B cell lymphoma 10 (BCL-10) and mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) were originally identified from oncogenic chromosomal translocations that occur in mucosa-associated lymphoid tissue (MALT) lymphomas, which are B cell malignancies that typically arise in the context of chronic inflammation1,2,3,4,5. Subsequent studies in gene-targeted mice showed that BCL-10 and MALT1 are key regulators of physiological antigen receptor signalling in B cells and T cells, which is required for adaptive immunity6,7,8. During lymphocyte activation, the two proteins bind to each other9,10 and inducibly interact with the caspase-recruitment domain (CARD)-containing coiled-coil protein 11 (CARD11; also known as CARMA1)11,12 to form the tripartite CARD11–BCL-10–MALT1 signalosome, which activates nuclear factor-κB (NF-κB) signalling and other inflammatory pathways in activated lymphocytes.

Related CARD protein–BCL-10–MALT1 (CBM) signalosomes have since been discovered in almost all tissues. The CARD11 homologues CARD9, CARD10 (also known as CARMA3) and CARD14 (also known as CARMA2) engage BCL-10 and MALT1 in response to a wide range of stimuli to form CBM complexes, which mediate cell type-dependent and context-dependent inflammatory responses. Not only are these signals essential for host defence and tissue homeostasis, but also genetic alterations in these pathways have emerged as key determinants of various human immunodeficiencies, inherited lymphoproliferative disorders, psoriasis, inflammatory bowel disease (IBD) and other autoinflammatory disorders, as well as leukaemias and lymphomas of B cell and T cell origin. This Review summarizes our current knowledge of the molecular regulation and physiological functions of the individual CBM signalosomes. It provides a framework to understand their pathophysiological roles and therapeutic implications for human immune-mediated diseases.

Structures and expression patterns

All CBM signalosomes have a modular composition with tissue-restricted expression patterns. BCL-10 and MALT1, which are ubiquitously expressed, are constitutively pre-assembled and are, by definition, included in all CBM complexes. BCL-10 is the central adaptor protein within these structures. It contains an amino-terminal CARD1 that mediates homophilic interactions with the CARDs of CARD9, CARD10, CARD11, CARD14 and BCL-10 itself13,14,15 (Fig. 1). In addition, BCL-10 has a serine/threonine-rich region that can be phosphorylated and mediates the interaction with MALT1 (ref.10). In turn, MALT1 has both a scaffolding function and proteolytic activity (and is therefore also referred to as paracaspase)9. It contains a death domain, three immunoglobulin-like domains and the catalytic caspase-like domain that is unique to the mammalian proteome.

Fig. 1: Structural organization of CBM complex components, showing sites of post-translational modification.
figure 1

B cell lymphoma 10 (BCL-10) consists of an amino-terminal caspase-recruitment domain (CARD) and a serine/threonine-rich (S/T-rich) carboxy-terminal region. Mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) comprises an amino-terminal death domain (DD), two immunoglobulin-like (Ig) domains adjacent to a caspase-like catalytic domain and a third, carboxy-terminal Ig domain. CARD-containing coiled-coil protein 9 (CARD9), CARD10 (also known as CARMA3), CARD11 (also known as CARMA1) and CARD14 (also known as CARMA2) consist of homologous amino-terminal CARD and coiled-coil (CC) regions. In addition, CARD10, CARD11 and CARD14 contain a tripartite membrane-associated guanylate kinase (MAGUK) domain, consisting of a PDZ domain, a SRC homology 3 (SH3) domain and a carboxy-terminal guanylate kinase-like (GUK) domain. Target sites for ubiquitylation are depicted with the respective linkage type of ubiquitin (Ub) chains (K63-linked or K27-linked). MALT1 paracaspase cleavage sites within BCL-10 and MALT1 are symbolized by the scissors. Phosphorylation sites and the known kinases mediating activating phosphorylation (shown in green) or inhibitory phosphorylation (shown in red) are also indicated. CAMK2, calcium/calmodulin-dependent protein kinase type II; CBM, CARD protein–BCL-10–MALT1; CK1α, casein kinase 1α; CK2, casein kinase 2; IKKβ, IκB kinase-β; PKC, protein kinase C.

The BCL-10–MALT1 module is rapidly engaged by either CARD9, CARD10, CARD11 or CARD14 (Fig. 2). All four CARD proteins have highly homologous CARD and coiled-coil domains for protein oligomerization. In addition, CARD10, CARD11 and CARD14 have PDZ, SRC homology 3 (SH3) and guanylate kinase-like (GUK) domains, which together define the membrane-associated guanylate kinase (MAGUK) region. Therefore, these three molecules are also known as CARMA (CARD and MAGUK) proteins. MAGUK domains typically organize signalling machineries at cell membranes; for example, the MAGUK domain of CARD11 mediates assembly of CBM signalosomes at the immune synapse in activated T cells11,12,16. The MAGUK domains of CARD10 and CARD14 probably carry out similar functions in other cell types. The coiled-coil and MAGUK domains of CARD10, CARD11 and CARD14 are connected by an autoinhibitory linker region that maintains these molecules in an inactive state in non-stimulated cells. CARD9 is unique within this family, as it lacks the MAGUK and linker regions. The mechanisms that maintain CARD9 in the inactive state and recruit it to upstream receptors upon stimulation currently remain unclear.

Fig. 2: Cell type-specific activation of CBM complexes.
figure 2

a | Caspase-recruitment domain (CARD)-containing coiled-coil protein 9 (CARD9) is expressed in myeloid cells and activated by pattern-recognition receptors (PRRs) of the SYK-coupled C-type lectin receptor (CLR) family (such as dectin 1, dectin 2 and mincle) in response to fungi or mycobacteria or by the IgG-binding Fcγ receptor (FcγR), which use protein kinase Cδ (PKCδ) for CARD9 phosphorylation. In addition, CARD9 is also activated by cytoplasmic PRRs, including nucleotide-binding oligomerization domain-containing protein 2 (NOD2), a sensor for intracellular bacteria; RAD50, which senses viral DNA; and retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated gene 5 (MDA5), receptors for viral RNA. CARD10 is expressed in various non-haematopoietic cell types, including airway epithelial cells, endothelial cells and cardiomyocytes, and is activated by receptors of the epidermal growth factor receptor (EGFR) and G protein-coupled receptor (GPCR) families, which both use PKC for CARD10 phosphorylation. GPCRs activate PKC via the coupled Gα subunit. CARD11 is expressed in lymphoid cells and mast cells and is activated upon T cell receptor (TCR) stimulation via ZAP70 and PKCθ and upon B cell receptor (BCR) stimulation via SYK and PKCβ. CARD11 is also activated upon stimulation of activating natural killer (NK) cell receptors (for example, NKGD2) and the mast cell FcεRI receptor, which use SYK and/or ZAP70 and unspecified PKC members. CARD14 is mainly expressed in keratinocytes and is activated by dectin 1 via PKC or by the IL-17 receptor (IL-17R) through a mechanism involving nuclear factor-κB (NF-κB) activator 1 (ACT1). b | The activated CARD proteins engage B cell lymphoma 10 (BCL-10) and mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) and induce the formation of filaments that activate NF-κB and mitogen-activated protein kinase (MAPK) pathways for inflammatory gene transcription and induce MALT1 paracaspase activity, which modulates NF-κB and MAPK feedback and inactivates mRNA regulators to control inflammatory gene expression. CBM, CARD protein–BCL-10–MALT1.

CARD9 and CARD11 are specifically expressed by haematopoietic cells, and CARD9 expression is restricted to myeloid cells17,18 (Fig. 2a). CARD9 couples several pattern-recognition receptors (PRRs) and other innate immune receptors to BCL-10 and MALT1 in dendritic cells (DCs), macrophages and neutrophils19. These receptors include immunoreceptor tyrosine-based activation motif (ITAM)-containing and SYK-coupled C-type lectin receptors (CLRs) such as dectin 1 (ref.17), dectin 2 (ref.20) and mincle (also known as CLEC4E)21, the Fcγ receptor FcγRIII, myeloid-associated immunoglobulin-like receptor II (MAIR-II; also known as CD300D), osteoclast-associated immunoglobulin-like receptor (OSCAR) and triggering receptor expressed on myeloid cells 1 (TREM1)18. CARD9 is also activated by cytosolic PRRs, specifically, the nucleic acid sensors retinoic acid-inducible gene I (RIG-I), melanoma differentiation-associated gene 5 (MDA5; also known as IFIH1)22 and RAD50 (ref.23), as well as by nucleotide-binding oligomerization domain-containing protein 2 (NOD2)24. CARD11 is expressed at high levels by B cells and T cells but is also expressed by natural killer (NK) cells and mast cells18. CARD11 signals from the B cell receptor (BCR), T cell receptor (TCR)11,12,25,26, activating NK cell receptors27 and the high-affinity receptor for IgE, FcεRI28,29 (Fig. 2a). CARD11 is also expressed by DCs and macrophages, but in contrast to CARD9, CARD11 is dispensable for cytokine production upon stimulation of ITAM-coupled receptors in these cells18.

CARD10 and CARD14 are broadly expressed in non-haematopoietic tissues, with the highest levels of CARD10 observed in the heart, kidney, lungs, intestine and endothelia30,31. CARD10 is activated by G protein-coupled receptors (GPCRs), such as the receptors for angiotensin II32 or lysophosphatidic acid (LPA)33, and by growth factor receptor tyrosine kinases (RTKs) of the epidermal growth factor receptor (EGFR) family34 (Fig. 2a). CARD14 is expressed at high levels in the epidermis, specifically in keratinocytes30,35, dermal endothelial cells36 and Langerhans cells37, and it has also been detected in the colon, lung and placenta38,39. In keratinocytes, CARD14 signalling is triggered by the IL-17 receptor40, dectin 1 (ref.41) and possibly other PRRs that remain to be identified (Fig. 2a).

Orthologues of the CARD-containing coiled-coil proteins, as well as of BCL-10 and MALT1, have been identified in ray-finned fishes, which indicates that CBM signalling is an evolutionarily conserved mechanism beyond the mammalian system42.

CBM complex assembly and NF-κB activation

The activation of CBM signalling is a highly regulated process that requires multiple post-translational modifications of CBM complex components (Fig. 3). The activation of CARD11–BCL-10–MALT1 signalling in T cells is the best understood pathway, and high-resolution structures of these complexes are available43,44 that function as models to facilitate our understanding of CBM signalling systems. Briefly, upon co-stimulation through the TCR and CD28, a series of steps initiated by the protein tyrosine kinase ZAP70 (reviewed previously45) lead to the activation of protein kinase Cθ (PKCθ), which directly phosphorylates CARD11 within the linker region46,47 through a mechanism that is controlled by the adaptor protein ADAP (also known as FYB1)48 and phosphoinositide-dependent kinase 1 (PDK1; also known as PDPK1), which bridges PKCθ to CARD11 (ref.49). In addition, casein kinase 1α (CK1α; also known as CSNK1A1) interacts with CARD11 (ref.50), and AKT and calcium/calmodulin-dependent protein kinase type II (CAMK2) further phosphorylate CARD11 (refs51,52) to positively regulate the pathway. Together, the activating phosphorylations of CARD11 neutralize four repressive elements within the linker region and thereby relieve intramolecular autoinhibition53.

Fig. 3: Molecular mechanisms activating CBM signalling in T cells.
figure 3

Upon T cell receptor (TCR) ligation and CD28 co-stimulation, the caspase-recruitment domain (CARD)-containing coiled-coil protein 11 (CARD11) linker region is phosphorylated by protein kinase Cθ (PKCθ) to activate CARD11. Phosphoinositide 3-kinase (PI3K) activity induces CARD11 phosphorylation at additional sites by phosphoinositide-dependent kinase 1 (PDK1) and AKT. IκB kinase-β (IKKβ) also phosphorylates CARD11 to amplify the signal. Interactions with casein kinase 1α (CK1α) and the adaptor protein ADAP support CARD11 activation. Activated CARD11 oligomerizes and nucleates B cell lymphoma 10 (BCL-10)–mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) units, resulting in the assembly of filamentous CARD11–BCL-10–MALT1 complexes that recruit E3 ubiquitin ligases — including tumour necrosis factor (TNF) receptor-associated factor 6 (TRAF6), cellular inhibitor of apoptosis 1 (cIAP1), cIAP2 and linear ubiquitin chain assembly complex (LUBAC) — to mediate K63-linked and M1-linked (linear) ubiquitylation (Ub) of the CARD protein–BCL-10–MALT1 (CBM) signalosome. The IKK complex is recruited to the signalosome and activated by TGFβ-associated kinase 1 (TAK1), together with TAK1-binding protein 2 (TAB2) and TAB3, and via IKK trans-autophosphorylation. Active IKK phosphorylates IκBα to induce its proteasomal degradation, releasing nuclear factor-κB (NF-κB) dimers that translocate to the nucleus. In addition, CBM signalling activates the stress kinases p38 and JNK through TAK1 and MKK7, promotes upregulation of the amino acid transporter ASCT2 and induces the activation of MALT1 proteolytic activity. Activating phosphorylation is shown in green; inhibitory phosphorylation is shown in red. DD, death domain; Ig, immunoglobulin-like; MAGUK, membrane-associated guanylate kinase; S/T-rich, serine/threonine-rich.

Activated CARD11 molecules oligomerize through their coiled-coil, SH3 and GUK domains and subsequently nucleate BCL-10–MALT1 heterodimers through CARD–CARD interactions. This triggers a prion-like process that results in the self-polymerizing assembly of BCL-10–MALT1 modules43 with a helical BCL-10 filament in the core and MALT1 molecules in the periphery at a 1:1 molar ratio with BCL-10 (ref.43). Because somewhat few activated CARD11 molecules are sufficient to induce the polymerization of a large number of BCL-10 and MALT1 molecules in an energetically favoured process, the assembly of the CARD11–BCL-10–MALT1 signalosome enables robust amplification of the initial immunoreceptor signal, which is comparable to the prion-like polymerization of innate immune signalling complexes containing mitochondrial antiviral signalling protein (MAVS) after RNA virus detection or of the adaptor protein ASC (also known as PYCARD) for inflammasome activation54.

The assembled CARD11–BCL-10–MALT1 signalosomes function as central scaffolds in activated lymphocytes that position ubiquitin regulators and protein kinases in close proximity for the activation of NF-κB and mitogen-activated protein kinase (MAPK) cascades. The E2 ubiquitin-conjugating enzymes UBC13 (also known as UBE2N) and UEV1A (also known as UBE2V1), the E3 ubiquitin ligases tumour necrosis factor (TNF) receptor-associated factor 2 (TRAF2), TRAF6, cellular inhibitor of apoptosis protein 1 (cIAP1; also known as BIRC2) and cIAP2 (also known as BIRC3), and the linear ubiquitin chain assembly complex (LUBAC) are all recruited to the CBM complex and catalyse regulatory mono-ubiquitylation and poly-ubiquitylation of multiple proteins, including BCL-10 and MALT1 themselves. BCL-10 undergoes K63-linked ubiquitylation mediated by cIAP1 and cIAP2 (refs55,56) and linear ubiquitylation mediated by LUBAC57, whereas MALT1 undergoes K63-linked ubiquitylation mediated by TRAF6 and possibly TRAF2 (refs58,59). Ubiquitylation of CARD11–BCL-10–MALT1 signalosomes provides platforms for the ubiquitin-binding proteins TAK1-binding protein 2 (TAB2) and TAB3 and the recruitment of TGFβ-associated kinase 1 (TAK1; also known as MAP3K7)58. In addition, IκB kinase-α (IKKα) and IKKβ are recruited through the ubiquitin-binding IKKγ subunit (also known as NEMO), which is further ubiquitylated by E3 ligases, including TRAF6 and possibly LUBAC60,61. Together, these mechanisms lead to the catalytic activation of IKKs by TAK1 and by trans-autophosphorylation58 and the subsequent activation of canonical NF-κB signalling pathways through the phosphorylation and degradation of IκBα. Active IKKβ can provide additional positive feedback at early stages of CBM signalling by phosphorylating CARD11 (ref.62). In addition, the activated CARD11–BCL-10–MALT1 complex activates the stress-activated MAPKs p38 and JNK, the latter of which occurs through the recruitment of the MAPK kinase MKK7 (also known as MAP2K7) and TAK1, which mediate activation of the JUN subunit of the activator protein 1 (AP-1) transcription factor29. However, several of the CBM-associated E3 ligases that positively contribute to T cell activation have at least partially redundant functions or might function during only specific stages of T cell development, as TRAF6-deficient T cells activate NF-κB in a normal manner63, and the catalytic activity of LUBAC is dispensable for TCR-induced NF-κB signalling64 under certain conditions.

Although the specific intermediates differ between BCR signalling and TCR signalling, BCR ligation activates CARD11–BCL-10–MALT1 signalosomes in an analogous manner to TCR ligation, whereby SYK and PKCβ activation are required to induce CARD11 phosphorylation46,47 (Fig. 2a). In addition, CARD10 and CARD14 are similarly activated by PKC isoforms, and the resulting CBM complexes also use E3 ligases, including TRAF2 and TRAF6, for their effector functions33,39,41,65,66,67. Moreover, although CARD9 lacks the MAGUK and linker domains, the CARD9 coiled-coil domain is phosphorylated by PKCδ68 in a VAV-dependent manner69 after SYK-coupled CLR signalling, and this phosphorylation is required for assembly of the CARD9–BCL-10–MALT1 complex, which then requires K27-linked ubiquitylation of CARD9 by TRIM62 (ref.70) for NF-κB activation.

Thus, although structural data for CBM signalosomes containing CARD9, CARD10 or CARD14 are currently unavailable and many of the individual mediators that control CARD9, CARD10 and CARD14 activation in innate immune signalling through CBM complexes remain to be defined, all CBM complexes are activated by PKC isoforms and assemble functionally — and presumably architecturally — highly homologous CBM complexes that use ubiquitin ligases for the activation of NF-κB and MAPK signalling pathways (Fig. 2b). These pathways are a major output of CBM signalling, which, depending on the chromatin accessibility of a given cell type, induce the expression of genes encoding a large number of inflammatory cytokines (such as TNF, IL1B and IL6), chemokines (such as CXCL8 and CCL20) and factors that control cellular proliferation, differentiation and survival (such as CCND1, BCL2L and IRF4) to orchestrate innate and adaptive immune responses to pathological threats71. As a result of the pro-inflammatory potency of activated CBM signalosomes, these complexes require stringent counter-regulation to ensure homeostasis of the organism. These mechanisms are summarized in Box 1.

Activation of MALT1 catalytic activity

The second major output of activated CBM signalosomes is activation of the proteolytic activity of MALT1, the biological importance of which is increasingly being recognized (Fig. 4). This activity can be detected 10 minutes after immediate IKK activation and requires MALT1 dimerization within the BCL-10–MALT1 filaments43,72 and MALT1 monoubiquitylation73.

Fig. 4: MALT1 paracaspase substrates and their molecular functions.
figure 4

Nine proteolytic substrates of mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) have been identified. These include regulators of nuclear factor-κB (NF-κB) and mitogen-activated protein kinase (MAPK) pathways (namely, A20, HOIL1, CYLD and RELB), repressors of mRNA stability (namely, regnase 1, roquin 1 and roquin 2) and components of the CARD protein–BCL-10–MALT1 (CBM) complex (namely, B cell lymphoma 10 (BCL-10) and MALT1). The NF-κB subunit RELB represses RELA and REL activity by heterodimerization. The deubiquitinases CYLD and A20 remove K63-linked or M1-linked (linear) ubiquitin (Ub) chains from target proteins and thereby attenuate NF-κB and MAPK signalling. HOIL1 is a subunit of the linear ubiquitin chain assembly complex (LUBAC) and supports HOIP-mediated linear ubiquitylation. Regnase 1, roquin 1 and roquin 2 bind to immunomodulatory mRNAs and mediate their degradation.

MALT1 cleaves a small set of substrates after arginine residues within sequences related to its consensus peptide sequence LVSRG72. So far, nine substrates have been identified that are cleaved by MALT1 after TCR ligation, including MALT1 (ref.74) and BCL-10 (ref.75) themselves, the NF-κB regulators A20 (also known as TNFAIP3)76, CYLD77, RELB78 and HOIL1 (also known as RBCK1)79,80,81, and the mRNA-binding proteins regnase 1 (also known as ZC3H12A)82, roquin 1 (also known as RC3H1) and roquin 2 (also known as RC3H2)83 (Fig. 4). Although the biological consequences of MALT1 paracaspase activity are best explored in the context of T cell signalling, MALT1 is also activated within CARD9-nucleated, CARD10-nucleated and CARD14-nucleated CBM complexes41,69,84, but it has not been determined as yet whether the same substrates are cleaved in all cell types.

The precise biological effects of MALT1 autoproteolysis and MALT1-mediated cleavage of BCL-10 also remain unclear. Nevertheless, MALT1 autocleavage has been reported to support optimal gene transcription74, and proteolysis of BCL-10 contributes to cell adhesion75 after T cell activation. The effects of A20, CYLD, RELB and HOIL1 cleavage are better understood. A20 and CYLD are deubiquitinases that negatively regulate NF-κB and JNK signalling. The cleavage of A20 and CYLD by MALT1 inactivates these deubiquitinases in the vicinity of active CBM complexes and thereby further promotes NF-κB and JNK signalling76,77. Likewise, the degradation of RELB by MALT1 increases NF-κB activity by eliminating transcriptionally inactive RELB–RELA and RELB–REL dimers78. The cleavage of HOIL1 can further amplify NF-κB signalling by enhancing the association between BCL-10 and IKKs81. Therefore, the MALT1-mediated proteolysis of A20, CYLD, RELB and HOIL1 fine-tunes, amplifies and/or prolongs NF-κB and JNK signalling after the initial pulse of CBM-mediated IKK and JNK activation. However, HOIL1 cleavage may also provide negative feedback at later stages of signal transduction79,80.

The control of inflammatory gene expression is mediated in an additional manner by MALT1-mediated cleavage of the ribonuclease regnase 1 (ref.82) and the mRNA decay regulators roquin 1 and roquin 2 (ref.83). These molecules cooperate to degrade overlapping sets of mRNAs by binding to a common stem–loop element within their 3ʹ untranslated regions85. Transcripts that are negatively regulated by regnase 1, roquin 1 and roquin 2 encode inflammatory cytokines and chemokines, including IL-6, TNF, IL-1β and CXC-chemokine ligand 1 (CXCL1), immunoregulatory surface receptors such as cytotoxic T lymphocyte antigen 4 (CTLA4), inducible T cell co-stimulator (ICOS) and OX40 (also known as TNFRSF4), and intracellular signalling molecules with key roles in inflammatory pathways, such as NF-κB inhibitor-ζ (NFKBIZ), A20 and interferon regulatory factor 4 (IRF4)83,86. By cleaving and inactivating regnase 1, roquin 1 and roquin 2, MALT1 protease activity stabilizes the target mRNAs, which upregulates the expression of an entire set of immunoregulatory transcripts in activated T cells and strengthens the inflammatory reaction. The MALT1 protease probably has additional functions, such as upregulation of the glutamine transporter ASCT2 (also known as SLC1A5) and activation of mechanistic target of rapamycin (mTOR) complex 1 (mTORC1) during TCR signalling to mediate metabolic reprogramming of activated lymphocytes, but the MALT1 substrates and mechanisms that control these pathways are currently unknown87,88.

Physiological functions of CBM complexes

The in vivo functions of the individual CBM complexes have been extensively studied in genetically manipulated mouse models. These models have established multiple non-redundant functions of the distinct signalosomes in the immune system (Fig. 5) and have provided the basis for an understanding of the roles of these complexes in human disease.

Fig. 5: Cell type-specific physiological functions of CBM complex signalling.
figure 5

In B cells and T cells, CARD11–BCL-10–MALT1 complexes mediate lymphocyte proliferation, differentiation, metabolic reprogramming and survival after antigen recognition. In B cells, signalling through this CARD protein–BCL-10–MALT1 (CBM) complex controls the differentiation of peritoneal B1 B cells, marginal zone B (MZB) cells and plasma cells. In T cells, this CBM complex controls the differentiation of regulatory T (Treg) cells, T helper 17 (TH17) cells and T follicular helper (TFH) cells, the last of which support plasma cell differentiation. In myeloid cells, CARD9–BCL-10–MALT1 complexes control innate immunity upon the recognition of fungal, bacterial and viral pathogen-associated molecular patterns (PAMPs) or sterile danger-associated molecular patterns (DAMPs) and induce inflammatory cytokine production. These signals support the priming of TH17 cells (IL-6, IL-1β and IL-23), recruitment of neutrophils (granulocyte–macrophage colony-stimulating factor (GM-CSF) and CXC-chemokine ligand 1 (CXCL1)) and generation of neutrophilic myeloid-derived suppressor cells (N-MDSCs). Caspase-recruitment domain (CARD)-containing coiled-coil protein 9 (CARD9) signalling in myeloid cells regulates pathways that are required for intestinal epithelial regeneration and controls the gut microbiota to maintain intestinal homeostasis. In epithelial cells from the airway and other tissues, the CARD10–BCL-10–MALT1 complex is activated in response to G protein-coupled receptor (GPCR) ligands (such as lysophosphatidic acid (LPA), endothelin 1 (ET1) and allergens) and epidermal growth factor (EGF) and has been implicated in airway inflammation. The CARD10–BCL-10–MALT1 complex also regulates cardiovascular inflammation and remodelling by inducing cytokine and chemokine production in endothelial cells, vascular smooth muscle cells and cardiomyocytes. In endothelial cells, this CBM complex mediates the disruption of the endothelial barrier. CBM signalling in hepatocytes drives inflammation and insulin resistance in response to free fatty acids (FFAs). CBM signalling and CARD14 control the expression of cytokines and chemokines in keratinocytes in response to PAMPs and IL-17 in the skin. BCL-10, B cell lymphoma 10; CCL, CC-chemokine ligand; ICAM1, intercellular adhesion molecule 1; MALT1, mucosa-associated lymphoid tissue lymphoma translocation protein 1; TNF, tumour necrosis factor; VCAM1, vascular cell adhesion molecule 1.

CARD11–BCL-10–MALT1 signalling in lymphocytes

As indicated above, the CARD11–BCL-10–MALT1 signalosome is absolutely essential for BCR and TCR signalling6,7,8,25,26. Although genetic deficiencies in CARD11, BCL-10 or MALT1 in mice do not affect overall lymphocyte numbers, the differentiation of specific lymphocyte subsets depends crucially on CARD11–BCL-10–MALT1 (Fig. 5). In the T cell lineage, this CBM complex is required for the development of thymus-derived regulatory T (Treg) cells and peripherally derived Treg cells under steady-state conditions89,90. CBM signalling also controls the differentiation of T follicular helper (TFH) cells, which are required for the germinal centre reaction and the production of high-affinity antibodies by B cells91, and the differentiation of T helper 17 (TH17) cells92,93, which control immune responses to extracellular pathogens. In the B cell lineage, CARD11–BCL-10–MALT1 signalling controls the generation of marginal zone B (MZB) cells and peritoneal B1 B cells94, as well as germinal centre formation and plasma cell differentiation91.

Conventional B cells and T cells require CARD11–BCL-10–MALT1 signalling for activation, proliferation and survival upon encountering antigens. Consequently, animals that lack any of the genes encoding these proteins have severe defects in adaptive immunity6,7,8,25,26. For example, BCL-10-deficient mice have severe defects in cellular and humoral immunity in response to lymphocytic choriomeningitis virus or vesicular stomatitis virus (VSV) infection6, and MALT1-deficient mice succumb to infections with strains of rabies virus that are usually non-lethal95. However, MALT1-deficient animals are protected from experimental autoimmune encephalomyelitis96, which indicates that CBM signalling mediates both protective and pathological functions of lymphocytes.

Studies with MALT1 knock-in mice that harbour an inactivating point mutation in the catalytic caspase-like domain of MALT1 — paracaspase mutant (Malt1PM) mice — provided the first genetic proof that individual biological functions of CBM signalling are differentially regulated by the specific outputs97,98,99. The engineered mice express MALT1PM proteins that maintain the scaffold function required for IKK and MAPK activation but are defective in the cleavage of MALT1 substrates. Whereas the differentiation of lymphocytes into the Treg cell, MZB cell and B1 B cell lineages is absolutely dependent on MALT1 proteolytic activity, the paracaspase activity of MALT1 is dispensable for conventional lymphocyte activation to a considerable extent. Consequently, the absence of the immunosuppressive Treg cell population in the presence of pathologically activated B cells and T cells results in a lethal autoimmune inflammatory syndrome in Malt1PM mice97,98,99. However, the MALT1 substrates that control the differentiation of Treg cells, MZB cells and B1 B cells are undefined. Moreover, because most lymphocyte differentiation and activation steps are controlled by a complex interplay between various B cell and T cell subsets, the available studies in the germline mutant mice have currently failed to dissect the cell-intrinsic contributions of CARD11–BCL-10–MALT1 signals in defined lymphocyte lineages for lymphocyte homeostasis.

CARD9–BCL-10–MALT1 signalling in myeloid cells

Studies in CARD9-deficient mice have defined the essential functions of the CARD9–BCL-10–MALT1 complex in the innate immune system17,24 (Fig. 5). Infection experiments showed that this CBM signalosome controls what is thought to be the most important mammalian defence pathway for antifungal immunity. The complex is required for innate immune protection against Candida albicans17,100, Aspergillus fumigatus101 and Cryptococcus neoformans102, which are all sensed by SYK-coupled CLRs. In addition, CARD9 signalling in antigen-presenting cells links innate sensing of fungi to the activation of adaptive immunity and provides a cytokine milieu that induces the differentiation of TH17 cells103. CARD9 also drives the induction of granulocyte–macrophage colony-stimulating factor (GM-CSF) expression after dectin 1 engagement100,104. Owing to their failure to provide these signals, CARD9-deficient mice have impaired recruitment of neutrophils, which are crucial for fungal engulfment and killing, to the site of infection101,105. The failure to produce GM-CSF is particularly important, as it permits invasive fungal growth into tissues104. At later phases of the fungal infection, CARD9 signalling triggers the differentiation of neutrophilic myeloid-derived suppressor cells (N-MDSCs), which prevent hyperinflammatory pathological T cell and NK cell responses106.

CLRs sense not only fungi but also pathogen-associated molecular patterns (PAMPs) from bacteria, viruses and parasites and endogenous danger-associated molecular patterns (DAMPs) that trigger sterile inflammation107. Mincle is the mammalian receptor for the cord factor (trehalose-6,6-dimycolate) from Mycobacterium tuberculosis, and CARD9–BCL-10–MALT1 activation in experimental models of tuberculosis promotes an IL-10-mediated anti-inflammatory feedback loop to prevent pathological, neutrophil-mediated systemic inflammation108.

CARD9 signalling also controls the production of inflammatory cytokines such as IL-1β, TNF and IL-6 in response to the recognition of cytosolic RNA or DNA by RIG-I, MDA5 or RAD50 (refs22,23,109). These pathways are engaged by viruses such as VSV and vaccinia virus or by cytosolic nucleic acids from bacteria or potentially endogenous sources. However, in contrast to pathways leading to the production of type I interferons, the nucleic acid-induced CARD9-dependent inflammatory pathways do not mediate immediate antiviral host protection but rather shape the subsequent adaptive immune response.

The cytosolic PRR NOD2 senses intracellular bacteria, such as Listeria monocytogenes, and its signalling through CARD9 has been implicated in host defence against this pathogen24. NOD2 is a risk gene for Crohn’s disease, in which an imbalance in the host–microbiota equilibrium shaped by the innate immune system has a particularly important role. Interestingly, CARD9-deficient animals have an altered intestinal microbiota that fails to metabolize tryptophan into aryl hydrocarbon receptor ligands110; these ligands normally induce T cells and innate lymphoid cells to produce IL-22, which promotes the regenerative proliferation of epithelial cells after tissue injury. As a consequence of impaired IL-22-mediated regenerative signals, CARD9-deficient animals are hypersusceptible to experimental colitis induced by chemicals or Citrobacter rodentium110,111, and they develop smaller and less proliferative tumours in models of colitis-associated cancer112.

Although several DAMPs, including cholesterol crystals and vimentin, induce inflammatory responses through CLRs such as dectin 1 and mincle107 — and thus, potentially, through CARD9 — the in vivo functions of CARD9–BCL-10–MALT1 signalling during sterile inflammation are not well explored. Nevertheless, the first mouse models of sterile autoantibody-induced rheumatoid arthritis or dermatitis showed that these pathologies are mediated by CARD9–BCL-10–MALT1 signalling113. In summary, these studies not only established that the CARD9–BCL-10–MALT1 signalosome has crucial roles in protection against pathogens, particularly fungi, but also provided first insights into the roles of this CBM complex in homeostatic pathways.

CARD10–BCL-10–MALT1 and CARD14–BCL-10–MALT1 signalling in non-haematopoietic cells

The CARD10–BCL-10–MALT1 complex controls inflammatory pathways in airway epithelia, the cardiovascular system and the liver in response to infection, hormones or metabolic cues31,114,115,116 (Fig. 5). Cell type-specific disruption of CARD10 in the respiratory epithelium showed that this CBM signalosome has non-redundant roles in the recruitment and activation of DCs and pathogenic T cells in a model of acute allergic airway inflammation114,117. Moreover, the CARD10–BCL-10–MALT1 complex controls the production of inflammatory cytokines and chemokines, such as CXCL8, GM-CSF and CC-chemokine ligand 20 (CCL20), induced by asthma-associated allergens that activate GPCRs, such as Alternaria alternata and house dust mites, or by messenger molecules such as LPA or ATP114.

Endothelial cells, smooth muscle cells and cardiomyocytes use the CARD10–BCL-10–MALT1 signalosome to induce the production of pro-inflammatory cytokines and chemokines, including IL-6, CXCL1, CXCL2 and CCL2, after stimulation with the GPCR agonists angiotensin II, thrombin or endothelin 1 (ET1), which are major triggers of cardiovascular inflammation31,65,115. The proteolytic function of MALT1 is also important in this context, as it mediates acute endothelial permeability during vascular inflammation84. When stimulated with thrombin, CARD10–BCL-10–MALT1 complexes in endothelial cells also upregulate expression of the adhesion molecules intercellular adhesion molecule 1 (ICAM1) and vascular cell adhesion molecule 1 (VCAM1) to promote leukocyte adhesion after local injury, which further exacerbates atherosclerotic inflammation118. Consistent with the pathophysiological relevance of these mediators, BCL-10-deficient mice are protected from atherosclerosis, aortic aneurysms and pathogenic fibrotic heart remodelling induced by chronic infusion of angiotensin II31,115.

In hepatocytes, the CARD10–BCL-10–MALT1 module is activated by saturated free fatty acids (FFAs) through receptor-independent mechanisms to trigger NF-κB-mediated inflammation. FFA levels are increased in serum from individuals who are obese, and the deletion of Bcl10 in mice abolishes hepatic NF-κB activation induced by a high-fat diet, leading to reduced inflammatory responses and protection from insulin resistance116.

Although the homeostatic functions of CARD10–BCL-10–MALT1 engagement by EGFR family RTKs are largely unclear, oncogenic RTK signalling, which frequently occurs in breast and lung cancers119, also activates the CARD10–BCL-10–MALT1 signalosome34,66,120. CARD10 activation by EGFR (also known as ERBB1) or ERBB2 (also known as HER2 or NEU) in breast cancer accelerates tumour progression by increasing cell proliferation and resistance to cell death34. In addition, the CARD10–BCL-10–MALT1 signalling axis promotes tumour cell invasion, migration and angiogenesis through the transcriptional control of matrix metalloproteinases, cathepsin B, vascular endothelial growth factor and inflammatory cytokines in the tumour microenvironment66,121. Comparable to its role in breast cancer, the EGFR–CBM–NF-κB axis also controls the proliferation, migration and invasion of lung cancer cells in vitro and in vivo66,120.

The physiological functions of CARD14 are somewhat unexplored. However, in keratinocytes, CARD14–BCL-10–MALT1 signalling is activated by microorganisms such as Staphylococcus aureus and presumably fungi through receptors such as dectin 1 and potentially other PRRs. This induces the expression of pro-inflammatory genes, including IL17C and TNF, through NF-κB signalling and mechanisms that involve MALT1 protease activity41 (Fig. 5). It is not only the production of IL-17 that is mediated by CARD14 signalling but also the response to IL-17, as the IL-17 receptor signals through an NF-κB activator 1 (ACT1; also known as TRAF3IP2)–CARD14–TRAF6 complex to induce the production of IL-23, CCL20 and other cytokines that recruit and stimulate immune cells and thereby shape the immune microenvironment during skin inflammation40. It is currently unknown whether BCL-10 and MALT1 also participate in the complex with CARD14 and ACT1.

Genetic CBM alterations in human disease

Considering the pivotal roles of CBM complexes in immunity and tissue homeostasis, it is not surprising that genetic alterations affecting these signalosomes are frequently observed in human pathology. Indeed, following the original discovery of the BCL10 translocation in lymphoma1, pathogenic gain-of-function (GOF) and loss-of-function (LOF) variants have been identified in almost all CBM components in patients (Fig. 6; Table 1).

Fig. 6: Somatic and germline alterations and coding risk variants in CARD9, CARD10, CARD11, CARD14 and MALT1.
figure 6

Somatic caspase-recruitment domain (CARD)-containing coiled-coil protein 11 (CARD11) mutations in B cell lymphoma (diffuse large B cell lymphoma (DLBCL), mantle cell lymphoma and follicular lymphoma) and in T cell lymphoma (Sezary syndrome and adult T cell leukaemia/lymphoma (ATL)) are depicted as circles, each representing one patient. The red box indicates a frequent deletion in ATL. The most frequent cellular inhibitor of apoptosis protein 2 (cIAP2)– mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) variant in gastric mucosa-associated lymphoid tissue (MALT) lymphoma is depicted, where dark blue indicates the cIAP2 moiety and light blue indicates the MALT1 moiety. For the germline CARD9, CARD10, CARD11, CARD14 and MALT1 gain-of-function and loss-of-function alterations, the affected amino acids or splice variants are indicated in colours that correspond to the colour of the box that indicates the respective human condition. Blue indicates rare variants with unknown disease penetrance or risk single-nucleotide polymorphisms. The asterisk in the inflammatory bowel disease (IBD) protective variant represents a stop codon that results from exon 11 skipping and produces an alternative protein sequence (blue line). BENTA, B cell expansion with nuclear factor-κB (NF-κB) and T cell anergy; BIR, baculovirus inhibitor of apoptosis protein repeat; CC, coiled-coil; DD, death domain; GUK, guanylate kinase-like; Ig, immunoglobulin-like; MCI, mild cognitive impairment; POAG, primary open-angle glaucoma; PRP, pityriasis rubra pilaris; SH3, SRC homology 3; UBA, ubiquitin-associated.

Table 1 Germline and somatic genetic alterations in CBM complex components and associated human diseases

Somatic GOF alterations in CARD11, BCL-10 and MALT1 in lymphoma

In MALT lymphoma, the translocation t(1;14)(p22;q32) fuses the intact BCL10 gene to the immunoglobulin heavy chain (IGH) locus1,2, whereas t(1;2)(p22;p12) brings BCL10 under the control of the immunoglobulin light chain locus122, and t(14;18)(q32;q21) juxtaposes MALT1 to IGH123. All three translocations result in the overexpression of BCL-10 or MALT1 and thereby enforce the BCL-10–MALT1 module to drive aberrant NF-κB activation124, which is a major oncogenic pathway in malignant B cells. The most frequent translocation in MALT lymphoma, which is detected in up to 50% of patients, is t(11;18)(q21;q21), which generates a chimeric GOF cIAP2–MALT1 fusion protein3,4,5. Although several breakpoints exist, all cIAP2–MALT1 fusion proteins that are found in patients contain the amino-terminal baculovirus inhibitor of apoptosis protein repeat (BIR) domains of cIAP2 and the MALT1 caspase-like domain (Fig. 6). These molecules autonomously activate both NF-κB and MALT1 paracaspase activity10,125. Intriguingly, the cIAP2–MALT1 fusion protein cleaves not only the classical MALT1 substrates but also NF-κB-inducing kinase (NIK; also known as MAP3K14)125, which liberates the kinase domain of NIK from the region that mediates degradation, thereby stabilizing the kinase to chronically activate oncogenic non-canonical NF-κB125. Another new substrate of cIAP1–MALT1 is the tumour suppressor protein LIM domain and actin-binding protein 1 (LIMA1). Cleavage of LIMA1 disrupts its capacity to inhibit lymphoma cell growth and generates a LIM domain-only fragment with tumour growth-promoting properties126. However, although the BCL-10 and MALT1 translocations enforce tumour cell-intrinsic survival and proliferation programmes124, they are not sufficient for lymphomagenesis by themselves; transgenic mice that overexpress BCL-10 or cIAP2–MALT1 in the B cell lineage develop polyclonal lymphoproliferation127,128 but not overt lymphomas, which indicates that further alterations are necessary for malignant B cell transformation.

Translocations involving CARD11 have not been described, but somatic GOF mutations in CARD11 are frequently detected in both B cell and T cell malignancies129,130,131,132,133 (Fig. 6). In B cell lymphoma, the CARD11 mutations typically affect the coiled-coil domain or the region between the CARD and coiled-coil domains and disrupt CARD11 autoinhibition, which renders the protein either independent from upstream signals or hyperresponsive to stimulation134. These variants are detected in diffuse large B cell lymphoma (DLBCL)129, mantle cell lymphoma130 and follicular lymphoma131. B cell-specific expression of a human GOF CARD11-L225LI lymphoma variant in mice leads to constitutive NF-κB and JNK activity that cooperatively drive lethal B cell lymphoproliferation. Likewise, the survival of human DLBCL cells depends on not only NF-κB but also JNK activity135,136, which indicates that the oncogenic output of aberrant CARD11 signalling is mediated by NF-κB and JNK, at least in a subset of human lymphomas. In addition, the proteolytic activity of MALT1 is required for DLBCL cell survival, which has led to the development of MALT1 inhibitors as potential therapeutics for lymphoma137,138,139,140. Furthermore, the wild-type CBM complex colocalizes with mutated MYD88 in DLBCL cells as part of a tumorigenic multiprotein supercomplex on endolysosomes that activates NF-κB and mTOR signalling141. Thus, even when the CBM complex is not mutated, it is an integral component of lymphomagenic pathways in human aggressive lymphomas.

CARD11 alterations are observed in many types of T cell non-Hodgkin lymphoma (T-NHL), including adult T cell leukaemia/lymphoma (ATL) associated with human T lymphotropic virus 1 (HTLV‐1) infection, Sezary syndrome, a leukaemic cutaneous T-NHL, other cutaneous T-NHLs and TFH cell-derived lymphomas such as angioimmunoblastic lymphoma132,133,142. In these malignancies, CARD11 GOF point mutations, intragenic deletions and gene amplifications are recurrently observed. Interestingly, the CARD11 mutations in T-NHL cluster in the coiled-coil domain or linker region, the latter of which is rarely affected in B cell lymphoma (Fig. 6), which indicates that the molecular wiring of pathological CBM signalling in B cell and T cell malignancies might be distinct. Nevertheless, overexpressed CARD11 T-NHL variants also constitutively activate NF-κB and MALT1 proteolytic activity, and the inhibition of these CBM outputs is toxic to T-NHL cell lines142, which indicates that strategies targeting these effector pathways should be explored for T-NHL therapy.

Germline CARD11 variants that enforce or inhibit CBM signalling

Germline GOF variants in CARD11 are detected in patients with B cell expansion with NF-κB and T cell anergy (BENTA), an inherited lymphoproliferative immunodeficiency143 that is characterized by lymphocytosis and recurrent infections. The CARD11 variants found in patients with BENTA are autosomal dominant, leading to pathological CARD11–BCL-10–MALT1 complex formation and increased NF-κB activity143 (Fig. 6). Despite their vigorous proliferation, BENTA B cells do not differentiate into plasma cells, and T cells have an anergic phenotype, which explains the recurrent opportunistic infections in patients143,144. Interestingly, two of the CARD11 germline mutations in patients with BENTA (G123S and C49Y) have also been detected as somatic mutations in patients with DLBCL129.

An independent set of germline mutations creates hypomorphic CARD11 variants that impair but do not entirely abolish CARD11 function145. The affected patients develop severe atopic allergic diseases and autoimmunity and are also hypersusceptible to infections146. The hypomorphic CARD11 variants suppress wild-type CARD11 in a dominant manner by interfering with proper assembly of the CBM complex and the activation of NF-κB, p38 and MALT1 activity as well as with mTORC1 signalling145. The altered TCR signalling pathways impair TH1 cell responses while enforcing pathogenic TH2 cell-mediated immunity, resulting in atopic dermatitis, food and environmental allergies, asthma and impaired protection against infection. Consistent with a broader role for pathophysiological CARD11 signalling in atopy, a single-nucleotide polymorphism (SNP) in the genomic CARD11 region was identified as being associated with atopic dermatitis by a genome-wide association study in the Japanese population147.

Inherited LOF mutations in CARD11, BCL-10 or MALT1

Germline LOF mutations have been described in all CBM components in lymphocytes, leading to combined immunodeficiency in children (Fig. 6). Destructive CARD11 mutations result in a complete loss of the CARD11 protein or the generation of a nonfunctional variant148,149. The affected patients have normal overall B cell and T cell counts, but they lack Treg cells148,149. The TCR-induced proliferation of conventional T cells is markedly reduced, and B cell development is blocked at the transitional B cell stage. Consequently, the patients lack immunoglobulins and develop severe respiratory tract infections with opportunistic pathogens very early in life148,149.

Patients deficient in BCL-10 or MALT1 have similar impairments in adaptive immunity150,151,152,153,154. The loss of BCL-10 expression has fatal consequences caused by uncontrolled gastrointestinal, central nervous system (CNS) and respiratory tract infections with various bacteria and viruses153. MALT1 LOF mutations in humans lead to severely reduced or absent MALT1 protein expression or a dysfunctional protein. Patients with homozygous MALT1 deficiency suffer from gastrointestinal, respiratory tract and systemic infections with multiple bacterial, viral and fungal species150,151,152,154,155. The transplantation of healthy donor haematopoietic stem cell grafts to MALT1-deficient patients successfully reversed the clinical and immunological phenotypes of the MALT1 deficiency152,154,155. Interestingly, in contrast to CARD11-deficient or BCL-10-deficient individuals, several patients with a MALT1 LOF mutation also developed autoimmunity and autoinflammation in the skin and intestine, which was possibly caused by an absence of Treg cells paired with only partially defective conventional T cell responses, similar to the phenotype observed in Malt1PM mice97,98,99.

Germline CARD9 LOF defects

The discovery of familial homozygous CARD9 LOF mutations led to the identification of an innate immunodeficiency with high susceptibility to fungal infections156. The originally isolated autosomal recessive CARD9-Q295* variant is defective in signal transduction downstream of dectin 1. TH17 cells were almost absent in the CARD9-mutant patients, whereas the frequencies of other lymphocytes and neutrophils were within normal ranges. Additional autosomal recessive CARD9 LOF mutations were subsequently identified in multiple individuals (Fig. 6) who suffered from severe and, in some cases, lethal infections with various fungal species, including C. albicans, A. fumigatus and dermatophytes157. The CARD9-Y91H LOF variant abolishes GM-CSF secretion by monocytes after fungal recognition104, resulting in impaired neutrophil recruitment to sites of infection, particularly the CNS, which is consistent with the role of CARD9 for neutrophil attraction to mediate antifungal defence in mice105. On the basis of these insights, GM-CSF has been used successfully to treat a CARD9-mutant patient with CNS candidiasis158.

CARD9 disease-risk polymorphisms

The importance of deviated CARD9–BCL-10–MALT1 signalling in the pathogenesis of human inflammatory disorders is underscored by the SNPs within the CARD9 locus that are recurrently associated with IBD159, ankylosing spondylitis160, primary sclerosing cholangitis161 and IgA nephropathy162. Several of the disease-associated CARD9 SNPs affect non-coding regions and presumably influence CARD9 expression163. However, a prevalent CARD9 SNP that correlates with an increased risk162,164 for inflammatory disease results in an amino acid substitution (S12N) in the CARD of CARD9. Knock-in mice with the respective alteration in the endogenous Card9 locus develop pulmonary allergic responses when challenged with A. fumigatus as a result of increased TH2 cell priming165 caused by aberrantly increased RELB activity. Consistent with this, the human CARD9-S12N variant also increases the risk of allergic bronchopulmonary aspergillosis165. It is likely that related mechanisms contribute to other CARD9-S12N-associated inflammatory pathologies either directly or indirectly — for example, by changing the composition of the microbiota towards a pathogenic state.

Rare, protective CARD9 isoform variants with truncations in the carboxy-terminal region decrease the risk of IBD70,166. One of these is a splice variant that excludes exon 11, creating a CARD9-ΔE11 protein that associates with BCL-10 and MALT1 in the normal manner but is impaired in its ability to recruit TRIM62 for NF-κB activation. Small-molecule inhibitors that bind full-length CARD9 and disrupt TRIM62 binding have been developed as candidate therapeutics for IBD167.

Germline CARD10 risk variants

Rare coding variants in CARD10 are enriched in a cohort of patients with primary open-angle glaucoma168 (Fig. 6). Furthermore, a CARD10 variant that alters the coiled-coil domain (R289Q) is associated with faster progression of amnestic mild cognitive impairment, which is a precursor of Alzheimer disease169. Together, these studies suggest that aberrant inflammatory signals through CARD10–BCL-10–MALT1 complexes might contribute to human neurodegeneration. However, mechanistic data supporting this hypothesis are currently unavailable.

Germline CARD14 GOF variants

CARD14 GOF variants have received increasing attention owing to their pathogenic association with psoriasis170. The CARD14 gene lies within a known psoriasis susceptibility locus, PSORS2, on chromosome 17q171. Several CARD14 GOF mutations affecting the coiled-coil domain or the region between the CARD and coiled-coil domains were identified in familial cases of psoriasis and have been mechanistically linked to PSORS2 (ref.171). A series of additional CARD14 GOF mutations have been identified in patients with inherited pityriasis rubra pilaris (PRP)35 (Fig. 6). These highly penetrant, autosomal dominant CARD14 GOF alterations are associated with early-onset manifestations of psoriasis or PRP38, but additional rare variants have been identified in sporadic cases, and a coding SNP in the GUK domain of CARD14 increases the risk of developing psoriasis38. The pattern of psoriasis-associated CARD14 mutations correlates with the distribution of CARD11 mutations in patients with BENTA or B cell lymphoma129,143, and, likewise, familial CARD14 mutations disrupt the CARD14 autoinhibitory conformation172 and induce constitutive assembly of CARD14–BCL-10–MALT1 signalosomes. These changes drive NF-κB and MALT1 paracaspase activation in keratinocytes and trigger the production of cytokines and chemokines, including TNF, IL-36γ, CXCL8 and IL-17C, all of which are key factors in the pathogenesis of psoriasis41. Knock-in mice that express the psoriasis-associated CARD14-E138del or CARD14-E138A variants in the germ line40,173 conclusively confirm that enforced CARD14 signalling is sufficient to drive psoriatic immune cell infiltration with epidermal thickening and scaling of the skin through the IL-23–IL-17 axis. Therapeutic IL-17 blockade alleviated the disease in CARD14 knock-in models40,173, similarly to the beneficial effects of IL-17 antagonists in human psoriasis.

Conclusions and perspectives

Since the original cloning of BCL10 approximately two decades ago1,2, CBM signalosomes have emerged as key signalling platforms for innate and adaptive immunity and inflammation. The earlier lessons obtained from engineered BCL-10-deficient, MALT1-deficient, CARD11-deficient and CARD9-deficient mice were recently corroborated by the identification of the respective primary immunodeficiencies in humans, which — although individually rare — confirmed that the functions of CBM signalling are highly conserved between mice and humans. Mechanistic insights into these pathways have already led to the development of rational strategies to reconstitute or strengthen these signals to boost immunity. These strategies include haematopoietic stem cell transplantation for inherited immune defects152,154,155, GM-CSF supplementation to improve neutrophil function in the absence of CARD9 (ref.158), novel experimental approaches to enforce CLR–CARD9 signalling by inhibiting the negative regulator CBLB as a potential treatment for disseminated fungal infections174,175 and the selective triggering of CLR–CARD9 signalling with defined adjuvants in vaccination regimens for the generation of long-lasting, protective TH1 cell-mediated and TH17 cell-mediated responses against tuberculosis176,177. Conversely, in inflammatory or malignant pathology with aberrant CBM signalling, the inhibition of CBM complex effector pathways has been explored to kill lymphoma cells or ameliorate inflammation in prevalent diseases, such as psoriasis or IBD. Here, in addition to general blockade of NF-κB or JNK signalling, specific inhibition of MALT1 paracaspase function and selective interference with the recruitment of E3 ubiquitin ligases or with CBM-mediated non-proteolytic protein ubiquitylation are promising stategies56,139,167,178. Nevertheless, pharmacological interference with CBM signalling could compromise host defence mechanisms, and prolonged MALT1 paracaspase inhibition carries the risk of unwanted inflammation owing to potential inhibitory effects on Treg cells.

So far, most of our mechanistic knowledge has been obtained from the CARD11–BCL-10–MALT1 complex in lymphocytes. Although the general principles of CBM signalling are shared between the CARD9–BCL-10–MALT1, CARD10–BCL-10–MALT1, CARD11–BCL-10–MALT1 and CARD14–BCL-10–MALT1 signalosomes, the precise regulation and effector functions of the other CBM signalosomes are much less well defined. As one example, although all CBM signalosomes activate the proteolytic activity of MALT1, the functional identities and biological effects of the specific MALT1 substrates in the innate immune system or in non-haematopoietic cells remain largely unclear. Thus, further explorations into the physiological and pathological functions of CBM signalling are expected to provide additional insights into the context-specific functions of immune and inflammatory pathways in health and disease and lead to the development of further therapeutic strategies targeting the immune system.