{"paper_id":"2ee6db52-d2a2-482a-bc2e-5d51b6da20d1","body_text":"1 \nCytokines control the physical state of immune tissue \nDebraj Ghose1, Thomas C. Ferrante1, Donald E. Ingber1-3* \n1Wyss Institute for Biologically Inspired Engineering, Harvard University, Boston, MA 02215, USA \n2 Vascular Biology Program and Department of Surgery, Boston Children's Hospital and Harvard Medical \nSchool, Boston, MA 02115, USA \n3 Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Boston, MA \n02134, USA \n \n*Corresponding author: Donald E. Ingber MD, PhD (Email: don.ingber@wyss.harvard.edu) \n \n Unlike most solid -like tissues, immune tissue is protean and reconfigurable—its component \ncells can patrol vast territories, find rare cellular partners, and dynamically self -organize into \nfunctional structures like germinal centers or tertiary lymphoid structures. The ability of \nimmune cells to collectively behave as an active material that switches physical state is essential \nfor adaptive immunity. Yet the principles governing the control of the physical properties of \nimmune tissues, and how chemical and mechanical forms of control interplay remain obscure. \nInspired by minimal immune cell components that recapitulate germinal center (GC) dynamics, \nwe developed an agent-based model that predicted that a tissue’s ability to program its own \nphysical properties governs the efficiency of antibody evolution. This motivated our search for \nbiomolecular forms of control capable of tuning this material property. Using a tractable in vitro \nmodel, we found that CD40L -stimulated human B cells self -organize into active, liquid -like \ncondensates that round, fuse, and internally mix.  GC cytokines IL -4 and IL -21 tuned this \neffective temperature, driving phase transitions from cohesive, liquid-like states to dispersed, \ngas-like morphologies. These findings demonstrate that cytokines control tissue fluidity , \nsuggesting a previously unrecognized biophysical form of immune regulation. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n2 \nSignificance Statement \nThe body's immune system is uniquely flexible, allowing cells to organize into temporary structure s, like \ngerminal centers  or tertiary lymphoid structures, to fight intruders.  How this physical flexibility is \ncontrolled is a major question. We discovered that the physical state of immune tissue is chemically \nprogrammable. We show that self-organized multicellular B cell structures behave like liquid droplets, and \nwe used this framework to demonstrate that signals (cytokines) can modulate their physical properties.  \nThis finding identifies a new biophysical axis of immune control, suggesting the body actively tunes its \ntissue fluidity to optimize the immune response. \nIntroduction \nBiological tissues are active living materials, whose physical states affect emergent function. The physical \nprinciples of liquid -like systems influence function at multiple biological scales, from the dynamics  of \nintracellular biomolecular condensates that organize the cell's interior (1) to the self-organization of cells \ninto tissues that drive morphogenesis in the embryo  (2, 3) . However, while most  developing tissues \ntransition from liquid-like to solid-like states as they mature (4), immune tissue is a notable exception: it \nmust remain perpetually fluid and protean to carry out its functions of surveillance and response. \n This lifelong fluidity is exemplified in  germinal center s (GCs) , which can form trans iently in native \nsecondary lymphoid organs, such as lymph nodes (5), or ectopic tertiary lymphoid structures that can form \nat cancer or chronic inflammation sites (6–10). In GCs, B cells undergo Darwinian selection to enrich for a \nsubset of B ce lls that produce high-affinity antibodies (11, 12). Live imaging has shown GCs to be highly \ndynamic  tissue structures that undergo constant cellular rearrangements (13, 14). This is likely because \nthe efficiency of antibody selection depends on how quickly B cells can find antigens, compete for them, \nlocate helper T cells , and shuttle between light and dark zones —all processes fundamentally limited by \ncellular reorganization dynamics and tissue material properties . The regulation of these dynamics is \ncomplex and not fully understood. While B cells can be steered when migrating by chemokine gradients \n(15) and \"decelerate\" during direct physical tethering to  T follicular helper  (TFH) cells (16), it remains \nunknown if chemical cues could also tune the intrinsic way in which B cells interact with their neighbors \nin three  dimensions. Such a mechanism would represent a distinct and higher-level mode of control, \nprogramming the material properties of the tissue by altering multicellular interaction dynamics. \nTo address th e possible existence of this type of physical mechanism of controlling immune cell self -\nassembly, we first developed a computational model to test the theoretical importance of such regulation \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n3 \nwithin the context of the GC. We then established a simplified experimental system to provide a direct, \nproof-of-principle test for whether key immune signals can, in fact, reprogram the physical properties of \nimmune tissue. Our findings reveal a previously unrecognized layer of biophysical control in the immune \nsystem, where the same signals that dictate immune cell fate also program the active material state of the \nmulticellular collective. \nResults \nTissue fluidity as a control parameter for B cell evolution \nThe GC is a transient yet highly organized microanatomical niche within lymphoid organs that \nfunctions as an evolutionary machine for B cells  (5, 17 –20). Within these structures, B cells \nundergo iterative cycles of mutation, proliferation, and affinity-based selection through dynamic \ninteractions with T follicular helper (TFH) cells and follicular dendritic cells (FDCs) that present \nantigen (Fig. 1A ) (12, 13, 17, 20 –23). A rich body of theoretical and computational work has \nprovided invaluable insights into the cellular logic and chemokine -driven dynamics of the GC \nreaction (24–28). Building on this foundation, we sought to first establish the theoretical \nimportance of the tissue's collective physical properties —a factor distinct from individual cell \nbehaviors. To test the potential link between the physical state of the GC environmen t and the \nefficiency of B cell evolution, we developed a minimal spatiotemporal agent -based model that \ncaptures the essential features of a GC multicellular reaction. \nOur computational model populates a hexagonal lattice with B cells in various differentiation \nstates and TFH cells, each simulated as individual agents. The model implements the core GC \ncycle: B cells in the light zone compete for limited antigen based on their receptor affinity—with \nhigher-affinity cells having a competitive advantage (18, 29, 30) . Successfully acquiring antigen \nand receiving help from TFH cells triggers B cells to migrate to another region of the GC (dark \nzone), where they proliferate and mutate their receptors before returning to the light zone for \nanother round of selection. This iterative process progressively enriches for high-affinity clones. \nIn our computational model of the GC, t he behavior of each agent is governed by two coupled \nrule sets. Physical rules dictate tissue rearrangement through Kawasaki Ising dynamics (31–33), \nwhere adjacent agents attempt to swap positions with probability determined by the change in \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n4 \nsystem energy. Specifically, the probability of accepting a swap is given by 𝑃𝑠𝑤𝑎𝑝 =\nmin⁡(1, 𝑒𝛥𝐸/𝑇𝑒𝑓𝑓)  (swaps that decrease energy are always accepted). The change in system \nenergy incorporates multiple contribution s: Δ𝐸 = Δ𝐸𝑏𝑎𝑠𝑒𝑙𝑖𝑛𝑒 + Δ𝐸𝑐ℎ𝑒𝑚𝑜𝑡𝑎𝑥𝑖𝑠 + Δ𝐸𝑎𝑑ℎ𝑒𝑠𝑖𝑜𝑛 , \nwhere baseline energy represents the cost of movement, chemotactic energy drives zone-specific \nmigration toward CXCL13 (light zone) or CXCL12 (dark zone) sources, and adhesion energy \npromotes cell-cell cohesion. Biological rules control the GC reaction through probabilistic events: \nantigen acquisition occurs competitively with probability proportional to local CXCL13 \nconcentration, state transitions between B cell subtypes occur at defined rates, cells divide with \nmutation in regions enriched for CXCL12, and cells undergo apoptosis stochastically. \nThe overall material state of this simulated tissue emerges from the interplay of these physical \nand biological processes. While we hold biological rates and interaction strengths constant, we \ncan tune the system's effective temperature (𝑇𝑒𝑓𝑓)—the parameter in the Metropolis acceptance \ncriterion that scales the energetic penalties for movement and thus controls the tissue's collective \nfluidity. \nWe first confirmed that the model recapitulates fundamental Darwinian selection. Over \nsimulated time, the B cell population progressively evolves toward higher affinity as low -affinity \nclones are outcompeted and eliminated (Fig. 1B, C). While not designed for precise quantitative \npredictions, the model serves as a powerful tool for testing qualitative hypotheses about GC \ndynamics. \nWe then used the model to test our central hypothesis: does the physical state of the tissue \nimpact selection efficiency? Running multiple simulations while systematically varying the \neffective temperature revealed that, at very low temperatures when cells are less motile, affinity \nmaturation proceeds slowly and inefficiently. As temperature increases, selection efficiency \nimproves dramatically, reaching an optimal regime before declining again at very high \ntemperatures, where the system becomes too disordered (Fig. 1D). \nThis theoretical finding, that B cell evolution is sensitive to the system's effective temperature, is \nsignificant. In biological systems refined by natural selection, such parameters are rarely left \nunregulated. Our model's prediction thus identifies tissue fluidity as a potential control knob for \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n5 \nGC reactions, motivating our subsequent search for biological mechanisms capable of modulating \nthe material state of immune cell collectives. \nB cells stimulated by CD40L form multicellular condensates \n Our model's prediction that B cell evolution is sensitive to effective temperature identifies it as a \nhypothetical control parameter in the GC multicellular organization reaction. However, the model \ndoes not prove that such regulation exists, only that it would be functionally important if the \nimmune system evolved a way to control it. Therefore, the following experiments were designed \nspecifically to test the hypothesis that such biological control knobs exist, by isolating the direct \neffect of cytokines on the material state of a human immune cell collective. \nAddressing this problem within a live GC is currently intractable. In that environment, the tight \ncoupling of chemical signaling with physical cell -cell contact and the simultaneous triggering of \ncomplex cell fate programs  (e.g., proliferation and selection) make it challenging to isolate the \ndirect effect of any single cytokine on the material state of the entire tissue. Instead, we required \na simplified system where we could directly measure the material state of a n immune  cell \ncollective in response to s pecific chemical cues. We therefore leveraged a well -described \nphenomenon where stimulating CD40 receptors in primary human B cells, which mimics the \ninitial TFH engagement, causes them to spontaneously self-organize into multicellular aggregates \nin suspension culture (Fig. 2A) (34–36). While these aggregates are known to form, their physical \nproperties are poorly understood. \nTo adopt this simplified in vitro model system, we first set out to characterize its physical nature \nwithin the framework of soft matter physics. Given that germinal centers are known to be highly \ndynamic sites of constant cellular rearrangement, we hypothesized that these self -organized \nhomotypic B cell aggregates may also rest within a similar liquid-like physical state. To explore this \npossibility theoretically, we used a simplified computational model that considers only a single \ncell type and homotypic  adhesion, representing the aggregates formed by B cells. This model \npredicts a phase diagram where the collective state of the cells is governed by the interplay \nbetween cell-cell adhesion energy (Δ𝐽𝑎𝑑ℎ𝑒𝑠𝑖𝑜𝑛) and an effective temperature (𝑇𝑒𝑓𝑓) that controls \ncell motility ( Fig. 2B). At the extremes, low temperatures and high adhesion promote solid -like \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n6 \ncrystalline states, while high temperatures and low adhesion lead to dispersed, gas -like states. \nBetween these regimes, the model predicts the existence of liquid -like condensates, which are \ncharacterized by their ability to exhibit rounding behavior and internal cellular rearrangements.  \nTo experimentally characterize multicellular B cell structures in the context of the phase diagram \nin Fig. 2B, we used multimerized CD40L to induce B cell aggregate formation and imaged their \ndynamics over time using live -cell microscopy (37). Tracking the boundaries of these structures \nrevealed that irregularly shaped structures rounded over time, a hallmark of surface -tension-\ndriven rounding observed in liquid droplets (Fig. 2C). This rounding behavior was quantified by a \nsteady increase in circularity over time, approaching a fluid equilibrium state rather than a \njammed solid one ( Fig. 2D ), suggesting that multicellular B cell structures behave like liquid \ndroplets. \nAnother defining characteristic of a liquid is internal fluidity, which allows constituent  \ncomponents to mix and rearrange. To test this, we prepared two populations of B cells labeled \nwith distinct fluorescent dyes. When co -cultured, aggregates of different colors readily fused \nupon contact, and their constituent cells thoroughly intermixed over several hours, confirming \ntheir liquid -like fluidity ( Fig. 2E and Supp. Video 1 ). This internal mixing was confirmed by a \ndecrease in the coefficient of variation over time, again consistent with a liquid-like system (Fig. \n2F). \nTaken together, direct experimental  evidence of both effective surface tension and internal \nfluidity establishes that stimulated B cells self -organize into active liquid-like droplets. We term \nthese structures “B-Lymphocyte Organized Multicellular Blobs ” (BLOMBs) for simplicity of \ndescription. This characterization  provided us with the ideal platform to test  our central \nhypothesis that TFH cytokines which are known to alter B cell state in the GC may also tune the \nmaterial properties of immune tissue. \nTFH-derived cytokines tune the physical state of BLOMBs \nHaving established BLOMBs as multicellular condensates, we next sought to identify the biological \ncontrol knobs capable of tuning their physical state. Our computational model predicted that \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n7 \ndecreasing Δ𝐸/𝑇𝑒𝑓𝑓  by increasing 𝑇𝑒𝑓𝑓  would decrease the circularity of multicellular \ncondensates (Fig. 3A). This provided a clear visual signature to look for in our experiments. \nThus, we sought to identify biological factors that could induce this predicted physical change —\nnamely, a transition to a more dispersed, less circular state. Our BLOMB system is formed by \nstimulating B cells with CD40L, which mimics the survival signal B cells receive from TFH cells (Fig. \n1A). We reasoned that the other key signals delivered by TFH cells during this interaction might \nserve as the 'control knobs' capable of tuning the collective's physical state. We therefore focused \non IL -4 and IL -21, two canonical TFH -derived cytokines that are known to act in concert with \nCD40L to direct B cell fate, driving proliferation, survival, and differentiation (38–43). The signaling \npathways downstream of the IL-4 and IL-21 receptors are known to intersect with pathways that \nregulate the actin cytoskeleton , contractility, and cell adhesion —the very components that \ngovern a cell's mechanical properties and biophysical interactions with other cells  (44–47). We \ntherefore hypothesized that these potent, fate-directing signals might also function to modulate \nthe collective physical state.  \nTo test this hypothesis, we perturbed BLOMBs from two donors with IL-4 (2.5 μg/mL) and IL -21 \n(1.25 μ g/mL). These saturating concentrations were chosen to ensure maximal signaling and \nprovide a clear proof-of-principle test for whether these cytokines could modulate the collective's \nphysical state. Imaging revealed striking morphological changes: while BLOMBs treated with \nCD40L alone or with IL -4 remained cohesive, the addition of IL -21 induced a transition to a \ndispersed, gas-like state. This was quantified by a significant decrease in aggregate circularit y, a \nphenotype that was amplified when both cytokines were present ( Fig. 3B). This cytokine-driven \nshift from a cohesive 'liquid -like' to a dispersed 'gas -like' state phenocopies the effect of  \ndecreasing Δ𝐸/𝑇𝑒𝑓𝑓⁡ in our model (Fig. 3A). \nTo further investigate the system dynamics, we compared time-lapse data from both simulations \nand experiments. Our model shows that increasing the effective temperature not only changes \nthe final state but also accelerates the dynamics towards a more dispersed configuration, as seen \nin simulation snapshots ( Fig. 3C) and quantified by a more rapid decrease in the coefficient of \nvariation (Fig. 3D ). This theoretical prediction was mirrored in our experiments. Time -lapse \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n8 \nmicroscopy revealed the structural evolution of the BLOMBs under different cytokine conditions \n(Fig. 3E). Quantification confirmed that the coefficient of variation decreased most rapidly and \nreached a lower plateau in the presence of IL -21, indicating that this cytokine increases the \nsystem's effective temperature and drives the collective toward a more fluid, gas -like state (Fig. \n3F). \nThese results demonstrate that TFH-derived cytokines can modulate the physical state of B-cell \ncollectives. By tuning  Δ𝐸/𝑇𝑒𝑓𝑓 , these signals fluidize the tissue, directly linking the chemical \nsignaling environment to the material properties our model identified as cr ucial for the GC \nreaction. \nDiscussion \nOur findings reveal a biophysical layer of control  in immune tissue regulation  where the same \ncytokine signals that determine B cell fate and promote Darwinian selection in the germinal \ncenter (GC) also control the physical state of cell collectives. We established this principle by \ncombining computational modeling with live cell imaging . First, our in silico  model, which \nsimulates the essential features of the GC, predicted that the efficiency of B cell evolution is \noptimally tuned by the system's 'effecti ve temperature'. Second, to determine if a biological \nmechanism for tuning this physical parameter exists, we experimentally perturbed a reductionist \nin vitro system of B-cell condensates, which we term B-Lymphocyte Organized Multicellular Blobs \n(BLOMBs). This in vitro experimental system is not intended to replicate the complex architecture \nof an in vivo GC. Rather, its value lies in its simplicity, which enabled a direct, proof -of-principle \ndemonstration that the TFH cytokines IL-4 and IL-21 can reprogram the material properties of a \nB cell collective, tuning its state from cohesive to dispersed form. Thus, our work establishes the \nexistence of a new form of biophysical control in immune tissues (Fig. 4). How the immune system \nprecisely leverages this tunability to optimize functions like affinity maturation in vivo remains a \nkey question for future investigation. \nThis work builds on the growing understanding that tissues can behave as active materials with \nliquid-like properties, a concept well-established in the study of embryonic development (2). The \nkey distinction, however, is that while embryonic tissues typically solidify upon maturation  (4), \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n9 \nour findings demonstrate that immune tissue not only maintains a fluid state throughout life but \nthat this state is actively and chemically programmable. This concept of a tunable material state \ndraws a powerful parallel to the regulation of intracellular biomolecular condensates  (1). \nAlthough the underlying molecular drivers and physical scales are distinct, the analogy is \ninstructive: in both systems, the collective interactions of individual components give rise to a \ncondensed phase whose material properties are tuned to regulate a specific biological function. \nWe therefore propose that immune cell assemblies are a form of programmable active matter, \ndistinct from other tissues that typically solidify during development. Having identified a core \nprinciple of this programmability in our simplified system, a key future challenge is to understand \nhow these rules apply within the complex 3D architecture of diverse lymphoid structures in vivo. \nFor example, the fluidity of the GC may be actively regulated to enhance the search-and-selection \nprocess during affinity maturation, ensuring that B cells can efficiently find antigen and connect \nwith helper T cells. \nBeyond the GC, this principle may govern the dynamics of ectopic tertiary lymphoid structures \n(TLSs) that form at sites of chronic inflammation or cancer  (6–10). The ability to tune the local \ntissue fluidity could impact immune cell infiltration, residency, and anti-tumor activity, as well as \nlead to the identification of new therapeutic targets . Similarly, this biophysical control may be \ncrucial in shaping mucosal immunity within specialized niches like inducible bronchus-associated \nlymphoid tissue or IgA -producing intestinal cryptopatches  (48, 49) . In these environments, \ncontrolling the collective state of lymphocytes could be essential for maintaining barrier integrity \nwhile allowing for rapid and flexible immune responses. This perspective shifts the focus from \npurely biochemical signaling to a model where the physical state of the tissue is itself a regulated \nvariable critical for immune function. \nThe principles uncovered in this study also suggest new avenues for the rational design of immune \ntissues. The self-organizing and physically tunable nature of BLOMBs provides a foundation for a \n\"bottom-up\" approach to tissue engineering, where cytokine inputs can be used to program the \ncollective state and geometry of a multicellular tissue construct. For example, by harnessing this \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n10 \ncytokine-mediated control over the material state, it may be possible to engineer synthetic \nlymphoid organoids ex vivo. \nSuch engineered immune tissues could serve as powerful platforms for fundamental research, \nallowing for the deconstruction of complex immune processes in a highly controlled environment. \nThey could also accelerate translational work by enabling high -throughput screening of \nimmunomodulatory drugs or serving as systems to produce  therapeutic antibodies. Looking \nforward, this work paves the way for fabricating active biological materials with prescribed and \nreconfigurable functions, hinting at a future where engineered immune tissues could be designed \nfor therapeutic applications. In short, o ur work clarifies that  immune tissue is a protean, \nprogrammable active material,  and that  tuning its physical state represents a powerful and \npreviously unrecognized axis for immunomodulation. \nMaterials and Methods \nPrimary human B lymphocytes were isolated from PBMCs from donor leukocyte collars via \nimmunomagnetic isolation and cultured as described in SI Appendix. Live-cell time-lapse imaging \nwas performed using brightfield and fluorescence microscopy, and image ana lysis used custom \nscripts in MATLAB 2024a, as detailed in SI Appendix. The agent -based and Kawasaki -Ising \ncomputational models were implemented in MATLAB 2024a as described in SI Appendix. \nAcknowledgments \nWe thank Aditya Patil, Pranav Prabhala, Min Wu, Girija Goyal, Yunhao Zhai and members of the \nIngber lab  for generously sharing reagents and techniques.  We are also grateful to Duanne \nWesemann and Harikesh Wong for valuable feedback and discussions. This work was supported \nby the Sloan Matter-to-Life program and the Wyss Institute for Biologically Inspired Engineering \nat Harvard University. \nAuthor Contributions \nD.G. conceptualized the project, designed the methodology, performed the investigation and \nformal analysis, and wrote the original draft and subsequent revisions. T.F. contributed to the \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n11 \nmethodology. D.E.I. acquired funding, contributed to conceptual discussions, and participated in \nwriting, review, and editing. \nCompeting Interests \nThe authors declare no competing interests. \nData Availability Statement \nAll primary data and analysis code supporting the findings of this study are publicly available on \nGitHub at https://github.com/DebrajGhose/cytokine_control_blomb. Further information and \nrequests for resources or materials should be directed to the corresponding author (D.E.I.).\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \n \n \nReferences \n \n1.  C. P . Brangwynne, et al., Germline P Granules Are Liquid Droplets That Localize by Controlled \nDissolution/Condensation. Science 324, 1729–1732 (2009). \n2.  M. S. Steinberg, The problem of adhesive selectivity in cellular interactions. Cellular \nMembranes in Development 321–434 (1964). \n3.  P. L. Townes, J. Holtfreter, Directed movements and selective adhesion of embryonic \namphibian cells. (1955). https://doi.org/10.1002/jez.1401280105. \n4.  A. Mongera, et al. , A fluid -to-solid jamming transition underlies vertebrate body axis \nelongation. Nature 561, 401–405 (2018). \n5.  W. Flemming, Beiträge zur Kenntniss der Zelle und ihrer Lebenserscheinungen. Archiv für \nmikroskopische Anatomie 16, 302–436 (1879). \n6.  I. Kobayashi, M. Ziff, Electron microscopic studies of lymphoid cells in the rheumatoid \nsynovial membrane. Arthritis Rheum 16, 471–486 (1973). \n7.  A. E. Schröder, A. Greiner, C. Seyfert, C. Berek, Differentiation of B Cells in the Nonlymphoid \nTissue of the Synovial Membrane of Patients with Rheumatoid Arthritis. Proceedings of the \nNational Academy of Sciences of the United States of America 93, 221–225 (1996). \n8.  J. Rangel-Moreno, et al., Inducible bronchus-associated lymphoid tissue (iBALT) in patients \nwith pulmonary complications of rheumatoid arthritis. J Clin Invest 116, 3183–3194 (2006). \n9.  W. Michie, A. Gunk, The Thyroid, the Thymus and Autoimmunity. International Journal of \nClinical Practice 20, 9–13 (1966). \n10.  D. M. Graham, H. D. Appelman, Crohn’s-like lymphoid reaction and colorectal carcinoma: a \npotential histologic prognosticator. Mod Pathol 3, 332–335 (1990). \n11.  J. M. J. Tas, et al., Visualizing antibody affinity maturation in germinal centers. Science 351, \n1048–1054 (2016). \n12.  J. Jacob, G. Kelsoe, K. Rajewsky, U. Weiss, Intraclonal generation of antibody mutants in \ngerminal centres. Nature 354, 389–392 (1991). \n13.  C. D. C. Allen, T. Okada, H. L. Tang, J. G. Cyster, Imaging of Germinal Center Selection Events \nDuring Affinity Maturation. Science 315, 528–531 (2007). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \n14.  T. A. Schwickert, et al., In vivo imaging of germinal centres reveals a dynamic open structure. \nNature 446, 83–87 (2007). \n15.  C. D. C. Allen, et al., Germinal center dark and light zone organization is mediated by CXCR4 \nand CXCR5. Nat Immunol 5, 943–952 (2004). \n16.  Z. Shulman, et al., Dynamic signaling by T follicular helper cells during germinal center B cell \nselection. Science 345, 1058–1062 (2014). \n17.  C. Berek, A. Berger, M. Apel, Maturation of the immune response in germinal centers. Cell \n67, 1121–1129 (1991). \n18.  J. Jacob, R. Kassir, G. Kelsoe, In situ studies of the primary immune response to (4 -hydroxy-\n3-nitrophenyl)acetyl. I. The architecture and dynamics of responding cell populations. J Exp \nMed 173, 1165–1175 (1991). \n19.  J. Jacob, G. Kelsoe, In situ studies of the primary immune response to (4 -hydroxy-3-\nnitrophenyl)acetyl. II. A common clonal origin for periarteriolar lymphoid sheath-associated \nfoci and germinal centers. J Exp Med 176, 679–687 (1992). \n20.  Y . J. Liu, J. Zhang, P . J. Lane, E. Y . Chan, I. C. MacLennan, Sites of specific B cell activation in \nprimary and secondary responses to T cell-dependent and T cell-independent antigens. Eur \nJ Immunol 21, 2951–2962 (1991). \n21.  J. Mitchell, A. Abbot, Ultrastructure of the Antigen -retaining Reticulum of Lymph Node \nFollicles as shown by High-resolution Autoradiography. Nature 208, 500–502 (1965). \n22.  K. Suzuki, I. Grigorova, T. G. Phan, L. M. Kelly, J. G. Cyster, Visualizing B cell capture of cognate \nantigen from follicular dendritic cells. J Exp Med 206, 1485–1493 (2009). \n23.  G. D. Victora, et al., Germinal Center Dynamics Revealed by Multiphoton Microscopy with a \nPhotoactivatable Fluorescent Reporter. Cell 143, 592–605 (2010). \n24.  T. B. Kepler, A. S. Perelson, Cyclic re -entry of germinal center B cells and the efficiency of \naffinity maturation. Immunol Today 14, 412–415 (1993). \n25.  D. Iber, P . K. Maini, A mathematical model for germinal centre kinetics and affinity \nmaturation. J Theor Biol 219, 153–175 (2002). \n26.  M. Meyer-Hermann, A mathematical model for the germinal center morphology and affinity \nmaturation. J Theor Biol 216, 273–300 (2002). \n27.  M. Meyer-Hermann, A molecular theory of germinal center B cell selection and division. Cell \nRep 36, 109552 (2021). \n28.  M. Oprea, A. S. Perelson, Somatic mutation leads to efficient affinity maturation when \ncentrocytes recycle back to centroblasts. J Immunol 158, 5155–5162 (1997). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \n29.  A. K. Szakal, M. H. Kosco, J. G. Tew, A novel in vivo follicular dendritic cell -dependent \niccosome-mediated mechanism for delivery of antigen to antigen -processing cells. J \nImmunol 140, 341–353 (1988). \n30.  M. Ziegner, G. Steinhauser, C. Berek, Development of antibody diversity in single germinal \ncenters: selective expansion of high-affinity variants. \nhttps://doi.org/10.1002/eji.1830241020. \n31.  K. Kawasaki, Diffusion Constants near the Critical Point for Time-Dependent Ising Models. II. \nPhys. Rev. 148, 375–381 (1966). \n32.  K. Kawasaki, Diffusion Constants near the Critical Point for Time-Dependent Ising Models. I. \nPhys. Rev. 145, 224–230 (1966). \n33.  R. M. Garner, S. E. McGeary, A. M. Klein, S. G. Megason, Tissue Fluidity Mediates a Trade-off \nBetween the Speed and Accuracy of Multicellular Patterning by Cell Sorting. Biophysical \nJournal 0 (2025). \n34.  J. M. Cliff, G. G. B. Klaus, A method for investigating the role of homotypic adhesion in \nlymphocyte activation. Journal of Immunological Methods 246, 51–59 (2000). \n35.  G. G. B. Klaus, M. Holman, J. Hasbold, Properties of mouse CD40: the role of homotypic \nadhesion in the activation of B cells via CD40. European Journal of Immunology  24, 2714–\n2719 (1994). \n36.  D. Possamaï, G. Pagé, R. Panès, É. Gagnon, R. Lapointe, CD40L -Stimulated B Lymphocytes \nAre Polarized toward APC Functions after Exposure to IL -4 and IL -21. The Journal of \nImmunology 207, 77–89 (2021). \n37.  L. E. Haswell, M. J. Glennie, A. Al -Shamkhani, Analysis of the oligomeric requirement for \nsignaling by CD40 using soluble multimeric forms of its ligand, CD154. Eur J Immunol  31, \n3094–3100 (2001). \n38.  J. Banchereau, et al., The CD40 Antigen and its Ligand. Annual Review of Immunology  12, \n881–926 (1994). \n39.  H. Gascan, J. F. Gauchat, G. Aversa, P . Van Vlasselaer, J. E. de Vries, Anti -CD40 monoclonal \nantibodies or CD4+ T cell clones and IL-4 induce IgG4 and IgE switching in purified human B \ncells via different signaling pathways. J Immunol 147, 8–13 (1991). \n40.  R. Ettinger, et al. , IL -21 induces differentiation of human naive and memory B cells into \nantibody-secreting plasma cells. J Immunol 175, 7867–7879 (2005). \n41.  K. Ozaki, et al., Regulation of B cell differentiation and plasma cell generation by IL -21, a \nnovel inducer of Blimp-1 and Bcl-6. J Immunol 173, 5361–5371 (2004). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \n42.  B. B. Ding, E. Bi, H. Chen, J. J. Yu, B. H. Ye, IL -21 and CD40L synergistically promote plasma \ncell differentiation through upregulation of Blimp-1 in human B cells. J Immunol 190, 1827–\n1836 (2013). \n43.  D. Zotos, et al. , IL -21 regulates germinal center B cell differentiation and proliferation \nthrough a B cell –intrinsic mechanism. Journal of Experimental Medicine  207, 365 –378 \n(2010). \n44.  E. J. Davey, G. Greicius, J. Thyberg, E. Severinson, STAT6 is required for the regulation of IL-\n4-induced cytoskeletal events in B cells. Int Immunol 12, 995–1003 (2000). \n45.  E. J. Davey, J. Thyberg, D. H. Conrad, E. Severinson, Regulation of cell morphology in B \nlymphocytes by IL-4: evidence for induced cytoskeletal changes. J Immunol 160, 5366–5373 \n(1998). \n46.  M. S. Naradikian, et al., IL4, IL21, and IFNγ interact to govern TBET and CD11c expression in \nTLR-activated B cells. J Immunol 197, 1023–1028 (2016). \n47.  S. Wang, et al. , IL -21 drives expansion and plasma cell differentiation of autoreactive \nCD11chiT-bet+ B cells in SLE. Nat Commun 9, 1758 (2018). \n48.  M. Tsuji, et al., Requirement for lymphoid tissue -inducer cells in isolated follicle formation \nand T cell -independent immunoglobulin A generation in the gut. Immunity 29, 261 –271 \n(2008). \n49.  J. E. Moyron-Quiroz, et al., Role of inducible bronchus associated lymphoid tissue (iBALT) in \nrespiratory immunity. Nat Med 10, 927–934 (2004). \n \n \n  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \nFigures \n \nFigure 1. A computational model predicts that B cell evolution is sensitive to tissue fluidity. (A) Schematic of the germinal center \n(GC) reaction, where B cells undergo iterative cycles of mutation and affinity-based selection through interactions with follicular \ndendritic cells (FDCs) and T follicular helper (TFH) cells. (B) Snapshots from an agent-based model simulating the GC, showing the \nprogressive enrichment of high -affinity B cells (blue) over lower -affinity B cells (yellow/green) over time. (C) The model \nrecapitulates Darwinian selection, as shown by the cumulative frequency plot where the B -cell population shifts toward higher \nreceptor affinity over simulated time. (D) The model predicts that the efficiency of affinity maturation is dependent on the \nsystem's effective temperature (𝑇𝑒𝑓𝑓), a parameter controlling tissue fluidity. Affinity maturation is optimal within a specific range, \nsuggesting that the physical state of immune tissue may be a tunable \"control knob\" for regulating adaptive immunity. \n  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \n \n \nFigure 2. CD40L-stimulated B cells self-organize into liquid-like multicellular condensates. (A) A representative brightfield image \nof multicellular aggregates formed by primary human B cells stimulated with CD40L. (B) A computational simulation shows two \ndistinct populations of cells (dark gray and light gray) spontaneously intermixing over time, a characteristic of a fluid system. (C) \nTime-lapse outlines of three representative B-cell aggregates show that initially irregular structures round up over 100 minutes, \nindicating the presence of an effective surface tension that minimizes surface area. (D) Quantification of rounding behavior. A plot \nof circularity versus time shows a steady increase, consistent with liquid-like relaxation dynamics. n = 11 and standard error of the \nmean is shown gray.  (E) Time-lapse microscopy reveals that separate B-cell aggregates fuse upon contact and their constituent \ncells intermix, further demonstrating their liquid-like nature. (F) Quantification of the intermixing process shows a decrease in the \ncoefficient of variation over time, confirming the internal fluidity of the B-cell condensates. n = 17 and standard error of the mean \nis shown gray. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \n \n \nFigure 3 TFH-derived cytokines tune the physical state of B -cell collectives. (A) Agent-based Kawasaki-Ising model that only \nconsiders one cell type and adhesion. The model predicts that increasing Teff, which lowers ΔJ/Teff, leads to a loss of cohesion, \ncausing simulated cell aggregates to become more dispersed and less circular. (B) Experimental validation shows that TFH \ncytokines modulate the physical state of B -cell aggregates across two different human donors . While aggregates treated with \nCD40L alone or with IL -4 remain cohesive, the addition of IL -21 induces a transition to a dispersed, gas -like state, characterized \nby a significant decrease in circularity. This phenocopies the effect of increasing effective temperature in the model. The horizontal \nblack lines are the bootstrapped mean s, and the black vertical bar s show the 95% bootstrap confidence interval of the mean \n(nboot = 1000, resampling particles within each condition) . (C) Snapshots from the agent -based model at different effective \ntemperatures ( Teff ) illustrate the transition from a condensed to a dispersed state over time. (D) Quantification from the \nsimulation shows that the coefficient of variation , a measure of intermixing, decreases more rapidly at higher effective \ntemperatures. (E) Experimental time -lapse microscopy of B -cell aggregates shows their structural evolution over 375 minutes \nunder different cytokine conditions. CD40L n=17, CD40L+IL4 n=17, CD40L+IL4+IL21=21, and shaded regions are 95% confidence \nintervals of mean.  (F) Experimental quantification confirms the model's prediction: the coefficient of variation decreases most \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n \nrapidly in the presence of IL-21, indicating that this cytokine increases the system's effective temperature and drives the collective \ntoward a more fluid state. \n \n \nFigure 4. A model for cytokine control of immune tissue physical state. This schematic summarizes the paper's central findings. \nInitially dispersed B cells, upon stimulation with CD40L, undergo a form of cellular condensation to self -organize into cohesive, \nliquid-like multicellular droplets. These droplets represent a baselin e physical state. The introduction of T follicular helper (TFH) \ncell-derived cytokines, such as IL -4 and IL-21, acts as biological control knobs, tuning ΔJ/Teff. This cytokine signaling can drive a \nphase transition, altering the collective's material properties from a cohesive liquid-like state to a more dispersed, gas-like state, \nthereby providing a mechanism for physically programming immune tissue. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\nLow affinity High affinity\nLow affinity\nHigh affinity\nBCC BCCA BCB-Inter BCB TFH 0 100ΔBCR\n← Higher affinity\n0\n1Cum. Freq.\nt = 20000\nt = 2000\nt = 39000\n0 1 2 3\nTeff\nΔE/Teff10-2\n10-1Affintiy (a.u.)\nComputational model prediction\ncontrol \nknobs\nimplies existence of\nphysical state of \nimmune tissue\nthat tune\nthat alters\nDiverse B cells (centrocytes) \ncompete for antigen\nApoptosisB cell internalizes  \nand processes \nantigen\nTFH signals enable \nsurvival and alter state\nCD40L, IL4, IL21\nPeptide \npresentation\nHigh affinity binding\nSearches for \ncognate TFH\nB cells (centroblasts) \ndivide and mutate\nB cells differentiate into \nmemory or plasma cells\nFDC\nTFH\n0 min\n1000 min\n2000 min\n3000 min\n4000 min\n5000 min\n?\nA\nB C\nD\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\n100 μm\n110 min 225 min0 min\n100 μm\n0 min 25 min 50 min 75 min 100 min\nTime\n0 50 100 150\nTime [minutes]\nTheoretical solid-like behavior\nLiquid-like rounding\nLiquid-like mixing\nTheoretical solid-like behavior\n0\n0.1\n0.2\n0.3\n0.4\n0.5\n0.6Circularity \n0 125 250 375Time [minutes]\n0.2\n0.4Coefficient of variation\nAdhesion (ΔJadh)\nEffective temperature (Teff)\nTime\n100 μm\n0 min 2500 min\n3750 min 5000 min\nA B\nC D\nE F\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\nCD40L CD40L+IL4 CD40L+IL21\nTeff=0.025\nTeff=0.04\nTeff=1\nTeff=0.025\nTeff=0.04\nTeff=1\nCD40L+IL4+IL21\n0.26 0.35 0.46 0.61 0.80\nEffective Temperature\nCircularity\n0\n0.5\n200 μm\nCD40L\nCD40L+IL4\nCD40L+IL21\n0 125 250 375\nTime [minutes]\n0.2\n0.4\nCoiffecient of Variance\n0\nt=0 t=300\n50 100 150 200 250 300\nTimesteps\n0.4\n0.8\nCoefficient of Variation\nCD40L\nCD40L+IL4\nCD40L+IL21\n0 min 375 min\n200 μm\n0.5\n0\nCircularity\nA B\nC\nD\nE\nF\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint \n\nIL4, IL21 tune ΔE/Teff CD40L\nB cells Multicellular\ndroplet\nAltered physical \nstate\nCellular\nCondensation\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted October 30, 2025. ; https://doi.org/10.1101/2025.10.28.685215doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}