The RTgutGC cell line was kindly gifted by Tania Rodríguez-Ramos and Prof. Brian Dixon (University of Waterloo, Canada). Routine cultivation followed the protocol described by Kawano et al., 2011. In brief, the cells were grown in 75 cm² culture flasks (13-680-59, Fisher Scientific) containing Leibovitz’s L-15 complete medium (L-15/C) with Phenol red and 2.05 mM L-glutamine (L1518, Sigma-Aldrich) supplemented with 10 % fetal bovine serum (FBS) (97068-085P, Avator) and 1 % Penicillin-Streptomycin (P/S) (P4333, Sigma-Aldrich). The cells were cultured at 20°C in an incubator with a normal atmosphere, and the media was changed twice weekly. Upon reaching confluency, determined by visually inspecting the flask using an inverted microscope, the cells were pretreated with 0.8 mL Versene solution (15040066, Gibco, Thermo Fisher) for 5 min to facilitate dissociation, then trypsinized with 0.25 % trypsin in PBS without Ca²⁺ and Mg²⁺ (392-0436, VWR) for 2 min. Confluency was typically reached after 7–9 days and the cells were subsequently sub-cultured in a 1:2 ratio.

The RTgutGC cells have demonstrated the ability to form a polarized epithelium when cultured on permeable supports in a two-compartment intestinal barrier model, as documented in previous studies (44, 45, 48). To mimic the intestinal lumen (apical/upper chamber) and portal blood (basolateral/lower chamber) environments, 24-well plates (734-0067, Falcon, Corning) with 33.6 mm² transparent transwell inserts (734-0036, Falcon, Corning) and 6-well plates (734-0065, Falcon, Corning) with 425.4 mm² transparent transwell inserts (734-0032, Falcon, Corning) with pore size 0.4 µm were utilized. Cells were counted using a Bright-Line Hemacytometer (Z359629 Merk) and seeded at a density of 62500 cells/cm². Specifically, 300 µl of cell suspension at 68750 cells/mL was added to the apical chamber of 24-well inserts, and 3.5 mL of suspension at 80714 cell/mL was used for 6-well inserts. Before reaching confluency, the media in both compartments were filled with L-15/C ( Figure 1A ). To better resemble the low nutrient concentration in the intestinal lumen versus the high nutrient concentration in the blood stream, the L-15/C in the apical chamber was replaced by L-15/ex during the experiment ( Figure 1B ) (44). L-15/ex is a reduced medium containing only inorganic salts, galactose and pyruvate concentrations of L-15, as formulated by Schirmer et al., 1997. The basolateral chamber contained 1 mL and 3.5 mL media for 24- and 6-well plates, respectively. The media in both compartments were changed 2 times per week over a period of 21 days.

Following seeding of RTgutGC cells in 24-well culture plates with membrane inserts, cells were allowed to attach and acclimate for 3 days prior to measurement of transepithelial electrical resistance (TEER). TEER measurements were performed on days 3, 7, 14 and 20 using an EVOM 3 Voltohmmeter and Endohm-6 electrode (World Precision Instruments, Berlin, Germany). This was conducted to confirm monolayer formation (data not shown) (45). On day 20, after TEER measurement, apical media was removed and replaced with L-15/ex containing functional ingredients in quadruplicate inserts. Following 24 hours of exposure to the functional ingredients 40 µg/mL and 60 µg/mL β-glucan (BG40 and BG60), 60 µg/mL and 90 µg/mL laminarin (Lam60 and Lam90), 100 mM and 120 mM L-Carnosine (Carn100 and Carn120) and 300 µg/mL and 600 µg/mL salmon hydrolysate (FPH300 and FPH600), TEER was measured again. Subsequently, cells were exposed to 60 µg/mL LPS for 6 hours and TEER was measured once more. Experimental setup is illustrated in Supplementary Figure 1 . TEER values were calculated by subtracting the background resistance (membrane inserts without cells) from the inserts containing cells, and results were reported as Ω x cm².

RTgutGC cells seeded in 6-well culture plates with membrane inserts were cultured for 20 days. Media was weekly changed in the apical and basolateral compartments. After this period, the functional feed ingredients; 40 µg/mL and 60 µg/mL β-glucan (BG40 and BG60), 60 µg/mL and 90 µg/mL laminarin (Lam60 and Lam90), 100 mM and 120 mM L-Carnosine (Carn100 and Carn120) and 300 µg/mL and 600 µg/mL salmon hydrolysate (FPH300 and FPH600) were added to the six replicate wells. After a 24-hour incubation, the media was removed from three of the replicates per treatment. The apical membrane was washed with 1 mL PBS before 3 mL Lucifer yellow (LY, potassium salt, MW = 522 Da; Thermo Fisher Scientific, Switzerland) dye mix (50 µg/mL) was added. The media in the basolateral compartment was changed with 3 mL PBS. The plates were incubated in the dark at 20°C, and 100 µl samples were taken from the basolateral compartment after 20, 30, 60, 90 and 120 min and added to conventional 96-well plates without inserts. Additionally, 100 µl samples were taken from the apical compartment after 20 and 120 min. When only sampling from the basolateral compartment, the volume was corrected by adding 100 µl PBS. Fluorescence was measured immediately after sampling using a microplate reader (VLB000 Varioskan™ LUX mulitimode) with excitation/emission wavelengths of 450/520 nm. Fluoresence readings were recorded as relative fluorescent units (RFUs) based on a single read and converted to the concentration of passed LY based on a separately made standard curve. Background fluorescence from wells without cells (blank) was subtracted from both control and stimulated cells. After the last sampling of LY, the LY dye mix was removed, and the cells washed twice with 1 mL PBS before cells were sampled for qPCR. The apical compartment of the last three replicates were washed with 1 mL PBS and then incubated with 60 µg/mL 3 mL LPS for 6 hours. After this period, the LY sampling protocol was conducted as previously described, followed by cell collection for qPCR analysis ( Supplementary Figure 2 ). Apparent permeability (P_app) for LY was calculated according to Hubatsch et al., 2007 (protocol 2) using Equation 1.

where ( d Q d t ) is the quantity of LY on the basolateral side as a function of time (µmol s^-1), C₀ the initial concentration of LY on the apical side (µmol/mL), and A the surface area of exposed RTgutGC monolayer on the transwell inserts (4.520 cm²).

After conducting the LY translocation assay, the cells were harvested for qPCR analysis at day 21 ( Supplementary Figure 2 ). Briefly, after rinsing with PBS, 600 µl TRI Reagent solution (Direct-zol RNA Miniprep Plus kit, Zymo Research Corporation) was added directly to the RTgutGC monolayer. The cells were detached and lysed by repeated pipetting and scraping of the insert bottom 8–10 times. The resulting cell homogenates were transferred to 2 mL cryogenic tubes (479-1220, VWR), placed on ice, and then stored at -80°C until further processing. Total RNA was subsequently extracted using the Direct-zol RNA Mini Plus kit (Zymo Research Corporation) according to the manufacturer’s protocol, which included DNase I treatment step to remove any residual genomic DNA that could interfere with the PCR reactions. The RNA was eluted in 50 µl of RNase-free water. To ensure complete elution, the eluate was reapplied to the column and centrifuged at 12 000 x g. RNA purity and concentration were measured using a NanoDrop ND-1000 spectrometer (ThermoFisher Scientific), with all samples showing a 260:280 greater than 1.9. RNA integrity was confirmed using an Aligent 2100 Bioanalyzer with a DNA Nano Chip (Agilent Technologies, CA, USA) yielding a RIN value of greater than 9. Total RNA was stored at -80°C until further use. First-strand complementary DNA (cDNA) was synthesized from 1.0 µg of total RNA using LunaScript RT SuperMix kit (New England Biosystems). Negative controls were performed in parallel by excluding either RNA or the enzyme.

The 13 chosen target genes had key roles in immune signaling and inflammatory responses; il6 (interleukin-6, IL-6), il8 (interleukin-8, IL-8), il1b (interleukin-1β, IL-1β), tnfa (tumor necrosis factor-alpha, TNF-a), myd88 (myeloid differentiation primary response 88, MyD88), barrier function and integrity; cdh1 (E-cadherin, CDH1), cldn3 (claudin-3), cldn12 (claudin-12), zo-1 (zonula occludens-1, ZO-1) and metabolic processes; tgfb (transforming growth factor-beta, TGF-β), pcna (proliferating cell nuclear antigen, PCNA), ialp (intestinal alkaline phosphatase, IALP) and slc10a2 (apical sodium-dependent bile acid transporter, SLC10A2). Primers for these genes were designed using NCBI’s Primer-BLAST or sourced from existing literature ( Supplementary Table 3 ). All primers were ordered from Invitrogen by Thermo Fisher Scientific in a hydrophilized state. The optimal annealing temperatures for each gene’s primer pair (forward and reverse) were tested on a pool of cDNA from all samples using gradient PCR method with temperatures ranging from 54.0 – 62.0°C. Final annealing temperatures were chosen based on amplification curves and melting peak analyses. PCR efficiency was calculated for each gene assay via two-fold serial dilutions of pooled cDNA, using LinReg PCR software (54, 55). Gene expression (quantified by Cq values) was analyzed on a LightCycler 96 system (Roche Diagnostics, Basel, Switzerland) using LightCycler 480 SYBR Green I Master Mix. Each 20 µL PCR reaction contained 3 µL PCR-grade water, 2 µL of forward and reverse primers (10 µM), 10 µL SYBR Green I Master Mix, and 5 µL cDNA template diluted at either 1:10 (il8, il1b, tnfa, myd88, tgfb, pcna, ialp, cdh1, cldn3, cldn12 and zo-1) or 1:5 (il6 and slc10a2). Negative controls (without cDNA) and no-RT controls were included. After detecting a low amount of gDNA in the samples, the samples were cleaned using RNA Clean & Concentrator-5 (Biosite-R1013, Zymo Research). The qPCR protocol consisted of an enzyme activation step at 95°C for 10 minutes, followed by 45 amplification cycles: denaturation at 95°C for 15 seconds, annealing at 55-58°C (temperature optimized per gene, see Supplementary Table 3 ) for 15 seconds, and elongation at 72°C for 15 seconds. Target gene expression was normalized to the geometric mean of actin beta (actb) and elongation factor 1α (EF1a), selected for their stable intra- and interspecific expression as confirmed using the geNORM function (56) in qbase. Mean normalized fold change was calculated from raw Cq values by relative quantification using Equation 2 (57).

All statistical analysis was performed in R Studio (R version 2024.04.1, ^©2009-2024, RStudio, PCB). Normal distribution was visualized using Q-Q plot and Shapiro-Wilks normality test, and homogeneity of variance was tested using Levene’s test. Data were analyzed using two-way ANOVA with (LPS) challenge and treatment as class variables followed by Dunnett’s multiple comparisons test within each challenge group (before and after LPS challenge) ( Supplementary Tables 4 and 5 ). The Dunnett’s test was chosen to compare multiple treatments to its respective control group while controlling the overall Type I error rate (58). Data were calculated as mean ± SEM with 3 or 4 technical well or insert replicates (given in the figure legends). Asteriks denote treatment groups that are statistically different from the control (*** p < 0.001, ** p < 0.01, * p < 0.05), whereas trends are denoted by “.” (0.05 ≤ p < 0.10).

For most functional feed ingredients, the cell viability measured by a quantitative measure of cell metabolic activity (AB assay) and cell membrane integrity (CFDA-AM assay) decreased with increasing inclusion levels. Except for carnosine, where the cell metabolic activity decreased while the cell membrane integrity did not – these two measurements followed each other. For further testing, two levels of inclusion were chosen, the highest level of an ingredient giving 100 % viability, and the second level giving 80 % viability. This gave the following working concentrations; β-glucan with 40 µg/mL and 60 µg/mL (BG40 and BG60) ( Figure 2A ), laminarin with 60 µg/mL and 90 µg/mL (Lam60 and Lam90) ( Figure 2B ), L-Carnosine with 100 mM and 120 mM (Carn100 and Carn120) ( Figure 2C ) and salmon hydrolysate with 300 µg/mL and 600 µg/mL (FPH300 and FPH600) ( Figure 2D ). LPS did not affect viability ( Figure 2E ) and a concentration of 60 µg/mL were chosen based on literature data (48). All test concentrations are shown in Supplementary Table 2 .

Based on the cell viability assay, two concentrations of each functional feed ingredient were selected. The first was a concentration that maintained 100 % viability, ensuring no adverse effects on cell functionality. This concentration could be safely incorporated into diet formulations during challenging farming conditions. The second concentration resulted in approximately 80 % viability, balancing minimal toxicity with a sufficient response to evaluate the ingredient’s effects. The assay revealed that increasing concentrations of all functional feed ingredients lowered the cell metabolic activity, while the cell membrane integrity was lowered by all, except carnosine. This suggests that while increasing concentrations of carnosine reduces metabolic activity, it is not directly cytotoxic. The observed effect may be due to a combination of carnosine’s antioxidant properties, which can modulate mitochondrial function and reduce oxidative stress without inducing membrane damage (59, 60), and the fact that high concentrations of carnosine are found in the gut, brain and muscle of fish and other vertebrates (61). In contrast, the other ingredients that reduced both metabolic activity and membrane integrity likely triggered apoptotic or necrotic pathways (62). However, the specific mechanisms were not the focus of this study, as our main emphasis was on immune- and barrier regulation. The viability of the RTgutGC cells was not affected by any of the tested LPS concentrations, consistent with previous studies (47, 48). While fish are generally considered less responsive to LPS than mammals due to the absence of the LPS-specific receptor TLR4 and co-stimulatory molecules, the chosen concentration has been shown to induce significant immunomodulatory effects (43, 47, 63, 64). Moreover, crude LPS preparations often contain contaminants such as peptidoglycans, nucleic acids and lipoproteins which are potent inducers of pro-inflammatory cytokine gene expression despite having no apparent impact on cell viability (65).

β-glucans are known to be very potent immune stimulants in aquaculture, with beneficial effects on growth, immune modulation and disease resistance (27, 66, 67). In the present trial, the highest concentration of laminarin (90 µg/mL) from L. hyperborea upregulated all measured pro-inflammatory cytokines, including il6, il8, il1b and tnfa compared to the control. This was not seen with the lower concentration of laminarin or any of the concentrations of β-glucan from S. cerevisiae, indicating a higher immunomodulating potency in laminarin. In mammals, the more structurally complex β-glucans, such as β-glucan from S. cerevisiae generally exhibit stronger immunomodulating effects compared to the lesser structurally complex β-glucans, such as laminarin (68, 69). This is because their structure usually interact more effectively with pattern recognition receptors (PRRs) such as Dectin-1-like receptors and Toll-like receptors (TLRs) expressed on immune cells (68–70). While the exact mechanism in fish is less well-defined, studies have suggested that fish possess analogous PPRs, such as C-type lectins, which may recognize β-glucans and modulate immune responses in gut epithelial cells (71). This interaction can enhance the expression of immune-related genes and promote the secretion of cytokines. In the current experiment, the higher potency of response by laminarin on the pro-inflammatory il6, il8, il1b and tnfa, than conventional β-glucan was unexpected and needs further examination. The result nevertheless does support notions that that the immune modulation of β-glucans is highly variable depending on a lot of factors like structure, solubility and exposure time (26, 27, 69, 70, 72, 73).

Bioactive peptides have also shown to have different antioxidant and immunomodulatory properties depending on the source and structure (32, 41, 74, 75). In the present work, salmon hydrolysate increased the expression of all pro-inflammatory cytokines in a dose dependent manner compared to the control, while carnosine was found to increase the expression of il1b. Although the concentration of carnosine is much higher than that of salmon hydrolysate (100 mM carnosine = 22 626 µg/mL), RTgutGC cells seem to be much more responsive to salmon hydrolysate. This is likely because of the natural occurrence of carnosine in the fish gut (61), and because of the complex composition of salmon hydrolysate. Unlike carnosine, which only consists of the dipeptide β-alanine and histidine, the diverse range of bioactive peptides in addition to a broader amino acid composition in salmon hydrolysate may have synergistic effects that provides enhanced immune signaling (32, 76). Moreover, it is also likely that hydrolysates exhibit higher bioavailability due to their smaller peptide fractions, facilitating more efficient cellular uptake and utilization compared to carnosine (32, 77). Fish protein hydrolysate has previously been shown to enhance the expression of pro-inflammatory cytokines in mice while still maintaining the intestinal homeostasis (78). Similar results were found in Mallet et al., 2014, where the increased expression of pro-inflammatory cytokines after feeding mice with shark-derived protein hydrolysate resulted in reduced inflammation in response to E. coli infection (79). While the effect of fish protein hydrolysate on the gut barrier- and immune function of fish is relatively scarce, Sandbakken et al., 2024, found that salmon hydrolysate downregulated the expression of immune related genes in Atlantic salmon (Salmo salar) intestinal segments compared to in fish fed fish meal. Nonetheless, controlled stimulation of immune cells or in this case, fish gut epithelial cells, to produce pro-inflammatory cytokines may support a more effective immune response during exposure to pathogens (80).

In the present trial, both β-glucan from S. cerevisiae, laminarin from L. hyperborea and salmon hydrolysate reduced the TEER in a dose dependent manner compared to the control, indicating a decrease in epithelial barrier integrity (81). This reduction of TEER in the intestine, could facilitate increased antigen influx and further trigger mucosal immune responses possibly leading to excessive inflammation (82). For the highest concentrations of β-glucan and laminarin, the reduction in TEER was accompanied by a decreased expression in adherence junction-related E-cadherin (cdh1) indicating weakened adherent junctions, and zo-1, which is essential for tight junction structure. Furthermore, they reduced the expression of intestinal alkaline phosphatase (ialp), which is an important brush border enzyme that destroys bacterial LPS (83). Addition of both salmon hydrolysate and carnosine also reduced the expression of the important barrier genes cldn3 and zo-1, however, salmon hydrolysate did not affect expression of cdh1 while carnosine, in fact, increased the expression. Interestingly, the lowest concentration of carnosine (100 mM) also tended to increase the expression of cldn12 involved in ion permeability regulation (84). As such, this concentration of carnosine may be promoting tight junction remodeling, potentially altering epithelial barrier permeability.

Tight junction selectivity and permeability toward ions are regulated by the specific types of claudins present in the junctional pores (85). In contrast, the apparent permeability (P_app) to larger molecules is primarily influenced by pore size, which can be modulated by the activity of myosin light chain kinase (MLCK) (86). Importantly, permeability via the pore and leak pathways can be regulated independently (86–89). This is seen by the effects observed with the low-dose laminarin, where both TEER and P_app of Lucifer yellow decreased. A reduction in TEER, indicating increased ion permeability, may reflect a shift in claudin isoform expression within tight junctions (85). On the other hand, a decrease in P_app, reflecting reduced permeability to larger molecules, could result from increased expression of occludin and/or decreased MLCK activity (86, 87). Such dual regulation, where both TEER and P_app are reduced has also been observed in the proximal intestine of Atlantic salmon (S. salar) (11). Further evidence of selective regulation of the leak pathway is provided by the low-dose carnosine treatment, which decreased P_app without affecting TEER (44, 82).The lowest concentrations of laminarin and carnosine significantly reduced P_app, suggesting formation of smaller pores that limit the passage of larger, uncharged molecules. These findings indicate that compensatory mechanisms are activated that enhance the paracellular barrier function by tightening junctions or selectively restricting permeability to certain solutes. Tight junction regulation by feed additives in fish are limited and warrants further investigation, however it has been seen that supplementation of laminarin in appropriate concentrations improves the gut health and digestive function of juvenile spotted seabass (Lateolabrax maculatus) (31).

The 6-hour LPS exposure caused a combination of inflammatory response, barrier dysfunction and compensatory mechanisms. More specifically, LPS triggered an inflammatory response leading to upregulation of genes coding for cytokines that are key mediators of inflammation (il6, il8, tnfa and il1b), a result that is in accordance with other studies (43, 48, 90). While these cytokines contribute to the cells’ immune defense against bacteria, their overproduction can cause local inflammation and tissue stress leading to cellular damage (91). This in turn weaken the epithelial barrier as reflected in the decreased TEER and disruptions of tight junction proteins like zonula occludens-1 (ZO-1) on a molecular level in the present study. Cytokines can induce the downregulation or removal of tight junction proteins, or activate myosin light kinase, which contracts the actin cytoskeleton. This strain on the tight junction can create cracks in the barrier, further compromising epithelial integrity (92). LPS also caused a decrease in the gene coding for proliferating cell nuclear antigen (pcna), an important marker of cell proliferation (93). This could indicate that the cells shift cellular resources away from proliferation toward managing damage. Exposure to LPS led to an increase in E-cadherin (cdh1) expression, a key protein involved in maintaining epithelial cell adhesion. This aligns with the findings of Wang et al. (2019), who observed a similar response in RTgutGC cells exposed to LPS. In contrast, studies in mammals generally report a downregulation of E-cadherin following LPS exposure (94, 95). This species-specific difference may be attributed to the unique challenges faced by fish, whose intestines are continuously exposed to waterborne pathogens, including LPS from abundant aquatic Gram-negative bacteria. While mammals rely more on their immune system to clear infections, fish may prioritize maintaining epithelial integrity as a frontline defense (4). Upregulating E-cadherin could serve as a protective strategy to reinforce the intestinal barrier and counteract the disruptive effects of LPS on cell adhesion.

Interestingly, pretreatment with the high-dose of β-glucan and both the concentrations of laminarin reduced the expression of il6 compared to the LPS-control. Pretreatment with salmon hydrolysate also showed the same trend. This could support the idea that these functional feed ingredients had priming effects by which they enhanced the immune system’s readiness to respond to pathogens while also preserving excessive inflammation as supported by other studies (40, 69). For instance, lentinan, a β-glucan extracted from the mushroom Lenitula edodens reduced the expression of genes associated with LPS-induced inflammation in the head kidney of rainbow trout (O. mykiss) (80). The anti-inflammatory properties of β-glucan from various sources was also revealed in an LPS-induced RAW 264.7 cell model (96). In another study the LPS-induced RAW 264.7 cell model was used after pretreatment with salmon hydrolysate, which resulted in reduced expression of pro-inflammatory cytokines (97). The potential beneficial effect of β-glucan and laminarin on LPS-induced inflammation is further supported by downregulation in ialp. This suggests that when these ingredients are effective, the cells may not need to upregulate ialp as much for the detoxification of LPS, implying a reduction in the inflammatory burden (98). The lowest concentration of carnosine, on the other hand, increased the expression of the pro-inflammatory cytokines il8 and il1b and showed a tendency to increase the expression of tnfa. This is rather surprising as other studies have found that carnosine protects against the oxidative damage induced by LPS and H₂O₂ (36, 99–101).

The lowest concentration of laminarin significantly upregulated the expression of the myeloid differentiation primary response protein (myd88) in response to LPS stimulation, compared to the LPS-control group. MyD88, an adaptor protein, is critical for initiating downstream signaling cascades such as NF-κB activation. This pathway facilitates a robust cellular response to LPS by promoting the release of pro- and anti-inflammatory cytokines, thereby modulating the immune response to bacterial challenges (102). The same treatment also increased the cldn3 expression suggesting a potential protective or regulatory role in maintaining epithelial barrier integrity during inflammation (82). This could be related to laminarins’ antioxidant capacity in the fish gut possibly reducing the damage by reactive oxygen species (ROS) produced by LPS exposure, as reported by Qiu et al. (2023). The highest concentration of carnosine and both concentrations of salmon hydrolysate, on the other hand, downregulated myd88 which could indicate an immune-modulatory effect, possibly to ensure a more controlled immune response after the inflammatory activation caused by LPS (103). Both concentrations of carnosine also upregulated cldn3, and the highest concentration of salmon hydrolysate counteracted the effect of LPS on cell proliferation by upregulating pcna. This could be beneficial for tissue recovery, especially where the barrier function may be compromised after LPS exposure (104).