Rebuilding Human Germinal Center B Cells In Vitro: From Antibody Training Ground to a Manipulable Research Platform

Human peripheral-blood B cells can be guided toward an induced germinal center B-like state through sequential signaling. Illustration: LuTra Studio.

Executive Summary

This article introduces the paper “In vitro induction of human germinal center B cells,” published by David G. Priest, Melike Fusun Demir, Wataru Ise, and James B. Wing in Science Immunology on September 18, 2026. Starting with human peripheral-blood B cells, the researchers used sequential CD40 ligand (CD40L), interleukin-4 (IL-4), and interleukin-21 (IL-21) signaling to establish a two-step culture system that generates induced germinal center B-like cells (iGCB cells), providing a controllable and experimentally accessible model for studying human germinal center biology.

When the human body encounters a pathogen for the first time, the antibodies produced during the early response are often not yet fully optimized. B cells must undergo mutation, competition, and selection within germinal centers in lymph nodes, tonsils, and other lymphoid tissues. This process gradually produces antibodies with stronger antigen binding, along with memory B cells and plasma cells that can support longer-term immunity.

The challenge is that human germinal center B cells are located primarily inside lymphoid tissues. They are far less accessible than cells circulating in peripheral blood and are difficult to manipulate experimentally. As a result, much of the field has relied on mouse models, surgically obtained tonsil samples, organoids, or complex three-dimensional culture systems.

The importance of this study goes beyond the appearance of a few germinal center-associated markers. Using cytometry by time-of-flight (CyTOF), Cellular Indexing of Transcriptomes and Epitopes by Sequencing (CITE-seq), and single-cell B-cell receptor sequencing (scBCR-seq), the authors showed that iGCB cells reproduce core protein and transcriptional features of human tonsillar germinal center B cells. The cultures also contained dark zone-like and light zone-like states and showed activation-induced cytidine deaminase (AID)-dependent somatic hypermutation (SHM).

This is not yet a complete artificial germinal center. The system does not reproduce the spatial organization of lymphoid tissue, antigen-specific selection, or repeated long-term cycles of affinity maturation. A more accurate interpretation is that the researchers developed a simplified and experimentally accessible model of human germinal center B-cell biology using cells obtained from peripheral blood.

What Question Did This Study Address?

The study addressed a deceptively simple question: Can researchers start with B cells from human blood and generate cells in vitro that faithfully reproduce the core features of germinal center B cells?

Previous studies have used cytokines to stimulate B-cell proliferation, induce antibody class switching, and generate antibody-secreting plasmablasts. Some culture systems have also induced BCL6 and other germinal center-associated markers. However, the expression of one or two markers does not demonstrate that a cell has activated the broader germinal center program.

The advance in this study is that the authors did not stop at asking whether the cells looked similar. They also examined whether these cells activated the functions and molecular programs expected of germinal center B cells.

For General Readers: The Germinal Center Is an Antibody Training and Selection Site

After infection or vaccination, a subset of activated B cells enters germinal centers located in lymph nodes, the spleen, tonsils, and other lymphoid tissues. Germinal centers do more than expand B-cell numbers. They provide an environment in which antibodies can be modified, tested, and selected.

One way to understand the germinal center is to think of it as an antibody training ground.

B cells accumulate mutations in their immunoglobulin variable-region genes, generating many related versions of an antibody with slightly different binding properties. These cells then compete for survival. B cells that capture antigen more effectively and receive sufficient T-cell help are more likely to survive and continue. After repeated cycles, the immune system can retain cells that recognize the antigen more effectively.

This is one reason antibody responses can improve in quality over time after infection or vaccination, rather than merely increasing in quantity.

Dark Zone and Light Zone: Generating Variation and Selecting the Winners

A mature germinal center is broadly divided into a dark zone (DZ) and a light zone (LZ).

In the dark zone, B cells proliferate rapidly. AID drives somatic hypermutation, introducing mutations into antibody variable-region genes.

In the light zone, B cells use their mutated B-cell receptors (BCRs) to compete for and capture antigen. They then present antigen-derived peptides to T follicular helper (Tfh) cells. B cells that capture antigen effectively and receive sufficient T-cell help may return to the dark zone for another round of mutation. Others may leave the germinal center and differentiate into memory B cells or plasma cells, including long-lived plasma cells capable of sustained antibody secretion.

This process is not a single on-off switch. It is an iterative cycle of variation, competition, and selection.

In the dark zone, B cells proliferate and undergo somatic hypermutation. In the light zone, they compete for antigen and Tfh-cell help before recycling or differentiating. Illustration: LuTra Studio.

Why Are Human Germinal Center B Cells So Difficult to Study?

Peripheral blood is relatively easy to collect, but active germinal centers are located mainly inside lymphoid tissues. Direct access to human germinal center B cells generally requires tonsil or lymph-node surgical specimens, or fine-needle aspiration (FNA).

These approaches allow investigators to observe genuine human immune responses, but they are less suited to repeatedly adding or removing one signal at a time and then measuring how that change alters B-cell fate.

Mouse models, tonsil organoids, and three-dimensional hydrogel cultures each provide important advantages. Animal models, however, do not fully reproduce human biology. Organoids better preserve multicellular interactions but are more complex. Three-dimensional cultures can provide spatial and matrix-derived cues, yet they do not always reproduce the full transcriptional program of germinal center B cells.

A system that starts with human blood, uses a limited number of defined components, and allows signals to be added or removed at specific times therefore fills an important gap between existing models.

The Main Finding: The Sequence of the Signals Matters

The core signals used in the system were multimeric CD40L (mCD40L), IL-4, and IL-21.

CD40L models a contact-dependent signal that B cells normally receive from helper T cells. IL-4 and IL-21 are cytokines involved in B-cell activation, fate determination, and antibody responses.

The researchers first stimulated naive B cells with mCD40L and IL-4 for four days. They then washed the cells and replaced the initial condition with a second stimulation using mCD40L and IL-21. This sequence caused a large fraction of cells to up-regulate B-cell lymphoma 6 (BCL6) and develop an iGCB phenotype characterized by high BCL6 expression together with CD38, AID, and other germinal center-associated features.

Reversing the order, or continuing IL-4 during the second phase, produced a substantially different result. IL-4 was therefore not simply beneficial in proportion to its concentration or duration. It supported early priming, but it had to be removed before IL-21 could drive cells more fully into the iGCB state.

The result is an important reminder that cell fate depends not only on which signals are present, but also on when they arrive and how long they remain available.

The two-step workflow uses CD40L plus IL-4 for early priming, followed by washing and CD40L plus IL-21 to induce iGCB cells for single-cell analysis. Illustration: LuTra Studio.

How Did the Researchers Show That These Cells Resemble Germinal Center B Cells?

The authors did not rely on BCL6 alone. They used several complementary layers of evidence.

First, CyTOF showed that iGCB cells shared a marker profile with tonsillar germinal center B cells, including BCL6, CD38, AID, high peanut agglutinin reactivity, and low CD45RB.

Second, CITE-seq measured single-cell RNA and surface proteins together. iGCB cells expressed germinal center-associated genes including BCL6, AICDA, MEF2B, MYBL1, LMO2, and S1PR2. They lacked key programs associated with memory, atypical B cells, and plasmablasts, represented by genes such as KLF2, ZEB2, and PRDM1.

Third, the culture contained dark zone-like and light zone-like populations. Dark zone-like cells expressed higher levels of BCL6 and AICDA and showed proliferative features. Light zone-like cells expressed genes associated with positive selection, including MYC, BATF, BCL2A1, and CD83.

Fourth, and most importantly, scBCR-seq showed that iGCB cells accumulated somatic mutations. These mutations were enriched in sequence motifs recognized as AID hotspots. Lineage-tree analysis also showed ongoing diversification within individual B-cell clones. Together, these findings indicate that the culture did more than reproduce a superficial marker pattern: functional SHM machinery was active.

The iGCB identity was supported by converging protein, transcriptomic, cell-state, and functional mutation evidence. Illustration: LuTra Studio.

Experimental Results: From Signal Screening to Functional Mutation Evidence

The results can be organized into six connected experiments. The investigators first identified a signaling sequence capable of initiating a germinal center program, then progressively tested cell identity, function, developmental potential, and experimental manipulability.

The first experiment compared how different stimuli affected naive B cells. CyTOF tracking on days 2, 4, and 7 showed that IL-21 drove stronger cell division, IL-4 produced intermediate proliferation, and IFN-γ inhibited proliferation. The cytokines also generated distinct phenotypes: IL-4 strongly increased CD23, IL-21 more readily promoted plasmablast formation, and IFN-γ induced T-bet and CXCR3. The signals therefore changed more than the rate of expansion; they directed the cells toward different fates.

The second experiment demonstrated that signal order was critical for producing iGCB cells. When cells received mCD40L plus IL-4 for four days, were washed on day 4, and then received mCD40L plus IL-21, BCL6 expression rose markedly on day 5 and persisted through day 8. Starting with IL-21 and switching to IL-4, or retaining IL-4 during the second phase, did not produce the same outcome. Temporal analysis showed that cells on days 4 and 5 remained in a CD23-positive pre-iGCB state. CD38 began to rise on day 6, and a prominent BCL6-high, CD38-positive, AID-positive iGCB population appeared on days 7 and 8. Some cells simultaneously moved toward a CCR6-positive memory-like fate, revealing continued fate divergence within the culture.

The third experiment tested whether the induced cells retained developmental plasticity. Using a CD20-positive, CD23-low, CCR6-negative, CD21-positive, CD38-positive sorting strategy, the investigators enriched the culture to approximately 76% iGCB cells. Restimulation with IL-21 drove many of these cells toward plasmablast differentiation. Reintroducing IL-4 reduced BCL6 and restored a CD23-positive, CCR6-negative phenotype. After additional rounds of stimulation, cultures could regenerate iGCB cells and contained a higher proportion of IgG-positive class-switched cells at later time points. The induced state was therefore not a fixed terminal phenotype; subsequent signals could redirect cell fate.

The fourth experiment directly compared cultured cells with human tonsillar B cells in the same analytical framework. The CITE-seq dataset contained 31,544 quality-control-passed cells, including 3,116 cells from tonsil samples. iGCB and tonsillar GCB cells shared core germinal center features, including CD38, MME, BCL6, MEF2B, MYBL1, LMO2, AICDA, HMCES, UNG, MSH2, and S1PR2. The cultures also separated into a dark zone-like population with higher BCL6, AICDA, and proliferation signals and a light zone-like population expressing MYC, BATF, BCL2A1, and CD83. Parallel CyTOF analysis showed that iGCB cells carried a BCL6-positive, AID-positive, PNA-positive, CD45RB-low protein phenotype resembling tonsillar GCB cells. The two populations did not completely overlap, however, reflecting differences between an in vitro culture and a lymphoid-tissue environment.

The fifth experiment tested the defining functional question: whether the cells actually underwent somatic hypermutation (SHM). The investigators cultured FACS-purified naive B cells, which were expected to begin with largely unmutated receptors, for nine days and analyzed them by scBCR-seq. Most cells remained unmutated, as expected for a short culture, but the fraction of mutated cells was significantly higher in iGCB and plasmablast clusters than in the in vitro memory-like cluster. Mutations were concentrated in the AID-preferred WRC/GYW hotspot motifs, and lineage trees showed continued diversification within clones spanning light zone-like and dark zone-like clusters. Tonsillar GCB cells still carried more SHM than iGCB cells, consistent with weeks of in vivo evolution compared with nine days in culture. Adding the cell-cycle inhibitor palbociclib removed Ki67-positive proliferating clusters but did not further increase the proportion of mutated cells.

The sixth experiment established the platform's value for studying cell-of-origin effects and inhibitory mechanisms. CyTOF experiments across five donors showed that naive B cells formed iGCB cells most efficiently. cMZP and cMZ cells retained this capacity at progressively lower frequencies and produced more plasmablasts, whereas classical class-switched memory B cells differentiated almost entirely into plasmablasts. CITE-seq across four donors further showed that, once cells entered the iGCB state, their phenotype was similar regardless of whether they originated from naive, cMZP, or cMZ populations. In a separate four-donor perturbation experiment, IFN-γ, IL-2, the BCL6 inhibitors FX-1 and 79-6, and the EZH2 inhibitor tazemetostat all significantly reduced the iGCB fraction, but each redirected the remaining cells toward a different alternative fate.

Together, these six experiments form a coherent evidence chain: sequential signaling produces a reproducible iGCB phenotype; the cells resemble human tonsillar GCB cells at both protein and transcriptomic levels; they activate AID-directed SHM; and their formation and subsequent fate can be redirected by cytokines and small molecules.

For Specialists: What Did the iGCB Platform Actually Establish?

From a technical perspective, the study established a minimal, feeder-free, two-step system for inducing human iGCB cells.

The standard workflow started with naive B cells isolated from peripheral blood. The first phase used mCD40L at 150 ng/ml and IL-4 at 50 ng/ml. On day 4, the cells were washed and moved into a second phase containing mCD40L and IL-21. Most of the major analyses were performed between days 7 and 9. The authors also showed that CD40L feeder cells could generate a similar phenotype, but soluble multimeric CD40L avoids the ongoing cell-line culture, passage, and irradiation conditions that can introduce practical variability.

One strength of the work is that cultured cells and adult tonsillar B cells were compared within the same single-cell analytical framework rather than being described through unrelated marker panels. The investigators integrated CITE-seq, CyTOF, gene-signature scoring, projection onto public datasets, and BCR repertoire analysis. This allowed benchmarking across protein expression, transcriptional programs, cell states, and functional mutation.

iGCB cells did not completely overlap with ex vivo tonsillar germinal center cells. Tonsil-derived cells showed stronger immediate-early gene and stress signatures, which may partly reflect tissue dissociation. The cultured iGCB cells were more enriched for cholesterol biosynthesis, ribosomal genes, and antioxidant pathways, potentially reflecting adaptation to a nutrient-rich and relatively hyperoxic culture environment. These differences do not mean that the core germinal center program was absent from iGCB cells.

The Timing of IL-4 and IL-21 Matters More Than the Presence of Either Cytokine Alone

One of the most interesting biological findings is the requirement for sequential signaling.

mCD40L plus IL-4 first established a pre-iGCB state. IL-4 was then withdrawn and IL-21 was added, producing a much larger population of BCL6-high iGCB cells. When IL-4 remained present, BCL6 declined and the cells were more likely to move toward alternative fates. CD23 expression also changed according to the recency of IL-4 exposure, leading the authors to describe it as a molecular timer of IL-4 signaling history.

This is consistent with the idea that germinal centers are not formed by exposing cells to a uniform mixture of cytokines. In vivo, B cells migrate, contact different cell types, and encounter signals at different times and locations. Although this two-dimensional culture lacks genuine spatial organization, the timed sequence of signals partially reconstructs this changing environment.

It is also notable that BCR cross-linking with anti-IgM could not replace CD40L and did not meaningfully increase iGCB formation. At least in this polyclonal system, T cell-derived CD40 signaling remained essential.

Which B-Cell Populations Can Enter the iGCB State?

The investigators compared naive B cells, circulating marginal zone precursor B cells (cMZP), circulating marginal zone B cells (cMZ), early switched memory B cells, and classical switched memory B cells.

Naive B cells generated iGCB cells most efficiently. cMZP and cMZ populations retained the ability to enter the iGCB state, but at lower frequencies and with greater plasmablast production. In contrast, class-switched memory B cells differentiated almost exclusively toward plasmablasts.

These findings suggest that antibody class switching marks an important functional boundary. Although the efficiency of iGCB formation differed among the starting populations, naive-derived and unswitched-memory-derived cells became phenotypically similar once they entered the iGCB state.

For the study of human B-cell heterogeneity, this provides direct functional evidence that circulating marginal zone and unswitched memory compartments have not completely lost the capacity to enter a germinal center-like program.

Inhibition Experiments Show the Platform’s Value for Mechanistic and Drug Studies

During the second phase, the authors added individual cytokines or small molecules to determine whether iGCB formation could be redirected or inhibited.

Interferon-gamma (IFN-γ) diverted nearly all cells toward a T-bet-positive, CXCR3-positive phenotype. Interleukin-2 (IL-2) increased memory-like populations and plasmablasts. The BCL6 inhibitors FX-1 and 79-6, as well as the EZH2 inhibitor tazemetostat, significantly reduced the proportion of iGCB cells, although they did not redirect the cells toward identical alternative fates.

This is where the platform becomes particularly useful. Investigators can add a compound at a defined time, remove a cytokine, or compare donor-specific responses while directly measuring how each intervention changes human B-cell fate.

Potential applications include research on vaccine adjuvants, immunodeficiencies, autoimmune disease mechanisms, B-cell lymphoma pathways, and interindividual differences in the capacity to induce a germinal center response. These are future opportunities for the platform, not clinical applications demonstrated by the present study.

What Is Still Missing From the Model?

First, the system lacks genuine spatial organization. It contains light zone-like and dark zone-like states, but it does not reproduce the anatomical zones created by chemokines, follicular dendritic cells, Tfh cells, and coordinated cellular migration in lymphoid tissue.

Second, it does not establish antigen-specific selection. Although the cells undergo AID-dependent SHM, they do not compete according to their affinity for a defined antigen. The system therefore cannot be considered a complete model of affinity maturation.

Third, the in vitro experiment lasted only about nine days, and the mutation burden remained lower than that of tonsillar germinal center B cells that had participated in sustained responses for several weeks. The system also did not maintain repeated light zone-dark zone cycles over the long term. With prolonged IL-21 exposure, some cells shifted toward a plasmablast fate.

Fourth, several characteristic germinal center B-cell genes, including RGS13 and SERPINA9, remained lower in iGCB cells. CD40L, IL-4, and IL-21 were sufficient to activate much of the core program, but they were not sufficient to reproduce the complete tissue-resident germinal center niche.

The iGCB system is therefore best suited for rapid and controlled signaling and perturbation studies. When the research question centers on antigen-driven selection, multicellular interactions, or long-term affinity maturation, tonsil organoids and more complex three-dimensional models remain essential. These platforms should be viewed as complementary rather than interchangeable.

The current iGCB model enables controlled perturbation studies but does not yet reproduce spatial organization or antigen-driven selection. Future systems may integrate Tfh cells, follicular dendritic cells, microfluidics, and repeated germinal center cycling. Illustration: LuTra Studio.

Jason’s Takeaway: The Real Advance Is Turning Complex Biology Into a Testable System

To me, the most important message is not the simplified headline that “three factors can generate germinal center B cells.” The real achievement is the construction of an intermediate model that preserves a meaningful portion of the underlying biology while remaining experimentally manipulable.

If a model is too simple, it may reproduce only a few markers while losing the relevant function. If it is too complex, it becomes difficult to determine which signal produced the observed outcome. This iGCB culture narrows the gap between those two extremes.

The requirement for IL-4 to appear first, then be removed before IL-21 drives the next stage, is also a reminder that immune engineering and cell-therapy development cannot be reduced to a list of ingredients. Signal order, exposure duration, and the state of the cell at the time of stimulation can be equally important.

The next step may be to add antigen-specific selection, controllable Tfh-cell interactions, follicular dendritic cell signals, or microfluidic and organoid engineering that recreates distinct spatial regions. These additions could move the field from inducing GC-like cells toward a programmable human antibody-maturation platform.

At this stage, however, the most accurate interpretation remains straightforward: this is not a complete artificial germinal center, but it is an important new tool for dissecting the formation of human germinal center B cells using an accessible peripheral blood starting material.

About LuTra Studio

LuTra Studio works at the intersection of biotechnology, drug development, translational research, and U.S.-Taiwan life-science collaboration. We provide scientific strategy, technology assessment, scientific communication, and market-entry consulting services.

If your team is evaluating an immunology research platform, a cell or gene therapy technology, or needs to translate complex scientific findings into content that industry partners and investors can understand, please contact LuTra Studio.

References

1. Priest DG, Demir MF, Ise W, Wing JB. In vitro induction of human germinal center B cells. Science Immunology. 2026;11:eaeg5223. Published September 18, 2026. https://doi.org/10.1126/sciimmunol.aeg5223

2. De Silva NS, Klein U. Dynamics of B cells in germinal centres. Nature Reviews Immunology. 2015;15:137–148. https://doi.org/10.1038/nri3804

3. Victora GD, Nussenzweig MC. Germinal centers. Annual Review of Immunology. 2012;30:429–457. https://doi.org/10.1146/annurev-immunol-020711-075032

4. Wagar LE, et al. Modeling human adaptive immune responses with tonsil organoids. Nature Medicine. 2021;27:125–135. https://doi.org/10.1038/s41591-020-01145-0