Fig 2: The CARD11 CARD-coiled-coil interface is weaker than that of CARD9, but is reinforced by the ID. a Cartoon schematic of the domain architecture of CARD9, CARD11-?ID, and CARD11 along with the chimeric constructs CARD91-131/CARD11-?ID and CARD111-143/CARD9, wherein the N-terminal residues comprising the CARD, linker, and CARD-interacting stretch of the coil-coil are swapped between the two proteins (See Supplementary Fig. 1B). b Normalized NF-?B activation in HEK-293 cells upon overexpression of the indicated constructs. Three biologically independent experiments were undertaken, the outcomes of which are shown as dots. The mean value for each construct is depicted as a gray bar. Dotted line represents the value of the CARD9WT construct, to which the other values were normalized. Representative western blot is shown for CARD9 and CARD11 constructs, using antibodies specific for regions outside of the N-terminal swapped domain or ID. Source data are provided in Source Data file

Fig 3: CARD-coiled-coil disruption promotes Bcl10-nucleating CARD9 helical assembly. a NS-EM micrographs of CARD92-152/WT, CARD92-152/I107E, and CARD92-152/L115I after removal of Zn2+ by EDTA and incubation at the indicated concentration, time, and temperature. Incubation of CARD92-152/WT for shorter times, at lower temperature (25 °C) or lower concentration (0.5 mM) also failed to produce filaments by NS-EM. Scale bars of all panels are 200 nm. b DAmFRET assay performed on CARD91-97, CARD91-142/WT, or CARD91-142/I107E. Inset depicts AmFRET binned at 50–60 µM for each construct, with 50 µM indicated by vertical dashed lines. See Supplementary Fig. 4D for replicate data, statistical analysis, and data binned at 100–200 µM. c Bcl10 FP nucleation assay. MBP-Bcl10 was mixed with CARD92–152 preparations as indicated. EDTA addition and incubation of CARD92–152 constructs occurred prior to initiation of nucleation assay; 1:1 and 5:1 refers to the molar ratio during the FP assay, with the final concentration of Bcl10 always set to 2 µM. TEV protease was added at t = 0 to indicated samples. Vertical bars represent the s.d. of three technical replicates. The data shown in the two plots were collected in the same experiment and were plotted separately to aid in visualization; hence, the ‘Bcl10’’ and ‘Bcl10 + TEV’’ data are identical in both plots. Source data are provided in Source Data file. d Bcl10, CARD92-152/I107E, or CARD92-152/L115I filaments, 2 min after addition of TEV protease and with addition of anti-HA antibody during NS-EM grid preparation. e, f Direct templating of Bcl10 by CARD92-152/I107E (e) or CARD92-152/L115I (f) filaments. Grids were prepared 2 min after addition of TEV protease and with anti-HA antibody added during NS-EM grid preparation. White arrows indicate CARD92–152-to-Bcl10-filament transitions. For panels d–f, samples were prepared under identical conditions and concentrations as in c with all grids prepared 2 min after addition of TEV protease. g Schematic of genetically encoded seeds. DAmFRET assay for Bcl10-mEos3.1 alone (left), showing that it partitions cells into overlapping populations that contain either monomer alone or polymer; or for Bcl10-mEos3.1 with co-expressed with µNS-Bcl10 (right), which shifts all cells to the polymer-containing population. h DAmFRET assay for Bcl10-mEos3.1 with co-expressed CARD91-97, CARD91-142/WT, or CARD91-142/I107E either alone or tethered to µNS

Fig 4: The structural basis for unidirectional polymerization of CARD9. a Cryo-EM density for CARD92-152/I107E filaments. Density is only present for the CARD core of the assembly. Seven CARDs of the modeled structure are shown, depicting the six CARDs contacted by any one CARD, colored as in b. Arrow indicates the direction of polymerization. b Schematic of the type I, II, and III interaction interfaces utilized by the CARD9 CARD. c Detail of the CARD interaction type I, II, and III interfaces, with residues involved in the interaction shown as sticks. CARDs are colored as in b. d CARD92--97 polymerization assay. At time t = 0, a super-stoichiometric concentration of EDTA was added to each CARD, which were pre-saturated with 1:1 Zn2+. Polymerization was monitored by absorbance at 350 nm. Vertical bars represent the s.d. of three technical replicates. See NS-EM endpoint images in e and Supplementary Fig. 6A. For mutants D66R and R35E, Zn2+ removal at higher concentrations does lead to increased A350, but without visible filaments by NS-EM (Supplementary Fig. 6B), indicating that the mutations are destabilizing and lead to non-specific aggregation upon Zn2+ removal at the higher concentrations. Source data are provided in Source Data file. e NS-EM micrographs of CARD92-97/WT and CARD92-97/D43R/K58D generated at the endpoint of the assay in d. Scale bars are 500 nm. f Alignment of the apo CARD9 CARD NMR structure (black, PDBID 6E26 [http://dx.doi.org/10.2210/pdb6E26/pdb]) and CARD Cryo-EM structure (red)

Fig 5: CARD92-142 forms a dimer with an extensive CARD-linker-coiled-coil interface. a Coiled-coil propensity for full-length CARD9, as predicted by the COILS server73. Schematic of the construct composition is shown above. b 15N-TROSY of CARD118–172, collected at 37 °C. c 15N-TROSY of the CARD92–152 dimer (black) and 15N-HSQC of CARD92-97 (orange) at 37 °C. The CARD92-97 spectrum was shifted to account for the difference in chemical shifts recorded by TROSY and HSQC experiments. d Ribbon diagram of the 20 lowest energy CARD2-142 dimer structures calculated. One of the two monomers is colored as a rainbow from N- to C-terminus. e Lowest energy CARD92-142 structure, with sub-domains of one monomer colored as in Fig. 1a. f “Open-face” depiction of the CARD-coiled-coil interface. Residues of the CARD, linker, and coiled-coil for which heavy atoms are within 2.5 Å of another sub-domain in any of the lowest energy structures are shown as sticks and colored in red, cyan, and blue, respectively. Residues forming the hydrophobic core of the interface are noted