The extensive intergenerational molecular effects of ocean acidication on the olfactory epithelium transcriptome of a marine sh are associated with a better viral resistance

Progressive climate-induced ocean acidication (OA) impacts marine life in ways that are dicult to predict but are likely to become exacerbated over generations. Although marine shes can balance internal acid-base homeostasis eciently, indirect ionic regulation effects that alter neurosensory systems can result in behavioural abnormalities. In marine invertebrates, OA can also affect immune system function, but whether this is the case in marine shes of ecological and commercial importance is not yet understood. Farmed sh are highly susceptible to disease outbreak yet strategies for overcoming such threats in the wake of OA are wanting. Here, we exposed two generations of the European sea bass (Dicentrarchus labrax) to end-of-century predicted CO 2 levels (IPCC RCP8.5), with parents being exposed for four years and their offspring for two years. Our design included a transcriptomic analysis of the olfactory rosette (collected from the F1 offspring) and a viral challenge (exposing F1 offspring to betanodavirus) where we assessed survival rates.


Results
Here, we designed a long-term experimental paradigm in which parental Dicentrarchus labrax (the F0 generation), were reared in the same pH conditions as their offspring (the F1 generation) (control treatment pH8.0, pCO 2 = 590 µatm or OA treatment pH7.6, pCO 2 = 1520 µatm, as predicted by 2100 in the IPCC RCP8.5 scenario). To generate the F1 generation, and as previously described, we collected gametes from 4-year-old F0 sh that had been reared from larval stage to adulthood in control or OA conditions (Crespel et al., 2017;Cominassi et al., 2019Cominassi et al., , 2020Howald et al., 2019;Mazurais et al., 2020). For each pH treatment, we collected and pooled sperm and eggs from 20 males and 6 females (see methodological summary in gure 1). We then performed in vitro fertilization, after which the resulting F1 generation was returned to OA or control conditions (Crespel et al. 2017;Cominassi et al. 2019;Howald et al. 2019;Cominassi et al. 2020;Mazurais et al. 2020) ( gure 1). After 2 years, we split our experimental F1 cohorts into two experimental arms; one in which we performed gene expression pro ling of olfactory rosette tissue from F1 juveniles (7 each from control and OA rearing conditions, gure 1), and a second arm in which we conducted a betanodavirus challenge, as previously described, to investigate how long-term OA conditions impact on viral infection parameters in D. labrax ( gure 1).

Viral challenge
In non-infected sh, and as expected, we did not detect any mortality, clinical symptoms or lesions. However, in infected sh, symptoms of VNN disease i.e. darkening of the body, whirling swimming and hyperactivity, were evident after ve days in the control treatment (dpi 5, pH8.0) but appeared three days later in sh exposed to the OA treatment (dpi 8, pH7.6). Similarly, mortality on-set was staggered such that sh started dying at 7 dpi in the control treatment and at 9 dpi in the OA treatment), but after 25 days, no additional mortality was observed ( gure 2). When we performed a Kruskal-Wallis test together with a Dunn's post hoc test to determine the Mean Day to Death (MDD), we found this to be 10.7 dpi in control and 12.8 dpi in OA sea water conditions. At the peak of mortality, the presence of the virus in the pool of organs was indistinguishable between treatments. At the end of the challenge, more sh survived infection in OA relative to control conditions, as re ected by a signi cantly higher survival rate (Kruskal-Wallis, H = 3.84, p-value = 0.05 in OA treatment (68%) than in the control (38%) ( gure 2).

Global gene expression pro ling
After a 24h fasting period, we dissected out the olfactory rosette from F1 juveniles (7 each, reared for 2 years in either OA and control conditions), extracted RNA and veri ed RNA integrity as described in Materials and Methods. cDNA library construction for the 14 samples and RNA-sequencing was performed by the GenomiX sequencing platform (MGX, Montpellier, France). Across the 14 libraries, Illumina sequencing generated a total of 772 449 468 reads with an average of 55 450 000 reads per library. More than 90% of the sequencing reads for each library (except for one sample that was 72%) could be mapped to D. labrax reference genome. Of the total 26 721 transcripts, 9 112 were differentially expressed (with an adjusted p-value ≤0.01), with 4 515 (49.6%) up-regulated and 4 598 (50.4%) downregulated in the OA-exposed cohort (listed in additional le 2). To validate the RNA-Seq results, we performed qPCR analysis of the samhd1, gvinp1, tlr3, cxcl14 and nfat5 genes through which we obtained robust coe cient of correlation metrics (r samhd1 :0.90, r gvinp1 :0.74, r tlr3 :0.91, r cxcl14 :0.91 and r nfat5 :0.89).

Discussion
Because the olfactory epithelium in sh is in direct contact with water, its cells are particularly exposed to environmental changes. However, in the context of global climate change and its longer-term impact of marine life, it's di cult to predict how sh and other species will adapt to an increasingly acidic marine environment. Indeed, impacts of olfactory systems in the context of OA were reported to impact on olfactory behaviour in broadly related work (Munday 2014;Porteus et al. 2018;Jiahuan et al. 2018;Williams et al. 2019;Porteus et al. 2021). For instance, Dicentrarchus labrax, when exposed to predicted end-of-century OA (around 1 000 µatm) for up to two weeks, experienced declined capacity for detecting food sources and avoiding predator-associated scents (Porteus et al. 2021). However, these relatively short exposures resulted in marginal effects on gene expression of the olfactory bulb (Porteus et al. 2018;Williams et al. 2019).
In this study, we experimentally investigated intergenerational long-term consequences of OA on the olfactory epithelium transcriptome in F1 D. labrax. Speci cally, we sought to understand how predicted end-of-century OA alters the olfactory system in a way that re ects long-term adaptations in F1 offspring, and how such changes might impact on overall robustness, such as survival after infection, a common problem in aquaculture. Cumulatively, we identi ed ~9,000 OA-induced differentially expressed transcripts, representing roughly a third of the total number of transcripts (34%), that revealed a profound modi cation of the D. labrax' olfactory epithelium transcriptomic pro le.
Overall, among differently expressed genes, we identi ed genes broadly related to olfactory function. These included genes involved in ion balance, GABA signalling, olfactory neuronal signal transduction and neuron excitability, and their regulation re ected signi cant adaptations to synaptic plasticity ( gure 4). We also observed transcriptome changes indicative of a metabolic depression in the olfactory epithelium, as exempli ed by down-regulation of genes involved in ATP synthesis and the mitochondrial electron transport system [cytochrome c oxidase (cox) genes] in combination with lower expression of genes involved in energetically costly processes (e.g. macromolecules biosynthesis). At a general level, we also detected higher expression of genes related to the AMPK signalling pathway, consistent with related work in the brain of sh exposed to OA a well as in other marine species (Schunter et al. 2018;Cao et al. 2018). Extending from these ndings, and unexpectedly, we also discovered that D. labrax, upon intergenerational exposure to OA, up-regulated gene programs conferring elevated resistance to virus infection. Consistent with these molecular changes, we found in our betanodavirus infection study that the OA-exposed F1 cohort succumbed to infection later and that fewer sh died than control F1 sh ( gure 4).
Across different organs in OA-exposed and control sh, virus presence and concentrations were similar, indicating that the viral infection in and of itself was similar, and that the difference in disease resistance was due to a superior defence relative to sh not exposed to OA. In related work, Bresolin de Souza and collaborators found that a three-month exposure to OA in Atlantic juvenile halibuts was associated with increased activity of complement component C3, lysozyme and brinogen, known factors in innate and complement systems though antiviral and in ammatory activities (Bresolin De Souza et al. 2014;Bresolin de Souza et al. 2016). Apart from this work, OA effects linked to immune function has been reported for shell sh (e.g. Leite Figueiredo et al., 2016;Castillo et al., 2017;Sun et al., 2017), but our insight into long-term adaptations for marine sh, and their potential rami cations for aquaculture remain largely unknown (Baag and Mandal 2022). As a progressively acidic marine environment will inevitably become a reality in the future, such knowledge is urgently needed. Here, we reveal for the rst time that sh may acquire an improved capacity to resist a viral infection upon years of intergenerational exposure to OA. In our transcriptomic characterization of the olfactory epithelium of sh exposed to intergenerational OA, we detected up-regulation of gene programs linked to innate antiviral immunity; pathogen receptors genes, Interferon-Stimulated Genes (ISG) and ribosome-related genes.
Detection of an invading pathogen is a critical rst step in the initiation of a robust immune response The second PRR class, C-type lectin receptors (CLRs), mediate bacteria-and fungi-associated responses, but following viral recognition, can induce both protective and detrimental effects depending on the pathogen (Hoving et al. 2014). Members of the mannose receptor, the asialoglycoprotein receptor and ctype lectin domain families, were represented among the up-regulated CLR genes (additional le 2).
Interestingly, the melanoma differentiation-associated protein 5 (mda5 or i h1) gene, which encodes for a member of RLRs and several members of the TLRs gene family, both considered as key virus sensors were also up-regulated (Takeuchi and Akira 2009;Ranjan et al. 2009). Stimulation of MDA5 or TLRs by RNA viruses is known to result in type I interferon response activation (Langevin et al. 2013).
Interferons are signalling cytokines secreted by every type of cell in order to interfere with viral progression by regulating the expression of more than 1000 genes at the transcriptional level (Klamp et al. 2003;Paul et al. 2007;Li et al. 2009;Kuenzel et al. 2010). These genes are called Interferon-induced genes (ISG) and exhibit a wide array of antiviral properties (de Veer et al. 2001;Wang et al. 2017). A large proportion of orthologous ISGs have been identi ed in sh and humans and very few sh ISG have no human ortholog (e.g. gig1, gig2, vig-B319 (Levraud et al. 2019b)). The mechanisms of antiviral activity of sh-speci c ISGs remain poorly understood (Schneider et al. 2014;Poynter and DeWitte-Orr 2016). Here, we found that OA-exposed sh expressed several up-regulated ISGs whose functions may partially explain their enhanced resistance to viral infection. These include the IFN-induced double-stranded RNAactivated protein kinases (pkr1, prkrir), a particular ISG subset that functions as PRR, antiviral effector, and inhibitors of virus translation and replication combined (Clemens and Elia 1997;Schneider et al. 2014;Poynter and DeWitte-Orr 2016); genes from the GTPase imap family (gimap8, 4, 7), known to mediate cell-autonomous resistance against pathogens (Haller and Kochs 2002;Klamp et al. 2003;Haller et al. 2007); and genes from the TRIM family (trim16, 39) that play pivotal role in viral restriction, modulation of immune signalling, autophagy and formation of cellular structures (Langevin et al. 2019).
The regulation of ISG transcription can take place via the classical interferon mediated JAK-STAT signalling pathway (Darnell et al. 1994;Wang et al. 2017) and/or through a variety of non-canonical pathways independent of interferon induction (Mostafavi et al. 2016;Wang et al. 2017). Here, we found that genes involved in the JAK-STAT pathway were both up-and down-regulated in the OA-exposed sh cohort. To better understand ISG signalling pathway regulation in the context of viral infection in sh, analyses at the post transcriptional levels (e.g. phosphorylation status) for JAK-STAT and non-canonical pathways would be necessary.
Finally, we found that intergenerational exposure to OA triggered a down-regulation of genes coding for ribosome biogenesis, ribosomal proteins and eukaryotic translation initiation factors, indicative of reduced ribosome abundance and malfunction. This is con rmed by the GO analyses that show strong down-regulation of both translation and macromolecule biosynthetic process. During an infection, the virus hijacks the host ribosomes to produce new viral particles along with host cellular factors to initiate viral translation (Lee et al. 2013;Li 2019). We hypothesised that as OA induced a malfunction of the host's ribosomes, viruses may lack the biosynthetic machinery to translate and transcribe their nucleic acids. This, together with the up-regulation of PRRs and ISGs might explain the observed superior resistance during the viral challenge.
VNN outbreaks are considered one of the most relevant infectious constraint for the culture of a variety of sh species. Hence, the new insights we provide into how a marine teleost of economic interest undergoes intergenerational acclimation could be of precious concern to aquaculture. Although juvenile sh exposed to intergenerational acclimation were more resistant to VNN, based on the changed transcriptome, their metabolic and odour transduction programs were altered, which may in turn alter the olfactory perception of a wide array of chemical cues that may consequently impact odour-mediated behaviours including feeding, homing and other inter-or intra-speci c interactions such as sociability, mating, competition and predator detection ( gure 4). Understanding how the interplay between acidi cation and warming over generations modulate olfactory behaviour and viral resistance could be useful to develop new strategies for maintaining the health and production in aquaculture facilities for either commercial or scienti c purposes.

Conclusions
We report here that intergenerational exposure to OA induces a deep modi cation of the transcriptomic pro le in the olfactory epithelium of the D. labrax that include plastic responses related to ion balance and transport, neuronal activity and plasticity, energy metabolism and innate immunity ( gure 4). This intergenerational plasticity may be considered as an acclimation (adaptive plasticity) to prevent more severe OA-induced physiological disruption at the whole organism level (Bailey et al. 2017;Schunter et al. 2018). It is noteworthy that immune system factors such as cytokines, interferons and interleukins also play a role in central nervous system and brain development, and can induce changes in neural network activity, supporting the intricated interplay between the immune and the nervous system (Mousa and Bakhiet 2013;Clarkson et al. 2017). Additional experimentations based on electrophysiology and behavioural tests would help determine whether the regulation that we observed in the neuronal plasticity and activity gene programs are associated with a perturbation of the olfactory function, as suggested by the regulation of key processes associated with energy metabolism. Likewise, further studies will be necessary to more globally characterize OA-induced effects on immune status in both cultured and wild sh and its capacity to resist the most sh pathogens in a changing ocean.  Howald et al. 2019;Cominassi et al. 2020;Mazurais et al. 2020). Sperm and eggs were collected and pooled from 20 males and 6 females of each pH-treatment (see methodological summary in gure 1).

Methods
Luteinizing hormone releasing hormone (LHRH) was injected to stimulate synchrony in oocytes full maturation. Taking great care of maintaining parental pH conditions, eggs were hatched and the resultant F1 offspring were reared in water at the same pH as their parents. Rearing conditions during larval and juvenile stages were similar to those described in previous studies (Howald et al. 2019;Cominassi et al. 2020;Mazurais et al. 2020). For both treatments, seawater temperature and salinity followed seasonality of the Bay of Brest. 402 juveniles (201 per treatment) were distributed evenly in 6 culture tanks (400 L, three tanks = three replicas per treatment) that were part of an open-circuit system. To guarantee high quality, seawater pumped 500 m off the coastline at a depth of 20 m passed through a sand lter, a tungsten heater, a degassing column packed with plastic rings, a 2-µm lter membrane, and a UV lamp. Seawater for the control treatment was then poured into each of the three replicas tanks. Seawater for the OA treatment was injected with CO 2 (through manipulation of a owmeter (Aalborg, USA) connected to a CO 2 bottle (Air Liquide, France)) in a header tank equipped with a degassing CO 2 column to favour mixing. Then, low pH seawater was poured into each of the 3 replicas tanks. pH in NIST scale and temperature in the six tanks were daily measured with a WTW 3110 pH meter (Xylem Analytics Germany, Weilheim, Germany; with electrode: WTW Sentix 41) calibrated daily with pH4.0 and pH7.0 buffers (WTW, Germany). Total alkalinity was measured once a week following the protocol of Strickland and Parsons (Strickland and Parsons 1972): a 50 ml sample of tank water was mixed with 15 ml HCl (0.01 M) and pH was measured immediately. Total alkalinity was then calculated with the following formula: with, total alkalinity (TA, mol l -1 ), volume (V, l) of HCl or of the sample, concentration (C, mol l -1 ) of HCl, hydrogen activity (H + , 10 -pH ) and hydrogen activity coe cient (УH + , here = 0.758). F1 juveniles were fed ad libitum with a diet that meets their nutritional requirements (Vitalis Cal, Skretting, Stavanger, Norway). No signi cant difference was observed in the mean weights (t test, t=0.02, df=64.52, p-value = 0.98) of the 2-year-old juveniles between the two treatments (n=35). F1 juveniles from the two treatments behaved and fed in a relatively similar way. Mortality events were not detected.

RNA extraction
Prior to sampling, sh were fasted for 24 h. Then, 2-year-old juveniles were rst anesthetized (20 mg L -1 ), and then euthanized with a lethal dose (200 mg L -1 ) of tricaine methane sulfonate 222 (MS222, Pharmaq, Fordingbridge, Hampshire, UK). Sampling consisted in dissecting the olfactory rosette from 7 individuals per treatment that were quickly stored in RNA Stabilization Reagent (RNAlater, Qiagen, Hilden, Germany) following recommendations from the supplier. Total RNA was extracted using Extract-All reagent ( ) { } (Eurobio, Courtaboeuf, Essonne, France) combined with Nucleospin RNA column that includes one step of DNase treatment (Macherey-Nagel, Düren, Germany) according to the manufacturer's instructions.
The concentration and purity of extracted RNA were veri ed (ratio>2) using an ND-1000 NanoDrop® spectrophotometer (Thermo Scienti c Inc., Waltham, MA, USA). The integrity of RNA was checked by electrophoresis using an Agilent Bioanalyzer 2100 (Agilent Technologies Inc., Santa Clara, CA, USA

Transcriptomic and Gene Ontology (GO) analysis
Raw reads mapping and gene quanti cation were done using STAR aligner (v2.7.2c) (Dobin et al. 2013) to the D. labrax reference genome guided by the reference gene annotation (Tine et al. 2014). Differential expression analysis was performed with the DESeq2 package (Bioconductor) (Love et al. 2014) using an adjusted p-value cut-off of 0.01. The SRR accession numbers for the raw sequence data are SRR15222852-65. Cbln11 structural annotation was carried out manually in GenomeView (Abeel et al. 2012) based on BlastHit of Cbln11 sequences available at NCBI and from mapped RNA-Seq reads. Gene structure was exported in gtf format and added to the current annotation. To extract biological meaningfulness and to visualize potentially affected pathways, a Gene ontology (GO) enrichment analyses was performed separately on signi cantly up-or down-regulated genes under OA. Raw reads from RNA-Seq were imported into Galaxy instance of Ifremer (Augusto et al. 2017). An obo GO le and the product annotation le for D. labrax (Tine et al. 2014) were used to analyse signi cantly (p<0.01) up or down-regulated genes. FDR was corrected with the Benjamini-Hochberg test. The corrected p-value to apply to the graph output was set to 0.01.

qPCR validation
To validate some results from the transcriptomic analysis, qPCRs were performed. iScript™ cDNA Synthesis kit (Bio-Rad Laboratories Inc., Hercules, CA, USA) was used to reverse transcribe in duplicate 500 ng of DNase-treated RNA samples into rst strand cDNA by following conditions provided by the manufacturer. Negative RT consisting in RT reaction without retro-transcriptase enzyme were also performed for all samples. The relative quanti cation of mRNA exhibiting differential expression through the RNA-Seq approach [SAM and HD domain containing deoxynucleoside triphosphate triphosphohydrolase 1 (samhd1), Interferon-induced very large GTPase 1 (gvinp1), Toll-Like Receptor 3 (tlr3), Chemokine C-X-C motif ligand 14 (cxcl14), Nuclear factor of activated T-cells 5 (nfat5)] was performed by qPCR using primers designed for the assembled transcripts and blasted using the NCBI BLAST tool to verify speci city (additional le 1). Elongation factor 1-alpha (ef1α) and Ribosomal Protein L13a (rpl13a) were tested as housekeeping genes for normalization but only rpl13a was used since ef1α did not meet the acceptance criteria in terms of gene expression stability measure (M) and coe cient of variation (CV).
The e ciency of the qPCR reaction tested for each primer pair through standard curves were around 100% with R2 >0.999. Transcript expression was quanti ed using the CFX96 Touch Real-Time PCR Detection system (Bio-Rad Laboratories Inc.) and the protocol previously described (Mazurais et al. 2020).

Viral Challenge experiment
Betanodavirus strain W80 isolated from diseased D. labrax displaying typical signs of VNN was used in this study. According to Castri et al. (Castri et al. 2001), a stock of virus was produced at 24°C on the SSN-1 (Striped Snakehead sh; Ophicephalus striatus) cell line (L15 medium, 10% FBS, pH7.6) and frozen. Cell debris was removed by centrifugation for 15 min at 2000 g; the virus was then aliquoted and stored at −80°C. Before the challenge, viral titration was carried out on one of the aliquots after a single freeze-thaw cycle based on the tissue culture infectious dose technique (TCID 50 ) described by (Dussauze et al. 2015). The infectious titer of the viral production was calculated according to the method of Kärber (Kärber 1931), and was found to be 1x10 8 TCID 50 /mL.
After 2-years rearing of the F1, 148 juveniles per treatment were divided into four ow through tanks of 400L ( gure 1) and were challenged by immersion in a bath with W80 at 25°C to mimic the environmental conditions where the clinical signs of disease were reported. For the viral challenge, the water ow was interrupted, the oxygenation was increased and the water volume was reduced to 100 L. Three out of four tanks per treatment were exposed for 3 h with an infectious dose of 5.10 4 TCID 50 /mL −1 of W80. The fourth tank per treatment was exposed in the same conditions to SSN-1 cell supernatant free from the virus and used as a negative control without virus. After a 3-hours bath exposure, the water ow was restored to slowly dilute the virus titration. The system was maintained open and the water temperature (25°C ± 2°C) was continually measured and recorded with a wireless probe (Cobalt, Oceasoft®) coupled to an acquisition system (ThermoClient 4.1.0.24). Juveniles were fed once a day with commercial pellets (Neo Start Coul 2 from Le Gouessant Aquaculture) except on the day of the viral infection.
Mortality was recorded twice a day during 45 days post-infection (dpi). Dead sh were stored at −20°C until viral examination. The presence and concentration of the virus was checked on a pool of organs (spleen, kidney, heart, brain). Virus concentrations were determined in six sh that died at the peak of mortality during the viral challenge (9 dpi in the control treatment; 11 dpi in OA treatment) and in ve survivor sh per treatment by immuno uorescence assay on the SSN-1 and E11 cell lines according to the adapted protocol by Dussauze et al. (Dussauze et al. 2015).
Survival curves were estimated using the Kaplan-Meier method and were compared by a log rank analysis using the online platform BiostaTGV. Firstly, survival curves for each replica within each treatment were compared, and if no signi cant difference was observed, results from the 3 replicas per treatment were pooled. In case of signi cant difference, each replica was analysed separately and compared to the other treatment. In addition, a Kruskal-Wallis test together with a Dunn's post hoc test were performed (XLSTAT software 2020) to determine whether the Mean Day to Death (MDD) and the nal survival rate varied with treatment.

Declarations
Ethics approval and consent to participate: As indicated in the Methods section, our two experiments followed European Commission recommendations on sh ethics and well-being and were authorized by the French Ethics Committee for animal testing (APAFIS #2018032209421223 and APAFIS #202001161613768).
Consent for publication: Not applicable.
Competing interests: The authors declare that they have no competing interests. Methodological summary. It is shown the rearing times and procedures applied on Dicentrarchus labrax parental linage (F0) and their offspring (F1) exposed to either the control (pH8.0) or the acidi ed (pH7.6) treatments; r: replica; nc: negative control.

Figure 2
Survival rate (%) of F1 Dicentrarchus labrax after viral challenge. Curves indicate the survival rate of noninfected juveniles (white circle) and infected juveniles (black circle) with W80 strain either from the control treatment (pH8.0, continuous line) or the acidi ed (pH7.6, broken line) treatment. Each challenge was performed in triplicate with 37 sh per replica tank. Survival was monitored during 45 days postinfection.

Figure 3
Gene ontology enrichments for biological processes. Enriched (q-value≤0.01) biological processes related to the (A) up-regulated and (B) the down-regulated genes in F1 Dicentrarchus labrax exposed to ocean acidi cation.