Rainbow trout (O. mykiss) with a mean body weight of approximately 100 g were obtained from the Cifuentes fish farm (Cifuentes, Guadalajara, Spain). Fish were maintained at the animal facilities of the Animal Health Research Centre (CISA-INIA-CSIC) in an aerated recirculating system at 15°C, with a photoperiod of 12:12 h light/dark. Fish were fed once a day with a commercial diet (Skretting, Norway) and acclimatized to laboratory conditions for at least two weeks prior to sampling. During this period, no mortalities were experienced and no pathological signs were ever observed. Ammonia and nitrite levels were measured daily and maintained below 0.025 and 0.3 mg/l, respectively.

Fish were sacrificed by benzocaine (Sigma) overdose using a water bath with approximately 150 ppm of the anaesthetic following the recommendations of Zahl et al. (43) and the CNS including the hypothalamus and the pituitary carefully dissected. Rainbow trout brains were fixed in Bouin solution for 24 h at room temperature (RT) and then embedded in paraffin. Subsequently, 3.5 μm-thick tissue sections were mounted on Superfrost Plus slides (MenzelGläser). Endogenous peroxidase was quenched with Bloxall blocking solution (Palex) and antigens retrieved by heating in Tris–EDTA buffer (10 mM Tris base, 1 mM EDTA, pH 9) in a microwave oven for 5 min at 800 W and another 5 min at 450 W. Thereafter, non-specific binding was blocked with 5% bovine serum albumin (BSA) in Tris-buffered saline (TBS). Brain slides were then incubated with the different specific monoclonal antibodies (mAbs) against rainbow trout Igs. All these mAbs had been previously characterized and used for immunohistochemistry (44, 45). Anti-IgM was diluted 1:500, anti-trout IgT was diluted 1:50 and anti-trout IgD was used at 15 µg/ml, all in blocking solution (TBS with 5% BSA). The incubation with anti-trout IgM and anti-trout IgD was performed for 1 h at RT, whereas in the case of anti-trout IgT, the incubation was performed overnight (ON) at 4°C. In all cases, after the incubation, slides were washed and a secondary anti-mouse IgG mAb conjugated with horseradish peroxidase added. After repeated washing, the slides were incubated with IMPRESS KIT VR anti-mouse IgG (Palex). The specificity of the reactions was determined by omitting the primary antibodies. Sections were counterstained with Gill´s hematoxylin, dehydrated and mounted in DPX (di-N-butyl-phthalate in xylene). Images were acquired with a Leica DFC320 digital camera connected to a Leica DM LS optic microscope.

Fish were euthanized by benzocaine overdose and blood was extracted (approximately 1 ml) with a heparinized needle from the caudal vein and diluted 40 times with Leibovitz medium (L-15, Life Technologies) supplemented with 100 IU/ml penicillin, 100 μg/ml streptomycin (P/S), 10 units/ml heparin and 5% fetal calf serum (FCS) (all supplements also obtained from Life Technologies). Blood suspensions were layered onto 51% Percoll cushions and centrifuged at 400 x g for 30 min at 4°C, without brake. Leukocytes were collected from the interface and washed in L-15 containing P/S and 5% FCS. Brains were then sampled, after blood extraction, to avoid the contamination of samples with blood leukocytes. The tissues were then transferred into a tube containing L-15 supplemented with P/S, 2% FCS and 10 U/ml heparin, and cell suspensions were then obtained by pressing samples through a 100 μm nylon cell strainer (BD Biosciences). Cell suspensions were placed onto 30/51% discontinuous Percoll (GE Healthcare) density gradients and centrifuged at 400 x g for 30 min at 4°C, without brake. Cells at the interface, corresponding to leukocytes, were collected and washed in L-15 containing P/S and 5% FCS. In all cases, counting and cell viability were determined by trypan blue (Sigma) exclusion, mixing the 0.4% trypan blue solution with the cells at a 1:1 ratio.

Brain and blood leukocytes isolated as described above were seeded on a poly-L-lysine (0.01% solution, Sigma-Aldrich)-coated slide and incubated at RT for 1 h in a humidified chamber. The cell samples were then fixed in 4% paraformaldehyde solution for 30 min at RT and subsequently incubated at RT for 1 h with blocking solution (TBS with 5% BSA and 0.5% saponin) in a humidified chamber. Unless otherwise specified, all the following incubation steps were performed for 1 h at RT in a humidified chamber. After the blocking step, samples were incubated with specific mouse mAbs recognizing the different Igs. Hence, samples were incubated individually with anti-trout IgM coupled to FITC (17 µg/ml), anti-trout IgD coupled to APC (50 μg/ml) or anti-trout IgT diluted 1:100 in blocking buffer. In the case of IgT, cells were subsequently stained with the secondary antibody anti-mouse IgG1 coupled to Alexa Fluor 488 (20 µg/ml). Samples incubated only with secondary anti-mouse IgG1 coupled to Alexa Fluor 488 and FITC or APC-conjugated mouse IgG1 isotypes were also included at the same concentrations (20, 17 and 50 μg/ml respectively) to confirm the specificity of the antibody signals. On the other hand, a double immunofluorescent detection of IgD and IgM was also performed. For this, the antibodies anti-trout IgM coupled to FITC (17 μg/ml) and anti-trout IgD coupled to APC (50 μg/ml) were used simultaneously on the cell samples. All the samples were counterstained with 1 μg/ml of DAPI (Sigma-Aldrich) for 10 min at RT, then rinsed with PBS 1x and mounted with Fluoromount (Sigma-Aldrich) for microscopy. Laser scanning confocal microscopy images were acquired with an inverted Zeiss Axiovert LSM 880 microscope with Zeiss Zen software. Tissue images were processed with Zeiss Zen and Adobe Photoshop CS6 software packages.

Brain and blood leukocytes isolated were also analyzed by flow cytometry. For this, cells were stained with anti-trout IgM [1.14 mAb mouse IgG1 coupled to R-phycoerythrin (R-PE); 0.25 μg/ml] (46) and anti-trout IgD [mAb mouse IgG1 coupled to allophycocyanin (APC); 5 μg/ml] (47) diluted in staining buffer (phenol red-free L-15 medium supplemented with 2% FCS) for 1 h in darkness at 4°C. Both antibodies were fluorescently labelled using R-PE or APC Lightning-Link labelling kits (Innova Biosciences) following the manufacturer’s instructions. Cells were also incubated in parallel with R-PE or APC-conjugated mouse IgG1 isotypes (clone MOPC-21, Biolegend) to confirm the specificity of the mAbs used. After staining, cells were washed twice in staining buffer and resuspended in staining buffer for their analysis in BD FACS Celesta with BD FACSDiva software (BD Biosciences). Doublets and dead cells were excluded from the flow cytometry analysis following the gating strategy described in Supplementary Figure S4 . The data obtained was analyzed using the FlowJo^® v.10 software (FlowJo LLC, Tree Star). In all cases, cell viability was checked using 4-,6-diamine-20-phenylindole dihydrochlorid (DAPI) at 0.2 μg/ml.

Brain and blood leukocytes (5x10⁴ cells) were seeded in 96-well plates and cultured for 48 h at 20°C in L-15 supplemented with P/S and 5% FCS. After this time, supernatants were collected to evaluate the concentration of IgM and IgD on the culture supernatants by ELISA. For this, 96 well-ELISA plates were coated overnight (ON) at 4°C with 100 μl of a mixture of anti-light chain (IgL) mAbs (2A1, 2H9, 2D12, 1B4, 3E4 and 1A6) (5 μg/ml) (48, 49) for detection of secreted IgD, or with 100 μl of anti-trout IgM (clone 1.14) (2 µg/ml) for detection of secreted IgM. In both cases, antibodies were diluted in 0.05 M carbonate buffer pH 9.7. Thereafter, three washing steps were carried out with PBS containing 0.05% Tween-20 (PBS-T) (Sigma-Aldrich) and non-specific binding sites were blocked with PBS containing 5% BSA for 1 h at RT. After three additional washing steps with PBS-T, 100 μl of the culture supernatants, diluted 1:2 in PBS supplemented with 1% BSA, were added to the wells and incubated for 1 h at RT. At this point, plates were washed with PBS-T and 100 µl of biotinylated anti-trout IgD or anti-trout IgM mAb added to the wells at a final concentration of 1 µg/ml in PBS with 1% BSA. Plates were then incubated for 1 h at RT. After three washing steps with PBS-T, the wells were incubated with 100 μl of Streptavidin-HRP (Thermo Fisher Scientific) at 100 ng/ml in PBS with 1% BSA for 1 h at RT. After another three washes, 100 μl of OPD (O-phenylenediamine Dihydrochloride) (Sigma) were added (1 mg/ml). The reaction was stopped by adding 50 μl of 2.5 M H₂SO₄ and absorbance at 490 nm measured in a FLUO Star Omega Microplate Reader (BMG Labtech). Positive and negative controls were included in all cases.

Brain and blood leukocytes (5x10⁴ cells) were cultured for 48 h as described above. After this time, supernatants were collected to evaluate the concentration of secreted IgT by Western blot. For that, supernatants were mixed 1:2 with 2x loading buffer [10% glycerol, 10% 2-mercaptoethanol, 100 mM dithiothreitol (DTT), 4% SDS, 0.05% bromophenol blue and 0.125 M Tris] and heated at 90°C for 5 min. Proteins were then resolved by SDS-PAGE on a 15% acrylamide gel containing 6 M of urea. Secreted IgT was detected after electrotransfer to a 0.2 µm nitrocellulose membrane and incubation with the specific anti- IgT mAb (1:1000) ON at 4°C with continuous agitation. Thereafter, a secondary goat anti-mouse IgG-HRP (GE Healthcare) was added (1:5000) and incubated for 1 h at RT with continuous agitation. Visualization was carried out with a ChemiDoc XRS Gel Imagin System (Bio-Rad) and analyzed with the Image Lab Software (Bio-Rad).

For the repertoire analysis total RNA was extracted from brain samples using TRI Reagent solution following the manufacturer´s instructions. An IgM, IgD and IgT repertoire analysis was performed in 6 brain samples. In all cases, RNA integrity was determined with a 2100 bioanalyzer using the Eukaryote Total RNA Nano assay (Agilent). Library construction was carried out using a custom protocol as follows. The cDNA synthesis from each sample was performed using 600 ng of total RNA in a reaction including an oligodT primer ( Supplementary Table S1 ) and a template switch oligo (TSO) which included a unique molecular identifier (UMI) ( Supplementary Table S1 ) that hybridizes to untemplated C nucleotides added by the reverse transcriptase during reverse transcription. The reaction was performed with SMARTScribe Reverse Transcriptase (Takara Bio) at 42°C for 90 minutes followed by incubation with 1 µl of uracil DNA glycosylase (5 U/µl, New England Biolabs) at 37°C for 40 min. The resultant cDNA was purified using NucleoSpin Gel and PCR Clean-up (Macherey-Nagel) at an elution volume of 20 µl. Target enrichment of 5’-end from IgM, IgD and IgT encoding transcripts were performed in two rounds of PCR followed by a third PCR for dual-indexation of PCR products, using in all reactions the Q5 High-Fidelity DNA Polymerase (New England Biolabs). The first PCR was performed with 1 µl of purified cDNA as template, the primer “Target_Enrichment_FW1” ( Supplementary Table S1 ) partially complementary to a region of TSO sequence, and as reverse the primers R1 of each IgH “IgX_R1” ( Supplementary Table S1 ), located in the constant region of the corresponding IgH, at a final concentration of 0.2 µM each. The PCR program was performed for 17 cycles (95°C for 10 s, 60°C for 20 s, and 72°C for 40 s) in a final volume reaction of 25 µl. PCR products were purified using NucleoSpin Gel and PCR Clean-up (Macherey-Nagel) at an elution volume of 25 µl. A second PCR was performed with 1 µl of purified PCR product as template, the primer “Target_Enrichment_FW2” ( Supplementary Table S1 ) partially complementary to Target_Enrichment_FW1, and as reverse the primers R2 of each IgH “IgX_R2” located inner in the constant region of the corresponding IgH which include an additional tail corresponding with first part of P7 Illumina adaptor ( Supplementary Table S1 ), both primers at a final concentration 0.2 µM. The PCR program was performed for 13 cycles (95°C for 10 s, 60°C for 20 s, and 72°C for 40 s) in a final volume reaction of 25 µl. PCR product was purified using NucleoSpin Gel and PCR Clean-up (Macherey-Nagel) at an elution volume of 25 µl. Finally, 1 µl of purified products from second PCR followed to a PCR round for indexing by which two different 8 nt sample-specific index, one in both side of amplicons, were incorporated together with the final region of P5 And P7 Illumina adaptors ( Supplementary Table S1 ). PCR were performed with sample-specific primers (final concentration 0.2 µM each) in a final volume reaction of 25 µl. The PCR program was performed for 10 cycles (95°C for 30 s, 54°C for 30 s, and 72°C for 40 s). The PCR results were visualized in a 1.4% agarose gel to corroborate the positive results and the correct size of amplicons. Then, equal volume (10 µl) of PCR product from each sample was mixed together following with two steps of purification; the first one using NucleoSpin Gel and PCR Clean-up (Macherey-Nagel) with NTI buffer diluted 1/3 at an elution volume of 25 µl. This purified product was cleaned and size selected by using SPRIselect beads (Beckman Coulter) at a concentration of 0.6x. At this stage, the ready-to-sequencing library was checked using a tapestation with DNA 1,000 kit (Agilent), quantified with the Qubit ™ dsDNA HS Assay Kit (Invitrogen, Life Technologies) and sequenced with NextSeq™ 1000/2000 P1 Reagents (600 cycles) in a NextSeq1000 instrument (Illumina) with protocol 308:8:8:308 cycles and 30% of Phix.

Raw data consensus paired reads were merged using the PEAR software (50), with a minimum overlap size of 10 nt. These merged reads were demultiplexed using the FASTX-Barcode-Splitter tool from FASTX-Toolkit, allowing up to 5 mismatches and a partial overlap of 3. The first 20 nt from reverse primer used in the PCRs were used as barcodes for the identification of 3’ ends corresponding to the constant gene of each Ig. Then, all reads were grouped by their UMI sequences thanks to the MIGEC software (51). This tool also assembles the resulting grouped reads into unique sequences using only the first read as template and the collision filter on. UMI groups with less than 5 reads were discarded. As both IgM and IgD sequences shared the primer R2, absence of cross-assignment was checked using the UMI barcode. Finally, the assembled sequences were trimmed 14 nt at the beginning and 20 nt at the end to remove lower quality regions. The IgH sequences obtained were compared with the available information from Oncorhynchus mykiss contained in the international ImMunoGeneTics information system database (52) using the IMGT/HighV-QUEST tool v1.9.5 (53).

All data were analyzed and handled with GraphPad Software (GraphPad Prism v8.0.1, La Jolla California, USA). Prior to analysis, data were checked for normality using the Saphiro-Wilk test. An unpaired or paired two-tailed Student´s t-test was used in case of normally distributed data, whereas non-normal data was tested with the non-parametric Wilcoxon matched-pairs signed-rank test or the non-parametric Mann-Whitney test for unpaired data. To check the distribution of the CDR3 spectratyping, a Shapiro-Wilk test was performed. Data were presented as mean + SEM, and significance between means was stablished at p ≤ 0.05.

As in other vertebrates, the teleost brain is subdivided into five regions along the rostro-caudal and dorso-ventral axes: the telencephalon (Tel), the optic tectum (OT), the rest of the Forebrain/Midbrain (rForeMid; which includes the inferior lobe), the cerebellum (Cb) and the rest of the Hindbrain (rHind) (54). Hence, we investigated Ig reactivity throughout Tel, OT and Cb frontal sections in rainbow trout by immunohistochemistry. The Tel was the brain region with the lowest positivity for all three Igs (IgM, IgD, and IgT), showing only a few positive cells located inside small blood vessels ( Figure 1a ).

The OT is the largest region in the brain; consequently, we analyzed three different frontal sections corresponding to the anterior (b), medium (c) and posterior (d) OT to account for potential rostro-caudal heterogeneity. Across all OT areas, IgM reactivity was mainly found inside blood vessels ( Figure 1 , Supplementary Figure S1 ), whereas IgD and IgT exhibited significantly higher overall reactivity ( Figures 1b-d ). IgT⁺ cells were the dominant B cell population broadly distributed throughout the parenchyma, in both anterior ( Figure 1b ) and medium ( Figure 1c ) OT, along with high overall IgT reactivity suggesting some secreted IgT in these regions. Conversely, IgD⁺ cells were observed at significantly lower rates compared to IgT⁺ counterparts ( Figures 1b-d ). We also studied two different frontal sections of the Cb, corresponding to the most anterior (e) and posterior (f) sections. Within these sections, different Cb areas could be identified. In the upper part ( Figure 1e ), a molecular layer (ML) and a granular layer (GrL) were differentiated, whereas the bottom part ( Figure 1f ) was characterized by the presence of the rHind. In all these areas, IgM reactivity was consistently low, with positivity mostly confined to blood vessels ( Figures 1e, f , Supplementary Figure S1 ). In contrast, both IgD⁺ and IgT⁺ cells were clearly identified within the cerebellar and the rHind parenchyma, with a pronounced predominance of IgD reactivity, reversing the OT’s IgT-dominant pattern ( Figures 1e, f ).

Given that B cells were mostly absent from the Tel brain area, we focused on studying in more detail the distribution of these cells in the OT and Cb. As mentioned above, some IgD reactivity was observed throughout the different sections of the OT, albeit at different frequencies ( Figures 2b–d ). Notably, IgD⁺ B cells were scarcely present in the anterior OT ( Figure 2b ), with only some IgD reactivity within the torus longitudinalis (TL) and the meninx, moderately abundant in the middle section ( Figure 2c ) and densely distributed in the most caudal region ( Figure 2d ). Especially in this caudal region, IgD⁺ B cells were clearly identified in the OT parenchyma as well as in the outermost layer of the OT, near the meninx, forming a barrier-like structure ( Figure 2d ). In the most posterior sections, IgD⁺ B cells were also clearly identified within the periventricular gray zone (PGZ) that englobes the tectal ventricle (TeV) ( Figure 2c ).

In the Cb, IgD⁺ B cells were analyzed across two different sections, an anterior and a more posterior section ( Figure 3 , Supplementary Figure S2 , respectively). As mentioned above, within both sections, in the upper part, the ML and GrL were differentiated, whereas the bottom part of the sections corresponded to the rHind. In the most anterior section, IgD reactivity was clear in the meninx, with high amount of IgD⁺ B cells forming an external layer ( Figure 3 , upper panel). Although only a few IgD⁺ B cells were identified in the ML, there was an overall background staining in this area that suggested the presence of secreted IgD in the extracellular space ( Figure 3 , upper panel). In contrast, numerous IgD⁺ B cells were identified in the GrL, often in organized clusters in which IgD⁺ B cells surrounded other IgD^- B cells ( Figure 3 , upper panel). Interestingly, this cluster organization pattern was also observed in the bottom area of the Cb, where IgD⁺ cells appeared organized in a clusters in the crista cerebellaris (CC) ( Figure 3 , bottom panel). Additionally, IgD⁺ cells forming a barrier were also visualized in the rHind, surrounding the rhomboencephalic ventricle (RV) ( Figure 3 , bottom panel).

Less overall IgD positivity was observed in the more posterior section of the Cb analyzed ( Supplementary Figure S2 ). In this cases, most of the IgD⁺ cells identified were found in the meninx forming a clear barrier that was more evident in the lower areas ( Supplementary Figure S2 ). Additionally, some scattered IgD⁺ cells were visualized in the rHind parenchyma ( Supplementary Figure S2 , bottom panel).

In the rainbow trout OT, IgT was the more predominant Ig expressed. Beyond IgT⁺ B cells, we observed extensive tissue-wide reactivity, especially in anterior and middle OT ( Figures 4b, c ). This intense staining pattern suggests IgT secretion into the extracellular spaces. In these most anterior regions of the OT, a high number of IgT⁺ B cells were distributed along the parenchyma ( Figures 4b, c ), but also associated to the meninx ( Figure 4b ). Notably, as described before for the IgD⁺ B cell population in the Cb, the population of IgT⁺ B cells identified in the external part of the meninx formed a dense barrier-like structure, as clearly visualized in ( Figure 4b ). In these anterior OT sections, IgT⁺ B cells were also identified in the most internal layers of the OT ( Figure 4c ), namely in the PGZ surrounding the TeV. In the posterior OT, the abundance of IgT⁺ B cells in the parenchyma decreased, whereas still some small groups of cells were found in the PGZ ( Figure 4d ).

The overall IgT reactivity of the Cb was much lower than that of the OT ( Figure 5 ), suggesting a lower amount of extracellular IgT in this tissue. Nonetheless, still numerous groups of IgT⁺ B cells could be clearly identified in the anterior part of Cb within both the GrL and the ML ( Figure 5 , upper panel). Some IgT⁺ cells were also found in the meninx layer, but in this case, their presence was more discontinuous than that observed in the OT ( Figure 5 , upper panel). In the bottom part of this anterior section of the Cb, high numbers of IgT⁺ B cells were distributed along the rHind parenchyma ( Figure 5 , bottom panel). When a more posterior section of the Cb was analyzed, a similar distribution of the IgT reactivity was observed in the rHind ( Supplementary Figure S3 ).

To further investigate the phenotype of the B cells present in the rainbow trout brain, we performed an isolation of leukocyte populations by means of Percoll gradients. For this, complete brains, including all areas analyzed by immunohistochemistry, were dissected and used in the isolation procedure. The fish sampled were previously thoroughly bled (to reduce to a minimum blood contamination in brain samples). Within these total brain leukocyte populations, we could clearly identify B cells expressing the three different rainbow trout Ig isotypes by confocal microscopy ( Figure 6A ). In these cultures, IgD⁺ B cells were more abundant, followed by IgT⁺ B cells and lastly by IgM⁺ B cells ( Figure 6A ). These total brain leukocyte populations were also used to perform an analysis by flow cytometry, staining the cells with mAbs specific for IgM and IgD and using the gating strategy described in Supplementary Figure S4 . In contrast to what occurs in blood, where most B cells expressing IgM co-express IgD on the cell surface (and are thus IgM⁺IgD⁺ B cells) (28), IgM⁺ B cells in the brain do not co-express IgD and are therefore IgM⁺IgD^- cells, such as those found in trout mucosal surfaces (33) ( Figures 6B, C ). In addition to these cells, brain leukocyte cultures contained a large population of cells exclusively expressing IgD on the cell surface (IgD⁺IgM^- B cells) ( Figures 6B, C ). This B cell subset accounted for up to 28% of all leukocytes in some fish, while it was barely present in systemic compartments, including blood ( Figure 6B ). The presence and prevalence of these different B cell subsets based on IgM and IgD surface expression was also confirmed by performing a double staining and examining the cells under the microscope. Through confocal microscopy, we could again confirm that IgD⁺IgM^- B cells were the most frequent B cell population within the IgM/D lineage in the rainbow trout brain ( Supplementary Figure S5 ). We could not carry out a flow cytometry analysis of IgT⁺ B cells because the anti-IgT mAb we have available in our laboratory does not work in flow cytometry.

Cells that exclusively express IgM or IgD on the cell membrane are usually B cells that have started a differentiation program towards Ig-secreting cells, namely plasmablasts and eventually plasma cells (33). Hence, to investigate the Ig-secreting capacity of the brain B cell subsets, we cultured total brain leukocytes for 48 h and then collected the supernatants to determine the presence of secreted IgM or IgD by ELISA and of IgT through Western blot. We compared the results obtained to those obtained in supernatants derived from blood leukocyte cultures from the same fish. As observed in Figure 6D , the amount of secreted IgD observed in cultures from brain leukocytes was significantly higher than that of blood leukocytes. On the contrary, the amount of secreted IgM detected in brain leukocyte cultures was significantly lower than that detected on supernatants from blood leukocytes ( Figure 6E ). In the case of IgT, again the amount of secreted IgT observed in brain leukocyte cultures was significantly higher than that of blood leukocyte cultures ( Figure 6F ).

To provide us with a better understanding of the B cell receptor (BCR) diversity within the rainbow trout brain in homeostasis, we performed a repertoire analysis of the different Ig heavy chains (IgH) in RNA samples obtained from the CNS. First, IgH sequences for each Ig isotype were grouped into unique junction sequence types (JSTs), described as a V_HDJ_H rearrangement together with a specific CDR3 amino acid sequence (55). In contrast to what is observed by immunohistochemistry or flow cytometry, the number of unique IgM sequences in the trout CNS significantly prevailed over the number of IgD and IgT sequences ( Figure 7A ). For all three Igs, we found sequences for which only one copy was obtained and also sequences that were repeated in variable numbers ( Figure 7B ). It is generally accepted that JSTs detected less than 3-5 times in a given tissue correspond to non-expanded antigen-naive B cells whereas JSTs detected more times reflect expanded antigen-experienced B cell clones, including antibody-secreting cells (39). Although the frequencies of expanded JSTs were similar for all three Igs, the frequency of IgM sequences repeated more than 11 times was slightly higher for IgM than for the other Igs ( Figure 7B ).

To gain additional insight into the molecular configuration of Igs in the CNS, we dissected the heavy chain variable region (V_HDJ_H) family usage. Interestingly, in the CNS, the pattern of V_H gene usage was very similar for the three Igs. The three Ig isotypes preferentially used IGHV1 or its duplicated gene IGHV1D, followed by IGHV11 ( Figure 8A ). IGHV6 was also frequently used by IgM and IgD (~ 7.6% and 11.7%, respectively), but in this case, less used by IgT (~ 4.4%) ( Figure 8A ). On the contrary, IgT used IGHV9 at higher levels (~ 14%) than IgM and IgD (~ 6.8 and 8.5%) ( Figure 8A ).

In contrast, the pattern of D segment usage was more variable between IgT and that of IgM and IgD in the rainbow trout brain. In this case, IgM and IgD preferentially used IGHD1 followed by IGHD2, IGHD3, IGHD3D, IGHD4 and IGHD5, being all D segments close to the IgM/D constant regions ( Figure 8B ). IgT, on the other hand, preferentially used D segments usually associated with IgT such as IGHD1T1, IGHD1T2, IGHD2T1, IGHD2T1D, IGHD2T2, IGHD2T2D, IGHD2T3D ( Figure 8B ). Surprisingly, some of these D segments associated in the genome to IgT were also used to a moderate extent by IgM and IgD at variable percentages (from 0.2 to 6.1%) ( Figure 8B ). The same occurred with IgM/D-associated D genes which were mildly used by IgT (1.1 to 6.2%) ( Figure 8B ). Notably, a high percentage of sequences from all three Igs lacked a D segment (~ 8.8, 7.5 and 7.6 for IgM, IgD and IgT respectively).

A strong selection for defined J_H segments were found for Ig sequences in the brain. Hence, IgM and IgD used IGHJ3 at very high percentages (~ 37.8% and 36.2% for IgM and IgD respectively), followed by IGHJ2 (~ 27.1% and 27.7% for IgM and IgD respectively). Other J_H segments such as IGHJ1 and IGHJ6D were used by these two Igs at lower percentages ( Figure 8C ). In contrast, IgT sequences preferentially used IGHJ1T1 (~ 57.7%) followed by IGHJ2T2 (~ 21.3%) and IGHJ1T2 (~ 15.2%) ( Figure 8C ).

To study the degree to which these B cells in the brain of non-immunized fish had experienced clonal selection, we also undertook a spectratyping analysis of the CDR3 regions of the different BCRs in which the distribution of BCR lengths is analyzed to provide with an indication of how BCR repertoire change (56). Hence, selections and expansions of B cells expressing a specific BCR with a defined CDR3 sequence lead to the modification of the Gaussian-like CDR3 length distribution that is found in physiological conditions (56). For each IgH, we first performed the analysis with all CDR3 sequences, not taking into account the V_H genes used. In this case, IgT CDR3 length distributions seemed to be the ones that differed more from a normal Gaussian-like distribution ( Supplementary Figure S6 ). We then performed the CDR3 spectratyping analysis focusing on CDR3 sequences that used a defined V_H segment, namely IGHV1, IGHV1D, IGHV9, and IGHV11, those mostly used in the brain. In these analyses, it seemed more clear that IgM CDR3 length distributions were much closer to normality than those of IgD and IgT ( Figure 9 ). For IgD, the highest selection seemed to occur when the IGHV9 was used, whereas for IgT, this happened when either the IGHV9 or the IGHV11 fragments were used ( Figure 9 ). Important differences were found among the individual fish, with some fish having selected one or two CDR3 lengths for IgD or IgT ( Figure 9 ).