2.2.1. DNA extraction.

DNA of the sediment bacteria was isolated using NucleoSpin Soil (TaKaRa, Kyoto, Japan) following the manufacturer’s instructions. The sample (250 mg) was first lysed with SL1 and Enhancer SX buffer, horizontally vortexed for 5 min at RT, and centrifuged (11,000 × g, 1 min) to obtain the supernatant containing DNA. The residue was again dissolved in SL2 buffer and vortexed to obtain the supernatant. Genomic DNA was isolated from mixed solutions. The intestinal contents of bacterial DNA were isolated using the same methods, except for a lower amount of starting material (100 mg).

2.2.2. PCR amplification.

The protocol for metagenome analysis followed that reported by Kim (2013) [54]. The forward primer 27Fmod (5’-AGRGTTTGATYMTGGCTCAG-3’) and reverse primer 338R (5’-TGCTGCCTCCCGTAGGAGT-3’) containing 10 bp multiplex identifiers and the Roche GS junior adaptors were used to amplify the V1-V2 region of the 16S rRNA gene from the metagenomic DNA using Astec PC-320. The PCR mixture contained 0.5 U of Extaq (TaKaRa), 2.5 μL of reaction buffer, 2.5 μL of dNTPs (10 mM), 0.2 μM of each primer, and 4 ng of gel-purified genomic DNA, and the total volume was adjusted to 25 μL with double distilled water. The cycling conditions were as follows: initial denaturation at 98°C for 1 min, followed by 20 cycles at 95°C for 15 s, 55°C for 30 s, and 72°C for 60 s, followed by a final 2-min extension at 72°C. The reaction was performed in triplicate for each sample. The products were checked by electrophoresis on a 1% (w/v) agarose gel. The DNA amplicons were gel-purified using Agencourt AMPure XP beads (Beckman Coulter, Brea, CA, USA) following the manufacturer’s instructions, and the concentrations were determined using an Agilent BioAnalyzer 2100 (Invitrogen, Waltham, MA, USA) and a Nanodrop spectrophotometer (Thermo Scientific, Waltham, MA, USA). The triplicate amplicons for the same sample were pooled together at equimolar ratios before sequencing.

2.2.3. Pyrosequencing of the 16S rRNA gene amplicons.

Amplicons were combined in a single tube at equimolar concentrations. The pooled amplicon mixture was purified twice (AMPure XP kit, Agencourt, Takeley, United Kingdom), and the cleaned pool was purified using the PicoGreen assay (Thermo Scientific, Waltham, MA, USA). The pool was then diluted in TE to 10⁵ molecules/mL, and 30 μL of this pool was added to the emulsion PCR reaction to attain a ratio of 0.3 molecules of amplicon per bead. Pyrosequencing was performed using a 454 Life Sciences GS Junior (Roche, Basel, Switzerland).

2.2.4. Sequence data analysis.

Sequence filtration was performed by evaluating sequence lengths larger than 200 bp and matching the barcode sequences of 16S rRNA pyrosequences obtained from the samples. Reads were assigned to relevant samples based on barcode sequences prior to the data analysis. Sequence data analysis was performed using the Quantitative Insight Into Microbial Ecology (QIIME) package [55]. This package enables operational taxonomic unit (OTU)-based community identification (97% similarity), picking representative sequences, taxonomic assignment using the RDP classifier [56], and construction of a phylogenetic tree using the FastTree method. The analysis results were imported to a developed pyrosequencing workflow program (OTUMAMi), which automatically computes the individual read counts and cumulative counts per sample. Microbial community statistics based on different levels of phylogeny were also computed. Taxonomic grouping of operational taxonomic units with read counts greater than nine in total was extracted as the major microbial community. Then, the reads were merged according to their phylogenetic affiliation, and the community compositions were interpreted with representative sequences. A phylogenetic tree was constructed using the same set of sequences using the QIIME. Sequence data from this study have been deposited in the DDBJ DRA database (https://ddbj.nig.ac.jp/DRASearch) under accession numbers DRR299057-DRR299088.

2.2.5. Data processing.

The rarefaction curves and related parameters were calculated using EstimateS. The abundance of OTUs was square-root transformed to calculate similarity. Generalized linear models with negative binomial families were used to assess the differential abundances of OTUs for crabs and sediment with the ’DESeq2’ package within R [57, 58]. The differences in OTUs among each sample were statistically analyzed with a Wald test with the Benjamini-Hochberg method to correct p-values for false discovery rate at alpha values of 0.1. Non-metric multidimensional scaling (nMDS) was conducted for sediment and crab intestinal bacteria based on the dissimilarity among OTUs assessed by Bray-Curtis similarities, followed by the analysis of similarity (ANOSIM) to test for significant differences between the bacterial communities. The nMDS and ANOSIM were performed using the PRIMER v5 software package (PRIMER-E Ltd, Ivy Bridge, UK). Phenotype analysis was performed using METAGENassist and PICRUST2 MetaCyc (over 1% frequency on average) [59, 60]. The assigned OTUs were statistically analyzed using the Wald test with the Benjamini–Hochberg method to correct p-values for false discovery rate at alpha values of 0.1.

4.1.1. Crab intestinal microbiome.

At the phylum level, Proteobacteria was dominant regardless of the species. The levels of Firmicutes of N. smithi and E. versicolor were significantly higher than those of P. indiarum which is mainly attributable to Lachnospiraceae (Fig 2A; in Wald test (alpha<0.1)). At the family level, the cellulolytic Demequinceae was notably higher than 3%, regardless of the species (S4 and S5 Files). Higher amounts of Spirochaetaceae (5.60% on average) were enriched in N. smithi in the Wald test (alpha<0.1; S4 and S5 Files). In fiddler crabs that live in mangroves, Proteobacteria and Bacteroides have been found to be enriched [28] In the present study, we found that photosynthetic Rhodobacteraceae was enriched in the intestine (3.7%, 4.5%, and 10.8% in N. smithi, E. versicolor, and P. indiarum, respectively). Bacteroidetes are highly abundant in the better-known litter feeder termites [63]. Similarly, Firmicutes are represented by common intestinal bacteria that are shared between termites and crabs [63–65]. Venn diagram and MDS analysis identified some common OTUs among crab species; however, clear species specificity, particularly for N. smithi, was found in ANOSIM test (Fig 3A and S3 File). In our field observations, the sesarmid crabs seemingly had different habitat preferences around the mounds; for example, N. smithi was found mostly in or near their burrow openings in the middle to upper part of the mound, and rarely in the flats (lower region of the mound). In contrast, E. versicolor and P. indiarum walked around the forest floor and were mostly found in the lower region of the mound. The sediment microbiome profile gradually varied depending on the region of the mound (Figs 2B and 3B and S3 File). Thus, the metagenomic profiles of the crabs are species-specific; however, the environment of the surroundings may also affect the profiles. This is supported by the fact that a particular individual of E. versicolor (EV4) collected in the sandier substrate showed a distinct intestinal microbiome (Fig 2A). The effects of the environment may occur through the microbiome of the habitat (pools of bacteria may colonize the crab intestines), as suggested in fiddler crabs [28]. In termites, however, phylogeny affects the intestinal microbial profile rather than the environment or diets [66–68]. Nevertheless, there is scope for further investigation in the future.

4.1.2. Sediment microbiome.

Sediment microbiome analysis revealed that Proteobacteria was the most dominant phylum, especially containing sulfur-related genera, according to previous reports (Fig 2B) [22–25]. Notably, some OTUs (Demequinceae) were found specifically in the burrows. In the MDS analysis, a regionally dependent cluster was observed in ANOSIM test, but the profile was not perfectly conserved between years (2013–2014) (Fig 3B). The OTUs were highly shared between the middle and lower regions, followed by the middle region and burrow, which were physically close (S3 File). This distribution pattern is presumably regulated by key physicochemical factors in the mound. For example, the upper region is rarely immersed by tides, except for the timing of high tides, while the flat region was regularly submerged twice per day. This suggests that the upper region is more arid and provides breathable and salty conditions for bacteria [5, 28, 69]. In addition, the upper region is more easily affected by sunlight because it has fewer obstacles apart from mangrove trees and leaves. Thus, it is reasonable that the upper region showed specific profiles that were statistically characterized by Hyphomicrobiaceae, Mariprofundaceae, and Rhodobacteraceae in the Wald test (alpha<0.1). These three families were aerobic. Zhu et al. (2018) reported that the upper zone was occupied by higher proportions of heterotrophic bacteria, such as Desulfobacterales, Anaerolineae, and Acidobacteria, while the proportion of Rhodobacterales and Xenococcaceae was greatly increased in the lower zone [69]. Thus, a distinct microbiome pattern may ubiquitously occur between the upper region and flat zones in the intertidal region, but the profile may vary depending on the environment and field [69]. In the upper region, sporadically high values were observed in some families (Ignavibacteriaceae, Ectothiorhodospiraceae, Acetobacteraceae, Acidobacteriaceae, and Conexibacteraceae).

4.2.1. Metabolism of carbon.

It is generally accepted that sediment microbes decompose cellulose in litter before transportation of organic matter by tides [5, 11, 14, 70]. In a previous report, we identified β-1,4-endoglucanse, β-glucosidase, and total cellulase activity in E. versicolor, P. indiarum, and Episesarma palawanense [39]. Bui and Lee (2015) reported the same enzymatic activities in Parasesarma erythodactyla and determined the complete/partial cDNA sequences of putative endo- β -1,4-glucanase in nine mangrove crabs [38]. These data suggest that mangrove crabs can endogenously digest cellulose, regardless of the species. These results are supported by the present results (Fig 6A and 6B). The litter is first shredded by the mouth/stomach teeth, which can promote the easy accessibility of the enzyme, and the cellulose involved is degraded in the stomach by endo- β-1,4-glucanase and β-glucosidase in the digestive juice. Partially degraded cellulose is then degraded by endo-β-1,4-glucanase from intestinal bacteria. After being released as feces in the sediment, cellulose was further degraded by cellulase from the sediment bacteria.

Kristensen and Pilgaard (2001) reported the degrading effect of deposited fecal pellets from N. versicolor fed with green R. apiculata leaves on the heterogeneity of anaerobic microbial activity in sediment [17]. The solid fecal material supported a 55-fold faster microbial decay than the solid leaf material. Similar studies have also been conducted, showing that sesarmid crabs reduce mangrove leaf litter to fecal fragments, resulting in faster microbial colonization, which enhances the breakdown of mangrove detritus, nutrient recycling, and retention in mangrove sediment [16, 71]. Our results support these data from the viewpoint of cellulose degradation in litter. During the degradation in the sediment, the more degraded the cellulose in the feces, the higher the water solubility because the molecular weight and solubility of cellulose are highly correlated [33]. Considering that dissolved organic carbon (DOC) derived from litter plays a key role in the food web as a main source after being transported offshore, mangrove crabs and sediments have an immense impact on the outwelling patterns of a mangrove ecosystem [14, 72–75].

Stable isotope data in the present study showed that both the δ¹³C values of the stomach contents of N. smithi and N. versicolor were approximately -29‰ (Fig 7). In Kristensen et al. (2017), various environments worldwide were tested for stable isotope analysis: δ¹⁵N of the mangrove litter ranged from 1.6 to 6.6‰, δ¹³C ranged from -30.8 to -25.0‰, whereas in MPB, the values were 2.3‰ and ranged from -20 to -19‰ [11]. Applying these data to the present study, the main food and carbon sources in the stomach contents of these crabs were litter. The difference in δ¹³C between stomach content and muscle was approximately 3.41±0.71 and 4.26±0.27‰ in N. smithi and E. versicolor, which was much higher than that of universally used discrimination values (0.4‰) per trophic level, but similar to previous studies (about 5‰) [11, 41, 45, 76]. This indicated that carbon metabolism, including assimilation, was similar to that reported previously. The link between this high discrimination value for carbon in the digestive manner of cellulose in the digestive juice and intestinal bacteria remains unclear. Kawaida et al. (2019) reported a close relationship between the assimilation efficiency of plant materials as a carbon source and the level of cellulase activity in the hepatopancreas using six crab species in the mangrove [45]. If cellulose is degraded by digestive enzymes and the intestine before assimilation, further examination is necessary.

Demequinaceae, found in crabs and burrows, is the most potent candidate for cellulose-degrading bacteria, and is closely related to Cellulomonadaceae, a family of cellulolytic bacteria (S4, S5 and S7 Files). This family (Demequinaceae) was highly and uniformly distributed in the three mangrove crabs (over 3% regardless of the species) and was enriched in the mound burrow (1.6% and 1.3% in B1 and B3, respectively). The Demequinaceae family of the order Actinomycetales was first proposed by Yi et al. (2007), who described a single recognized species, Demequina aestuarii [77, 78]. To date, this genus contains eight species; however, little information is available on the cellulose degradation activity of this genus, except for D. aestuarii. Nevertheless, from the genomes of six Demequina species, proteins belonging to glycoside hydrolase (GH) families were annotated (not shown). In the present study, Demequinaceae was also detected in the sediment, although the percentage (average 0.15%) was much lower than that in the intestinal microbiome (average 3.97%) (S4, S5 and S7 Files). However, the percentages in the burrows were relatively high (1.6% and 1.3% in B1 and B3). The crabs normally live in the burrow and regularly defecate. The lack of Demequinaceae detected in B2 might be ascribed to the low coverage of this sample or the unequal distribution of feces in the wall of the burrow. According to the phenotypes in the metagenome analysis, the cellulose-degrading activity in the intestine can be attributed partly to the Bacteroides or Clostridium genera in METAGENassist. In the phenotype prediction (Fig 4), a very low percentage of OTUs was enriched in “cellulose degrader” in the sediment, but this was not confirmed by the cellulase assay (Figs 4B and 6). It should be noted that these Demequinaceae reads were not assigned to cellulose degraders in this system because Demequinaceae was a newly identified family (Fig 4). Although cellulose degradation by fungi in mangrove sediments has been described [35], fungi were not detected through 16S metagenome analysis in this study. However, to the best of our knowledge, little is established about their symbiosis with crabs.

Thus, the present study shows the cooperative occurrence of cellulose degradation in both the intestine and sediment microbiome as well as endogenous cellulase in the digestive juice. The public carbohydrate-active enzyme (CAZyme) database includes 156 different families of GHs based on their structure and function (Lombard et al., 2014). In ten GH families (GH1, 3, 5, 6, 7, 8, 9, 12, 45, and 48), cellulolytic activity was found (15 families for hemicellulose). Thus, the robust cellulose molecule is degraded by a vast array of specific activities of highly diverse enzymes [79]. Moreover, sequentially associated reactions by a variety of GHs in endogenous enzymes in the intestinal and sediment microbiomes may affect the rate of degradation (solubility) of cellulose before transporting organic compounds into the ocean. Bui and Lee (2014) reported that cellulolytic enzymes in mangrove crabs were classified as GH9 [38] using the genome of Demequinaceae available in the database, GH3 was found. The classification of GHs in the present study samples remains unclear, but several GHs may be involved sequentially in cellulose degradation in the mangrove ecosystem.

4.2.2. Metabolism of nitrogen.

To date, N₂-fixation in mangrove ecosystems has been detected mainly in sediments [5, 46, 49–53]. Similarly, we also detected nitrogenase activity in the upper, burrow, and root regions of the mound (Fig 6D). Based on the data in the phenotype prediction analysis of the intestinal microbiome of the crabs, we also detected apparent nitrogenase activity in the intestinal bacteria of E. versicolor and N. smithi (Fig 6C). In animals, nitrogenase activity has been reported in the symbiotic microbes in termites, wood-boring beetles, shipworms, sponges, corals, and fiddler crabs, which allow them to survive in nitrogen-poor environments [26, 80–83]. Generally, nitrogen-containing compounds are partly released as feces, while other compounds are incorporated into the body of the host animal [65]. According to Tayasu et al. (1994), at least 30%–60% of the nitrogen content of the wood-feeding termite Neotermes koshunensis (Kalotermitidae, Isoptera) is derived from the atmosphere via nitrogen fixation [84]. Moreover, Thompson (2013) estimated that the wood-feeding sawfly S. noctilio F. derives >90% of its larval N budget from N₂ fixation [85].

However, it is still premature to apply these cases directly to sesarmid crabs because an earlier study ruled out the assimilation of symbiotic nitrogen-fixing bacteria in the intestine, based on data obtained by a CHN analyzer after a breeding experiment of N. versicolor [86]. The prevailing opinion so far, based on analysis including stable isotope analysis, is that most leaf-eating sesarmid crabs compensate for their N supply partly by occasional consumption of animal tissues, such as carcasses of fish, crustaceans, and mollusks [87, 88], or through predation/cannibalism, and partly by sediment-associated bacteria, MPBs, fungi, and meiofauna [89–91].

In the present study, the stable isotope analysis did not support the prediction that crab muscle shows lower δ¹⁵N than gut contents (food) of crabs, possibly due to nitrogen fixation in intestinal microbes and assimilation of microbes (i.e., δ¹⁵N value was close to zero) (Fig 7). In the biofilm on the carapace of the fiddler crabs, δ¹⁵N values were significantly lower than in primary producers as potential sources of detritus (phytoplankton, MPBs, litter, etc.), although the values from muscle and other tissues were approximately 8% [26], which is similar to or higher than the values reported in the present study. The present findings indicate the contribution of nitrogen fixation from the intestinal bacteria of sesarmid crabs; false positives from the assay are rare because the reaction requires high activation energy. Further investigation is needed to analyze the contribution of the intestinal microbiome to the body.

In N. versicolor and N. smithi, a clear difference in Δ¹⁵N values between the stomach content and muscle was observed, with values of 1.10±0.20‰ and 4.07±0.38‰, respectively (Fig 7; p<0.01 in t-test). This may reflect the digestive systems of both the crab species. Using the isotope mixing model, Kristensen et al. (2017) reported that Δ¹⁵N in leaf-eating crabs may vary up to 3‰ without serious problems [11]. The δ¹⁵N values of the stomach contents of N. versicolor were significantly higher than those of N. smithi (Fig 7; p<0.01 in t-test). This may be because N. smithi are more herbivorous, but the aging of litter and the crab preference for leaf color should be considered. The litter changes in color from green to yellow-brown, and there have been several reports about crab preference for this color/aging [43, 88, 92]. N. versicolor reportedly prefers brown litter to other colors, whereas N. smithi shows little preference for leaf color. In earlier studies, δ¹³N values generally decreased with aging, suggesting that we cannot easily compare these values between species to determine their food composition [88, 92].

In the phenotype prediction for nitrogen fixation, the enriched OTUs were mainly derived from the families Spirochaetaceae, Shewanellaceae, Vibrionaceae, and Rhodobacteraceae (Fig 4). The presence of Spirochaetaceae is noteworthy. In the termite gut, nitrogen-fixing and ammonia-oxidizing bacteria play a crucial role in providing termite bodies with nitrogen from the low nitrogen content of the wood diet [93]. It is widely accepted that Spirochetes, the most conspicuous bacterial group in lower termite guts, is capable of diverse metabolic processes, including nitrogen fixation, acetogenesis, and degradation of lignin phenolics (Lilburn et al., 2001; Warnecke et al., 2007; Yamada et al., 2007; Hongoh et al., 2008; Lucey and Leadbetter, 2014) [94–98]. It is important to note that a significantly higher percentage of Spirochaetaceae was observed in N. smithi (5.60%), than in E. versicolor (<1.1%) and P. indiarum (0.3%) (Wald test; alpha<0.1; S4 and S5 Files). In addition, the nitrogenase activity in N. smithi was higher than that in E. versicolor, with remarkably high activity in the two individual samples (p<0.05, Wilcoxon test; Fig 6C). It should also be noted that, based on previous reports and our visual observations, N. smithi is herbivorous, while the other two species are omnivorous [43, 92]. The function of Spirochaetaceae, especially N. smithi, should be further investigated in the future.

Nitrogen in the air can be fixed in both the sediment and the crab intestine. It should be noted that the activity of nitrogenase is generally quite fragile against oxygen and is easily inactivated within a few days [99, 100]. The intestinal condition is anaerobic; however, once it is released into the sediment, it is exposed to oxygen [100]. It has been generally suggested that crabs eat their feces mixed with soils, resulting in repeated processing in this step [18, 70, 101]. After the crabs eat the mixture containing the bacteria, the intestine can provide anaerobic conditions to the bacteria again. If this is repeated, it might be possible that the viability of the bacteria is maintained to fix nitrogen. The reason for the two outliers in nitrogenase activity in N. smithi remains unknown (Fig 6C), but in plants, the induction of nitrogenase at the transcriptional level under nitrogen-limited conditions has been reported [102]. Although the details remain unknown, nitrogenase activity might be suppressed under normal conditions but stimulated by unknown factors.

In the phenotype analysis of the sediment, nitrogen cycles such as ammonia oxidizer, nitrate reduction, and nitrogen fixation were enriched regardless of the mound regions (Fig 4B). Similar results were obtained from PICRUSTs (S6 File). These findings may be mainly attributable to Desulfobulbaceae, Rhodobacteraceae, Hyphomicrobiaceae, etc., in the analysis using METAGENassist. This partly follows Andreote et al. (2012), who found that the microbial core involved in methane, nitrogen, and sulfur metabolism consists mainly of Rhodobacteraceae, Desulfobacteraceae, Burkholderiaceae, and Planctomycetaceae [25]. The samples used to determine nitrogenase activity and perform metagenome analysis were obtained during different years and months (November 2017 and Dec 2013/2014, respectively); thus, environmental factors could differ and influence the results. However, the study area is characterized by a long rainy season from May to December, and the average temperature in November and December was almost constant (low-high: 21–32°C) [103]. Thus, temperature and precipitation conditions are unlikely to differ significantly between the samples. Spring-neap tidal cycle might play a role because the samples were taken at different periods of the cycle, where the frequency of inundation/exposure differs. Further studies considering temporal changes in the metagenome and enzymatic analyses are needed.

4.2.3. Metabolism of sulfur.

Under anaerobic conditions in mangrove sediment, it is widely accepted that oxygenic sulfur-oxidizing bacteria and strict anaerobic sulfate-reducing bacteria (SRBs) are very important in the processing [24] where SRBs degrade organic compounds, including cellulose, by using sulfate as a terminal electron acceptor [104–106]. In marine sediments from temperate climates, SRBs degrade 53% of organic matter, and this value varies between 70% and 90% in salt marsh plateaus [104]. There are many reports about the dominance of sulfur-reducing bacteria in mangrove sediments, which are classified as δ-Proteobacteria and ε-Proteobacteria [24, 104], although little is known about the function of intestinal bacteria in sulfur metabolism. In the present study, sulfur-related Desulfobulbaceae (δ-Proteobacteria; 16.2±5.0%), Desulfobulbaceae (δ-Proteobacteria; 10.2±5.4%), Desulfarculaceae (δ-Proteobacteria; 3.9±1.5%), and Desulfuromonadaceae (2.2±1.5%) were highly enriched in the sediment, whereas they were clearly less enriched in crab intestines (0.02±0.04%, 1.4±1.1%, 0.02±0.07%, and 0.10±0.2%, respectively) (S5 File). The completely inconsistent levels of Desulfobacteraceae, Desulfobulbaceae, Desulfarculaceae, and Desulfuromonadaceae may suggest mutual complementation of sulfur metabolism between the intestine and sediment, but this remains unclear.

4.2.4. Significance of the burrow.

From our observations during the field sampling, the burrows sampled in this study were considered N. smithi burrows. They showed higher cellulase activity than other regions in the Dunnett’s test (Fig 6B), and a significant number of OTUs were enriched in phenotype prediction (Fig 4B). N. simithi mainly inhabits the mid to upper part of the lobster mound, whereas burrows of E. versicolor and P. indiarum are usually on the flat or root area of the mangrove (not collected in this study). In addition, the body and corresponding burrow entrance size of N. smithi were larger than those of other species. As per the Venn diagram analysis of the burrow and three crab intestines (S8 File), a total of 349 burrow OTUs were shared between the crabs, including Demequinceae. Eighty OTUs were specifically shared with N. smithi, which included Spirochaetaceae and Desulfobulbaceae. Indeed, crabs mainly feed on the litter in the burrow, but it is still not wise to discuss the species-specific metabolic function of the burrow because we did not have the data of the burrow of E. versicolor and P. indiarum.

In the experimental design, the burrow in this study was expected to be most bioturbated, while lower (open space) regions were expected to be less affected by animals. However, 2045, 3071, 2524, and 2417 OTUs from burrows were shared with the upper, middle, lower mound, and dug regions, respectively, in the Venn diagram analysis (S3 File). Considering the OTU sharing with crab intestine (S7 and S8 Files), the burrow may be the most bioturbated, but the effects were not restricted to this region, possibly because of the mixing of the sediment by tidal effects or some other factors.