Targeting the immunological synapse in multiple myeloma Authors: Liu Sumei, Forouzan Bahmani Published: 2026 ABSTRACT The immunological synapse, a highly organized cell–cell junction, is an important fulcrum for effective T-cell-mediated killing of cancer cells. In multiple myeloma, however, therapeutic resistance reflects multiple interacting layers, including synaptic disruption, clonal evolution, metabolic reprogramming, and spatially heterogeneous bone-marrow niches, all of which can contribute to T-cell dysfunction and immune escape. This review examines the immunological synapse as a potentially modifiable intercellular interface: we describe how the bone marrow niche impairs synaptic function and summarize strategies that may help restore synaptic structure and signaling. We explore how the myeloma niche deranges synaptic function and then detail innovative strategies to restore its fidelity. We analyze how BiTEs establish synaptic contacts and how CAR-T cells assemble modified synaptic interfaces that have distinct structural and signaling properties; we then compare shared and divergent mechanisms of action. Looking forward, we discuss next-generation approaches, including trispecific antibodies, armored/logic-gated CAR-Ts, niche-modifying agents, and synaptic nanotechnology, which aim to create specific, resilient, and context-aware cytolytic junctions. By framing therapeutic progress through the lens of synaptic engineering, this narrative review selectively integrates mechanistic, translational, and clinical reports supporting synapse-directed interventions, while explicitly separating preclinical models from early-phase and randomized clinical evidence so that hypothesis-generating findings are not treated as equivalent to patient outcome data. Introduction: the synaptic core of cancer immunity The immunological synapse is a highly organized cell–cell junction that coordinates receptor engagement, adhesion, and directed secretion to support antigen-specific T-cell activation and effector function [1, 2]. The classical, concentric architecture of the synapse, often described as central and peripheral supramolecular activation clusters, enforces spatiotemporal ordering of T cell receptor microclusters, integrins and costimulatory molecules, thereby converting stochastic ligand encounters into deterministic intracellular signaling and polarized delivery of cytolytic cargo [2, 3]. Beyond simple receptor aggregation, the synapse operates as a mechanical and trafficking hub that orchestrates centrosome polarization, actin dynamics, and regulated exo- and endocytosis, all of which are essential for productive target cell killing [3, 4]. Because the synapse imposes important physical and signaling constraints on immune recognition, its integrity is one determinant of antitumor immunity, but not an exclusive explanation for response or resistance in multiple myeloma; clonal evolution, metabolic rewiring, and marrow spatial heterogeneity also shape immune escape and therapeutic outcome [5–9]. In hematologic malignancies that reside in structured niches, such as multiple myeloma, the bone marrow microenvironment remodels cell contacts and diffusible signals in ways that erode synapse fidelity, with direct consequences for immune surveillance, immunotherapy response and disease progression [5, 10]. Importantly, recent single-cell and spatial studies show that this niche is heterogeneous rather than uniform: early precursor lesions already display altered NK and cytotoxic T-cell compartments, focal lesions contain site-specific subclones and altered immune composition, and multiplex imaging can resolve exhausted T cells, macrophage-depleted regions, and prognostic plasma-cell/CD4 + T-cell neighborhoods within the same marrow [11–17]. Multiple myeloma establishes a multilayered immunosuppressive milieu that undermines synaptic competence by driving T cell exhaustion, expanding myeloid derived suppressor cells, augmenting regulatory T (Treg) cell activity and upregulating inhibitory ligands that interrupt activating receptor–ligand interactions [18–22]. However, synapse disruption is not the only, or always the primary, route to myeloma-associated immune dysfunction. Soluble immunomodulatory mediators released by malignant plasma cells and bone-marrow stromal cells (for example TGF-β, IL-10 and IL-6) can reprogram T-cell signaling programs, surface receptor expression and metabolic states and thereby alter synapse formation and function independent of physical receptor mis-localization [23, 24]. Similarly, expansion of suppressive cellular subsets including Tregs and MDSCs, and functional skewing of CD4 + subsets toward Th17 (IL-17 producing) phenotypes, have been linked to both enhanced myeloma growth and direct impairment of effector T-cell responses, effects that can occur whether or not classical synaptic geometry is altered [25–27]. T cells in the bone marrow of myeloma patients characteristically express high levels of coinhibitory receptors and transcriptional programs associated with terminal dysfunction, which correlates with diminished proliferative capacity, reduced cytokine production and impaired cytotoxicity [21, 28]. Simultaneously, myeloid-derived suppressor cells (MDSCs) in myeloma suppress T cell receptor signaling and metabolically constrain effector cells through arginase, nitric oxide and reactive oxygen species, thereby weakening the very signaling thresholds required for synapse maturation [19, 29]. Treg cells further blunt anti-myeloma responses by contact dependent and soluble mechanisms that limit T cell activation and by competing for costimulatory cues in the niche [18, 30]. Tumor cell intrinsic programs and microenvironmental mediators converge on inhibitory ligand expression, producing a molecular shield that sterically and functionally disrupts synaptic geometry [20, 31]. Programmed death-ligand 1 (PD-L1) expression is inducible in myeloma by niche cytokines and survival factors, and PD-L1 binding to programmed cell death protein 1 (PD-1) on T cells contributes to synaptic destabilization and aborted cytolytic synapse formation [20, 31]. Emerging clinical and translational data also implicate additional checkpoints such as T cell immunoreceptor with Ig and ITIM domains (TIGIT) and lymphocyte activation gene-3 (LAG-3) in myeloma associated T cell dysfunction, suggesting that redundant inhibitory axes cooperate to fragment synaptic signaling [32, 33]. There is also direct experimental evidence that malignant plasma cells actively interfere with productive cell contacts. Tumor expressed adhesion and immunomodulatory molecules can alter integrin engagement and antigen presentation, resulting in defective microcluster centripetal flow and mislocalization of lytic machinery, which together translate into failed target cell killing despite apparent antigen recognition [34, 35]. Clinically, this synaptic sabotage helps to explain why otherwise potent effectors, including endogenous tumor-reactive T cells and adoptively transferred cellular therapies, frequently display limited activity or develop resistance within the bone marrow niche [36, 37]. Reconciling the important role of the immunological synapse with the broader biological ecology of myeloma opens an actionable framework: restoring synaptic architecture and signaling is a mechanistic and translational priority, but one that should be interpreted alongside clonal evolution, metabolic state, and marrow spatial organization [5, 7, 8]. The aim of this review is to treat the immunological synapse in multiple myeloma as a modifiable contact area whose architecture and signaling can be restored or redesigned through molecular, cellular, and biophysical interventions. We explore recent mechanistic studies and clinical correlates that define how the myeloma niche deranges synaptic function, and we explicitly distinguish mechanistic, translational, and clinical evidence throughout the review so that preclinical modeling is presented as biologically informative but not equivalent to clinical efficacy data [38, 39]. To help readers navigate the review, Sect. 3 outlines the native synapse, Sects. 4 and 5 compare BiTE and CAR-T synapses, Sect. 6 contrasts convergent and divergent mechanisms, and Sect. 7 summarizes next-generation strategies. The key clinical take-home messages are summarized again at the end of the review. Methods This narrative review was designed and reported in line with SANRA principles for narrative reviews. Literature searches were conducted in PubMed/MEDLINE, Embase, and Web of Science up to 29 Feb 2026 using combinations of the terms ‘immunological synapse,’ ‘multiple myeloma,’ ‘bone marrow microenvironment,’ ‘CAR T,’ ‘bispecific antibody,’ ‘trispecific antibody,’ ‘BCMA,’ ‘GPRC5D,’ ‘TGF-β,’ and ‘VLA-4.’ We prioritized peer-reviewed mechanistic, translational, and clinical studies relevant to myeloma immunotherapy. Preclinical and modeling studies were used to support biological plausibility and are interpreted as hypothesis-generating, whereas clinical trial data were treated as the highest level of therapeutic evidence. Evidence was synthesized qualitatively, with mechanistic, translational, and clinical findings discussed separately wherever possible. The foundational blueprint: anatomy of a killer synapse The natural synapse against cancer: key components The immunological synapse is the nanoscale interface that organizes antigen recognition, adhesion and costimulation into a single signaling/secretory unit; its geometry and mechanics (T-cell receptor (TCR)– peptide-major histocompatibility complex (pMHC) bond lifetime, effective avidity, intermembrane gap and force transmission), together with adhesive and co-stimulatory inputs, determine whether a contact triggers productive cytotoxicity or a hyporesponsive program [1, 40–43]. Table 1 summarizes the core synapse-engineering variable, geometry, ligand kinetics, adhesion, actin flow, and costimulation, and their functional consequences when each parameter is favorable or unfavorable. Table 1Key physical and molecular variables and their functional consequencesVariable (mechanics/molecular)What it controlsFunctional consequence if favourableIf unfavourableIntermembrane gap / ligand lengthKinetic segregation (CD45 exclusion)Favors ITAM phosphorylation → productive signalingCD45 remains → signal dampeningTCR–pMHC bond lifetime, catch bondsMicrocluster nucleation & force stabilizationRobust microcluster formation, granule polarizationShort contacts → abortive signalingAdhesion strength (LFA-1/ICAM-1, VLA-4)Contact stability / dwell timeLonger dwell time → efficient killingLow adhesion → rapid disengagementActin retrograde flowMicrocluster centripetal transportProper sorting & centrosome dockingDisrupted actin → mislocalization of lytic machineryCo-stimulation (CD28, 4-1BB)Signal amplitude & metabolic programmingCD28 → fast, high amplitude; 4-1BB → sustained, memory phenotypeMismatched costim → burnout or poor persistence Table 1Key physical and molecular variables and their functional consequencesVariable (mechanics/molecular)What it controlsFunctional consequence if favourableIf unfavourableIntermembrane gap / ligand lengthKinetic segregation (CD45 exclusion)Favors ITAM phosphorylation → productive signalingCD45 remains → signal dampeningTCR–pMHC bond lifetime, catch bondsMicrocluster nucleation & force stabilizationRobust microcluster formation, granule polarizationShort contacts → abortive signalingAdhesion strength (LFA-1/ICAM-1, VLA-4)Contact stability / dwell timeLonger dwell time → efficient killingLow adhesion → rapid disengagementActin retrograde flowMicrocluster centripetal transportProper sorting & centrosome dockingDisrupted actin → mislocalization of lytic machineryCo-stimulation (CD28, 4-1BB)Signal amplitude & metabolic programmingCD28 → fast, high amplitude; 4-1BB → sustained, memory phenotypeMismatched costim → burnout or poor persistence Key physical and molecular variables and their functional consequences Co-stimulatory receptors critically gate the outcome of antigen recognition; CD28 accelerates early recruitment and activation of proximal kinases including Lck and ZAP-70 and thereby amplifies and prolongs productive signaling, whereas members of the TNFR superfamily such as 4-1BB contribute prosurvival and metabolic programs that sustain cytotoxicity and persistence [44, 45]. Adhesion molecules provide the mechanical and spatial scaffolding for synapse assembly: high-affinity lymphocyte function-associated antigen-1 (LFA-1) binding to intercellular adhesion molecule-1 (ICAM-1) stabilizes close membrane apposition, supports centripetal actin flows that consolidate signaling microclusters, and tunes synaptic dwell time in a manner that influences killing efficiency [46, 47]. At the mesoscale the synapse organizes into concentric supramolecular activation clusters: a central supramolecular activation cluster (SMAC) enriched for TCR and signaling adaptors, a peripheral SMAC containing integrins and adhesion complexes, and a distal SMAC dominated by actin dynamics and large ectodomain molecules [40, 48]. Microcluster nucleation at the periphery initiates signaling and is followed by actin-driven centripetal transport that sorts receptors into the central domain where signaling may be attenuated or terminated via ubiquitin-dependent sorting [43, 48]. In the cancer setting these spatial rules are co-opted and sometimes subverted: tumor cells and engineered agents such as bispecific T cell engagers or chimeric antigen receptor (CAR) constructs alter synaptic geometry, ligand density, or intermembrane spacing with direct consequences for T cell activation and myeloma cell killing [49, 50]. Key synapse-engineering components, receptor biophysics, co-stimulation, adhesion, SMAC organization, and actin polarity, are summarized in Table 2. Table 2Key structural & functional components of the killer immunological synapseComponentFunctional role (short)Refs.TCR–pMHC biophysics (affinity, off-rate, catch-bonds)Determines microcluster nucleation and trigger threshold for signaling. [41–43]Co-stimulatory receptors (CD28, 4-1BB)Gate signal amplitude, metabolic programming and persistence. [44, 45, 51, 52]Adhesion molecules (LFA-1 / ICAM-1, VLA-4)Mechanical scaffold — controls contact stability and dwell time. [46, 47, 53, 54]SMAC architecture (cSMAC/pSMAC/dSMAC)Spatial sorting of signaling vs. adhesion vs. actin flow; controls signal duration. [40, 43, 48]Actin dynamics & centrosome polarizationDrive microcluster centripetal flow, vesicle trafficking and granule delivery. [4, 43, 46] Table 2Key structural & functional components of the killer immunological synapseComponentFunctional role (short)Refs.TCR–pMHC biophysics (affinity, off-rate, catch-bonds)Determines microcluster nucleation and trigger threshold for signaling. [41–43]Co-stimulatory receptors (CD28, 4-1BB)Gate signal amplitude, metabolic programming and persistence. [44, 45, 51, 52]Adhesion molecules (LFA-1 / ICAM-1, VLA-4)Mechanical scaffold — controls contact stability and dwell time. [46, 47, 53, 54]SMAC architecture (cSMAC/pSMAC/dSMAC)Spatial sorting of signaling vs. adhesion vs. actin flow; controls signal duration. [40, 43, 48]Actin dynamics & centrosome polarizationDrive microcluster centripetal flow, vesicle trafficking and granule delivery. [4, 43, 46] Key structural & functional components of the killer immunological synapse Synaptic quality determines functional outcome Synaptic stability, defined by the lifetime and molecular composition of the contact, is a principal determinant of whether a T cell executes target lysis, enters a hyporesponsive anergic state, or progresses toward exhaustion under chronic antigen exposure [55, 56]. Strong, sustained TCR triggering delivered within a robust adhesive scaffold typically drives rapid cytotoxic granule polarization and target apoptosis, whereas weak or intermittent signaling in the absence of adequate co-stimulation favors incomplete effector differentiation and molecular programs associated with dysfunction [42, 57]. Physical sorting at the cleft (e.g., kinetic-segregation of CD45) and actin-driven microcluster flows jointly set local signaling thresholds: short intermembrane gaps and intact actin favor net ITAM phosphorylation and granule delivery, whereas increased spacing or disrupted actin destabilizes signaling and shortens synaptic dwell time (See Table 1) [43, 46, 58]. The actin cytoskeleton plays a dual role: it generates the retrograde flow that traffics microclusters to the central domain and it transduces mechanical forces that stabilize catch bonds and promote productive signaling [43, 46]. Disruption of actin dynamics reduces microcluster integrity, shortens synaptic dwell time, and converts otherwise productive contacts into abortive interactions that predispose to tolerance rather than killing [43, 59]. For engineered therapeutics in multiple myeloma such as CAR T cells or T cell engagers the topology of the synthetic synapse is equally decisive: co-stimulatory domain choice and spatial orientation of the antigen binding site modulate synapse architecture, with 4-1BB domains tending to produce synapses associated with sustained signaling and persistence and CD28 domains producing more immediate but sometimes shorter lived responses [44, 57]. Modeling and experimental dissection show that antigen density, epitope location, and intermembrane gap size influence synaptic quality and thus clinical efficacy and resistance, so that rational tuning of biophysical parameters can convert an unstable contact into an engineered bridge that restores durable cytotoxicity against myeloma [50, 60]. In sum, the killer synapse is not a binary structure but a tunable interface where ligand biophysics, adhesion, co-stimulation, and cytoskeletal mechanics are integrated to determine whether a T cell extinguishes a cancer cell or is diverted into dysfunction [1, 56]. Understanding and engineering these components to maximize synaptic stability while limiting chronic overstimulation will be important for next-generation immunotherapies for multiple myeloma. Take-home message: synaptic quality is necessary but not sufficient for antitumor function; adhesion, co-stimulation, and actin dynamics determine whether a contact becomes productive or abortive. Bispecific T-cell engagers (BiTEs): formation and properties of a synthetic synaptic interface The “Bridge” as a synaptic nucleator Bispecific T-cell engagers create membrane-anchored ternary complexes (CD3–BiTE–antigen) that can rapidly establish a functional cytolytic contact between T cells and tumor cells. Mechanistically, canonical BiTEs assemble ternary complexes on opposing cell membranes (CD3, BiTE, tumor antigen), and these membrane-anchored ternary complexes are sufficient to drive two-dimensional adhesion and the assembly of a functional cytolytic interface without peptide–MHC engagement or canonical co-stimulatory signaling [50, 61]. Empirical imaging and agent-based modeling work demonstrates that BiTEs drive a sequence of events, solution-phase binary binding, random cell-cell encounters, and 2D membrane binding leading to formation of compact synaptic contacts, thereby converting Brownian encounters into productive cytolytic contacts at the population level [50]. Structurally engineered, half-life-extended BiTE formats maintain the same fundamental geometry of the synapse while modulating residence time and systemic exposure, which in turn affect the kinetics of synapse nucleation and serial killing [62]. This framing suggests that a BiTE can be considered an engineered minimal synaptic program: a soluble scaffold that imposes a spatially constrained intermembrane gap, concentrates CD3 signaling units and antigenic ligands, and can trigger T-cell effector programs even when traditional antigen presentation pathways or co-stimulatory inputs are limited [50, 59]. Synaptic characteristics of BiTE action BiTE-induced synapses have biophysical and temporal signatures that distinguish them from canonical TCR–MHC synapses. Quantitative imaging flow cytometry and mechanistic modeling reveal that BiTE synapses form rapidly after cell-cell encounter and can take multiple stoichiometric configurations, including one-to-one and multi-cell cluster synapses, with ternary complex density on the contact plane determining adhesion probability and downstream killing [50]. The effective intermembrane distance created by BiTEs is comparable to physiologic TCR–pMHC spacing, enabling productive lytic granule polarization and calcium signaling despite the absence of peptide specificity [61, 62]. Because formation is governed by encounter probability and two-dimensional bonding kinetics rather than total systemic drug concentration alone, local cell densities, antigen density, and membrane mobility become dominant determinants of synaptic frequency and potency [50]. Synapses generated by different BiTE formats therefore vary chiefly in their temporal persistence and serial engagement capacity: small, non-Fc BiTEs produce very short systemic half-lives that favor continuous exposure strategies and rapid serial killing, while half-life extended constructs permit intermittent dosing yet preserve sufficient synaptic spacing to allow durable lytic events [62, 63]. Importantly, synaptic action is accompanied by stereotyped cytokine release and upregulation of T-cell activation markers, but the magnitude and duration of these signals are tightly linked to synapse abundance and duration rather than solely to peripheral concentration [50, 63]. Clinical implications of the BiTE synapse: pharmacokinetics, durability and antigen-escape as synaptic failure Preclinical imaging, agent-based models, and ex vivo assays suggest that BiTEs nucleate compact ternary complexes on cell membranes and that 2-D bonding kinetics, antigen density, and local encounter probability shape synapse formation in controlled systems. Translational evidence: These findings help explain why half-life, valency, and epitope geometry influence synaptic persistence and serial-killing capacity, but they should be interpreted as hypothesis-generating rather than equivalent to patient outcome data [38, 39, 63]. Clinical pharmacology and trial data support the translational relevance of BiTE format, but response and durability are not determined by synaptic geometry alone. First-generation small BiTEs (for example, blinatumomab) required continuous infusion because of short plasma half-life, whereas half-life–extended constructs can alter dosing logistics; however, clinical response patterns in multiple myeloma are also shaped by tumor burden, prior lines of therapy, marrow reserve, baseline T-cell fitness, and patient-specific factors, so synaptic properties should be presented as one contributor among several rather than the sole explanation for outcome [64–67]. Model-informed analyses link regimen, tumor burden, and local cellular composition to synaptic abundance and therefore to clinical pharmacodynamics, arguing for exposure profiles tailored to tumor kinetics and anatomical sanctuary sites such as the bone marrow [50]. Second, clinical resistance is often multifactorial. Antigen loss or alteration can abolish target engagement, but relapse also reflects tumor burden, prior treatment exposure, soluble antigen decoys, marrow microenvironmental suppression, and effector T-cell fitness; therefore, antigen escape should be described as one resistance mechanism rather than as synonymous with synaptic failure [64, 65, 67, 68]. Published genomic and translational analyses document examples of antigen loss or alteration after BCMA-directed therapies (e.g., TNFRSF17/BCMA deletions and extracellular domain changes) that have been linked to relapse in patient series, evidence that antigen escape is a demonstrated but not universally frequent mechanism of resistance [68, 69]. Third, soluble decoys such as elevated soluble BCMA produced by γ-secretase cleavage can act as sinks that attenuate effective bridging at the cell surface and thus reduce synaptic assembly; clinical and preclinical data indicate that modulation of antigen shedding, dual-antigen targeting, or trispecific designs can mitigate this failure mode [69, 70]. Finally, because synapse formation is a stochastic, cell-population phenomenon, pharmacologic strategies that optimize local, sustained encounter probabilities, either by continuous low-level exposure, compartmental targeting, or combination approaches that enhance T-cell fitness, are conceptually preferable to intermittent high-peak exposures that do not sufficiently increase productive synapse density across sanctuary niches [50, 63]. Take-home message: BiTE efficacy depends not only on target binding, but also on synapse frequency, antigen density, dosing exposure, and T-cell fitness; therefore, antigen escape and microenvironmental suppression should be interpreted as clinically relevant resistance pathways rather than as purely theoretical synaptic defects. Chimeric antigen receptor T-cells (CAR-Ts): building a stable, enhanced synaptic superstructure The CAR as a synthetic synaptic organizer Chimeric antigen receptors are engineered, modular receptors that fuse an antibody derived antigen recognition domain to membrane and intracellular signaling modules such that a single polypeptide both senses antigen and initiates T cell activation through CD3ζ and engineered costimulatory elements [71, 72]. The original proof of principle that antibody variable fragments linked to T cell signaling chains can convert antigen recognition into cytolytic activation was reported by Eshhar and colleagues and established the conceptual basis for the CAR as a synthetic synaptic organizer [71, 73]. This architecture was intentionally designed to collapse the normally distributed events of antigen presentation, co-stimulation and spatial reorganization into a single focal receptor complex that nucleates contact with target cells, clusters signaling motifs and directs polarized secretion [72, 74]. By combining an scFv targeting moiety with a signaling scaffold that commonly includes CD3ζ and one or more costimulatory domains, modern CARs recapitulate essential synaptic functions while also imposing synthetic constraints on geometry, valency and signaling kinetics [72, 75]. Because the CAR physically couples recognition and activation, it functions less like a simple ligand-binding receptor and more like an engineered supramolecular activation cluster that can redefine the local protein landscape at the T cell–target interface [76]. The “super-synapse” of CAR-T cells Stability and Potency. Imaging studies indicate that CAR-T cells can form nonclassical immune synapses with less stereotyped SMAC segregation than canonical TCR synapses, but this architecture is not uniformly superior: in a comparative study, carCTL synapses lacked distinct LFA-1 adhesion rings and were less reliant on LFA-1 for stable conjugate formation, highlighting that potency depends on receptor design and antigen context rather than on a single canonical synapse architecture [77]. Other constructs show faster granule polarization and target-cell disengagement; however, these observations remain construct dependent and have been reported in selected imaging systems and target-antigen settings, so they should not be generalized to all CAR platforms or all disease contexts [34, 77, 78]. The choice of costimulatory domain influences signaling tempo, metabolic programming, and persistence, but the magnitude of this effect varies with CAR scaffold design, antigen system, manufacturing conditions, and clinical context. Comparative studies show that costimulatory domains can alter CAR membrane organization, diffusion kinetics, and immune-synapse quality, and myeloma cohorts further show that in vivo expansion, leukapheresis composition, and lymphodepletion success all shape response and toxicity rather than costimulation alone [57, 79–81]. It follows that synaptic stability is a manipulable variable: tuning co-stimulation, spacer length, scFv affinity and anchor topology all alter contact lifetimes and thereby modulate the balance between immediate tumor clearance and long-term cellular fitness [76, 82]. Spatial Organization. How a CAR arranges proteins within a supramolecular activation cluster determines downstream signaling cascades and effector polarization. Classical TCR synapses achieve a conserved concentric SMAC organization that positions proximal kinases, adhesion molecules and actin flows to channel signals and secretory traffic [83, 84]. CARs, by contrast, frequently create a more disordered protein distribution while still accomplishing granule delivery; this altered topology is associated with differences in microcluster behavior, phosphorylation patterns and force transmission across the interface [76, 77]. Protein kinase C theta is an exemplar signaling node whose synaptic localization is governed by motifs in its V3 domain and by association with CD28, and its placement influences NF-κB activation and downstream transcriptional programs that regulate survival and cytokine production [84, 85]. Actin dynamics are equally central: actin remodeling both scaffolds receptor clustering and controls centrosome docking and granule release, and quantitative studies show that CAR engagement can modify actin retrograde flow and mitochondrial redistribution in ways that change killing efficiency and metabolic coupling [83, 86]. Understanding how CAR design dictates the localization of kinases such as PKCθ and the choreography of actin therefore provides a direct route to engineer synapses with desired signaling and secretory properties [76, 84]. The differences in immunological synapse formation between classical T cells and multiple myeloma cells versus CAR-T cells and multiple myeloma cells are illustrated in Fig. 1. Synaptic determinants of CAR-T efficacy and toxicity Chronic antigen exposure can drive progressive hypofunction in CAR-T cells, but in relapsed/refractory multiple myeloma the degree of exhaustion, persistence, and toxicity varies across products and patients. In a 27-patient ide-cel cohort, in vivo CD8 + CAR-T expansion correlated with response, whereas persistence alone was not essential for durable remission; separate studies also show that starting T-cell composition and manufacturing-related failure modes influence whether a product expands effectively after infusion [67, 79, 80]. Toxicity is likewise a synapse-centric phenomenon. On-target off-tumor toxicity results when CARs form productive synapses with healthy tissues that share antigen expression or when high-avidity synaptic contacts override the quantitative thresholds that normally protect normal cells [87, 88]. The field has therefore pursued multiple synapse-aware engineering solutions including affinity or avidity tuning of scFvs to favor high antigen density on malignant cells, protease-regulated or drug-controllable CARs that gate activity to permissive contexts, and synthetic AND/NOT logic circuits that require dual inputs to assemble a productive synaptic unit. These strategies explicitly treat the CAR as an engineerable synaptic scaffold whose geometry, signaling tempo and activation thresholds can be tuned to separate potent antitumor killing from systemic toxicity [82, 88]. In sum, viewing CARs through the lens of synapse biology reframes design priorities: beyond affinity and antigen selection, optimal CARs must organize a synaptic superstructure that balances immediate cytolysis with metabolic durability, local specificity with systemic safety, and signaling intensity with resistance to burnout [52, 76]. Integrating quantitative imaging, biochemical dissection of synaptic nodes such as PKCθ, and clinical metabolic phenotyping provides a high-resolution roadmap to build next generation CARs that are not only receptors but engineered bridges capable of restoring durable antitumor immunity [51, 77, 84]. Take-home message: CAR-T performance is shaped by synapse architecture, manufacturing quality, co-stimulatory design, and the marrow milieu, so durable benefit requires both a well-designed receptor and a permissive biological context. Fig. 1Natural versus CAR-T immunological synapses in myeloma killing. Comparison of natural T-cell and CAR-T synapses. Left: a canonical TCR–MHC synapse organizes into concentric zones (cSMAC/pSMAC/dSMAC) that coordinate receptor signaling, adhesion and actin dynamics. Right: CAR-T “super-synapses” often show less classical SMAC organization but still promote rapid granule polarization; CAR design (costimulatory domain, spacer/hinge and epitope targeting) strongly influences synapse stability and functional durability. Created by the authors using BioRender.com Natural versus CAR-T immunological synapses in myeloma killing. Comparison of natural T-cell and CAR-T synapses. Left: a canonical TCR–MHC synapse organizes into concentric zones (cSMAC/pSMAC/dSMAC) that coordinate receptor signaling, adhesion and actin dynamics. Right: CAR-T “super-synapses” often show less classical SMAC organization but still promote rapid granule polarization; CAR design (costimulatory domain, spacer/hinge and epitope targeting) strongly influences synapse stability and functional durability. Created by the authors using BioRender.com Convergence and divergence at the synaptic cleft Convergent mechanisms Both natural cytotoxic T lymphocytes redirected by bispecific T cell engagers and chimeric antigen receptor T cells culminate in formation of a lytic junction that concentrates lytic granules, enabling directional delivery of perforin and granzymes into the target cell [89, 90]. The core biochemical cascade at this lytic synapse follows conserved steps: polarization of the microtubule organizing center, focal accumulation of lytic granules, calcium triggered granule fusion and perforin pore formation, and granzyme driven apoptosis of the target [89, 91]. Bispecific antibodies create this outcome by physically bridging CD3 on a bystander T cell to a tumor antigen, thereby recreating an immunological synapse that functionally recapitulates the orientation and secretion machinery of a native cytotoxic synapse [50, 92]. CAR T cells achieve the same terminal effector program by providing antigen recognition and activation in cis on the engineered T cell surface, which initiates granule polarization through CAR proximal signaling modules [78, 93]. A second point of mechanistic convergence is the capacity of both modalities to bypass many of the proximal defects that impair natural anti-myeloma synapses within the bone marrow microenvironment [94, 95]. Whereas endogenous tumor reactive T cells in multiple myeloma are frequently impaired by low TCR precursor frequency, inhibitory checkpoint expression, and suppressive stromal cues, exogenous engagement by a high-affinity CAR or a high-avidity bispecific restores the immediate signaling threshold required for cytolytic polarization [50, 67]. Clinically, these response rates are consistent with the idea that redirected synapse formation can help overcome some preexisting forms of native tolerance and promote tumor debulking [93, 96]. Divergent synaptic philosophies Despite these shared endpoints, CAR T cells and bispecific engagers implement fundamentally different synaptic architectures and temporal programs that reflect divergent engineering philosophies [50, 78]. CAR T synapses are formed by a living, membrane anchored recognition receptor that couples antigen binding to a modular intracellular signaling cassette; this configuration produces a synapse in which co-stimulatory signals, tonic signaling, and feedback from the T cell metabolic state are endogenously integrated [93, 97]. By contrast, bispecific antibodies create a minimal adhesive bridge between two independent cells, relying on the native CD3 complex and the unmodified T cell signaling apparatus to transmit activation; the resulting synapse tends to be more transient and is repeatedly reformed as soluble engager molecules diffuse and rebind [50, 92]. These architectural differences translate to divergent signal kinetics and downstream biology [98, 99]. CARs that incorporate CD28 based co-stimulatory modules typically deliver rapid, high amplitude signaling that accelerates effector differentiation and early tumor clearance, whereas 4-1BB containing CARs bias toward slower kinetics, enhanced mitochondrial biogenesis and improved long term persistence [97, 98]. Bispecific engagers instead produce repeated pulses of physiological TCR initiated signaling that, depending on dosing schedules and T cell availability, can produce sustained cytotoxic pressure without the same single-cell programmed fate imprinting imposed by CAR intracellular domains [50, 96]. Spatial determinants at the nanoscale further differentiate the two strategies [78]. CAR constructs, their hinge length and epitope targeting define intermembrane gap and force transmission across the synapse, which in turn determine microcluster formation and lytic granule docking [78]. Bispecific molecules impose their own geometry; for some designs the induced intercellular spacing more closely resembles the native TCR interface and thus differentially engages kinetic segregation and exclusion of large phosphatases such as CD45 [50]. These design variables are consequential for efficacy and toxicity because they shape the magnitude, duration and localization of kinase activation within the synaptic cleft [78, 89]. Table 3 compares engineered modalities across synapse architecture, kinetic behavior, and translational implications. Table 3Comparative synaptic architectures of major engineered therapeuticsModalityHow it builds a synapse (architecture)Typical synaptic kinetics/phenotypeKey clinical/translational implicationsRefs.BiTEsSoluble bridge forms ternary complexes (CD3–BiTE–antigen) that nucleate compact synapses.Rapid formation, often transient; dependence on encounter probability.PK format (continuous vs. half-life extended) shapes serial killing and exposure; antigen shedding reduces efficacy.[50, 61–63, 70]CAR-T cellsMembrane-anchored synthetic receptor couples antigen binding to intracellular signaling (CD3ζ + costim).Larger or nonclassical protein patterns; can deliver rapid, high-potency killing at single-cell level.Costimulatory choice (CD28 vs. 4-1BB) → tradeoffs in speed vs. persistence; manufacture and tonic signaling affect durability.[51, 52, 71, 72, 76, 77, 86]Trispecific engagersTri-arm molecules add co-stimulation or dual tumor antigen binding to the bridge.Stronger activation pulses at the synapse; reduced single-antigen escape in models.Can localize signal-2 and reduce systemic costimulation; promising vs. antigen heterogeneity.[100–105]Armored/logic-gated CARsCARs engineered to secrete cytokines or require combinatorial antigen logic.Sustained local supportive environment; gated activation restricted to permissive contexts.Enhance persistence and safety (NOT/AND gating, dominant-negative receptors).[106–113]Nanoparticle/synapse-directed deliveryNanocarriers deliver co-stimulatory ligands or payloads to the contact area (T-cell bound).Focal payload release at moment of synapse formation; amplifies local signals without systemic exposure.Improves synapse stability and reduces systemic toxicity; modular co-delivery strategies.[114–117] Table 3Comparative synaptic architectures of major engineered therapeuticsModalityHow it builds a synapse (architecture)Typical synaptic kinetics/phenotypeKey clinical/translational implicationsRefs.BiTEsSoluble bridge forms ternary complexes (CD3–BiTE–antigen) that nucleate compact synapses.Rapid formation, often transient; dependence on encounter probability.PK format (continuous vs. half-life extended) shapes serial killing and exposure; antigen shedding reduces efficacy.[50, 61–63, 70]CAR-T cellsMembrane-anchored synthetic receptor couples antigen binding to intracellular signaling (CD3ζ + costim).Larger or nonclassical protein patterns; can deliver rapid, high-potency killing at single-cell level.Costimulatory choice (CD28 vs. 4-1BB) → tradeoffs in speed vs. persistence; manufacture and tonic signaling affect durability.[51, 52, 71, 72, 76, 77, 86]Trispecific engagersTri-arm molecules add co-stimulation or dual tumor antigen binding to the bridge.Stronger activation pulses at the synapse; reduced single-antigen escape in models.Can localize signal-2 and reduce systemic costimulation; promising vs. antigen heterogeneity.[100–105]Armored/logic-gated CARsCARs engineered to secrete cytokines or require combinatorial antigen logic.Sustained local supportive environment; gated activation restricted to permissive contexts.Enhance persistence and safety (NOT/AND gating, dominant-negative receptors).[106–113]Nanoparticle/synapse-directed deliveryNanocarriers deliver co-stimulatory ligands or payloads to the contact area (T-cell bound).Focal payload release at moment of synapse formation; amplifies local signals without systemic exposure.Improves synapse stability and reduces systemic toxicity; modular co-delivery strategies.[114–117] Comparative synaptic architectures of major engineered therapeutics The synaptic basis of resistance The clinical failure modes observed with both modalities highlight shared and modality-specific vulnerabilities, but these should be interpreted as multifactorial rather than purely synaptic [69, 118]. To keep the clinical hierarchy explicit, the strongest current human evidence sits with approved CAR-T and bispecific programs, including ide-cel (median PFS 8.8 months in KarMMa) [93], randomized phase 3 evaluation of cilta-cel [119], and teclistamab supported by MajesTEC-1 and FDA review [120]. The 2024 IMWG bispecific-antibody guidance provides a useful practice anchor for this part of the review [121]. Genomic studies and laboratory functional assays have identified tumor-intrinsic mechanisms that can eliminate or alter the targeted antigen (for example point mutations, extracellular domain changes, splicing, and homozygous deletions of TNFRSF17/BCMA) in small series and case reports. These mechanistic data demonstrate biologically plausible routes to therapy failure in experimental and sequencing datasets [68, 98]. Clinically, cohort and case reports confirm that antigen-negative relapse occurs after BCMA- or GPRC5D-targeted therapies, although it is not uniformly common; a comprehensive genomic/clinical study documenting mechanisms of antigen escape in patients previously treated with BCMA- or GPRC5D-targeted therapies underscores the translational importance of this resistance pathway [68, 98, 122]. Importantly, these mechanistic failures have concrete clinical manifestations. Large clinical programs show high initial response rates but frequently limited durability: in the KarMMa trial of idecabtagene vicleucel (ide-cel), the overall response rate was high but median progression-free survival was modest (median PFS ≈ 8.8 months) [93]. In contrast, some BCMA CAR products reached longer median PFS in selected cohorts (for example, CARTITUDE-1 reported substantially prolonged PFS in the primary analysis), but outcomes are heterogeneous across centers and follow-up intervals, and long-term durable remissions remain a minority in many series [119]. For bispecific T-cell engagers, teclistamab produced deep responses in heavily pretreated patients, yet follow-up shows that a non-negligible fraction of responders subsequently progress, underscoring that early depth of response does not always translate to durable disease control [120]. These clinical observations highlight that relapse (both antigen-positive and antigen-negative), inter-patient variability in durability, and sanctuary-site biology are recurring, clinically meaningful synaptic failure modes that must be acknowledged explicitly when claiming therapeutic promise [93, 119, 120]. Divergence emerges when examining T cell intrinsic failure modes [99]. CAR T products are uniquely susceptible to manufacturing dependent variability in the starting T cell repertoire and to single infusion kinetics that can drive replicative stress and terminal differentiation, processes that predispose to exhaustion and impaired long term persistence in vivo [123, 124]. Bispecific engagers by design enlist circulating unmodified T cells repeatedly over time and can be dosed to modulate exposure, which in many series has produced a different pattern of T cell phenotypes and an opportunity to recover function between dosing intervals [50, 92]. Rather than acting as a static suppressive bath, the marrow microenvironment is compartmentalized and dynamic [11–17, 94, 125]. Single-cell RNA sequencing in precursor disease revealed early NK-cell expansion, loss of GrK+ memory cytotoxic T cells, and MHC class II dysregulation in CD14 + monocytes, while multi-region single-cell RNA/ATAC profiling of myeloma lesions showed a median of six tumor subclones per patient, site-specific transcriptional plasticity, and macrophage depletion in focal lesions [11, 12, 17]. More recent multiplexed imaging and single-cell proteomic studies further map immune-metabolic neighborhoods and demonstrate that plasma-cell/CD4 + T-cell interactions and tumor-immune distributions vary by marrow site and disease stage, with implications for progression risk and therapeutic response [13, 14, 16]. In this context, soluble mediators produced by tumor and stromal cells (notably TGF-β, IL-10 and IL-6) and MDSC-derived metabolic suppressors (arginase, reactive oxygen species, adenosine pathway enzymes) can dampen TCR signaling, alter co-receptor expression, and impair cytolytic granule trafficking even when a physical contact is formed, thereby producing functional synaptic failure via biochemical rather than purely geometric mechanisms [11, 12, 14, 16, 17]. In addition to steric or adhesive disruption of synaptic architecture, soluble mediators produced by tumor and stromal cells (notably TGF-β, IL-10 and IL-6) and MDSC-derived metabolic suppressors (arginase, reactive oxygen species, adenosine pathway enzymes) can dampen TCR signaling, alter co-receptor expression, and impair cytolytic granule trafficking even when a physical contact is formed, thereby producing functional synaptic failure via biochemical rather than purely geometric mechanisms. Furthermore, Th17-derived IL-17 has been reported to promote myeloma growth and to modify local immune responses in ways that may reinforce immunosuppressive circuits, while Treg expansion contributes classic contact-dependent and cytokine-mediated suppression; both axes therefore deserve explicit consideration alongside synaptic geometry when interpreting T-cell dysfunction in myeloma [25, 27]. However, the resilience of CAR T cells to repeated antigen exposure depends on intrinsic persistence and memory formation, so exhaustion or senescence acquired during manufacturing or immediately post infusion can convert what would otherwise be a durable synaptic bridge into a transient contact incapable of durable tumor control [123, 124]. Finally, these mechanistic insights point to rational solutions that are already entering the clinic [126, 127]. Dual antigen targeting, sequential or combination use of different synapse forming agents, modulation of dosing schedules for bispecifics, and CAR engineering strategies that minimize tonic signaling while preserving co-stimulation are all designed to restore or prolong high quality synaptic contacts and to impede antigen negative relapse [50, 127]. Overcoming the synaptic determinants of resistance therefore requires an integrated strategy that addresses both the physical requirements for sustained lytic junction formation and the cellular fitness of the effector T cells that generate them [69, 78]. Table 4 organizes the major synaptic failure modes in myeloma by mechanism and pairs each with a pragmatic counter-strategy. Table 4Common synaptic failure modes in myeloma and pragmatic counter-strategiesSynaptic failure modeMechanistic effect on synapsePractical counter-strategies (summary)Refs.Antigen escape (loss/cleavage/mutation)Removes physical ligand required for ternary/synaptic complexes → synapse fails.Dual/tri-antigen targeting, sequential modalities, monitor genomics; reduce antigen shedding (γ-secretase inhibitors).[68, 69, 98, 103]Soluble antigen decoys (e.g., sBCMA)Acts as sink for soluble engagers → reduces surface bridging.Combine with shedding inhibitors; increase local concentration in marrow; dual-target engagers.[70, 128]Microenvironmental suppression (MDSCs, Tregs, TGFβ)Lowers T cell fitness, increases checkpoint ligands → unstable/aborted synapses.Combine with MDSC/Treg depletion, armored CARs (dnTGFβR), niche-modifying agents.[19, 29, 105, 108, 129]Synaptic burnout/T-cell exhaustionChronic high-intensity synapses → metabolic collapse and epigenetic remodeling.Tune costimulatory domains (CD28 vs. 4-1BB), affinity/tonicity adjustments, inducible control modules and intermittent dosing strategies.[13, 51, 52, 130, 131]On-target off-tumor synapse formation (toxicity)Productive synapses with healthy tissue → severe toxicity.Affinity tuning, AND/NOT logic gating, protease-regulated or small-molecule gated CARs.[82, 87, 88, 132] Table 4Common synaptic failure modes in myeloma and pragmatic counter-strategiesSynaptic failure modeMechanistic effect on synapsePractical counter-strategies (summary)Refs.Antigen escape (loss/cleavage/mutation)Removes physical ligand required for ternary/synaptic complexes → synapse fails.Dual/tri-antigen targeting, sequential modalities, monitor genomics; reduce antigen shedding (γ-secretase inhibitors).[68, 69, 98, 103]Soluble antigen decoys (e.g., sBCMA)Acts as sink for soluble engagers → reduces surface bridging.Combine with shedding inhibitors; increase local concentration in marrow; dual-target engagers.[70, 128]Microenvironmental suppression (MDSCs, Tregs, TGFβ)Lowers T cell fitness, increases checkpoint ligands → unstable/aborted synapses.Combine with MDSC/Treg depletion, armored CARs (dnTGFβR), niche-modifying agents.[19, 29, 105, 108, 129]Synaptic burnout/T-cell exhaustionChronic high-intensity synapses → metabolic collapse and epigenetic remodeling.Tune costimulatory domains (CD28 vs. 4-1BB), affinity/tonicity adjustments, inducible control modules and intermittent dosing strategies.[13, 51, 52, 130, 131]On-target off-tumor synapse formation (toxicity)Productive synapses with healthy tissue → severe toxicity.Affinity tuning, AND/NOT logic gating, protease-regulated or small-molecule gated CARs.[82, 87, 88, 132] Common synaptic failure modes in myeloma and pragmatic counter-strategies Beyond BiTEs and CAR-Ts: the next generation of synaptic engineering Trispecific antibodies: adding a third arm to recreate signal two and target niche cues Trispecific antibodies extend the minimalist bridge concept by adding a third functional arm that can provide local co-stimulation or target a microenvironmental cue, but this platform should still be described as investigational in multiple myeloma. Preclinical studies support the concept that trispecific designs can enhance T-cell activation and serial killing, yet the current human evidence remains limited to first-in-human or abstract-level reports in small, heavily pretreated cohorts, with short follow-up and no randomized comparative data. Accordingly, the clinical feasibility of trispecific antibodies remains promising, but it is not yet established in randomized myeloma data [100, 102]. Finally, trispecific formats permit pharmacologic tuning of synaptic signals: conditional CD28 provision limits systemic costimulation while maximizing localized signal two within the synaptic cleft, a property that both amplifies efficacy and aims to improve safety relative to systemic agonists [101]. “Armored” and logic-gated CAR-Ts: reinforcing the synapse from within Armored CAR-T cells that autonomously secrete cytokines or express dominant-negative receptors constitute an intracellular approach to synaptic engineering where the T cell becomes both the sensor and a self-contained amplifier of the synaptic milieu [106]. Multiple preclinical at CAR-T cells engineered to produce cytokines such as IL-7 and chemokines such as CCL19 improve T-cell persistence and recruit endogenous immune subsets in tumor models; these data support the concept of an intracellular “armoring” strategy to sustain synaptic function. Early human trials of armored CAR constructs have demonstrated feasibility in small cohorts, but durable clinical benefit remains unproven: available evidence is limited to early-phase, uncontrolled studies and requires confirmation in larger, randomized trials before durability can be claimed [108, 109]. Conversely, incorporation of dominant-negative transforming growth factor-β receptors (dnTGFβR) into CAR-T cells protects synaptic signaling from paracrine suppression by TGFβ abundant in the bone marrow niche and restores cytotoxic programs that would otherwise be attenuated [110, 113]. These armored strategies are relevant in multiple myeloma, where stromal TGF-β, adenosine and other suppressive mediators reduce synaptic efficacy. Armored CARs can resist these suppressive signals (for example via dominant-negative receptors or cytokine secretion), thereby preserving synapse formation and effector function in suppressive microenvironments [106, 113]. In parallel, logic-gated receptor circuits such as synNotch AND gates and NOT gate systems allow CAR-T cells to restrict full activation to spatially and antigenically defined contexts, thereby increasing synaptic specificity and reducing off-tumor synapse formation [107, 111]. Recent demonstrations of NOT-gated and combinatorial designs applied to myeloma antigen sets show that logical gating preserves potency while materially improving the safety margin of synapse formation in heterogeneous antigen landscapes [112]. Targeting the synaptic environment: making bone marrow real estate permissive Engineering the synapse is not limited to the receptor or the soluble bridge; altering the physical and adhesive properties of the bone marrow niche can materially increase productive encounter rates and synaptic durability [53]. Very late antigen 4, integrin α4β1, mediates firm adhesion of myeloma cells to stromal elements and thereby sequesters targets in protective microniches that reduce immune encounter probability [54]. Preclinical ablation or pharmacologic blockade of VLA-4 disrupts medullary retention, repositions malignant plasma cells and sensitizes them to immunotherapies, indicating that pharmacologic modulation of adhesion converts inaccessible targets into synapse-competent substrates [133, 134]. Small-molecule α4 antagonists and combinations with CXCR4 inhibitors have been pursued as mobilizing strategies to increase circulating tumor cell exposure and thereby raise the per-cell encounter frequency required for robust synapse nucleation by BiTEs or CAR-T cells [115, 116, 135]. Translationally, these tactics suggest combination regimens where transient niche disruption or mobilization precedes synapse-forming immunotherapies to improve access to sanctuary sites such as the endosteal niche and anatomical marrow compartments [114]. A schematic overview of different therapeutic strategies targeting the immunological synapse in multiple myeloma is shown in Fig. 2, illustrating how each modality engages T cells to promote activation, proliferation, and tumor cell killing. Fig. 2Targeting the immunological synapse in multiple myeloma. Therapeutic strategies that modify the T-cell–tumor synapse in multiple myeloma. Schematic showing how BiTEs, CAR-T cells, trispecific engagers, armored/logic-gated CARs and nanoparticle delivery each alter intercellular geometry, co-stimulation and local signaling to increase productive contacts, persistence or safety. The panel emphasizes the different engineering approaches and their principal clinical trade-offs. Created by the authors using BioRender.com Targeting the immunological synapse in multiple myeloma. Therapeutic strategies that modify the T-cell–tumor synapse in multiple myeloma. Schematic showing how BiTEs, CAR-T cells, trispecific engagers, armored/logic-gated CARs and nanoparticle delivery each alter intercellular geometry, co-stimulation and local signaling to increase productive contacts, persistence or safety. The panel emphasizes the different engineering approaches and their principal clinical trade-offs. Created by the authors using BioRender.com Nanotechnology and synaptic delivery: precision placement of signal two Nanoparticle platforms are best presented as an experimental delivery strategy that can localize immunomodulatory payloads near the synapse in preclinical systems. Foundational work showed that T-cell–linked nanoparticles can polarize toward the nascent immunological synapse and deliver payloads at the contact zone, but these approaches have not yet reached robust clinical validation in multiple myeloma. For that reason, nanoparticle-based synaptic engineering should be framed as a future platform with translational promise rather than as a clinically established therapy [136, 137]. Challenges and future direction The promise of restoring or engineering immunological synapses in multiple myeloma is biologically compelling, but the supporting evidence is uneven across strategies, spanning preclinical, early-phase, and randomized clinical data [69, 129]. Nevertheless, a synapse-centric framework has important limitations. In newly diagnosed multiple myeloma, terminal exhaustion and antigen-experience signatures were not consistently detected from diagnosis through maintenance therapy, which cautions against using exhaustion markers as universal surrogates for synaptic failure [138]. Likewise, response to BCMA×CD3 bispecifics depends strongly on the pre-existing T-cell landscape and baseline immune fitness, showing that T-cell state can be as important as synapse geometry for outcome [139, 140]. Finally, antigen escape is real but not the sole mechanism of relapse: BCMA extracellular-domain mutations, TNFRSF17 deletions, and GPRC5D loss have all been documented after targeted therapy, but these data also show heterogeneous tumor-intrinsic resistance pathways rather than one universal failure mode [68]. Even when productive synapses are restored, residual disease can persist through clonal diversification, altered metabolism, or spatially segregated marrow niches, so synapse-directed strategies should be developed as part of a multiscale control framework rather than as a stand-alone solution [5, 6, 8, 9]. Antigenic escape and lineage plasticity remain central obstacles for synapse-directed therapies because loss, downregulation or heterogenous expression of targets such as BCMA or other plasma cell antigens rapidly reduce the fraction of malignant cells capable of forming productive synaptic contacts with engineered effectors [69, 118]. The bone marrow niche is not a passive backdrop but an active architect of synaptic failure through suppressive myeloid populations, Treg cells, soluble mediators and a remodeled extracellular matrix that together impede receptor clustering, force transmission and metabolic support for sustained cytolytic activity [141, 142]. Importantly, distal signals such as gut-microbiota derived metabolites, especially short-chain fatty acids (SCFAs), are increasingly recognized modulators of systemic T-cell physiology and exhaustion states and may influence antitumor responses and immunotherapy outcomes; early myeloma-focused studies suggest that depletion of SCFA-producing bacteria correlates with disease status and that SCFA levels may predict treatment response, arguing that microbiome-derived biochemistry should be considered when studying T-cell dysfunction in myeloma [143–145]. Practically, this implies that synapse-centric mechanistic studies and clinical correlative studies should measure soluble mediators (TGF-β, IL-10, IL-6), MDSC functional markers (arginase, ROS, CD38/adenosine pathway), and, where feasible, gut microbiome composition or circulating SCFAs as part of an integrated biomarker strategy to distinguish geometric synaptic failure from biochemical suppression [24, 26]. To avoid conflating mechanistic promise with clinical maturity, the next-generation approaches below are graded according to the highest currently available evidence tier, with preclinical/modeling findings presented as hypothesis-generating and clinical trial data presented as outcome-level evidence (Table 5). Table 5Evidence grading of key synapse-engineering strategies in multiple myelomaStrategyHighest current evidence levelSuggested interpretationBiTEs/bispecific antibodiesPhase 1–2 clinical dataClinically active, but still largely supported by early-phase trials in myeloma.CAR-T cellsRandomized clinical data for cilta-cel; phase 2 for ide-celStrongest clinical evidence among synapse-engineering platforms.Trispecific antibodiesPreclinical and early-phase clinicalPromising, but still emerging; should be described as investigational.Armored/logic-gated CAR-TsPreclinical; limited early-phase feasibilityMechanistically compelling, but not yet established clinically.Niche-modifying agentsMostly preclinical / translationalBest presented as combination-enabling strategies.Synaptic nanotechnologyPreclinicalShould be framed as a future platform rather than a clinical approach. Table 5Evidence grading of key synapse-engineering strategies in multiple myelomaStrategyHighest current evidence levelSuggested interpretationBiTEs/bispecific antibodiesPhase 1–2 clinical dataClinically active, but still largely supported by early-phase trials in myeloma.CAR-T cellsRandomized clinical data for cilta-cel; phase 2 for ide-celStrongest clinical evidence among synapse-engineering platforms.Trispecific antibodiesPreclinical and early-phase clinicalPromising, but still emerging; should be described as investigational.Armored/logic-gated CAR-TsPreclinical; limited early-phase feasibilityMechanistically compelling, but not yet established clinically.Niche-modifying agentsMostly preclinical / translationalBest presented as combination-enabling strategies.Synaptic nanotechnologyPreclinicalShould be framed as a future platform rather than a clinical approach. Evidence grading of key synapse-engineering strategies in multiple myeloma The spatial and metabolic heterogeneity within the marrow compartment creates microdomains where antigen density, oxygen tension and nutrient availability diverge, producing pockets where even well-designed receptors cannot assemble or maintain an effective supramolecular activation cluster [146, 147]. These microdomains are now measurable with spatial transcriptomics, multiplex immunofluorescence, imaging mass cytometry, and single-cell RNA/ATAC or proteomic profiling, which together show that exhausted T cells, macrophage-depleted lesions, and distinct plasma-cell neighborhoods can be spatially segregated within the same patient and can differ in checkpoint expression, metabolic wiring and target-antigen abundance [11–14, 16, 17]. Translational experience to date underscores the clinical consequences of these mechanistic barriers because deep initial responses to CAR or bispecific platforms are commonly followed by relapse that reflects antigen loss, immune exclusion or functional attrition of effector cells [69, 129]. Understanding and predicting which of these failure modes will dominate in an individual patient is currently limited by insufficient spatially resolved biomarkers and by a paucity of longitudinal synapse-centric assays that can be deployed in clinical cohorts [13, 147]. Addressing this knowledge gap will require integrating multiplexed spatial proteomics, single-cell functional assays and in situ measures of synaptic nanostructure into prospective trials so that mechanistic endpoints can guide adaptive engineering strategies [13, 146]. A related technical frontier concerns the clinical durability of engineered synapses. Mechanistically, chronic or excessive synaptic engagement can produce ‘synaptic burnout’ signaling, metabolic strain, epigenetic remodeling) that impairs long-term effector function [129, 148]. Clinically, this mechanistic vulnerability translates into measurable relapse rates and variable durability across products and patients: for example, ide-cel produced high response rates but a median PFS of ~ 8.8 months in the KarMMa cohort, demonstrating that early depth of response often does not equate to long-term disease control in many patients [93]. Conversely, certain cohorts treated with cilta-cel achieved substantially longer PFS in controlled studies, illustrating how product design, patient selection, and follow-up time interact to shape durability [119]. For T-cell engagers, teclistamab and other BsAbs show promising deep responses, but a substantial minority of responders ultimately relapse and require salvage therapy, emphasizing inter-patient variability in both magnitude and duration of benefit [120, 149]. Together, these clinical patterns argue for (i) frank acknowledgment of relapse and variable durability as central ‘synaptic failure modes’, (ii) systematic incorporation of longitudinal, synapse-centric biomarkers into trials to prospectively capture these outcomes, and (iii) testing of combination or sequencing strategies aimed explicitly at preventing or reversing synaptic collapse [93–120, 149]. Rational design choices such as co-stimulatory motifs, spacer architecture, affinity tuning and inducible control modules influence not only initial potency but also long term metabolic fitness and transcriptional fate, and these design variables must be optimized with an eye to preventing premature synaptic collapse [51, 148]. Importantly, recent data caution that T cell dysfunction in myeloma is nuanced because classical hallmarks of terminal exhaustion may be absent at diagnosis for many patients, implying that therapeutic pressures and longitudinal antigen exposure rather than baseline disease state often precipitate irreversible synaptic failure [138, 150]. Safety concerns remain a dominant constraint on aggressive synapse engineering because on-target off-tumor synapse formation and bystander activation can produce severe, sometimes fatal toxicities as shown by early clinical reports and by the real world experience of cytokine release and neurotoxicity [132, 151, 152]. Engineering safety into synapse design will therefore be essential, and promising approaches include protease-regulated or small molecule gated receptors, combinatorial logic circuits that require dual antigen recognition to form a productive synapse, and affinity tuning to increase discrimination of high antigen density on malignant cells versus low density on normal tissues [69, 82]. These modalities must be evaluated not only for their ability to reduce acute toxicity but also for unintended effects on synaptic geometry and long term efficacy [82, 153]. Looking forward, several concrete priorities can accelerate progress [146, 147]. First, routine incorporation of spatially resolved, synapse-focused biomarker panels into early phase trials will enable correlation of synaptic architecture with outcome and toxicity [13, 146]. Second, combinatorial strategies that pair synapse-restorative microenvironmental interventions such as MDSC depletion or matrix remodeling with engineered receptors should be systematically tested because microenvironmental reprogramming can convert otherwise suppressive niches into permissive arenas for productive synapse formation [142, 154]. Third, adaptive receptor architectures that can modulate their signaling tempo or assembly state in vivo offer a path to maximize on-tumor killing while minimizing chronic overstimulation and burnout [82, 148]. Finally, deeper mechanistic interrogation of synaptic biomechanics and of key nodes such as PKCθ, actin regulators and mitochondrial coupling will provide tractable engineering targets to build synapses that are both potent and durable [96]. To explore how synapse engineering might eventually inform treatment decisions, we propose a pragmatic, three-part framework for future study. First, patient selection and sequencing should be guided by tumor burden and antigen integrity (for example soluble-BCMA and imaging-derived metabolic tumor volume), since lower burden with intact antigen appears to favor consolidation with autologous CAR-T while high burden or urgent need for rapid therapy favors off-the-shelf bispecifics [125]. Second, combination strategies should be used to preserve or restore synaptic competence: pair synapse-forming agents with niche-modifying measures (γ-secretase inhibition to reduce BCMA shedding, IMiD/CELMoD backbone or CD38 modulation to improve T-cell fitness, and selective MDSC/Treg targeting) and consider trispecific or dual-antigen formats to reduce single-antigen escape [96, 155, 156]. Third, adaptive monitoring and stratification are essential, longitudinal sBCMA, spatially resolved immune biomarkers and early measures of T-cell fitness should trigger sequencing (switch to alternative synapse platforms, add niche-directed agents, or modify dosing) to prevent synaptic burnout and to personalize therapy [157, 158]. Collectively, these principles provide a hypothesis-generating framework for future sequencing, combination selection, and patient stratification, but they require prospective clinical validation before they can inform routine practice. In conclusion, reframing therapeutic goals from simply delivering cytotoxic receptors to constructing engineered synaptic bridges that are specific, resilient and context aware offers a roadmap for durable immunologic control of multiple myeloma; realizing that roadmap will require cross-disciplinary investment in spatial technologies, rational receptor engineering and controlled clinical experimentation [82, 147]. Conclusion This review shows that the immunological synapse is a useful organizing concept for understanding both immune escape and therapeutic design in multiple myeloma. BiTEs and CAR-T cells differ in how they build a synapse, but both aim to restore directed cytotoxicity. Their clinical impact is limited by antigen loss, marrow niche suppression, and T-cell fitness, which means that synapse engineering should be combined with biomarker-guided patient selection, niche modulation, and evidence-based interpretation of preclinical findings. Next-generation approaches such as trispecific antibodies, armored CAR-T cells, and synapse-directed delivery remain promising, but they should be presented as emerging strategies rather than established standards until stronger clinical data are available. For multiple myeloma, a disease defined by an immunosuppressive niche that dismantles productive immune contacts, this shift is paramount. As this review has detailed, strategies like BiTEs and CAR-T cells are fundamentally synaptic interventions. BiTEs serve as soluble, minimalist bridges that nucleate cytolytic junctions, while CAR-T cells build enhanced, membrane-anchored superstructures that redefine T-cell activation. Both modalities converge on the essential goal of restoring directed cytotoxicity but diverge in their architectural philosophies and temporal dynamics, leading to distinct clinical profiles and resistance mechanisms. However, the clinical effect of these synapse-directed therapies remains conditional on disease burden, prior therapy, baseline immune fitness, and patient-specific factors, so synaptic engineering should be framed as a mechanistic framework rather than a complete explanation for outcome. The future of myeloma therapy may be informed by continued development of this synaptic engineering paradigm, provided that clinical validation continues to confirm safety, durability, and benefit. Next-generation strategies, such as trispecific antibodies with integrated co-stimulation, armored CAR-T cells designed to resist microenvironmental suppression, and nanotechnology for precision signal delivery, may help create synapses that are potent, durable, and specific. At present, however, trispecific antibodies and nanoparticle-based approaches should be described as promising but still experimental, because the human evidence is mainly limited to small, nonrandomized early-phase studies with short follow-up, whereas the most mature clinical evidence remains concentrated in approved CAR-T and bispecific antibody programs. Overcoming challenges like antigen escape, synaptic burnout, and on-target off-tumor toxicity will require a deep integration of biophysical principles, spatial biology, and rational receptor design. However, aspiration must be tempered with clinical realism: current CAR-T and T-cell engager programs demonstrate important successes but also frequent relapses and heterogeneous durability, and these outcomes are shaped not only by synapse quality but also by baseline immune fitness, tumor burden, and antigen landscape. Translating synaptic engineering into routine long-term control will require iterative clinical validation and close attention to these failure modes.