Impact of Down-Stream Processing on Functional Properties of Yeasts in Diets of Atlantic Salmon (Salmo Salar): Implications for Gut Health


 Yeasts are becoming popular as novel ingredients in fish feeds because of their potential to support better growth and concomitantly ensure good fish health. Here, three species of yeasts (Cyberlindnera jadinii, Blastobotrys adeninivorans and Wickerhamomyces anomalus), grown on wood sugars and hydrolysates of chicken were subjected to two down-stream processes, either direct heat-inactivation or autolysis, and the feed potential of the resulting yeast preparations was assessed through a feeding trial with Atlantic salmon fry. Histological examination of distal intestine based on widening of lamina propria, showed that autolyzed W. anomalus was effective in alleviating mild intestinal enteritis, while only limited effects were observed for other yeasts. Our results showed that the functionality of yeast in counteracting intestinal enteritis in Atlantic salmon was dependent on both the type of yeast and the down-stream processing method, and demonstrated that C. jadinii and W. anomalus have promising effects on gut health of Atlantic salmon.


Introduction
Future growth of salmon farming is highly dependent on sustainable feed ingredients that meet the nutritional needs and improve overall health status of sh, at low environmental cost. The growth of the salmon sector imposes demands on feed resources like wild sh stocks, which are under pressure 1,2 . This has led to a change in the salmon feed composition, from being mainly based on marine ingredients towards the use of more plant-based ingredients 3 . Studies have shown that high dietary inclusion of plant ingredients such as soybean meal (SBM) 4-10 , pea protein concentrate 11 , faba bean 12,13 and corn gluten meal 14 , is associated with a condition widely known as SBM induced enteritis (SBMIE) in sh, including Atlantic salmon, rainbow trout and sea bass.
Microbial ingredients such as yeasts 15,16 , bacterial meal [17][18][19] and microalgae 15 have been shown to counteract SBMIE in Atlantic salmon. However, question remains on whether this effect is primarily due to the intrinsic properties of microbial biomass itself, the type of processing or the combination of both.
Øverland and Skrede 20 suggested that down-stream processing of yeast after harvesting is imperative to preserve valuable nutrients and bioactive components, and to improve nutrient digestibility in sh.
Previously, chemical, enzymatic, physical, and mechanical treatments have been used to enhance the nutritional and functional values of yeasts for various applications [21][22][23][24] . Different down-stream processing strategies have shown varying impacts on the integrity and nutritional values of yeast 21,23 .
While down-stream processing may increase accessibility to contents of the yeasts cells, methods such as cell crushing using a micro uidizer may be excessively harsh, leading to alteration in bioavailability of the bioactive components 25,26 . Having this in mind and considering cost-effectiveness in terms of energy savings, scalability and commercialization, autolysis was selected as the down-stream processing method in the present study.
Autolysis is a slow process during which cell membrane permeability increases and endogenous lytic enzymes such as proteases, β-glucanases and chitinases are activated within the yeast cells 27,28 , leading to lysis of the intracellular components of the cell. Autolysis can be induced at low pH or high temperature 21,23,29−32 . Hernawan and Fleet 29 reported that the ultrastructure and content of yeast cell wall polysaccharides could be modi ed through autolysis. In addition, using atomic force microscopy (AFM), Schiavone, et al. 32 have shown that autolysis can be used to enhance the adhesive property of mannoprotein present on yeast cell wall. However, to our knowledge, no study has reported the relationship between changes induced by autolysis and the effectiveness of yeast in modulating gut health in Atlantic salmon fry.
The current study was designed to investigate whether the ability of yeast to counteract enteritis is linked to either the type of yeast, with its associated cell wall properties, or to the down-stream processing used during yeast production, or a combination of both factors. To address this question, we used three different non-Saccharomyces yeasts, namely Cyberlindnera jadinii (anamorph name Candida utilis), Blastobotrys adeninivorans (synonym Arxula adeninivorans) and Wickerhamomyces anomalus produced at laboratory scale. The yeasts were selected based on their ability to utilize hydrolyzates of wood and meat co-products, their high growth rate and high protein content as well as their low production of side products such as alcohol 22,33 . Yeasts were subjected to direct inactivation or autolysis and their functionality as feed ingredient was tested using 5% inclusion levels in diets for Atlantic salmon fry. The impact of the downstream processing on the yeast cells and the impact of the yeast cells on salmon performance were assessed using a variety of methods.

Results
Production of yeast.
The three types of yeast were produced by fermentation at 20 or 200 L scale using a growth medium based on wood-derived sugars and a hydrolysate of by-products from chicken 22 . The yeast cells were harvested and washed before either spray drying directly or autolysis followed by spray drying. Table 1 shows compositional data for the various yeast preparations. The contents (% of dry mass) of the cell wall components β-glucan, mannan, and chitin ranged between 9.6% and 20.4%, 8.7% and 17.8%, and 1.1 and 2.7%, respectively. The total glucan content of C. jadinii was 40-80% higher compared to B. adeninivorans and W. anomalus. On the other hand, the mannan and chitin contents of W. anomalus were 30-40% and 17-60% higher compared to C. jadinii and B. adeninivorans, respectively. Autolysis reduced the glucan content by 20%, 13% and 18% for C. jadinii, W. anomalus and B. adeninivorans, respectively. There was no reduction in mannan content after autolysis for C. jadinii, whereas the mannan contents of the other yeasts were reduced by 5 to 15%. The chitin content was lower in autolyzed yeasts compared to inactivated yeasts, except for B. adeninivorans. The contents of crude protein, crude lipids and ash were mostly unaffected by autolysis (Table 1). W. anomalus had the highest crude protein (52-53%) and crude lipids content (8-9%) compared to C. jadinii (45-48% for crude protein and 6-6.2% for crude lipids) and B. adeninivorans (37-39% for crude protein and 8.5-8.6% for crude lipids). The ash contents ranged from 3-8%. The sum of detected compounds showed values close to 100% for C. jadinii (ca. 93.2%) and W. anomalus (95.3%), whereas this value was clearly lower for B. adeninivorans (ca. 76.1%), suggesting that B. adeninivorans contained components that were underestimated and/or undetected by our analysis (Table 1).
Fish Growth Performance.
To assess the impact of the yeasts on SBMIE in Atlantic salmon fry, a feeding trial was conducted where sh were raised from an average initial weight of 5 to 25 g during the experimental period ( Supplementary Fig. S3a). During this period, no mortality or abnormal behaviour were observed. There were no statistically signi cant differences in feed intake, biomass gain and SGR (P > 0.05) between the various dietary treatments ( Supplementary Fig. S3b,d). The FM fed sh had lower FCR compared to the other dietary treatments ( Supplementary Fig. S3c).

Morphology and Histopathological Changes in Fish.
Fish fed with the FM diet showed normal distal intestine morphology, whereas sh fed the SBM diet developed mild signs of SBMIE ( Fig. 1b & Supplementary Fig. S2a-b). Widening of the lamina propria, sh fed the AWA or ICU diet showed only mild signs of SBMIE that were not statistically different (P > 0.05) from the FM treatment (Fig. 1b). Thus, based on this parameter, the AWA and ICU diets led to suppression of SBMIE in the SBM control diet. Fish fed either the ACJ (P = 0.072) or IWA (P = 0.067) diets showed mild signs of SBMIE in the distal intestine and showed a tendency to be statistically distinguishable from the SBM group (Fig. 1b). Fish fed either ICJ, IBA or ABA were not statistically different from the SBM control.
When considering changes in supranuclear vacuoles (Fig. 1a) and connective tissue (Fig. 1c) in absorptive enterocytes there were no differences between the diets. Morphometric measurements of villi length showed there was no signi cant difference among the diets (Fig. 1d). Similarly, there was no statistically signi cant variation in the morphological measurement of the enterocyte height of the pyloric caeca among the diets (Fig. 1e). Also, the number and size of mucous cells in the mucosal area of pyloric caeca were signi cantly different between the diets ( Fig. 1f-g).
Changes in Immune Response Parameters.
Measurements of protein expression of ve different markers indicative of immunological response in the distal intestine showed statistically signi cant differences between the FM and SBM control diets for TNFα and Annexin 1 ( Fig. 2a-b). There was a signi cant increase (P < 0.05) in the level of TNFα in sh fed the ICJ and ICU diets compared to the SBM control (Fig. 2a). In addition, sh fed the yeast diets ACJ and IBA showed signi cantly reduced levels of annexin 1, relative to the SBM control (Fig. 2b). No signi cant differences were observed in the level of CD83 for all the experimental diets, except for the IBA diet, which gave a signi cantly lower CD83 level compared to both the FM and SBM controls (Fig. 2d). No signi cant differences were detected in the levels of IFNγ and IgM (Fig. 2c,e).
Relationships among all diets based on the ve immune markers, showed a positive and signi cant correlation (P < 0.05) between the FM control diet and the ACJ diet (Fig. 2f). Also, several of the experimental diets (ICJ/ICU, IBA/IWA, ABA/AWA, ABA/ICU and AWA/ICU) showed signi cant positive correlations (P < 0.05) (Fig. 2f). Furthermore, we examined possible correlations between cell wall components of yeasts consumed by the sh and the measured immune parameters. The daily intake of glucans, mannans and chitin was calculated from average daily feed intake (dry matter) and the presence of the respective cell wall components in each yeast (Table 1). Based on these calculations, sh fed ICJ/ACJ consumed the highest amounts of glucans (3-4 mg per day), whereas sh fed IWA/AWA had the highest intake of mannans (3-3.2 mg per day)(Supplementary Table S3). The results of the correlation analysis showed a negative and signi cant relationship between annexin 1 expression in the distal intestine and glucan intake (Fig. 2g).
Yeast Ultrastructure and Cell Wall Composition.
SEM micrographs ( Fig. 3 and Fig. 4) showed that C. jadinii and W. anomalus have ovoid-like shape, whereas B. adeninivorans has a rod-like shape. The inactivated yeasts (Fig. 3a-c) appeared to have smooth surfaces with no wrinkles, whereas the autolyzed yeasts ( Fig. 4a-c) appeared shrivelled and partly broken, seemingly releasing their intracellular contents. These observations were further con rmed by the TEM micrographs, which showed that the intracellular compartment in inactivated cells ( Fig. 3d-f) is compact, with visible organelles, whereas autolyzed cells showed a destroyed intracellular structure ( Fig. 4d-f). Measurements of cell wall thickness showed that, among the three inactivated yeasts, W. anomalus had the thickest cell wall (ca. 160 nm), followed by B. adeninivorans (ca. 104 nm) and C. jadinii (ca. 96 nm), which both had considerably thinner cell walls (Table 1). Autolysis reduced the cell wall thickness of all the three yeasts, but the extent of this reduction varied. B. adeninivorans was mostly affected by autolysis, showing a reduction of nearly 40% in cell wall thickness. For W. anomalus and C. jadinii the reductions were 28% and 16%, respectively (Table 1). Confocal and AFM imaging showed that yeasts were able to retain their shape and structure after the drying process. The heat-inactivated yeasts regained their shape like normal yeast creams and appeared smooth with thicker and organised intracellular layers ( The elasticity of the yeast cells was determined by measuring the Young's modulus using AFM with a silicon nitrite cantilever. C. jadinii exhibited the lowest Young modulus, 254 ± 12 kPa, while B. adeninivorans showed an intermediate value of 509 ± 7 kPa and W. anomalus showed the highest value (1126 ± 29 kPa) (Fig. 5a-c). The distributions were bimodal for ICJ and ABA, indicating that the cell elasticity is not homogenous among the cell population of these species. However, for all three yeasts, autolysis reduced the Young modulus ( Fig. 6a-c) and this effect was most pronounced for B. adeninivorans. This implies that cell permeability was modi ed during autolysis in the three yeast species.
The experiments with Con A-functionalized tips at the surface vs the force needed to break the interaction are presented in Fig. 5d-l and Fig. 6d-l. The adhesion frequency for inactivated yeast was 65%, 86% and 52%, for C. jadinii, B. adeninivorans and W. anomalus, respectively ( Fig. 5j-l). Autolysis decreased the adhesion frequency for C. jadinii (22%) and B. adeninivorans (29%), but led to a minor increase in adhesion for autolyzed W. anomalus (57%) (Fig. 6j-l). The unbinding force or adhesion force of the interaction between the ConA-tip and the yeast cell surface was estimated to be in the range of 44-76 pN and there were no clear trends regarding the effect of autolysis on this force ( Fig. 5j-l and Fig. 6j-l).
The length of mannoprotein unfolded (nm) differed between the yeast species and declined upon autolysis for C. jadinii and B. adeninivorans, but not for W. anomalus, which had the highest length of mannoprotein unfolded to begin with (Fig. 5g-i; Fig. 6g-i). The length of mannoprotein unfolded for inactivated C. jadinii and B. adeninivorans were around 70% and 20%, lower compared to inactivated W. anomalus, respectively. For autolyzed yeasts, the length of mannoprotein unfolded were 78% and 55% lower in C. jadinii and B. adeninivorans, compared to W. anomalus, respectively. Based on adhesion frequency and length of mannoprotein unfolded, it appeared that B. adeninivorans was most signi cantly affected by autolysis. The AFM experiments with ConA-functionalized tips also provided insight into the rupture distance, as shown in Fig. 5d-f and Fig. 6d-f. The rupture distance ranged from 0-400 nm for C. jadinii and B. adeninivorans, with a numerically wider distribution towards larger lengths for the autolyzed yeasts ( Fig. 5d-e and 6d-e). In contrast, the rupture distance for W. anomalus ranged from 0-300 nm and became only slightly larger after autolysis ( Fig. 5f and 6f).
Confocal micrographs (Fig. 3g-i; Fig. 4g-i), con rmed the expected speci city of ConA for mannans, as shown by the localized green coloration along the most exterior part of the yeast cell walls, where mannoproteins are predominately expected. Potential correlations between chemical composition of the cell wall and AFM-probed cell surface properties are presented in Supplementary Fig. S4. The data showed that there are signi cant positive correlations (P < 0.05) between mannan content, cell wall thickness and Young modulus. Also, length of mannoprotein unfolded and adhesion force showed positive, but insigni cant correlations (P > 0.05) with the mannan contents of the yeasts. In contrast, there was a signi cant negative correlation between glucans and the AFM-derived parameters.

Discussion
The current study shows that all three yeasts could be used at 5% dietary inclusion level without compromising the performance of Atlantic salmon fry. Feed intake and biomass gain of sh were not affected by the dietary treatments; thus, we can assume that the observed differences in health-related parameters are mainly due to dietary treatments and not to differences in sh weight or feed consumption. The present study provides insight into the ability of the three yeast species to counteract SBMIE in Atlantic salmon fry, with particular focus on the effect of yeast autolysis.
Although dietary exposure to SBM is known to induce SBMIE in the seawater phase of the Atlantic salmon 4,9,15 , the effects are less severe during the freshwater phase 16,34,35 . In the current freshwater experiment, the histological ndings were in accordance with previous studies performed in juveniles 16,34,35 , with only mild SBMIE symptoms detected. It has been suggested that the immaturity of intestinal functions may be responsible for the mild in ammatory changes observed in juveniles, compared to post-smolt salmons 34,35 . Despite mild symptoms, our results show that both AWA and ICU were e cient in alleviating SBMIE, as demonstrated by changes associated with widening of lamina propria, similar to those fed the FM control diet. ACJ and IWA had smaller non-signi cant effects on preventing SBMIE, while ICJ, IBA and ABA had no effects. These observations are in agreement with an earlier study of Grammes, et al. 15 , which showed that inactive dry C. jadinii and Kluyveromyces marxianus can be used to mitigate SBMIE in Atlantic salmon reared in seawater. Interestingly, the same yeast (ICU) used in the study of Grammes, et al. 15 , showed similar effect in our study, reinforcing our choice of a positive control. Our data also show that the health bene cial effects of yeast depend on the type of yeast and the processing condition used after harvest.
Although the aetiology of SBMIE has been linked to the saponin content of SBM 4,6 , the exact mechanism of action is still debatable. Hitherto, amino acid and fatty acid metabolism, T-cell mediation, intestinal dysbiosis, and immune responses have been linked to SBMIE in Atlantic salmon 5,15,17,36,37 . Studies have shown a consistent abundancy of enteropathogenic bacteria 15,37 and revealed that NOD-like receptors 15 and Toll-like receptors 7 are activated in sh suffering from SBMIE. Based on this information, we propose two possible pathways through which AWA, ICU, IWA and ACJ could alleviate SBMIE in the present experiment.
The rst proposed mode of action is activation of the immune system by yeast cell wall components. In higher vertebrates, β-glucan exerts its mode of action by binding to dectin-1 receptors expressed on the surface of several innate immune cells such as dendritic cells, neutrophils, eosinophils, macrophages, monocytes and some T-cells 38 . It has been shown that the dectin-1 receptor synergizing with Toll-like receptors can modulate the production of TNFα in mice 39 . Similarly, the involvement of mannan in immune system activation has been reported in literature 40,41 Mannan activates the immune system through C-type lectin receptors such as mannose receptors, dectin-2, dectin-3, galectin-3 and Toll-like receptors present in several immune cells 40,41 . Previous study has shown that loss of mannan in mutant yeasts reduced production of TNFα and IL-6 in human monocytes, indicating the signi cant of mannan as an inducer of cytokine for monocytes 42 . However, in sh, the presence of the dectin-1 receptor, along with the entire superfamily V of C-type lectin receptors, is still debatable 43 . Our results showed that only sh fed ICJ, ACJ, AWA and ICU had increased TNFα levels compared to SBM. The other yeast treatments did not promote increased production of TNFα compared to the SBM control group. TNFα is a proin ammatory cytokine, involved in an early stage of the immune response and has a key role during the in ammatory process by regulating the proliferation, migration and phagocytic activity of leukocytes 44 . Increased levels of TNFα may be related to the health bene cial effect of yeast in functional feeds, as reported in previous study 45 .
Increased TNFα production was counterbalanced by reduced Annexin 1 production in FM/ICJ/ICU fed sh, indicating that the immune responses were normalized in sh fed these diets. Our results showed that Annexin 1 production reduced in sh fed ICJ, ACJ, IBA, IWA, AWA and ICU diets, compared to SBM fed sh. Annexin 1 is an important marker for anti-in ammatory responses and has protective properties in the gut, as indicated previously being up-regulated in distal intestine of sh suffering from SBMIE 46 . Similarly, it has been reported that annexin 1 was up-regulated during response to in ammatory bowel disease in humans 47 , which resemble SBMIE in sh, as reported in previous literature 15,48 The positive correlation in immunological responses between the FM, ACJ, AWA and ICU diets suggested that the ability of ACJ, AWA and ICU to counteract SBMIE was linked to immune responses. Furthermore, the difference between ICJ-ACJ and IWA-AWA, in alleviating SBMIE, may be linked to accessibility of speci c immune receptors (dectin-1, dectin-2 etc.) with yeast cell wall components 40,42 .
The second mechanism of action by which yeasts may counteract intestinal enteritis is through binding of its mannans with mannose-speci c lectin-type receptors of enteropathogenic bacteria, thereby preventing adhesion of these bacteria to the surface glycoproteins of intestinal villi 49 . This is supported by our AFM experiment with Con-A, which allow us to determine the adhesive (binding) capacity of mannan molecules on the surface of the yeast cells. The speci city of ConA for mannans was con rmed by the immuno uorescence analysis in this experiment, which is in contrast to the earlier approaches where D-mannose was used to antagonize the surface of the AFM functionalized tip 50,51 . The binding capacity of yeast could be associated with the amount of mannan present on its cell wall. Our results showed that W. anomalus contained the highest amount of mannan, which may account for the improved protection against SBMIE, compared to the other yeasts.
The adhesion frequency in uences the binding capacity of yeast cells and give an indication of distribution and accessibility of the mannoproteins on the surface of the yeast cell wall. High adhesion frequency suggests that mannoproteins are more accessible for interaction. In this study, adhesion frequency was reduced with autolysis in C. jadinii and B. adeninivorans yeast, but slightly increased in W. anomalus. This difference may explain the improved protection of AWA against SBMIE, compared with IWA. There is an indication that the length, type and exibility of mannoprotein unfolded 52 and the branching and structure of its α-mannoside residues 49,53 may contribute to the adhesive properties of the yeast cell. In the present study, the length of mannoprotein unfolded ranged between 45-150 nm for the three yeasts and was higher than those observed for different strains of S. cerevisiae 52 . The difference in length of mannoprotein unfolded is an indication that the length of the mannan chains that made up the cell wall protein differed among the three yeasts, with IWA/AWA having the longest. Positive correlation between mannan content and length of mannoprotein unfolded indicates that mannan composition is linked to the stretching of mannoprotein on the surface of the yeast cells. Furthermore, the rupture distance gives useful information on the exibility and extension of the mannoproteins 32,52 . The rupture distance ranged from 0-400 nm in C. jadinii and B. adeninivorans, with a slightly wider distribution towards larger lengths for the autolyzed yeast. In contrast, rupture distance in W. anomalus ranged from 0-300 nm, and became slightly longer due to autolysis. The variation in rupture distance suggested that anchorage of mannoprotein differed among the yeasts 52 . This may indicate that the health bene cial effects of AWA were linked to its capability to adhere better with enteropathogenic bacteria, compared to the other yeasts. Data from our AFM and cell wall thickness measurements indicated that W. anomalus had the highest mannoprotein levels, increased adhesion frequency when autolyzed and had the largest cell size compared to the other yeasts. Thus, the accessibility of mannoproteins can be a decisive factor for the protective effect of W. anomalus against SBMIE in Atlantic salmon fry. Moreover, the fact that the autolysis process increased this effect can be due to alteration of the cell wall surface, as shown by the change in the rupture distance. This observation support previous assertion of Firon, et al. 49 , who argued that the relationship between mannan concentration and pathogen adhesion is not always direct, indicating that involvement of other factors such as length, type and exibility of mannoproteins, are key to the binding capacity of yeast. Although the present study indicates positive effects of AWA and ICU on SBMIE, further in vivo experiments, using the SBMIE with Atlantic salmon in seawater is warranted to study the effect of different yeast strains and down-stream processing on gut health. Likewise, validation of these results with similar yeasts produced under industrial scale is recommended in the future.
In conclusion, this study demonstrates that the yeast strains C. jadinii and W. anomalous showed the most promising effect on gut health of Atlantic salmon, as demonstrated by histological changes based on widening of lamina propria, as well as changes in immune response parameters. Furthermore, processing by autolysis improved the health bene cial effect of the W. anomalus. Our data show that C. jadinii and W. anomalus, which has shown high productivity in previous fermentation studies, have potential for reducing SBMIE in Atlantic salmon. The results also showed that the amounts, length and accessibility of cell wall components (β-glucans and mannoproteins) could be decisive factors for the protective effects of yeast against SBMIE in Atlantic salmon fry. The functionality of yeast in counteracting intestinal enteritis in Atlantic salmon fry is dependent on the yeast species and the downstream processing used during yeast production.

Materials And Methods
Yeast Production and Processing.
anomalus yeast was cultivated in 20 L scale according to the protocols described in Lapeña, et al. 33 For washing, the yeasts were separated by centrifugation and re-suspended in the same volume of 7 o C deionized water in a 30 L EINAR bioreactor system (Belach Bioteknik, Sweden), equipped with a helical impeller. The washed yeasts were then again centrifuged to obtain yeast creams with 12.5%, 5.5% and 15% dry matter contents for C. jadinii (CJ), B. adeninivorans (BA) and W. anomalus (WA), respectively. Half of these microbial biomasses were dried and heat-inactivated by spray-drying using a SPX 150 MS (SPX Flow Technology, Denmark AS) spray-dryer with inlet and outlet temperatures of 180 o C and 80 o C, respectively. The spray-dryer was tted with a co-current nozzle and the pump speed was set to auto and stabilized at around 35% of maximum speed of the pump. The other half of the yeast creams underwent autolysis by incubating the creams at 50 o C for 16 h in a 30 L EINAR bioreactor system, with constant stirring at 50 rpm using a helical impeller, followed by spray-drying using the same conditions as for the untreated yeast. Dried yeasts were kept at 4 o C until use.
Formulation and Production of Fish Feeds.
Nine experimental diets were produced in this experiment. The diets were as follows: a shmeal (FM) control; a diet with 40% SBM as a positive control; 6 treatment diets containing 40% SBM and 5% yeast ingredients [inactivated CJ (ICJ), autolyzed CJ (ACJ), inactivated BA (IBA), autolyzed BA (ABA), inactivated WA (IWA) and autolyzed WA (AWA)], respectively. An extra control diet containing 40% SBM and 5% of a reference preparation of C. jadinii (ICU) already described for its ability to counteract enteritis 15 was also used in this trial. The feed formulation is as presented in supplementary Table S1.
The diets were formulated to have a similar ratio of digestible protein to digestible energy, and to meet the nutrient requirements of Atlantic salmon as recommended by NRC 55 . To meet sh amino acid requirements, crystalline lysine and methionine were added to the diets due to the high inclusion of plantbased ingredients. All dry ingredients were mixed in a Moretti Forni mixer (Spiry 25, Mondolfo, Italy). Gelatin was mixed in cold water and heated up to 60 °C in a microwave oven before mixing with dry ingredients and sh oil using the same mixer as above. The mash was cooled down to room temperature prior to cold-pelleting using a P35A pasta extruder (Italgi, Carasco, Italy). The pellets were dried in small experimental dryers at approximately 60 °C drying temperature and stored at 4 °C prior to feeding.
Fish Management and Feeding.
The sh experiment was conducted at the Fish Laboratory of Norwegian University of Life Sciences (NMBU, Ås, Norway), which is an experimental unit approved by the National Animal Research Authority, Norway (permit no. 174). The experimental procedures were performed in accordance with the institutional and national guidelines under the applicable laws and regulations controlling experiments with live animals in Norway (regulated by the "Norwegian Animal Welfare Act" and "The Norwegian Regulation on Animal Experimentation" derived from the "Directive 2010/63/EU" on the protection of animals used for scienti c purposes). The study was carried out in compliance with the ARRIVE guidelines.
In total, 1215 Atlantic salmon fry with an average start weight of 5.71 ± 0.05 g were sorted, batch weighed and randomly distributed into 27 berglass tanks (80 L) equipped with automatic feeders. Each tank was randomly stocked with 45 sh. Each diet was fed to triplicate tanks, 20% in excess based on feed consumption in each tank. Feeding was done twice a day with automatic feeders, and uneaten pellets were collected after each feeding from the outlet water settling on a screen for each tank. Daily feed intake was calculated from the dry weight of the feed given and the dry weight of recovered uneaten pellets, adjusted for feed recovery rate from sh tanks. Feeds were kept under refrigerated conditions (4 °C) throughout the experiment. Fish were exposed to a 24 h light regime and recirculated freshwater with an average temperature of 15.0 °C. The water ow was standardized to about 6 L min − 1 , and the oxygen content of the outlet water was kept within 8.2-10.1 mg L − 1 . The experiment lasted for 37 days, after which the sh were counted and grouped weighed to estimate the growth performance.
Sampling Procedure for Fish Tissue.
For tissue sampling, six sh per tank were randomly selected, anesthetized with metacaine (MS-222™; 50 mg L − 1 water) and killed with a gentle blow to the head. The individual body weight of each sh was recorded and included in the total tank mean. Distal intestine and pyloric caeca tissues were collected from each sh and further processed, as described below. The distal intestine was opened longitudinally, the content was removed and the tissue was carefully divided into two parts: one part was xed in 10% phosphate-buffered formalin for 24 h before storage in 70% ethanol until further processing for histological analysis; the second part was immediately snap-frozen in liquid nitrogen and stored at -80 °C for indirect enzyme-linked immunosorbent assays (ELISA). Pyloric caeca were treated in the same way as the distal intestine samples for histological analysis.

Morphometric and Histological Examination of Fish Tissues.
Formalin-xed distal intestine and pyloric caeca samples were dehydrated in ethanol, equilibrated in xylene and embedded in para n using standard histological techniques. Longitudinal sections of approximately 6 µm in thickness were prepared. The sections were stained with hematoxylin, eosin and Alcian blue 8 GX. Changes in villi length were captured using a DMLS light microscope (Leica Microsystems, Wetzlar, Germany) equipped with a Leica E3 digital imaging camera and LAS EZ v4.9 software. Randomly selected villus of 18 distal intestine tissues from each dietary group (at least 80 measurements per group) was measured from the stratum compactum to the tip of the fold by ImageJ software. For histological evaluation, changes associated with intestinal tissues were blindly evaluated with a focus on the characteristic changes known for SBMIE in Atlantic salmon. 9 The histological scores were obtained through a semi-quantitative scoring system measuring changes in three morphological parameters: loss of supranuclear vacuoles in absorptive enterocytes; widening of lamina propria in mucosal folds; and increase of connective tissue between the base of folds and stratum compactum.
Each parameter was given a score of 1-5, where 1-2 represents normal morphology; 3-4 mild and moderate enteritis; and 5 for severe enteritis ( Supplementary Fig. S2a-b). To measure changes associated with pyloric caeca, the longitudinal enterocytes area was selected (image 20x) and measured from the base to the apex. Measurement of enterocyte height was performed using the Easy Scan Software. The total number and average mucous cell size in the caeca mucosal area were measured using the ImageJ software. The number of mucous cells was counted per 1 mm 2 of the mucosal area ( Supplementary Fig.  S2c-e).
Indirect ELISA of Distal intestine Tissues.
Immunological parameters were analyzed using the distal intestine samples by indirect ELISA 56 . Brie y, distal intestine samples from nine sh per treatment were homogenized using metal beads and lysis buffer (Tris 20 mM, NaCl 100 mM, Triton X-100 0.05%, EDTA 5 mM, and protease inhibitor cocktail 1x, pH = 7.2). Subsequently, the homogenate was centrifuged at 12000 x g for 25 min at 4 °C. The supernatant containing soluble proteins was stored at -20 °C until use. The protein concentration was quanti ed using the BCA protein assay kit (Thermo Fisher Scienti c) following the manufacturer's instructions. Then, each sample was diluted in carbonate buffer (NaHCO 3 60 mM, pH 9.6) and seeded (in duplicate) in a 96-well plate (Maxisorp, Thermo Fisher Scienti c) at 50 ng µL − 1 (100 µL) for overnight incubation at 4 °C. After blocking with 5% Block solution (Biorad), diluted in PBS, for 2 h at 37 °C, the plates were incubated for 90 min at 37 °C with the rst antibody (Supplementary Table S2). Then, the second antibody-HRP (Thermo Fisher Scienti c), at 1:7000 dilution, was added, followed by incubation for 1 h at 37 °C. Finally, the chromogenic substrate 3,3′,5,5′-tetramethylbenzidine (Invitrogen) was added (100 µL) followed by incubation for 30 min at room temperature. The reaction was stopped with 50 µL of 1 N sulfuric acid and absorbance at 450 nm was measured using a Spectramax microplate reader (Molecular Devices).
Chemical Analysis of Yeast and Fish Feeds.

Calculations for Fish Growth Parameters.
The average biomass gain, feed conversion ratio (FCR), and speci c growth rate (SGR) were calculated according to the equations presented in Agboola, et al. 57 . Brie y, the biomass gain was calculated as the difference between the average nal weight and the average initial body weight of sh per tank. The FCR was calculated as the ratio between average feed consumption per day and average biomass gain per day. The SGR was calculated as logarithm differences between average nal and initial weight of sh divided by the experimental duration.
Morphology and Ultrastructure of the Yeast Cells.
The ultrastructure of yeast cells was examined using a scanning electron microscope (SEM) and a transmission electron microscope (TEM). For each yeast, the SEM and TEM samples were taken before and after the autolysis process, i.e. before spray drying. Three yeast samples per treatment were prepared according to the procedure described in Straume, et al. 58 for SEM and TEM imaging. Samples for SEM were coated with Pt-Pl and examined in a Zeiss EVO 50 EP (Zeiss International, Germany) scanning electron microscope at an accelerating voltage of 15 kV in the secondary emission mode. The sections for TEM were examined in a FEI MORGAGNI 268 (FEI, USA) transmission electron microscope, and photographs were recorded with a VELETA camera. The imaging was performed at the Imaging Centre, Faculty of Biosciences, Norwegian University of Life Sciences. The cell wall thickness was obtained by measuring the length of ve random locations on the cell wall surface of twenty TEM micrographs of each yeast using ImageJ.
Cell Surface Properties of Yeast as Determined by AFM.
Atomic force microscopy (AFM) measurements were done following the protocol described in Schiavone, et al. 32 and Schiavone, et al. 50 . Experiments were carried out with a Nanowizard III atomic force microscope (Bruker-JPK Instruments). The spring constants of each MLCT cantilever (Bruker) were determined using the thermal noise method 59 and were found to be in the range of 10-20 pN nm − 1 . Yeast sample preparation was done by re-suspending the dry yeast mass in sodium acetate buffer (18 mM CH 3 COONa, pH 5.2, 1 mM CaCl 2 and 1 mM MnCl 2 ) and immobilized on polydimethylsiloxane (PDMS) stamps, as described in Dague, et al. 60 . 100 µL of yeast suspension was deposited on the PDMS stamps by convective/capillary assembly. Using bare AFM tips, AFM heights (expressed in nm) were recorded in Quantitative Imaging mode 61 with a maximal force of 1 nN, at 20 °C in buffer solution.
Elasticity of cells was determined from 3072 force curves recorded in force volume mode at an applied force to the surface of 0.5 nN and speed of approach and retraction of 2 µm.s − 1 . Elasticity histograms were generated by analyzing the force-distance curves according to the Hertz model described in Schiavone, et al. 52 , with an indentation of 50 nm and considering a conical tip geometry with halfopening angle α of 0.31 rad and a Poisson ratio ν of 0.5.
To probe cell surface polysaccharides, AFM tips were functionalized with Concanavalin A (ConA) from Canavalia ensiformis (Sigma-Aldrich, L7647) via a silicon nitride dendritip as described in Jauvert, et al. 62 . To analyze the stretching of polysaccharides at the surface of the cell, elongation forces were stretched using the worm-like chain model introduced in Bustamante, et al. 63 , which describes the polymer as a curved lament. The contour length from this model represents the length of mannoprotein unfolded. At least three cells were analyzed for each treatment, representing a total of 3072 force curves for each treatment. The force curves were analyzed with the JPK data processing software (JPK BioAFM, Bruker Nano, Germany). All speci c adhesion peaks were considered for the histograms, which were generated using the Origin® 2020 software (OriginLab Northampton, MA, USA). A Gaussian distribution curve tted on the histogram was used to determine the maximal values of Young modulus, length of mannoprotein unfolded and adhesion force for each yeast group.
Quanti cation of Yeast Cell Wall Polysaccharides.
The total polysaccharide content of the yeast cell wall was estimated without prior cell wall isolation according to the protocol described by François 64 . Brie y, the yeast samples were hydrolyzed with sulphuric acid and the released sugar monomers (mannose, N-acetylglucosamine and glucose) were quanti ed by high-performance anion-exchange chromatography with pulsed amperometric detection as described in Dallies, et al. 65 and Hansen, et al. 23 . The content of β-glucan in the yeast samples was determined using a Megazyme® kit (reference K-YBGL) and α-glucan was calculated as the difference between total glucan and β-glucan.
Immuno uorescence Analysis of Yeast for Determining Mannan Speci city for ConA.
Approximately 200 mg of each spray-dried yeast was xed with 10% formalin for 30 min at room temperature in Eppendorf tubes. Thereafter, the sample was centrifuged at 1000 x g for 5 min at 4 °C and re-suspended in PBS. For uorescence detection of mannan with ConA lectin, the sample was blocked for 1 h at room temperature with PBS containing 1% bovine serum albumin. Subsequently, the sample was incubated with 5 mg mL − 1 of ConA-conjugated FITC (Sigma-Aldrich) for 1 h at room temperature in the dark. The samples were then gently layered on slides and allowed to dry for 10 min, before mounting in the Vectashield Medium (Vector Lab). Between all the steps of this procedure, the samples were washed in PBS. The slides were analyzed using a Zeiss LSM800 confocal microscope (Zeiss International, Germany).

Statistical Analysis.
Fish performance, morphometric, histological and immune parameters were analyzed using the SPSS statistical software package version 26 (IBM Institute, Armonk, NY, USA). Fish performance, morphometric and immune response data were tested for treatment effects using one-way ANOVA.
Signi cance difference (P < 0.05) between means for sh performance and morphometric data were detected using the Tukey HSD test, whereas, for immune response parameters, Dunnett's multiple comparison test was used for detecting signi cant differences. Data from morphometry measurements (villi length) was tested for normality by the Shapiro-Wilk test and homogeneity of variance using Levene's test. Data from the histological evaluation were analyzed using a non-parametric Kruskal-Wallis test by ranks followed by Dunn's multiple comparison test. Signi cance was set at P < 0.05. The tank effect was considered for all parameters and found to have no in uence on the statistical analyses.
Correlation coe cients between the diets using ve immune markers were examined using corrplot package in R. Likewise, correlations between dietary intake of yeast cell wall components and immune markers were determine using the same R package. Also, the correlations between cell wall components and AFM data were examined using the corrplot package in R (CRAN: http://cran.rproject.org/package=corrplot).

DATA AVAILABILITY
The datasets generated during and/or analysed during this study are available upon reasonable request from the corresponding authors.