Postnatal liver B cell precursors contribute to the establishment of a mature B cell pool in secondary lymphoid organs in mice

Human newborn and adult samples for immunohistochemistry

Archived formalin-fixed, paraffin-embedded liver tissue from perinatal and fetal autopsies was retrieved from the archives of Alberta Precision Laboratory (Calgary, Alberta, Canada). Pediatric tissue was obtained from pregnancy terminations and neonatal and pediatric deaths. Autopsy consent was obtained in all cases and ethics approval was sought for use of surplus diagnostic tissue. Formalin-fixed, paraffin-embedded adult liver tissue was obtained from surgical resections from Alberta Precision Laboratory. All relevant ethical regulations were followed in this study and the study was approved by The Conjoint Health Research Ethics Board, University of Calgary (ethic ID REB14-2296). Slides were fully scanned and analyzed using Aperio Imagescope Software (Leica Biosystems). Samples were stained using anti-CD19 (cat. number 90176S; clone D4V4B; Cell Signaling Technology; 1/800) at MCIC-Histology Core at the Trev and Joyce Deeley Research Centre, BC Cancer-Victoria. Data from deidentified patients are in Table S1.

Animals

Wild-type (WT) C57BL/6 mice were purchased from Centro de Bioterismo located at the Universidade Federal de Minas Gerais (Brazil) and bred in colonies, or purchased from the Jackson Laboratory. Excluded flora mice were purchased from Taconic Biosciences. CD19cre/TdTomato mice were obtained from crossing CD19^Cre/Cre (B6.129P2(C)-Cd19tm1(cre)Cgn/J) with Rosa26tdTomato/tdTomato lineage (B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J). Kaede mice were maintained in the BWH vivarium for several generations, and Rag1^−/− mice were purchased from the Jackson Laboratory. All animals were housed in microisolator acrylic cages with a filtered air system (Alesco; 5 mice/cage) under controlled conditions of temperature (23 °C) and dark/light cycle (12/12 h). Water and autoclaved mouse food were given ad libitum. All experiments with mice were approved by the Animal Ethics Committee from Universidade Federal de Minas Gerais (register number 289/2020) and by Institutional Animal Care and Use Committee from Brigham and Women’s Hospital (protocol number 2016N000230), following international guidelines for animal care.

RNA in situ hybridization

Deidentified human newborn, and adult liver samples, embedded in paraffin and 6 µm thick, were obtained from the Yale University Pathology Biobank, and de-identified human fetal liver samples were obtained from Liza Konnikova Lab, Yale University Department of Pediatrics after Institutional Review Board approval by Yale University (#2000028847). Fetal samples were obtained with approval of the University of Pittsburgh Institutional Review Board and signed informed consent from elective terminations. All samples were stained using RNAscope Target Probes for CD19 and TSA Vivid Dye 650, following the RNAscope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics) protocol. Briefly, the slides were baked at 60 °C for 1 h, deparaffinized with xylene and ethanol washes, and treated with hydrogen peroxide. Target retrieval was performed at 99 °C for 15 min, followed by the application of protease and the CD19 RNA probe. A series of amplification AMP 1–3 were conducted, and HRP-C1 was linked with TSA Vivid Dye 650 (1:1,200). Finally, the slides were mounted with DAPI-containing Mounting Medium (Abcam). Imaging of CD19 messenger RNA (mRNA) signals was carried out using a Nikon Eclipse Ti2 confocal microscope system at 20× magnification.

Intravital imaging

Confocal microscopy imaging was performed as previously described in Marques et al.¹¹ CD19^Cre/TdTomato mice received intravenously anti-CD31 BV421 (0.15 mg/kg, clone 390; BD; Cat. 563356) and were anesthetized (subcutaneously) with ketamine and xylazine (Syntec; 60 mg/kg and 15 mg/kg, respectively). Liver was surgically exposed and imaged using an inverted Nikon Eclipse Ti A1 confocal microscope. Movies were recorded taking 1 frame every 30 s for a total 1 h long. Images and videos rendering were made using NIS-Elements AR 5.41.01 (Nikon). Cell quantification was made by manual counting of cells in the image field (average 5 to 10 images per mice). Cellular tracking and morphologic analysis were performed semi automatically combining the software algorithm with post visual checking and refining. Image and video tracking analysis was performed using at least 3 animals per group.

Isolation of cells from the liver, spleen, bone marrow, blood, and peritoneal cavity

Following euthanasia by CO₂ asphyxiation, the right atrium of the heart was incised to collect blood, which was immediately treated with 50 μL of heparin (500 IU/mL). Preceding organ extraction, animals underwent perfusion with HBSS (Sigma-Aldrich) to eliminate residual blood within the organs. Perfusion volumes comprised 20 mL of HBSS for adult animals aged 9 wk and 5 to 10 mL for other age groups. The liver was macerated using a syringe plunger directly onto a 100 μm cell strainer positioned over a Petri dish. Similarly, the spleen was macerated using a syringe plunger onto a 70 μm cell strainer over a Petri dish. Bone marrow was harvested from the femur and tibia of both hind limbs. A specialized apparatus was utilized for bone marrow extraction, involving a perforated base in a 0.5 mL Eppendorf tube nested within a 2 mL Eppendorf tube. Following centrifugation at 10,000 g for 30 s, cells residing within the bone marrow cavity descended into the 2 mL Eppendorf for subsequent collection. Post–initial isolation, liver, spleen, bone marrow, and blood were individually transferred to 15 mL tubes and volume-adjusted to 10 mL using FACS buffer (HBSS + 2% fetal calf serum + 12.5 mM HEPES + 2 mM EDTA). Subsequent centrifugation at 500 g for 5 min at 4 °C marked the initiation of distinct processing paths for each organ. For spleen, bone marrow, and blood, the obtained pellet underwent lysis with 2 mL of ACK for 1 min at room temperature to eliminate erythrocytes. Lysis was terminated by the addition of 10 mL of FACS buffer, and subsequent centrifugation at 500 g for 5 min at 4 °C. The final pellet was then subjected to antibody panel labeling and analysis on the cytometer as described subsequently. For liver, pellets were resuspended in 5 mL of Percoll 80%, overlaid with Percoll 40%, and subjected to centrifugation at 800 g for 20 min at 25 °C. Leukocytes, separated through a density gradient, were localized at the Percoll interface, and were collected and centrifuged at 500 g for 5 min at 4 °C, and the resultant pellet was processed for antibody labeling and cytometer analysis as described subsequently. When included in the analysis, peritoneal lavage was the first procedure done, before blood collection and perfusion. Peritoneal cavities received 2 to 5 mL of FACS buffer, which was further recollected. Samples were adjusted to a volume of 10 mL, and underwent centrifugation at 500 g for 5 min at 4 °C. The final pellet was subsequently marked with the antibody panel and analyzed on the cytometer as described subsequently.

Staining of isolated cells and cytometer/sorter analysis

Cells isolated from liver, spleen, bone marrow, blood, and peritoneal cavity underwent labeling with specific antibodies. A total of 20 μL of the antibody panel mix or FMO controls were added to the cell pellet. Following a 20-min incubation at 4 °C, samples were washed with 200 μL of FACS buffer and centrifuged at 500 g for 5 min at 4 °C. The pellet was then incubated for 15 min at room temperature with 50 μL of the viability dye 7-AAD (1:20; BD Biosciences; 559925). Further washing with 200 μL of FACS buffer and centrifugation at 500 g for 5 min at 4 °C was performed before the final pellet was resuspended in 200 μL of FACS buffer for analysis on the Fortessa cytometer (BD) or the sorting of specific cell populations on the ARIA sorter (BD). The composition of antibody mixes varied based on the experimental requirements: for B cell subpopulation analysis, CD45 APC-Cy7 (BioLegend; Cat. 103116), CD19 BV421 (BioLegend; Cat. 115538), CD93 PE (BioLegend; Cat. 136503), IgM PE-Cy7 (BioLegend; Cat. 406513), CD21 BUV395 (BD, Cat. 740249), CD23 BUV496 (BD, Cat. 741058), CD43 PE-Dazzle (BioLegend; Cat. 143217), CD5 BV711 (BioLegend; Cat. 100639), and CD3 AF700 (BioLegend; Cat. 100216); for B cell precursor sorting for transfer, CD45 APC-Cy7 (BioLegend; Cat. 103116), CD19 BV421 (BioLegend; Cat. 115538), CD93 PE (BioLegend; Cat. 136503), IgM PE-Cy7 (BioLegend; Cat. 406513), CD3 PerCP (BioLegend; Cat. 100326), Ly6G PerCP (BioLegend; Cat. 127654), Ly6C PerCP (BioLegend; 128028) e NK1.1 PerCP (BioLegend; Cat. 108726); for B cell sorting for RNA sequencing (RNA-seq), CD45 APC-Cy7 (BioLegend; Cat. 103116), CD19 BV421 (BioLegend; Cat. 115538), CD93 PE (BioLegend; Cat. 136503), IgM PECy7 (BioLegend; Cat. 406513). Flow cytometry analyses were performed using FlowJo v10 software (BD), with subpopulations identified based on a specific combination of markers and a gating strategy detailed in the supplementary figures.

Adoptive transfer of B cell precursors to Rag1^−/− mice

B cell precursors were isolated from the liver or bone marrow as described above, using a consistent gating strategy for B cell precursor isolation (CD45⁺CD3⁻CD19⁺IgM⁻CD93⁺) across the organs (Fig. S1). In a first set of experiments, B cell precursors from the liver of 1-wk-old WT mice or B cell precursors from the bone marrow of 9-wk-old WT mice were isolated and transferred to 9-wk-old Rag1^−/− mice. One million precursors were intravenously transferred. The liver, spleen, bone marrow, and peritoneal lavage of Rag1^−/− animals were collected 4 wk post-transfer and processed for analysis of B cell subpopulations by flow cytometry, as previously described. Alternatively, B cell precursors from the liver or bone marrow of 1-wk-old WT mice were isolated and transferred to 1-wk-old Rag1^−/− mice. Three hundred thousand precursors were intravenously transferred. The liver, spleen, bone marrow, and peritoneal lavage of Rag1^−/− animals were collected 1-wk post-transfer and processed for analysis of B cell subpopulations by flow cytometry, as previously described.¹²

Estrous cycle monitoring of females

To plan and organize mating, as well as to ensure greater accuracy in determining gestational periods, the females used in this study were continuously monitored to assess their estrous cycle.¹³

Antibiotic treatment protocol

Pregnant females were subjected to an antibiotic treatment regimen from gestational day E15 until the third week of lactation. The antibiotics were administered ad libitum through the drinking water, supplemented with a solution containing ampicillin (1 g/L), metronidazole (1 g/L), neomycin (1 g/L), and vancomycin (0.5 g/L).¹⁴ The solution was prepared and replaced every 48 h to ensure compound stability.

Validation of microbiota alteration by antibiotic treatment

The overall alteration in gut microbiota composition to validate the antibiotic treatment was assessed using fecal samples cultured on sheep blood agar. After fecal collection, the samples were diluted (10⁶) and plated. The plates were incubated in anaerobic jars for 48 h at 37 °C.

Photoconversion of liver cells in Kaede mice

Liver cells from 1-wk-old Kaede mice were photoconverted through a 2-min exposure to a laser (405 nm; peak power <5 mW; sustained power: 0.5 to 4.9 mW; Electra Pro Portable Violet Laser; Laserglow Technologies). The laser was maintained at a fixed distance of 6 cm above the animal. The animal remained anesthetized throughout the procedure using isoflurane. An ink mark was made on the mouse skin over the liver region. The entire animal was then covered with aluminum foil, which had an opening only above the marked liver region. The aluminum foil blocked scattered laser, ensuring that only the liver region was exposed to the laser and photoconverted. After 2 min of laser exposure and recovery from anesthesia, the animals returned to their cages with their mothers until organ collection. Liver, spleen, lung, intestinal lamina propria, mesenteric lymph nodes, blood, and bone marrow were collected at different time points after photoconversion: 0, 1, 2, 3, and 7 d. These organs underwent the same procedures for cell isolation and labeling as described above for cytometry analysis. The samples were then analyzed for the presence of red (photoconverted) CD19⁺ cells. Kaede mice that did not undergo the photoconversion process served as a negative control for the experiments.

Sample collection for antibody quantification

To assess the establishment of individual antibody production and contributions from maternal milk antibodies, we conducted various crosses between Rag1^−/− and WT animals, combined with the use of adoptive mothers. The first group, a negative control with no antibody presence from any source, was formed by crossing Rag1^−/− mouse. The second group, a positive control with antibodies from all sources, resulted from crossing WT animals. The third group, in which animals only have antibodies from maternal milk, was formed by crossing Rag1^−/− mice, and transferring the Rag1^−/− offspring to adoptive WT mothers, resulting in pups that did not have their own antibody production but received antibodies from the milk of the adoptive WT mother. The fourth group resulted from crossing Rag1^−/− (female) with WT (male), giving rise to a Rag1^−/+ offspring. Consequently, these pups only possessed their own antibody production and did not receive antibodies from maternal milk. Serum from the pups belonging to the 4 cross groups was collected at 1, 2, 3, 4, and 9 wk of age for further ELISA analysis. Additionally, maternal milk was collected from WT littermates at different times (0 and 4 d and 1, 2, 3, and 4 wk) to assess the presence of antibodies in maternal milk. The littermate was separated from the pups for 2 h before milk collection. After this period, littermates were anesthetized intraperitoneally with ketamine (85 mg/kg) and xylazine (1 mg/kg) and received 1 IU of oxytocin intraperitoneally. Milk collection began 2 min after oxytocin administration using a pipette. After recovering from anesthesia, littermates were reunited with the pups. The same female was used for milk collection at all times, with 4 replicates performed. For ex vivo liver culture experiments, liver was perfused via portal vein to remove blood and incubated in William’s medium for sample collection. For serum, milk and supernatant samples, immunoglobulins IgG, IgA, and IgM were quantified using the ELISA method.

Antibody quantification using ELISA

For the ELISA protocol, 96-well plates were coated with 100 μL/well of capture antibody (goat anti-mouse Ig, Human ads, UNLB, 1010-01; SouthernBiotech) diluted in Coating Buffer at 1:5,000. The plates were kept at 4 °C in a humid chamber overnight, followed by washing with PBS-Tween. Blocking was performed with 200 μL/well of PBS-casein and incubated for 1 h at room temperature, followed by washing with PBS-Tween. The samples, diluted in PBS-casein, were then added to the plate. For the blank, only PBS-casein was added to the well. Standard curves were generated by serial dilution (12 points ranging from 1,000 ng/mL to 0.48 ng/mL) of the antibodies: mouse IgM isotype control UNLB (0101-01; SouthernBiotech) or mouse IgA isotype control UNLB (0106-01; SouthernBiotech) or mouse IgG isotype control UNLB (0107-01; SouthernBiotech). After adding the samples, blank, and standard, the plate was incubated at 37 °C for 1 h. After washing with PBS-Tween, the plates were incubated at 37 °C for 1 h with the detection antibodies, diluted in PBS-casein at 1:10,000: goat anti-mouse IgM, human ads, BIOT (1020-08; SouthernBiotech) or goat anti-mouse IgG, human ads, BIOT (1030-08; SouthernBiotech) or goat anti-mouse IgA, BIOT (1040-08; SouthernBiotech). After washing with PBS-Tween, another incubation at room temperature for 1 h was performed with streptavidin-HRP diluted in PBS-casein at 1:5,000. After washing with PBS-Tween, the plates were incubated at room temperature and protected from light for 5 min with the revelation solution (4 mg OPD, 10 ml citrate buffer, 2 μL H₂O₂ 30%). After 5 min, the revelation reaction was interrupted by adding 20 μL/well of 2 N sulfuric acid. The plates were then read in a spectrophotometer with at 492 nm.

Gut microbiota data processing

Fecal samples were collected under sterile conditions and stored to further DNA isolation using commercial kits. The raw sequence reads of the 16S rRNA gene processed, trimmed, and assembled, leading to the generation of amplicon sequence variants (ASVs). This process was accomplished using DADA2.¹⁵ During the processing, the primers used in amplification were systematically removed, and sequences with more than 2 expected errors were eliminated, ensuring data integrity. The remaining sequences were employed to create an error identification and correction model, enhancing data accuracy. The forward and reverse readings, already corrected, were concatenated to form the ASVs. This concatenated data was used to eliminate chimeric sequences and to quantify the ASVs accurately. Each ASV received a taxonomic classification via the TAG.ME package. This classification was based on the SILVA database as a reference, utilizing the specific model tailored for the amplicon corresponding to the 515F-806R region.

Gut microbiota analysis

The Beta diversity was evaluated by calculating the Bray-Curtis dissimilarity. The results were visualized in a principal coordinate analysis. A permutational analysis of variance test was employed to identify distinctive microbial patterns associated with each disease under study. To identify differentially abundant bacterial taxa within each clinical group (each unhealthy group compared with the healthy group), the Wald test from the DESeq2 package (version 1.30.1)¹⁶ was utilized. During this analysis, candidates were selected based on adjusted P values for multiple tests using the Benjamini-Hochberg method (adjusted P < 0.05).

Mapping of RNA-seq libraries and differential gene expression analysis

Raw sequencing reads were quality-filtered using Trimmomatic v0.40.¹⁷ Adapter sequences (Illumina) were trimmed, and reads containing low-quality bases (Phred < 20) or shorter than 36 bp were discarded, with the following parameters: PE -phred33 ILLUMINACLIP:TruSeq.fa:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:3:20 MINLEN:36. Trimming was followed by read error correction by SGA k-mer–based algorithm¹⁸ version 0.9.9 with standard settings. Next, the reads were mapped against the Genome Reference Consortium Mouse Build 38 (GRCm38), and the transcript abundance was quantified using Salmon v1.10.0.¹⁹ Finally, the count data were direct to differential analysis with DESeq2 (version 1.43.5) R 4.4.0 package¹⁶ with standard settings. Biological process Gene Ontology terms search was performed through online database PANTHER 18.0 with standard parameters.²⁰

Single cell transcriptomic profile analysis

Single-cell RNA-seq count matrixes from human surgically removed adult liver and aborted fetal liver assessed herein were obtained from the dataset GSE134355 on NCBI Gene Expression Omnibus public databank. These count matrixes served as input to downstream analysis in Seurat (version 5.1.0),²¹ initially filtering cells with unique molecular identifiers over 5,000 or <200, besides outliers with more than 20% mitochondrial counts. We performed normalization regressing out heterogeneity associated with mitochondrial contamination fitting the counts to a Gamma-Poisson General Linear Model, removing this confounding source of variation through scTransform Seurat’s function²² and glmGamPoi package.²³ Integration of adult and fetal data was executed through Harmony 1.2.0²⁴ grouping by samples. Dimensionality reduction was implemented by principal component analysis, followed by UMAP embedding using the first 30 principal components, the same parameter as used for clustering with both FindNeighbors and FindClusters (at 3 resolution) Seurat’s functions. Differential gene expression comparison between the fetal and adult liver cells was done based on the nonparametric Wilcoxon rank sum test. Genes with fold change >1.5 (absolute value), with adjusted P value (Bonferroni correction) <0.1, and that detected in a minimum fraction of 20% cells in either of the two populations were considered differentially expressed. Biological process Gene Ontology and Reactome terms search were performed through online database PANTHER 18.0 with standard parameters.²⁰

Statistical analysis

Unless stated in transcriptomic experiments, statistical analyses, and graph plotting were performed using GraphPad Prism 10 software (GraphPad Software). Unpaired Student’s t test was conducted for normally distributed data, and the Mann-Whitney test was applied for data not following a normal distribution. In cytometry experiments for subpopulation comparisons across all ages, 1-way analysis of variance was employed, with 9 wk as the control age, and Dunnett’s post-test for normally distributed data. For data not conforming to a normal distribution, the Kruskal-Wallis test with Dunnett’s post-test was used. Results were presented as mean ± SEM with a minimum of 4 animals per group. Statistically significant differences were considered when P < 0.05.

Neonatal liver has a large B cell population

To investigate the cellular organization within newborn livers and compare it with their adult counterparts, we analyzed human liver samples from biopsies. These samples underwent immunohistochemistry (anti-CD19) and in situ hybridization to identify B cells (CD19 mRNA), allowing also for general histomorphology evaluation. Adult liver samples exhibited classical histological features, characterized by massive cords of hepatocytes crossed by numerous sinusoids, with immune cells (mainly Kupffer cells; green arrow) primarily located inside the sinusoids (Fig. 1A). A few CD19⁺ cells were discernible across the field of view (Fig. 1A). However, newborn samples presented a stark contrast: we observed numerous clusters of cells, presumably leukocyte aggregates, dispersed throughout the liver (red arrows). These clusters were predominantly situated within the parenchyma, surrounded by hepatocyte cords (Fig. 1A). Morphological characteristics such as cell size, structure, and nuclear morphology indicated that these clusters predominantly consisted of immune cells. Immunohistochemistry analysis revealed a substantial population of CD19⁺ cells in newborn livers, significantly higher than the numbers found in adult livers (Fig. 1B, C). Notably, B cells comprised a significant proportion of the immune cell clusters found in newborn livers (Fig. 1A, black arrows), highlighting the significant presence of organized immune cells within the parenchyma, reminiscent of lymphoid follicles observed in other organs. To confirm the presence of B cells in our samples, we performed in situ hybridization for CD19 mRNA using a fluorescent probe (Fig. 1D) across different developmental stages: fetal, infant, and adult. CD19⁺ cells were observed in all human samples; however, only fetal samples exhibited a significant number of CD19⁺ cells organized into clusters, a feature not present in the other stages. After observing a large population of B cells within neonatal livers, we speculated that such unique microenvironment might induce different gene signatures in newborn cells compared with adults. To swiftly validate this hypothesis and guide subsequent in vivo experiments, we initially analyzed a publicly available single-cell RNA-seq dataset from Han et al.²⁵ which we filtered to isolate B cells from fetal (26 wk of gestation) and adult livers (21, 23 and 52 yr old) using selected cell surface markers. Initial clustering via dimensional reduction UMAP unveiled 47 clusters corresponding to various subsets of hepatic cells (including general B cell markers, naïve B cells, and plasma cells) (Fig. 2A). To establish a direct correlation with the cells seen in human samples, we further refined general B cells as belonging to the clusters 33 and 37 based on high CD19 expression for subsequent gene expression and enriched pathway analysis (Fig. 2A, B). This analysis revealed 128 differentially expressed genes (DEGs) (false discovery rate <0.05) in adult B cell samples compared with their fetal counterparts. Adult liver B cells exhibited several upregulated genes related to immunoglobulin processing and release and to major histocompatibility complex (MHC) classes I and II (Fig. 2C–E). In fact, pathway and Gene Ontology analyses demonstrated enrichment in adult liver B cells for antigen-receptor, response to biotic stimulus, lymphocyte activation, and various other processes related to immune response (Fig. 1E). Thus, B cells in neonatal and adult livers have significant differences that underscore the immaturity and less functional specialization of neonatal B cells upon birth.

Next, we refined the analysis of general B cell clusters to examine the genetic profiles of distinct B cell progenitor populations (Fig. 2F). This classification was based on the analysis of key expression markers: CD22, which is expressed in a cytoplasmic form during early stages of B cell differentiation, IGHM, encoding the mu heavy chain C region that defines the IgM isotype, and IGHD, which encodes IgD expression. Specifically, we identified a small subset of immature B cells (CD19⁺IGM⁺IGD^+/−), pre-B cells (CD19⁺IGM⁻IGD⁻), and pro-B cells (CD19⁺IGM⁻CD22⁺). However, due to their low abundance in the dataset, we opted to combine these subtypes into a larger progenitor B cell cluster for further analysis. When comparing the genetic profile of fetal B cell progenitors to their adult counterparts, we observed significant differences in DEGs (Fig. 2G). Despite being progenitors in adult livers, these B cells had significant upregulation of MHC class II genes as compared with their fetal counterpart (Fig. 2H), suggesting that they are already functioning as antigen presenting cells as indicated by enriched pathways associated with leukocyte cell-cell adhesion, regulation of CD4 T cells differentiation and antigen processing and presentation of antigen via MHC class II (Fig. 2I). Conversely, pathways enriched in precursor B cells from fetal livers were associated with metabolic processes (Fig. 2I). Thus, these findings support the notion that both fetal and adult livers harbor progenitor B cells, which are genetically distinct, suggesting that aging plays a critical role in the B cell maturation process.

Hepatic B cells gradually reduce their frequency within the liver during postnatal development

Validation of our hypothesis that liver B cells exhibit distinct profiles across developmental phases was constrained by the limited availability of public datasets, which included cells only from fetal and adult stages, without intermediate age groups²⁵ Therefore, to further understand the dynamics of B cell in their native environment, we generated a novel mouse strain that express Td-tomato under the regulatory control of CD19 (CD19^tdtomatoflox), allowing a precise and full chronological investigation of B cell biology within the liver. Using confocal intravital microscopy, we confirmed the significant presence of liver-resident B cells in fetal (embryonic day 19) and neonatal mice (1 and 2 wk old), mirroring the characteristics observed in human newborns (Fig. 3A;  Supplementary Movie 1). These cells did not exhibit a uniform distribution across the imaging field; instead, they congregated in discrete cellular clusters (Supplementary Movie 2), also resembling human biopsies. Dual staining with anti-CD31 for sinusoidal endothelial cells revealed that almost 60% of B cells were positioned extraneously to the vessels, forming cohesive aggregates within the parenchyma (Fig. 3B–E). We extended our study to include infant (3 wk old) and adult mice (9 wk old) (Fig. 3A; Supplementary Movie 2). Notably, large parenchymal B cell aggregates observed in infants were entirely absent in these mice, with B cells primarily located within the liver sinusoids (Fig. 3B–E; Supplementary Movies 3 and 4). This suggests a temporal change not only in B cell number, but also in their distribution within the liver. Time-lapse imaging captured in vivo revealed that these clusters, typically consisting of more than 5 cells, displayed dynamic and coordinated motility (Fig. 4A; Supplementary Movie 2). This was further corroborated through individual digital cell tracking and morphological analyses (Fig. 4B–G; Supplementary Movie 5). Neonatal B cells exhibited distinct morphological characteristics compared with their adult counterparts, displaying larger cell areas, diameters, perimeters, and increased elongation, which resulted in a more elongated cell profile (Fig. 4B–H). These distinctive features may reflect bona fide migratory behavior.²⁶ Our observations suggest that a subset of these neonatal B cells gradually reduces their presence within the liver parenchyma, potentially indicating their entry into systemic circulation or migration to other organs (Fig. 4H; Supplementary Movie 5). To determine whether the B cells in the sinusoids were liver-resident or circulating cells, we compared leukocytes from perfused and nonperfused liver tissues in both neonates and adults. We found no significant difference in the percentage of B cells between perfused and nonperfused livers at any age (Fig. 4I), indicating that these B cells are part of the hepatic resident immune cell pool. Overall, our findings suggest that B cells are an integral component of the liver, interacting with the organ throughout all stages of postnatal development.

The maturity profile of B cells changes during hepatic postnatal development

Both genetic and in vivo analysis suggested that B cells should not be perceived as a uniform cell population; rather, they constitute a diverse array of subpopulations distinguished by variations in phenotype, location, and function.⁸ Following our visualization and characterization of B cell residency in the liver across various developmental stages, we next focused on immunophenotyping the subpopulations of hepatic B cells and their potential alterations in subtype and frequency during postnatal development. Due to a lack of comprehensive immunophenotyping data for liver B cells in the existing literature, we conducted our analysis alongside well-characterized organs: the bone marrow (a primary lymphoid organ), spleen (a secondary lymphoid organ), and blood (which contains circulating cells). Our analytical framework categorized B cell subpopulations into 3 primary groups based on CD93 and IgM expression: precursor cells (CD19⁺CD93⁺IgM⁻, including pro-B and pre-B precursors), immature/transition cells (CD19⁺CD93⁺IgM⁺), and mature cells (CD19⁺CD93⁻IgM⁺, representing B1 and B2 lineages) (Fig. 5A;  Figs. S1 and S2). During fetal development, precursor cells were the dominant population, accounting for a substantial proportion of B cells in both the liver (∼95%) and spleen (∼75%) (Fig. 5B–H; and Figs. S1 and S2). These precursor cells predominantly manifest as pro-B cells (CD43⁺) (Fig. 5E). In 1-wk-old mice, we observed a significant reduction in precursor cells in both organs. The spleen exhibited approximately 20% of B cell precursors, while the liver retained a high frequency of these cells (∼65%) (Fig. 5B–H). The distribution between pro-B (CD43⁺) and pre-B (CD43⁻) cells in the liver became more balanced, with each accounting for approximately 30%, indicating a shift with reduced pro-B and increased pre-B cell frequencies (Fig. 5E, F). These findings suggest a retention of B cell precursors within the liver during the postnatal phase, contrasting with the spleen, which experiences a significant decline in precursor B cells and shifts toward a predominance of mature B cells (∼40%) (Fig. 5B–H). Notably, bone marrow exhibited a B cell subpopulation profile similar to that of the liver, with slightly higher percentages of precursors (∼85%, with pro-B and pre-B cells each comprising approximately 40%), while blood primarily contained immature B cells (∼55%) (Fig. 5B–H). By the second week postbirth, the B cell subpopulation profile remained largely unchanged from the first week. However, liver precursors continued to decline, reaching around 50%, while those in the bone marrow remained relatively stable at ∼80%. In both the liver and bone marrow, pro-B cells decreased and pre-B cells increased, reaching approximately 40% in the liver and ∼60% in the bone marrow. Additionally, the liver displayed numbers of immature B cells comparable to or higher than that in the bone marrow (approximately 20% to 30%) (Fig. 5B–H). However, by the third week postbirth, liver B cell precursors further diminished and nearly disappeared. Conversely, mature B cells steadily increased, reaching levels comparable to those in adult animals by the third week (∼40%). At this point, liver B cell profile now resembled that of the spleen, characterized by a virtual absence of precursors and a marked prevalence of mature cells. The bone marrow continued to harbor precursors, maintaining a predominance of pre-B over pro-B cells. By the fourth week postbirth, the B cell subpopulation profile across the body became firmly established, closely resembling that observed in adults. Both the liver and spleen housed approximately 40% mature B cells, predominantly of the B2 lineage (Fig. 5B–H). Interestingly, the bone marrow emerged as the exclusive site for B cell precursors, with the majority being pre-B cells, alongside approximately 20% immature cells. Finally, immunophenotyping analyses specifically of liver B cells across different ages highlighted the 3 primary B cell subpopulations: precursor, transitional/immature, and mature (Fig. 5I–N). Taken together, our immunophenotyping strategy provides a comprehensive and novel framework for studying liver B cells, demonstrating that subpopulation profiles are not fully defined at birth but become relatively stable starting from the fourth week onward.

The establishment of mature B cell population 4 wk after birth is confirmed by the increase in self-produced antibody

We next investigated the B cell function by accessing their ability to produce antibodies throughout the development of the mice. Given the contribution of milk to the circulating poll of immunoglobulins during the neonatal phase, we devised a comprehensive strategy that allowed us to measure self-derived and milk-derived antibodies separately. To assess individual antibody production and the influence of maternal milk antibodies, we implemented a sequential breeding protocol involving Rag1^−/− and WT animals, together with the use of foster mothers. Four distinct groups were designed for this purpose (Fig. 6A). The first group (negative control) lacked circulating antibodies and resulted from the breeding of Rag1^−/− mice (Fig. 6A). The second group (positive control) involved the reception of antibodies from all sources, including self-produced and milk-derived antibodies, and was achieved by crossing WT mice (Fig. 6A). The third group (milk-derived antibodies) comprised mice that exclusively received antibodies from maternal milk. This was achieved by crossing Rag1^−/− mice, and the resulting Rag1^−/− offspring were promptly transferred to foster WT mothers after birth (Fig. 6A). The fourth group (self-derived antibodies) comprised mice generated from the breeding of a Rag1^−/− (female) with a WT (male), yielding Rag1^+/− offspring (Fig. 6A). In this scenario, these offspring exclusively had endogenous antibody production and did not receive antibodies from maternal milk (Rag1^−/− mother). In all groups, serum samples from the offspring were collected throughout postnatal development for immunoglobulin quantification.

As expected, Rag1^−/− mice (negative control) had no detectable levels of immunoglobulins over life. In the positive control group, we found a progressive increase in IgM and IgA production throughout development, culminating in adult-like titers by the fourth week (Fig. 6B, C). No detectable levels of IgM and IgA originating from maternal milk were found in the serum (Fig. 6B, C). Importantly, the establishment of endogenous production for these immunoglobulins occurred earlier when maternal antibodies are absent and self-production is the only antibody source (Fig. 6B and C). Simultaneously, IgG was identified at elevated levels from early development (Fig. 6D). However, we found that neonatal serum IgG predominantly originated from maternal milk, with endogenous production progressively increasing with age, particularly in the fourth week postweaning (Fig. 6D), evidencing that antibodies obtained from maternal milk supplement the shortfall in immunoglobulin secretion in newborns. Despite the absence of serum IgM and IgA transferred from maternal milk, all tested isotypes—IgG, IgA, and IgM—were found in milk, exhibiting varying abundance, with IgG being the most abundant and IgM the least abundant (Fig. 6E).

To investigate whether the liver could also be the source of this circulating endogenous antibodies, we harvested livers from 1- and 8-wk-old animals after an extensive perfusion to remove antibodies from the blood, and culture them in specific media. Supernatants were collected at several time points (0 to 24 h) and IgA, IgM, and IgG measured by ELISA. We found that culture of adult liver produced all 3 antibodies during the time points evaluated (Fig. 6F–I), and regardless of stimulation with lipopolysaccharide (not shown). Conversely, newborn liver culture did not produce IgA or IgM, although containing levels of IgG (likely a remnant from the blood) (Fig. 6F–I), indicating that, as expected, B cell precursors present in the newborn liver cannot produce antibodies. Taken together, these findings underscore that self-antibody production is only established during postnatal development, which parallels the maturation of B cells.

Hepatic B cell precursors in newborns contribute to the pool of extrahepatic B cell populations

The high frequency of B cell precursors in the liver led us to hypothesize that newborn liver may actively contribute to the formation of B cell compartments within other lymphoid organs. To test whether postnatal hepatic B cell precursor could differentiate into mature B cells, we performed an adoptive transfer of sorted B cell precursors (CD19⁺CD93⁺IgM⁻) into Rag1^−/− mice (which lack T and B cells) and immunophenotyped engrafted B cell subpopulations in the liver, spleen, bone marrow, and peritoneal cavity using flow cytometry (Fig. 7A). Two distinct secondary lymphoid organs were chosen for assessment: the peritoneal cavity, which—under wild type circumstances—is enriched in B1 lineage cells; and the spleen, which is enriched in B2 lineage cells. In experiments with adult Rag1^−/− mice as recipients, we observed comparable engraftment levels of liver and bone marrow–derived B cells in both the bone marrow and spleen. However, there was a more robust engraftment of liver-derived B cells in the liver and peritoneal cavity (Fig. 7B). The distribution of total B cells in the examined organs revealed approximately 10% in the bone marrow and peritoneal cavity, 5% in the spleen, and 1% in the liver. (Fig. 7B). We found that liver B cell precursors from 1-wk-old mice had the same differentiation potential as compared with a bona fide 8-wk-old mouse bone marrow B precursor. Both precursor populations successfully differentiated into mature B cells, predominantly of the B2 lineage in the spleen and B1 lineage in the peritoneal cavity (Fig. 7C, D). Notably, precursors from both liver and bone marrow effectively colonized the bone marrow of the recipient Rag1^−/− mice but maintained their precursor phenotype (Fig. 7E). Although B cell precursors were detected in the liver, their percentages were lower compared with the bone marrow, and mature B cells were also identified in the liver, but at a lower frequency than in the spleen and peritoneal cavity (Fig. 7F).

In parallel experiments with 1-wk-old Rag1^−/− mice as recipients, we also found a similar engraftment between liver and bone marrow derived B cells. However, in newborns, B cells demonstrated enhanced engraftment compared with adult mice. The total B cell percentages were approximately 30% in the bone marrow and 20% in the liver, spleen, and peritoneal cavity (Fig. S3A-D). We similarly found that 1-wk-old liver B precursors had the same engraftment profile as compared with bone marrow B cell precursors from 1-wk-old mice (Fig. S3A–D). While both precursor populations successfully colonized all assessed organs, they did not undergo differentiation into mature B cells (Fig. S3A–D), suggesting that the maturation of the lymphoid organ is critical to drive B cell differentiation. Thus, these findings demonstrate that hepatic B cell precursors maintain the potential to differentiate into mature B cells, successfully establishing in various organs.

To assess the functionality of the mature pool of B cells derived from engrafted liver precursors and their ability to respond to an antigen stimulus, Rag1^−/− mice that had received precursor B cells were immunized with ovalbumin emulsified in complete Freund’s adjuvant. B cell differentiation was then quantified using flow cytometry (Fig. 7G). We found that the transferred B cell precursors underwent differentiation into mature B cells, responding effectively to the immunization. This differentiation encompassed the development of germinal center B cells, plasmablasts, plasmacytes, and memory B cells, with no discernible distinction between cells originating from the liver or bone marrow (Fig. 7J–M). Apart from the observed reduction in the overall number of B cells in the transferred mice compared with the WT control mice (Fig. 7H), the frequencies of differentiated cells displayed a comparable outcome between the two groups.

Taken together, we demonstrated that precursor B cells originated from liver are functional and can differentiate into effector B cells after an antigen challenge. However, these findings did not directly prove the ability of liver B cell precursors to migrate to lymphoid organs. Given the similarities of B cells across tissues, the only approach to investigate whether these cells could exit the liver to populate other organs is by selectively staining liver cells and tracking their trajectory. To delve into the in vivo migration potential of liver precursor B cells, we implemented a model utilizing 1-wk-old Kaede mice subjected to focal liver cell photoconversion. (Fig. 7N). In this model, Kaede mice consistently emit green fluorescence in all cells. Following local irradiation with violet light (405 nm), only the irradiated green cells undergo a color change, turning red through photoconversion.²⁷ Importantly, this specific photoconversion does not affect cellular functions.²⁸ After laser exposure, we assessed the presence of photoconverted B cells in the liver, spleen, and bone marrow on days 0, 1, 2, 3, and 7 postphotoconversion (Fig. 7O–R). Our results confirmed the specificity of photoconversion, with the liver exclusively possessing photoconverted B cells on day 0 (Fig. 7O). We found that the frequency of photoconverted B cells reduced in the liver on days 1 and 2, when they started to be detected in the bone marrow (Fig. 7Q) and predominantly in the spleen (Fig. 7R). The detection of photoconverted CD19⁺ B cells in the spleen or bone marrow on day 1 and subsequent days confirmed that these cells originated from the liver. Presumably, owing to several rounds of division and the subsequent decrease in their red fluorescence, no photoconverted cells were observed in liver, spleen, or bone marrow by day 3 postphotoconversion, a pattern that persisted until day 7 (Fig. 7O).

To determine whether liver-resident photoconverted immune cells disseminate to other immunologically relevant tissues, we assessed the presence of photoconverted cells in the lung, gut lamina propria, and mesenteric lymph nodes 48 h postconversion—a time point in our system corresponding to the peak abundance of photoconverted cells in peripheral tissues. As shown in Fig. 7S, liver-derived cells were readily detected in all assessed organs. Notably, when analyzing CD45⁺ photoconverted cells, we found that B cells comprised approximately 15% to 30% of the immune cells in the liver, lung, and gut, whereas CD19⁺ cells represented nearly 80% of the leukocytes in the mesenteric lymph nodes (Fig. 7T). These data suggest that liver-derived immune cells, including B cell subsets, are capable of trafficking to peripheral mucosal and lymphoid tissues, potentially contributing to the local immune cell pool. Given the consistent detection of photoconverted cells in distant tissues, we next investigated whether these cells could be identified in the circulation, thus supporting a hematogenous route of migration. Indeed, a substantial proportion (∼80%) of photoconverted CD45⁺ cells in the blood were CD19⁺ B cells (Fig. 7T), indicating that circulating B cells of hepatic origin likely mediate their redistribution to distal immunological niches via the bloodstream. Thus, these results demonstrate that liver B cell precursors retain a differentiation potential into functional mature B cells and can migrate to other organs, underscoring our hypothesis that the postnatal liver significantly contributes to the systemic formation of B cell compartment.

The liver has a transcriptionally distinct and unique B cell population

We next investigated the transcriptional landscape associated with precursor B cells from liver versus precursors B cells from the bone marrow and spleen of newborns at 1 wk of age. We found that DEGs of B cell precursors in these organs were only slightly different (Fig. 8A–F). Interestingly, pathways associated with leukocytes, FCGR, and the regulation of B cell activation, among others related to the immune response, were downregulated in the liver B cell precursors compared with their bone marrow and splenic counterparts, reinforcing the regulatory milieu that the liver imprints on B cells since the beginning of development.

We then investigated whether age was a major contributor to B cell transcriptional landscape in the liver. We performed RNA-seq on precursor and immature B cells isolated from fetal (E17), newborn (1 wk old), and adult (9 wk old) WT mice. When we compared newborn vs. fetal liver B cell precursors (Fig. 8G–I), we identified several genes and pathways with significant alterations. Notably, we found that pathways related to the cellular response to cytokines, defense against viruses, and interferon signaling were downregulated in B cell precursors from newborn livers, aligning with the expected antigen exposure postbirth. Conversely, genes and pathways associated with B cell-mediated immunity, immunoregulatory interactions between lymphoid organs, FCGR activation, and several others related to the immune response were significantly upregulated. These pathways were mainly connected via MHC class I–mediated antigen presentation, endoplasmic reticulum phagosome, and FCGR activation pathways (Fig. 8I). The complexity of this scenario increased at later stages of mouse development. By comparing immature B cells from adult mice with newborns, a more striking divergence was observed (Fig. 8J–L). Interestingly, pathways associated with FCGR receptors, B cell activation, and several other B cell–related pathways were enriched in B cells from adult mice in comparison to newborns (Fig. 8K). Likewise, the same immunologic hub, but now involving a broader network, connected all the enriched pathways (Fig. 8L). These findings suggest that, despite general classification based on cell surface markers, B cell states represent transcriptionally distinct cell populations dependent on the life stage, and aging emerges as a key factor in determining the B cell phenotype.

The gut microbiota diversity modulates early life liver B cell development

The potential link between aging and B cell maturation prompted us to hypothesize that postnatal environmental cues, particularly those coming from changes in diet and microbiota, could play a significant role in this context. Therefore, we further investigated whether the gut microbiota could act as a crucial factor by providing cues to liver B cells during early life, potentially driving their maturation process. We first performed a 16S rRNA sequencing of fecal samples during the postnatal development to determine how the microbiota changes after birth until the adult phase. We found a gradual increase of species richness and evenness in microbial community abundance throughout postnatal development, as indicated by the Shannon Diversity Index (Fig. 9A). Notably, a distinct shift in diversity was observed, with the first week of life demonstrating lower diversity compared with later developmental stages (beyond the third week up to adulthood), while the second week represented an intermediate stage (Fig. 9A). Interestingly, we found a direct correlation between B cell maturation status and postnatal development period. The microbiota dynamic coincided with the retention of B cell precursors in liver during the neonatal phase and the increase in B cell maturation in the body with age (Fig. 9B). Employing permutational analysis of variance, a nonparametric statistical test evaluating differences in multivariate data, we determined that the variable “aging” alone accounted for approximately 70% of the identified significant differences in the gut microbiota among the groups (Fig. 9C). Consistent with this, numerous genera and families, including Citrobacter, Escherichia, Shigella, Lactobacillus, Bacteroides, Parabacteroides, and Akkermansia, among several others, exhibited statistically significant differences between age groups (DESeq2, WaldTest, adjusted P < 0.01) (Fig. 9D;  Fig. S4). Thus, the dynamic of microbial composition shifts significantly during postnatal development, with the aging vector emerging as an essential factor influencing the gut microbiota diversity.

To directly assess the putative impact of gut microbiota diversity in B cell development, we next performed bulk RNA-seq on isolated hepatic precursor and immature B cells from mice exhibiting a restricted variability in their microbiota compared with WT controls (“Excluded Flora mice” from Taconic Bioscience, here “EF” for short). This unique mouse model lacks Proteus spp., segmented filamentous bacteria, and 6 opportunistic pathogens. It is important to note that germ-free mice were not employed in these studies given their underdeveloped immune system,^29,30 which would affect the interpretation of our results. In the liver, we found 332 DEGs (157 upregulated and 175 downregulated) in immature B cells from EF vs. WT mice (Fig. 9E) and 700 DEGs (396 upregulated and 304 downregulated) in precursors (Fig. 9F). Further analysis using of cell precursors from 1-wk-old mice (Fig. 9G, H) revealed that a large number of genes were significantly altered in the liver (352 DEGs) as compared with the bone marrow (221 DEGs) or spleen (226 DEGs). Network analysis of these altered pathways confirmed that the DEGs were related to immune response (Fig. 9G, H). Consistent with our hypothesis that the gut microbiota would mostly impact B cell development within the liver, B cell gene signature was only minimally altered in bone marrow (Fig. 9G, H). Although the significance of these altered genes and pathways found in precursor and immature B cells from EF vs. WT mouse livers do not precisely indicate how the gut microbiota regulates B cell development, our findings highlight the central role of microbial factors as determinants in guiding B cell maturation throughout the life, especially within the liver.

Depletion of newborn microbiota leads to major alterations in liver B cell profile, reducing serum levels of immunoglobulins

Building on our previous findings that the absence of specific bacterial families alters the transcriptional profile of hepatic B cell precursors, we developed an experimental protocol to directly test the role of neonatal gut microbiota in shaping liver B cell development and function. Recognizing that germ-free mice present widespread immunological deficiencies and that direct systemic antibiotic treatment in neonates is both technically challenging and potentially harmful, we opted for a maternal antibiotic treatment approach. Pregnant females received a combination of 4 broad-spectrum antibiotics in their drinking water beginning at embryonic day 15, ensuring that neonates would be exposed indirectly through maternal milk.³¹ This strategy allowed us to safely and effectively modulate the neonatal microbiota. As shown in Fig. 10A and B, cultures of whole gut content from 3-wk-old pups from litters born to antibiotic-treated females failed to yield any bacterial growth under anaerobic conditions, confirming a robust depletion of gut microbiota. In contrast, age-matched litters from females that received plain water displayed normal levels of culturable bacteria, as indicated by colony-forming unit counts in fecal samples (Fig. 10B). Next, to measure the impact of microbiota depletion on hepatic B cell development, we immunophenotyped liver leukocytes from 3-wk-old pups. Strikingly, pups from antibiotic-treated litters exhibited a marked reduction in specific hepatic B cell subsets—including pre-B, B1, and immature B cells—when compared with age-matched control animals with an intact microbiota (Fig. 10C–H;  Fig. S4). In contrast, the frequencies of mature B2 cells and mature B cells remained largely unaltered, suggesting a selective dependence of early B cell developmental stages on microbiota-derived signals (Fig. 10C–H). These changes were especially notable given our prior observations that hepatic B1 cells undergo a marked expansion during the third week of life, suggesting a developmentally regulated and microbiota-sensitive window of B cell maturation in the liver. A summary of all alterations on liver B cell compartment due to antibiotic treatment can be seen in Fig. 10I.

The significant loss of B1 cells, known to contribute to natural antibody production, led us to hypothesize that microbiota depletion might impair systemic immunoglobulin levels in early life. To directly test this, we quantified serum immunoglobulins in 3-wk-old pups using ELISA. Consistent with the reduction in B1 and immature B cells, neonates from antibiotic treated dams displayed a significant decrease in serum IgM and IgG levels, while IgA levels remained unchanged. These findings indicate that IgM- and IgG-producing B cells are particularly reliant on gut microbiota–derived signals that likely reach the liver during a critical developmental window. Together, these results reveal that early-life gut microbial cues are essential for the establishment of functional hepatic B cell compartments and for mounting effective humoral responses in the neonatal period.