Gnotobiotic rainbow trout (Oncorhynchus mykiss) model reveals endogenous bacteria that protect against Flavobacterium columnare infection

The health and environmental risks associated with antibiotic use in aquaculture have promoted bacterial probiotics as an alternative approach to control fish infections in vulnerable larval and juvenile stages. However, evidence-based identification of probiotics is often hindered by the complexity of bacteria-host interactions and host variability in microbiologically uncontrolled conditions. While these difficulties can be partially resolved using gnotobiotic models harboring no or reduced microbiota, most host-microbe interaction studies are carried out in animal models with little relevance for fish farming. Here we studied host-microbiota-pathogen interactions in a germ-free and gnotobiotic model of rainbow trout (Oncorhynchus mykiss), one of the most widely cultured salmonids. We demonstrated that germ-free larvae raised in sterile conditions displayed no significant difference in growth after 35 days compared to conventionally-raised larvae, but were extremely sensitive to infection by Flavobacterium columnare, a common freshwater fish pathogen causing major economic losses worldwide. Furthermore, re-conventionalization with 11 culturable species from the conventional trout microbiota conferred resistance to F. columnare infection. Using mono-re-conventionalized germ-free trout, we identified that this protection is determined by a commensal Flavobacterium strain displaying antibacterial activity against F. columnare. Finally, we demonstrated that use of gnotobiotic trout is a suitable approach for the systematic identification of both endogenous and exogenous probiotic bacterial strains that may protect teleostean hosts against F. columnare and other pathogens. This study establishes a novel and ecologically-relevant gnotobiotic model that will improve the sustainability and health of aquaculture.


51
As wild fish stock harvests have reached biologically unsustainable limits, aquaculture has 52 grown to provide over half of all fish consumed worldwide [1]. However, intensive aquaculture 53 facilities are prone to disease outbreaks and the high mortality rate in immunologically 54 immature juveniles, in which vaccination is unpractical, constitutes a primary bottleneck for 55 fish production [2][3][4]. These recurrent complications prompt the prophylactic or therapeutic use 56 of antibiotics and chemical disinfectants to prevent fish diseases [5,6] but may lead to final 57 consumer safety risks, environmental pollution and spread of antibiotic resistance [7]. In this 58 context, the use of bacterial probiotics to improve fish health and protect disease-susceptible 59 juveniles is an economic and ecological sensible alternative to antibiotic treatments [8,9]. 60 Probiotics are live microorganisms conferring health benefits on the host via promotion of 61 growth, immuno-stimulation or direct inhibition of pathogenic microorganisms [10,11]. The 62 native host microbiota plays a protective role against pathogenic microorganisms by a process 63 known as colonization resistance [12,13]. In fish, the endogenous microbial community, 64 whether residing in gastrointestinal tract or in the fish mucus, was early considered as a source 65 of protective bacteria [14][15][16][17][18]. However, selection of probiotic bacteria is often empirical or 66 hampered by the poor reproducibility of in vivo challenges, frequently performed in relatively 67 uncontrolled conditions with high inter-individual microbial compositions [15,19]. 68 To improve evidence-based identification of fish probiotics and their efficacy in disease 69 prevention, the use of germ-free (GF) or fully controlled gnotobiotic hosts is a promising 70 strategy [20,21]. In addition to laboratory fish models such as zebrafish (Danio rerio) [22][23][24], 71 several fish species have been successfully reared under sterile conditions to test probiotic-72 based protection against pathogenic bacteria, including Atlantic cod (Gadus morhua) [25], 73 Atlantic halibut (Hippoglossus hippoglossus) [26], European sea bass (Dicentrarchus labrax) 74 [19] and turbot (Scophthalmus maximus) [27] (for a review, see [28]). Salmonids, especially rainbow trout (Oncorhynchus mykiss) and Atlantic salmon (Salmo salar), 76 are economically important species, whose production in intensive farming is associated with 77 increased susceptibility to diseases caused by viruses, bacteria and parasites [29]. Here we 78 studied the probiotic potential of endogenous members of the rainbow trout microbiota to 79 protect against infection by Flavobacterium columnare, a fresh-water fish pathogen causing 80 major losses in aquaculture of fish such as Channel catfish, Nile tilapia and salmonids [30]. We 81 developed a new protocol to rear GF trout larvae and showed that GF larvae were extremely 82 sensitive to infection by F. columnare. We then identified two bacterial species originating 83 either from the trout microbiota (a commensal Flavobacterium sp.) or the zebrafish microbiota 84 (Chryseobacterium massiliae) that fully restored protection against F. columnare infection. Our 85 in vivo approach opens perspectives for the rational and high throughput identification of 86 as Conv larvae to F. columnare Fc7 infection, whereas those maintained in sterile conditions 143 died within the first 48h after infection (Fig. 4B). These results suggested that microbiota 144 associated with Conv rainbow trout provide protection against F. columnare Fc7 infection. To 145 identify culturable species potentially involved in this protection, we plated bacteria recovered 146 from 3 whole Conv rainbow trout larvae at 35 dph on various agar media. 16S rRNA-based 147 analysis of each isolated morphotype led to the identification of 11 different bacterial strains 148 corresponding to 9 different species that were isolated and stored individually (

152
We then re-conventionalized GF rainbow trout larvae at 22 dph with an equiratio mix of all 11 153 identified bacterial strains (hereafter called Mix11), each at a concentration of 5.10 5 CFU/ml. 154 After exposure to F. columnare strain Fc7, these Re-Conv Mix11 larvae survived as well as Conv 155 fish (Fig. 4C), demonstrating that the Mix11 isolated from the rainbow trout microbiota 156 recapitulates full protection against F. columnare infection observed in Conv larvae. 157 Resistance to F. columnare infection is conferred by one member of the trout 159 microbiota. 160 To determine whether some individual members of the protective Mix11 could play key roles 161 in infection resistance, we mono-re-conventionalized 22 dph GF trout by each of the 11 isolated 162 bacterial strains at 5.10 5 CFU/ml followed by challenge with F. columnare Fc7. We found that 163 only Flavobacterium sp. strain 4466 restored Conv-level protection, whereas the other 10 164 strains displayed no protection, whether added individually (Fig. 5A) Fig. S5B) and of all tested F. columnare strains (Supporting Fig. S5C), suggesting 169 a potential contact dependent inhibition. Consistently, we identified a cluster of 12 genes in the 170 Flavobacterium sp. strain 4466 genome (tssB, tssC, tssD, tssE, tssF, tssG, tssH, tssI, tssK, tssN, 171 tssP and tssQ) characteristic of type 6 secretion system (T6SS), T6SS iii , a contact-dependent 172 antagonistic system only present in phylum Bacteroidetes [33]. To improve the taxonomic 173 identification of the protective Flavobacterium isolated from the trout larvae microbiota, we 174 performed whole genome sequencing followed by Average Nucleotide Identity (ANI) analysis. 175 We determined that despite similarity with Flavobacterium spartansii (94.65 %) and 176 Flavobacterium tructae (94.62 %), these values are lower than the 95 % ANI needed to identify 177 two organisms as the same species [34]. Furthermore, full-length 16S rRNA and recA genes 178 comparisons also showed high similarity with F. spartansii and F. tructae, however, the 179 obtained values were also below the 99 % similarity threshold required to consider that two 180 organisms belong to the same species (Supporting Table S1). Similarly, a maximum likelihood 181 based phylogenetic tree (Supporting Fig.S6) generated from sequences of 15 bacterial strains clustered with sequences of F. spartansii and F. tructae, but did not allow the identification of 184  To determine whether our GF trout model could be used as a controlled gnotobiotic approach 204 to screen for trout probiotics, we pre-exposed 22 dph GF rainbow trout larvae to 205 Chryseobacterium massiliae, a bacterium that does not belong to trout microbiota but was 206 previously shown to protect larval stage and adult zebrafish from infection by F. columnare 207 [35]. After 48 h of bath in a C. massiliae suspension at 10 5 CFU/ml, we infected trout larvae 208 with F. columnare strains Fc7, ALG-00-530, IA-S-4 and Ms-Fc-4 and observed that C. massiliae protected against all tested F. columnare pathogens (Figure 7). These results showed 210 that the GF rainbow trout model enables the evaluation of bacterial species, endogenous to trout 211 or not, with probiotic potential against highly virulent F. columnare strains. 212

DISCUSSION 214
Although the use of probiotics is a promising approach to improve fish growth and reduce 215 disease outbreaks while limiting chemical and antibiotic treatments [17,36,37], rational and 216 evidence-based procedures for the identification of protective bacteria are limited. Here, we 217 established a controlled and robust model to study trout resistance to infection by bacterial 218 pathogens and to identify trout probiotics in microbiologically controlled conditions using GF 219 and gnotobiotic rainbow trout. 220 Our gnotobiotic protocol is based on the survival of rainbow trout eggs to chemical sterilization 221 eliminating the microbial community associated to the egg surface. Similarly to gnotobiotic 222 protocols used for zebrafish [24,38], cod [25] and stickleback (Gasterosteus aculeatus) [39], 223 our approach produced larvae that were GF up to 35 dph at 16°C without continued exposure 224 to antibiotics, therefore avoiding possible effects of prolonged antibiotic exposure on fish 225 development [40]. Similarly to GF stickleback larvae at 14 dph [39], we observed no 226 development or growth differences between GF and Conv trout larvae at 21 dph. In contrast, 227 GF sea bass (D. labrax L.) larvae grew faster and had a more developed gut compared to 228 conventionally raised larvae [41]. These discrepancies could come from the fact that, in our 229 study and in the GF stickleback study, anatomical analyses were performed before first-feeding, 230 whereas the GF sea bass were already fed when examined [41]. Indeed, trout larvae initially 231 acquire nutrients by absorbing their endogenous yolk until the intestinal track is open from the 232 mouth to the vent. We therefore cannot rule out that at later stages of development, when fish 233 begin to rely on external feeding, differences between GF and Conv fish may occur, especially 234 in the structure and size of organs or in body weight. However, the hurdles associated with 235 long-term fish husbandry while keeping effective sterility control, de facto limits our approach 236 to relatively short-term experiments on larvae with limited feeding time and low complexity While GF conditions cannot be compared to those prevailing in the wild or used in fish farming 239 [25], our results showed that GF rainbow trout larvae are highly susceptible to F. columnare, 240 the causative agent of columnaris disease affecting many aquaculture fish species [30,42]. 241 Interestingly, our GF rainbow trout larvae model also revealed the protective activity of C. 242 massiliae, a potential probiotic bacterium isolated from Conv zebrafish [35], against various F. 243 columnare strains from different fish host and geographical origins. These results demonstrate 244 that GF rainbow trout is a robust animal model for the study of F. columnare pathogenicity and 245 support C. massiliae as a potential probiotic to prevent columnaris diseases in teleost fish other 246 than its original host. 247 Furthermore, we demonstrated that the relatively simple culturable bacteria isolated from 248 microbiota harbored by Conv trout larvae effectively protect against F. columnare. 249 Interestingly, different studies have demonstrated that highly diverse gut communities are more 250 likely to protect the host from pathogens [43,44]. This constitutes the base for the paradoxical 251 negative health effect associated with the massive utilization of antibiotics in aquaculture: the 252 reduction in microbial diversity facilitates colonization by opportunistic pathogens [45]. While 253 this advocates for practices leading to enrichment of fish microbial communities to minimize 254 pathogenic invasions in aquaculture [16], our results demonstrate that resistance to a bacterial 255 pathogen can also be achieved by a single bacterial strain in a low complexity microbiota. 256 Moreover, previous studies of resistance to infection provided by controlled bacterial consortia 257 in gnotobiotic hosts often relied on community composition, rather than individual members of 258 the microbiota [46][47][48][49]. We showed that the observed protection in larvae is mainly due to the 259 presence of Flavobacterium sp. strain 4466. We cannot exclude, however, that at later 260 developmental stages, the presence of other bacterial species may be needed for more efficient 261 implantation or stability of protective members in the trout microbiota.
The molecular basis of F. columnare pathogenicity is poorly understood, but was recently 263 shown to rely on the secretion of largely uncharacterized virulence factors and toxins by the 264 In conclusion, we showed that germ-free and gnotobiotic trout larvae are an effective 302 experimental tool to study microbiota-determined sensitivity to major salmonid freshwater 303 pathogens, enabling the validation of endogenous and exogenous potential probiotic strains. 304 This approach will also be instrumental in studying the molecular basis of probiosis against fish 305 pathogens as well as host-pathogen mechanisms, ultimately contributing to the mitigation of 306 rainbow trout diseases in aquaculture. 307 The eyed rainbow trout eggs received at 210 dd were transferred to sterile Petri dishes (140 mm 329 diameter, 150 eggs/dish) and washed twice with a sterile methylene blue solution (0.05 mg/ml). 330

MATERIAL AND METHODS
The eggs, kept in 75 ml of methylene blue solution, were then exposed to a previously described 331 antibiotic cocktail [24] (750 µl penicillin G (10,000 U/ml), streptomycin (10 mg/ml); 300 µl of 332 filtered kanamycin sulfate (100 mg/ml) and 75 µl of the antifungal drug amphotericin B solution 333 (250 µg/ml)) for 24 hours by gentle agitation at 16°C. Eggs were then washed 3 times with fresh sterile water and treated with bleach (0.005 %) for 15 minutes. Following 3 washes with 335 sterile water, eggs were treated for 10 minutes with 10 ppm Romeiod (COFA, France), a 336 iodophor disinfection solution. Finally, eggs were washed 3 times and kept in a class II hood at 337 16°C in 75 ml of sterile water supplemented with the previously mentioned antibiotic cocktail 338 until hatching spontaneously 5 to 7 days following the disinfection process. Once hatched, fish 339 were immediately transferred to 75 cm 3 vented cap culture flasks containing 100 ml of fresh 340 sterile water without antibiotics (12 larvae/flask). The hatching percentage was determined by 341 comparing the number of hatched larvae in Petri dish relative to the total number of eggs. 342 Sterility: Sterility was monitored by culture-based and 16S rRNA PCR-based tests at 24 h, 7-343 and 21-day post-treatment. After feeding started, 50 µl of GF fish flask water was sampled 344 before each water change as well as one larva every week to perform culture-based and 16S 345 rRNA-based PCR sterility tests. 50 µl of rearing water from each flask was plated on LB, YPD 346 and TYES agar plates, all incubated at 16°C under aerobic conditions. Fish larvae were also 347 checked for bacterial contamination every week using the following methods. Randomly

Characterization of culturable conventional rainbow trout microbiota 387
To identify the species constituting the cultivable microbiota of Conv trout larvae, 3 individuals 388 were sacrificed with an overdose of tricaine at 35 dph, homogenized following the protocol 389 described above and serial dilutions of the homogenates were plated on TYES, LB, R2A and 390 TS agars. The plates were incubated a 16°C for 48 to 72 hours. All morphologically distinct 391 colonies (based on form, size, color, texture, elevation and margin) were then isolated and 392 conserved at -80°C in respective broth medium supplemented with 15 % (vol/vol) glycerol. 393 In order to identify individual morphotypes, individual colonies were picked for each 394 morphotype from each agar plates, vortexed in 200 µl DNA-free water and boiled for 20 min 395 at 90 o C. Five µl of this bacterial suspension was used as template for colony PCR to amplify 396 the 16S rRNA gene with the universal primer pair 27f and 1492R. 16S rRNA gene PCR 397 products were verified on 1% agarose gels, purified with the QIAquick ® PCR purification kit 398 and two PCR products for each morphotype were sent for sequencing (Eurofins,Ebersberg,399 Germany

Germ free rainbow trout microbial re-conventionalization 428
Each isolated bacterial species was grown for 24 hours in suitable medium at 150 rpm and 429 20°C. Bacteria were then pelleted, washed twice in sterile water and diluted to a final 430 concentration of 5.10 7 CFU/ml. At 22 dph, GF rainbow trout were mono-re-conventionalized 431 by adding 1 ml of each bacterial suspension per flask (5.10 5 CFU/ml, final concentration). In the case of fish re-conventionalization with bacterial consortia, individual bacterial strains were 433 washed, then mixed in the same aqueous suspension, each at a concentration of 5.10 7 CFU/ml. 434 The mixed bacterial suspension was then added to the flask containing GF rainbow trout as 435 previously described. In all cases, fish re-conventionalization was performed for 48 h and the 436 infection challenge with F. columnare was carried out immediately after water renewal. Semi-thin sections were stained with toluidine blue solution for 1 min at 60°C, washed with 456 distilled water for 5 seconds, ethanol 100 % for 10 seconds and distilled water again for 20 seconds, dried at 60°C and embedded in Epon resin which was allowed to polymerize for 48 458 hours at 60°C. Light microscopy images of semi-thin EPON sections were prepared with Nikon 459 Eclipse 80i microscope connected with Nikon DS-Vi1 camera driven by NIS-ELEMENTS 460 D4.4 (Nikon) software. 461 462

Whole fish clearing and 3D imaging 463
For a 3D imaging of cleared whole fish, fish were fixed with 4 % formaldehyde in phosphate-464 buffered saline (PBS) overnight at 4°C. Fixed samples were rinsed with PBS. To render tissue 465 transparent, fish were first depigmented by pretreatment in SSC 0.5X twice during 1 hour at 466 room temperature followed by an incubation in saline sodium citrate (SSC) 0.5X + KOH 0.5 % 467 + H2O2 3 % during 2 hours at room temperature. Depigmentation was stopped by incubation in 468 PBS twice for 15 minutes. Fish were then post-fixed with 2 % formaldehyde in PBS for 2 hours 469 at room temperature and then rinsed twice with PBS for 30 min. Depigmented fish were cleared 470 with the iDISCO+ protocol [80]. Briefly, samples were progressively dehydrated in ascending 471 methanol series (20, 40, 60 and 80 % in H2O, then twice in 100 % methanol) during 1 hour for 472 each step. The dehydrated samples were bleached by incubation in methanol + 5 % H2O2 at 473 4°C overnight, followed by incubation in methanol 100 % twice for 1 hour. They were then 474 successively incubated in 67 % dichloromethane + 33 % methanol for 3 hours, in 475 dichloromethane 100 % for 1 hour and finally in dibenzylether until fish became completely 476 transparent. Whole sample acquisition was performed on a light-sheet ultramicroscope 477 (LaVision Biotec, Bielefeld, Germany) with a 2X objective using a 0.63X zoom factor. 478 Autofluorescence was acquired by illuminating both sides of the sample with a 488 nm laser. 479 Z-stacks were acquired with a 2 µm z-step. 480 Statistical analyses were performed using unpaired, non-parametric Mann-Whitney test for 483 average survival analysis and the log rank (Mantel-Cox) test for Kaplan-Meier survival curves. 484 Analyses were performed using Prism v8.2 (GraphPad Software). A cut-off of p-value of 5 % 485 was used for all tests. * p<0.05; ** p<0.01; *** p<0.001, **** p<0.0001.