (molluscs, 16.7%), Polychaeta (bristleworms, 3.4%) and Actinopterygii (ray-finned fishes, 0.7%). Although relatively high in biovolume (11.2%) and occurrence (26.0%), detritus was low in weight (6.6% of total weight).

The variation of occurrences, weight and biovolume of food items when compared to the sex and size of the fishes, as well as the site and season of catches, are illustrated in Table 4. The diet composition of Ps. chrysospilos did not vary under season and size parameters (PERMANOVA, df = 1, Pseudo-F_season = 1.64, p = 0.16; df = 1, Pseudo-F_{fish size} = 0.54, p = 0.75) but did vary under the sex and site parameters (df = 1, Pseudo-F_gender = 3.86, p_gender = 0.003; df = 1, Pseudo-F_{sampling sites} = 3.80, p_{sampling sites} = 0.001). The only diet component that drove the variance between males and females was ‘Uca spp.’ (Mann-Whitney U, df = 1, U = -1.93, p = 0.05), with females consuming more of these crabs than their male counterparts (see Fig. 3). The variation observed between sites were driven by five of the diet components: Mollusca, Uca spp., Dolichoderus sp. and detritus (Kruskal Wallis H test, df = 3, χ²_Mollusca = 16.98, p_Mollusca = 0.001; df = 3, χ²_{Uca spp.} = 8.15, p_{Uca spp.} = 0.04; df = 3, χ²_Dolichoderus = 12.01, p_Dolichoderus = 0.007; df = 3, χ²_detritu = 65.17, p _detritus < 0.001). The three other food items, Acetes spp., Polychaeta and Actinopterygii, were not significant contributors for the variation observed between sites (df = 3, χ²_Acetes = 2.39, p_Polychaeta = 0.50; df = 3, χ²_Polychaeta = 7.51, p_Polychaeta = 0.06; df = 3, χ²_{Actinopterygii} = 3.63, p_Polychaeta = 0.30).

Visualisations on the Costello graph (see Fig. 3) revealed Acetes spp. and Uca spp. to be the most significant prey items for Ps. chrysospilos, followed by Dolichoderus sp. and Mollusca with equal significance. Detritus, also high in occurrence, was not considered significant based on the position of this diet component within the lower right quadrant of the Costello graph. Both Polychaeta and Actinopterygii were considered less important prey items reflected by their positions within the lower left quadrant of the Costello graph. The diet composition of individuals of Ps. chrysospilos at the four sites (see Fig. 4) were significantly different. At DHTV, Dolichoderus sp. was the most important food item; at TDST, Mollusca was the item with the most significant contribution. At DHBL, Acetes spp. accounted for 30.2% biovolume, and was also by far, the most important prey item to Ps. chrysospilos here. At DDCM, Uca spp. was the main prey item and accounted for 35.9% of the biovolume of the mudskippers occurring at this site.