European common carp (Cyprinus carpio carpio L.) of the R3 × R8 strain were used, which originated from a cross between the Hungarian R8 strain and the Polish R3 strain (38). Carp were bred and raised in the aquatic research facility of Wageningen University at 24°C in recirculating UV-treated water and fed pelleted dry food (Skretting, Nutreco) twice daily. Fish were housed in rectangular 60-l tanks connected to the same recirculation system, resulting in common water supply. All experiments were performed with the approval of the animal experiment committee of Wageningen University (DEC number 2015098).

Nine-month-old fish (20–40 g, n = 5) were starved for 24 h and subsequently euthanised in 0.3 g/l tricaine methane sulfonate (TMS) (Crescent Research Chemicals, Phoenix, USA) in aquarium water buffered with 0.6 g/l sodium bicarbonate and bled via the caudal vein. The third segment of the intestine was isolated, as this segment shows a stable microbiota composition (39), and intestinal content of these samples was collected by scraping. Samples were snap-frozen in liquid nitrogen and stored at -80°C.

Prior to isolation of RNA from intestinal content, all samples were weighed. Total RNA from intestinal content was isolated using the RNeasy PowerMicrobiome Kit (Qiagen, 26000-50), following the manufacturer’s instructions. Agarose gel electrophoresis and a ND1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) were used to control RNA yield and quality (260/280: >2.0, 260/230: >2.0). RNA samples were stored at -80°C until further use.

Total community analyses were performed on initially five replicates per treatment similar as described for Perez-Jaramillo et al. (40). Briefly, total community RNA was used for amplification and sequencing of the 16S rRNA, targeting the variable V3–V4 regions resulting in amplicons of approximately ~460 bp. Illumina 16S rRNA gene amplicon libraries were generated and sequenced at BaseClear (Leiden, the Netherlands). In short, barcoded amplicons from the V3–V4 region of 16S rRNA genes were generated using a 2-step approach. Ten ng RNA was used as template for the One-Step RT-PCR Kit (Qiagen^®) employed according to the manufacturer’s instructions with a total volume of 50 μl using the 341F (5′-CCTACGGGNGGCWGCAG-3′) and the 785R (5′-GACTACHVGGGTATCTAATCC-3′) primers appended with Illumina adaptor sequences. Control PCR reactions were performed alongside each separate amplification without addition of template. PCR products were purified using AMPure XP beads according to the manufacturer’s instructions, and the sizes of the PCR products were checked on a Fragment Analyzer (Advanced Analytical) and quantified by fluorometric analysis. Purified PCR products were used for the 2nd PCR in combination with sample-specific barcoded primers (Nextera XT Index Kit, Illumina). Subsequently, PCR products were purified, checked on a Fragment Analyzer (Advanced Analytical), and quantified, followed by multiplexing, clustering, and sequencing on an Illumina MiSeq with the paired-end (2×) 300-bp protocol and indexing. For each independent sample, at least 25,000 raw reads were sequenced. The sequencing run was analysed with the Illumina CASAVA pipeline (v1.8.3) with demultiplexing based on sample-specific barcodes. The raw sequencing data produced were processed removing the sequence reads of too low quality (based on “failed-chastity<=1”, using a chastity threshold of 0.6, on the first 25 cycles) and discarding reads containing adaptor sequences or PhiX control with an in-house filtering protocol. A quality assessment on the remaining reads was performed using the FASTQC quality control tool version 0.10.0.

The RDP extension to PANDASeq (41) named Assembler (42) was used to merge paired-end reads with a minimum overlap of 10 bp and at least a PHRED score of 25. Primer sequences were removed from the paired sample FASTQ files using FLEXBAR version 2.5 (43). Sequences were converted to FASTA format and concatenated into a single file. All reads were clustered at 97% similarity into OTUs using the UPARSE strategy by dereplication, sorting by abundance with at least two sequences and clustering using the UCLUST smallmem algorithm (44). These steps were performed with VSEARCH version 1.0.10 (45), which is an open-source and 64-bit multithreaded compatible alternative to USEARCH. Next, chimeric sequences were detected using the UCHIME algorithm (46) implemented in VSEARCH. All reads before the dereplication step were mapped to OTUs using the usearch_global method implemented in VSEARCH to create an OTU table and converted to BIOM-Format 1.3.1 (47). Finally, taxonomic information for each OTU was added to the BIOM file by using the RDP Classifier version 2.10 (48). Representative OTU sequences were assigned to a taxonomic classification via BLAST against the Silva database (version 128). All steps were implemented in a Snakemake workflow (49), which is available on GitHub ¹ . For downstream analysis, we took the obtained OTU table and prepared a “filtered table” using QIIME (1.9.1) custom scripts (50). First, we extracted from the OTU table the Bacteria domain using the command split_otu_table_by_taxonomy.py. Next, we discarded singletons, doubletons, Chloroplast, and Mitochondria sequences using the command filter_otus_from_otu_table.py.

Carp were euthanised with 0.3 g/l TMS (Crescent Research Chemicals) in aquarium water buffered with 0.6 g/l sodium bicarbonate and bled using vacuettes (BD Vacutainer SST with 21-G needles, Becton Dickinson) via the caudal vein. The head kidney was isolated, and total head kidney leukocytes (HKLs) were separated on a 1.02–1.08-g/ml Percoll (GE Healthcare, Thermo Fisher Scientific) density gradient, as previously described (13).

Production of NO was determined as nitrite accumulation using the Griess reaction, as previously described (57). HKLs were seeded at a density of 1 × 10⁶ per well in 96-well culture plates (CORN3596; Corning) and stimulated with one of the following: RPMI (control) or LPS (L2880, 30 µg/ml, Sigma-Aldrich) in the presence or absence of different concentrations of sodium acetate (0.5–25 mM, S2889-250G, Sigma-Aldrich), sodium butyrate (0.05–25 mM, 303410-100G, Sigma-Aldrich), or sodium propionate (0.05–25 mM, P1880-100G, Sigma-Aldrich) or in the presence or absence of a mix of acetate (62.5%), butyrate (10%), or propionate (27.5%) in the proportion of produced SCFAs after in vitro fermentation of MacroGard^® at 0.1–25 mM. After 96 h at 27°C in the presence of 5% CO₂, nitrite production was measured at OD540, using a FilterMax F5 Multi-Mode Microplate Reader and quantified using a sodium nitrite (NaNO₂) standard curve.

Production of reactive oxygen species (ROS) was determined by a real-time luminol-based ECL assay, as previously described (58). HKLs were seeded at a density of 1 × 10⁶ per well in white 96-well plates (CLS3912; Corning) and stimulated with one of the following: RPMI (control), zymosan (tlrl-zyd, 50 µg/ml, InvivoGen) in the presence or absence of different concentrations of sodium acetate (0.5 and 2.5 mM, S2889-250G, Sigma-Aldrich), sodium butyrate (0.1–0.5 mM, 303410-100G, Sigma-Aldrich), or sodium-propionate (0.1–0.5 mM, P1880-100G, Sigma-Aldrich) or in the presence or absence of a mix of acetate (62.5%), butyrate (10%), or propionate (27.5%) in the proportion of produced SCFAs after in vitro fermentation of MacroGard^® at 0.5–5 mM. Stimulation with PMA (P8139, 1 mg/ml, stimulated control; Sigma-Aldrich) was used as a technical control for the assay (data not shown). SCFA concentrations were based on the outcomes of the nitrogen radical production analysis, and thus fewer SCFA concentrations were included the ROS analysis. Chemiluminescence emission was measured in real time (every 2 min for 120 min) with a FilterMax F5 Multi-Mode Microplate Reader at 27°C and expressed as area under the curve, as previously described (56). Fold changes were calculated as the area under the curve of stimulated HKLs relative to unstimulated HKLs (RPMI without zymosan).

For gene expression analysis, HKLs were seeded at a density of 4.5 × 10⁶ per well in a 24-well plates (CORN3524; Corning) and stimulated with RPMI and LPS (L2880, 30 µg/ml, Sigma-Aldrich) in the presence or absence of a mix of acetate (62.5%), butyrate (10%), or propionate (27.5%) in the proportion of produced SCFAs after in vitro fermentation of MacroGard^® at 2.5 mM or of sodium acetate (1.55 mM, S2889-250G, Sigma-Aldrich), sodium butyrate (0.25 mM, 303410-100G, Sigma-Aldrich), or sodium propionate (0.765 mM, P1880-100G, Sigma-Aldrich); at the relative concentration, each was included in the 2.5-mM SCFA mix. Each stimulation was performed in technical triplicate and in n = 5 independent experiments. At 3 and 6 h post stimulation, 13.5 × 10⁶ cells were lysed in the lysis buffer from the RNeasy Kit (RLT buffer) and stored at -80°C until RNA isolation.

Carp of 9 months (20–40 g, n = 20 per group) were starved overnight and anaesthetised with one-third the killing dose of TMS (i.e., 0.10 g/l TMS in aquarium water buffered with 0.2 g/l sodium bicarbonate) to allow manipulation without killing. Anaesthetised fish received an oral gavage with 100 µl of PBS (n = 20, Lonza, LO BE17-516F) or MacroGard^® dissolved in PBS (n = 20, 10 mg/ml, 1 mg per fish) using a 200-μl pipette. Fish feeding (twice-daily) was resumed at 12 h post gavage. At 3, 7, 11, and 14 days post gavage, before morning feeding, n = 5 fish were euthanised in 0.3 g/l TMS in aquarium water buffered with 0.6 g/l sodium bicarbonate and bled via the caudal vein (data for 3, 11, and 14 days are not shown). The third segment of the intestine was isolated, and intestinal content of the third segment was collected by scraping, snap-frozen in liquid nitrogen, and stored at -80°C for 16S rRNA sequencing. A tissue sample of the third segment of the intestine was snap-frozen in liquid nitrogen and stored at -80°C for gene expression analysis.

Total RNA from isolated tissue samples (25–50 mg per sample) and cells (13.5 × 10⁶ cells per sample) was isolated using the RNeasy Mini Kit (Qiagen, 74106), including on-column DNase treatment, according to the manufacturer’s instructions, and stored at −80°C. Prior to cDNA synthesis, total RNA was treated with DNase I, Amplification Grade (Invitrogen), and cDNA was synthesised using random primers (300 ng) and Superscript III First-Strand Synthesis for RT-PCR (Invitrogen). cDNA samples were diluted 25× in nuclease-free water prior to real-time quantitative PCR (RT-qPCR) analysis.

Gene expression was measured with RT-qPCR using ABsolute qPCR SYBR Green Mix (Thermo Scientific) in a Rotor-Gene Q (Qiagen) as previously described (59). Fluorescence data were analysed using Rotor-Gene Analysis software version 1.7. The relative expression ratio (R) of each sample was calculated according to the Pfaffl method (60) based on the take-off deviation of sample versus each of the PBS controls and normalised relative to the s11 protein of the 40-s subunit as reference gene (see Table 1 for primer information).

Statistical analyses were performed using SPSS (v23.0), and differences were considered significant if p ≤ 0.05. Data presented as fold change (gene expression analyses) were first transformed using natural log transformation [Y = LN(Y)] before analysis. All data were tested for Gaussian distribution using the Shapiro–Wilk test.

Differences in cumulative gas production were analysed with two-way ANOVA followed by the Bonferroni post hoc test. Differences in pH of the fermentation liquid after in vitro fermentation were assessed by Welch’s ANOVA followed by the Games–Howell test, as the data did not have equal variances. Differences in NH₃ and short-chain fatty acid profiles in the fermentation liquid after in vitro fermentation were assessed by one-way ANOVA followed by Tukey post hoc test. Data for comparisons of NO production, ROS production, and gene expression in vitro were analysed using repeated measures and sometimes had some data points missing; therefore, analyses were performed with a linear mixed model, followed by a Bonferroni post hoc test. Comparison of gene expression in vivo was assessed by one-way ANOVA.

Analysis of 16S ribosomal RNA rather than DNA provided insight into the active intestinal microbial communities. Identification of bacterial microbiota up to family level allowed us to investigate the presence of bacteria with theoretical fermenting capacities. The intestinal content of the third segment of the intestine from n = 5 individual fish was analysed by 16S rRNA sequencing, revealing a total of n = 55 active operational taxonomic units (OTUs) in the intestinal content of unhandled carp.

By far, the most represented phyla in the carp intestinal microbiota were Fusobacteria (71% ± 5.8% (mean ± SD, n = 5) of total reads), whereas Bacteroidetes (21% ± 5.3%), Proteobacteria (5% ± 0.5%), and Firmicutes (2% ± 0.5%) were considerably less frequent ( Figure 1A ). The majority of Proteobacteria belonged to the order of Gammaproteobacteria (95%), while the rest belonged to Betaproteobacteria (4%). Looking at the distribution per family in the active microbiota, the most abundant family were Fusobacteriaceae (71% ± 5.8% [mean ± SD, n = 5)], followed by Bacteroidaceae (16% ± 4%), Porphyromonadaceae (5% ± 2%), Vibrionaceae (3% ± 0.75%), Aeromonadaceae (1.5% ± 0.4%), and Erysipelotrichaceae (1.4% ± 0.1%) ( Figure 1B ). Finally, looking at the putative genus level, it becomes clear that all Fusobacteriaceae are member of the Cetobacterium genus, as 71% ± 5.8% (1 OTU [mean ± SD, n = 5)] of the total reads belong to this genus. Following Cetobacterium genus, the most abundant genera based on 16S RNA sequencing are Bacteroides (16% ± 4.1%, 2 OTUs), Vibrio (3% ± 0.75%, 1 OTU), Aeromonas (1.5% ± 0.4%, 1 OTU), and two different uncultured genera of the families Porphyromonadaceae (4.5% ± 4%, 3 OTUs) and Erysipelotrichaceae (1.1% ± 0.1%, 1 OTU). Although the sequencing effort could not identify bacteria at the species level, the abundant presence of the Bacteroides genus well known to express β-glucan-degrading genes (61) suggested the ability of carp microbiota to ferment β-glucans.

In order to further investigate the ability of the naive intestinal microbiota of carp to ferment β-glucans, in vitro fermentation was performed ( Figure 2A ). Through an in vitro batch culture system, fermentation of different substrates could be analysed. As a measure for fermentation kinetics, cumulative gas production was measured over time. As a negative control, minor cumulative gas production was measured in the PBS group, similar to gas production in the negative control without substrate (data not shown). As a positive control, fermentation of the readily digestible monosaccharide glucose was measured, resulting in the highest cumulative gas production. Two different β-glucan preparations were studied: curdlan, a high molecular weight linear polymer consisting of β-1-3-linked glucose residues from Alcaligenes faecalis, and MacroGard^®, a branched polymer β-(1,3/1,6)-glucose feed additive. In comparison, fermentation of curdlan resulted in the most continuous and highest cumulative gas production, while fermentation of MacroGard^® showed an intermediate plateau around 24–36 h. Gas production differed significantly between MacroGard^® and curdlan from 44 to 86 h after start of the in vitro fermentation (as assessed with a two-way ANOVA followed by Bonferroni post hoc, p ≤ 0.05). No matter the difference in kinetics of gas production between curdlan and MacroGard^®, naive carp intestinal microbes were able to ferment β-glucans.

Finally, analysis of the fermentation liquid after in vitro fermentation revealed differences between substrates in pH and NH₃, whereas these parameters did not differ at start (data not shown). In comparison, a (non-significant) trend toward lower pH was observed in the glucose group, compared to the two β-glucan-treated groups. No difference in pH was observed between MacroGard^® and curdlan treatment ( Figure 2B ). Ammonia production was significantly higher in the PBS group compared to all other groups ( Figure 2C ). Interestingly, NH₃ production was different between curdlan and MacroGard^®, providing further evidence for differences in fermentation between the two β-glucans.

Results from the in vitro batch culture analysis provided evidence for the ability of intestinal microbiota of carp to ferment β-glucans, at least to a certain degree. In general, metabolites produced by microbiota during fermentation such as SCFA are often considered necessary components for immune homeostasis (62). We therefore measured the accumulation of SCFA metabolites in the in vitro fermentation liquid, produced during β-glucan fermentation. Significant differences in SCFA profiles between the readily fermentable substrate glucose and the two different β-glucan preparations could be observed ( Figures 3A–E ). Overall, a clear production of acetate, butyrate, and propionate was observed after fermenting either curdlan or MacroGard^® in vitro. Comparing the two different β-glucan preparations showed no differences in acetate production but revealed an interesting inverted trend for butyrate and propionate levels. After fermentation of curdlan, higher butyrate concentrations were measured, while after fermentation of MacroGard^® higher propionate concentrations were measured ( Figures 3B, C ). Profiles of other SCFA; formic, isobutyric, valeric, and iso-valeric acid were also measured, but concentrations were considered negligible (<0.1 mM, data not shown) and therefore excluded during the follow-up experiments. In both β-glucan groups, total SCFA levels were higher than in the glucose group; however, no differences were observed between both β-glucan groups ( Figure 3E ). Overall, differential production of SCFAs could be observed in both β-glucan groups, with fermentation of MacroGard^® specifically resulting in an increase in propionate. With the practical and commercial applicability of MacroGard^® being higher than of curdlan, our follow-up studies focused on the effects of fermentation of MacroGard^®.

In order to investigate the potential immunoregulatory effects of SCFAs, head kidney leukocytes (HKLs) were stimulated with LPS as a well-known stimulator of pro-inflammatory responses and stimulator of nitric oxide (NO) production or with zymosan as a well-known stimulator of pro-inflammatory responses and stimulator of ROS production in common carp, while different concentrations of SCFAs were added. Analysis of NO production at 96 h post stimulation revealed differential modulating potencies of the various SCFAs ( Figure 4 ). Clear dose-dependent modulating effects could be observed for the three SCFAs investigated, but only in the presence of LPS. Butyrate was the most potent SCFA, followed by propionate and lastly acetate.

Furthermore, adding the SCFAs mixed in the same ratio as measured after in vitro fermentation of MacroGard^® (see Figure 3 ) showed similar reductions in in vitro induced NO production in HKLs ( Figure 4D ). Finally, titrating down this SCFA mixture revealed a significant reduction of LPS-induced NO production even at a concentration as low as 2.5 mM.

Interestingly, when analysing the production of ROS, SCFAs appeared to increase the ROS production in a dose-dependent manner, although this trend was only significant for acetate and propionate ( Figure 5 ). Furthermore, adding the SCFAs mixed in the same ratio as measured after in vitro fermentation of MacroGard^® (see Figure 3 ) showed similar modulations of in vitro production of ROS by HKLs. Both with and without the presence of a strong ROS inducer (zymosan), a clear increase could be observed in the ROS production of HKLs.

Studying the gene expression can provide preliminary insights in the effects of SCFAs on inflammation, via gpr40L SCFA receptors. To this end, HKLs were stimulated in vitro with LPS, a mix of SCFAs in the same ratio as measured after in vitro fermentation of MacroGard^®, both LPS and the mix, or with different SCFAs (acetate, butyrate, or propionate) separately. Subsequently, at 3 and 6 h post stimulation, cytokine gene expression ( Figure 6 ) and gpr40L ( Figure 7 ) gene expression were studied.

Addition of SCFAs generally resulted in downregulation of cytokine gene expression ( Figure 6 ), either pro-inflammatory (il-1b, il-6, tnfα, cxca, p40) or anti-inflammatory (il-10), especially when first induced by presence of LPS. In the absence of LPS, addition of SCFAs resulted in a downregulation of cytokine gene expression of il-6, tnfa, and p40. Separately, SCFAs acetate, butyrate, and propionate showed slightly different levels of regulation, with acetate barely regulating gene expression. In general, addition of SCFAs as a mix showed comparable regulation as addition of SCFAs separately. Gene duplicates were also studied (il-6b, il-10b, tnfαa1, tnfαb1, tnfαb2, cxcb, p40b, and p40c, data not shown) and were regulated similarly to the genes presented in Figure 6 ).

Addition of SCFAs resulted in a differential regulation of gene expression of different grp40L receptors ( Figure 7 ). Here, we studied gene expression of ten gpr40L genes and found three (gpr40Lb-1.1, gpr40Lc-1.1, and gpr40Lc-2) not expressed in HKLs. For the seven expressed gpr40L receptors, four were upregulated by stimulation with LPS (grp40La-1.1, gpr40La-1.2, gpr40Lb-2.1, gpr40Lb-2.2).

Regulation of the gpr40L expression in HKL by SCFAs was only stimulatory for gpr40La-1.2 while inhibitory for gpr40La-1.1, gpr40Lb-2.1, and gpr40Lc-1.2. Again, of the different SCFAs analysed in this study, addition of acetate barely regulated gene expression.

To follow up on the above-described in vitro modulating effects of SCFAs, we subsequently analysed the in vivo gene expression locally in the posterior intestine in response to a single oral gavage with β-glucans (MacroGard^®). Based on the in vitro fermentation results (see Figures 2 , 3 ) and the assumption that also in vivo the active naive microbiota of common carp can ferment β-glucans and lead to the production of specific SCFAs, we measured whether the effects of an oral gavage with β-glucans could be measured by a change in gene expression of gpr40L and other genes. Analysis of microbiota composition on day 7 and day 14 showed no significant differences between the control (PBS) and the β-glucan-treated group (data not shown). Gene expression regulation was most clear at 7 days post gavage ( Figure 8 ), while gene expression profiles on 3, 11, and 14 days post gavage did not show distinct trends (data not shown). Similar to the in vitro observations on HKL, we found a significant downregulation in the posterior intestine of cxca, il-1b, il-6, and tnfa expression at 7 days post gavage ( Figure 8A ). Analysis of the expression response of type I IFN-related genes (tlr3.1, tlr3.2, and mx1-5) revealed a clear downregulation of tlr3.1 and a marginal but significant downregulation of mx3 expression ( Figure 8B ). Further, both in vitro and in vivo gpr40La-1.1 gene expression was downregulated ( Figure 8C ), but differently, gpr40La-2 expression was significantly upregulated only in vivo. Of interest, all ten gpr40L receptors were expressed in the posterior intestine.