IDRs promote CAR condensation

IDRs contain diverse sequence and structure features. To determine which IDR promotes the condensation of CAR, we selected candidates from six well-characterized IDRs that were previously shown to induce condensation in a cellular environment. These include IDRs from FUS, EWS, TAF15, Nup98, TDP43 and a synthetic IDR (synIDR)31,32,33,34,35,36,37. We chose CD19 CAR, which is commonly used in research and clinical practice, as an initial model. This CAR is composed of a single-chain variable fragment (scFv, FMC63) that targets CD19, a stalk and transmembrane domain from CD8α and cytosolic signaling domains from 41BB, CD28 and CD3ζ. The IDR was fused to the C terminus of CD3ζ. A superfolder GFP tag, which promotes the folding of fused client proteins38, was further attached on the C terminus of IDR for visualization of CAR condensation (Fig. 1a). The superfolder GFP tag enables live-cell imaging, which avoids potential fixation-induced artifacts in characterizing IDR condensation39. The DNA fragment encoding the control or IDR CAR was packaged into lentivirus and introduced into primary T cells purified from the human peripheral blood mononuclear cells (PBMCs). Flow cytometry revealed the total cellular expression (by GFP) versus cell surface localization (by FMC63) of individual CARs (Fig. 1b), demonstrating that fusion with IDR did not affect the trafficking of CAR to the cell surface. To visualize the condensation of CAR on the cell surface, we stained live CAR-T cells with a plasma membrane dye CellMask deep red and performed total internal reflection fluorescence (TIRF) microscopy, which effectively reduces the cytosolic background of fluorescence. We found that the CAR fused with FUS, EWS or TAF15 displayed enhanced condensation as compared to the control CAR (Fig. 1c). This is demonstrated by the overall clustering level as quantified by normalized variance under both resting (Fig. 1d) and stimulation conditions (Extended Data Fig. 1b). We scored cells with similar CAR surface expressions to minimize the influence from CAR expression on condensation (Extended Data Fig. 1a). The cluster number per cell was higher in these three IDR CARs as compared to the control CAR in resting but not stimulated cells (Extended Data Fig. 1c), suggesting that clusters might undergo fusion during CAR stimulation. The average fluorescence intensity of cluster per cell was higher in stimulated cells when comparing FUS and EWS CARs to control CAR (Extended Data Fig. 1d). However, because of large variance in the sample, this was not statistically significant. We also compared the mobility of CAR in IDR-promoted condensates by fluorescence recovery after photobleaching. This was performed in HEK293T cells because of their robust CAR expression that is more suitable for high-frequency time-lapse imaging. We found that EWS condensates showed slower recovery than FUS or TAF15 condensates (Extended Data Fig. 1e), suggesting that EWS condensates were more stable. Together, these data demonstrate that FUS, EWS and TAF15 promoted CAR condensation. Therefore, we focused on these three CARs in the subsequent functional assays.

Fig. 1: IDRs promote CAR condensation on the T cell surface.

a, Schematics of the CD19 CAR used in this study. b, Expression of control and IDR CARs measured by flow cytometry. The x axis displays total CAR expression using a GFP fusion. The y axis displays surface expression of CAR by an anti-FMC63 antibody. c, Condensation of CAR on the plasma membrane imaged by TIRF microscopy. CAR is tagged by GFP (cyan). The plasma membrane is labeled by a CellMask deep red dye (magenta). d, Quantification of CAR clustering by normalized variance (n = 50 cells for control, FUS, EWS, TAF15, Nup98 and TDP43 CAR; n = 43 for synIDR CAR). Shown are the means ± s.d. All comparisons were made between individual IDR CARs and the control CAR (P values determined using a Mann–Whitney U-test). e, CAR-T cell proliferation with or without IL-2. Quantification of the absolute cell number in the culture with or without IL-2 (n = 3 technical replicates). Shown are the means ± s.d.

Source data

Previous work showed that aggregation of CARs targeting GD2 or CSPG4 induces spontaneous activation in the absence of antigen29,30. Therefore, we assessed whether IDR-induced CAR condensation causes a similar spontaneous activation. We found that IDR CAR-Ts did not proliferate in the absence of interleukin 2 (IL-2) (Fig. 1e). The expression of CD69, a T cell activation marker, was similar between the control and IDR CAR-Ts in the absence of antigen (Extended Data Fig. 1f). Moreover, IDR CAR-Ts did not secret detectable tumor necrosis factor alpha (TNFα), interferon-γ (IFNγ) and IL-2 in the absence of antigen (Extended Data Fig. 1g–i). Together, assessed from cell proliferation, activation marker and cytokine production, IDR-induced CAR condensation did not trigger spontaneous signaling.

IDR from FUS enhances the cytotoxicity of CD19 CAR-T

To determine how IDRs affect the cytotoxicity of CAR-Ts against cancer cells, we cocultured the control or IDR CAR-Ts (Fig. 2a) with modified Nalm6, a B cell leukemia line that expresses either high or low CD19 (Fig. 2b). The Nalm6 cells express a luciferase reporter that enables the quantification of cytotoxicity by the luciferase assay. The FUS and EWS but not TAF15 CAR displayed a substantially higher cytotoxicity toward both CD19high and CD19low Nalm6 cells (Fig. 2c, d and Extended Data Fig. 2a,b). This result was recapitulated using another B cell line Raji as the target (Extended Data Fig. 2c–e). The higher cytotoxicity of FUS and EWS CAR could be explained by their higher secretion of cytotoxic factors including granzyme A, granzyme B, perforin, FasL and IFNγ, when CAR-Ts were engaged with CD19low Nalm6 cells (Fig. 2e–i and Extended Data Fig. 2f). Because superfolder GFP was included in the control CAR to balance the protein size increase in the IDR CARs (CAR expression is generally reduced with increased protein sizes), we tested whether superfolder GFP affects CAR-T function. We titrated the lentivirus concentrations to achieve a comparable CAR expression with and without GFP (Extended Data Fig. 2g,h). We found that the inclusion of GFP to the control CAR did not affect the expression of CD69 and the release of TNFα (Extended Data Fig. 2i, j). We also confirmed that the inclusion of GFP did not affect the cytotoxicity of the control or IDR CAR-Ts (Extended Data Fig. 2k–m). Together, these data showed that the FUS and EWS IDRs enhanced the cytotoxicity of CAR-Ts against CD19low cells, which was accompanied with a higher secretion of cytotoxic factors.

Fig. 2: IDRs enhance the cytotoxicity of CD19 CAR-T in vitro and in vivo.

a, Expression of the control or IDR CAR targeting CD19 (scFv FMC63) in human primary T cells measured by flow cytometry using an anti-FMC63 antibody. b, Expression of high (34,517 molecules per cell) or low (1,470 molecules per cell) CD19 in Nalm6 cells. The average number of CD19 antigens on each cell was determined by flow cytometry. c,d, Cytotoxicity of CD19 CAR-T targeting CD19high or CD19low Nalm6 cells in vitro (n = 3 technical replicates). Shown are the means ± s.d. (P values at various E:T ratios determined using an unpaired two-sided Student’s t-test). e–i, Production of cytotoxic factors by CD19 CAR-T cocultured with CD19low Nalm6 cells for 1 day at an E:T of 3:1 measured by flow cytometry (n = 3 technical replicates). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). j, Timeline of examination of the antitumor effect of CD19 CAR-Ts in a CD19low Raji B-derived xenograft model. k–m, Tumor progression (rainbow color) quantified by bioluminescent imaging. Both the averaged and individual traces are shown (n = 6 mice). Shown are the means ± s.d. A two-way ANOVA was used to compare IDR CAR and control CAR groups over time. An unpaired two-sided Student’s t-test was used to compare two groups at specific time points. n–q, Quantification of pan-T cell abundance, CD8+ T percentage, granzyme B expression and T cell exhaustion markers including TIM3, LAG3 and PD1 in total T cells in mouse blood on day 43 after CAR-T infusion analyzed by flow cytometry (n = 6 mice). MFI, median fluorescence intensity. Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test).

Source data

To assess the tumor-killing effect of IDR CAR-Ts toward CD19low cancer cells in vivo, CD19low Raji B cells expressing a luciferase reporter were injected into the immune-deficient NSG (NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ) mice intravenously. Three days later, the control, FUS or EWS CAR-T cells were infused through the tail vein. The cancer progression was monitored by bioluminescence imaging (Fig. 2j). We found that FUS CAR-T inhibited cancer proliferation better than the control CAR-T (Fig. 2k–m). This effect was repeated using T cells generated from a different donor (Extended Data Fig. 3a–g). EWS CAR-T, though displaying enhanced cytotoxicity in vitro (Fig. 2d and Extended Data Fig. 2b,e) and better tumor control at early time points in vivo (day 7; Fig. 2l), did not achieve a sustained antitumor effect (Fig. 2l). We will discuss reasons later. The enhanced antitumor effect of FUS CAR was accompanied by a mildly (not statistically significant) enhanced percentage of CD8+ T cells (Fig. 2o and Extended Data Fig. 3h) and granzyme B expression (Fig. 2p and Extended Data Fig. 3i) but reduced expression of exhaustion markers including LAG3 (mainly contributed by CD4+ T cells) and PD1 (mainly contributed by CD8+ T cells) (Fig. 2q and Extended Data Figs. 3j–n and 4a–d,k–l). No significant increase was detected in the T cell number (Fig. 2n and Extended Data Fig. 4e–h) or memory phenotype (Extended Data Fig. 4i, j). To further investigate the mechanism of improved antitumor efficacy in FUS CAR-T, we performed single-cell RNA sequencing (scRNA-seq) experiments on T cells isolated from the mouse blood. We found that cytotoxic T cells were more highly represented in the FUS group, as compared to the control or EWS group (Extended Data Fig. 5a–c). The abundance of cells expressing cytotoxic factors (GZMB, GZMK and NKG7) was also higher in the FUS group (Extended Data Fig. 5d–g). Moreover, the expressions of exhaustion markers including PD1 (PDCD1 as the gene name), CXCL13 and PRDM1 were lower in the FUS group (Extended Data Fig. 5h–k). Together, these data suggest that FUS promoted cytotoxicity but reduced exhaustion of CD19 CAR-T, which explains its enhanced antitumor activity in the blood cancer model.

IDRs from FUS and TAF15 enhance cytotoxicity of HER2 CAR-T

To determine whether the effect of IDR in promoting cytotoxicity applied to CARs beyond CD19, we constructed IDR CARs targeting HER2 (Fig. 3a), an antigen commonly overexpressed in multiple solid tumors including breast, ovarian, lung and colorectal cancers. Human primary T cells were infected with lentivirus encoding the control, FUS, TAF15 or EWS CAR (Fig. 3b). Similar to the case of CD19 CAR, IDR did not trigger spontaneous activation (without antigen) of HER2 CAR as assessed by CD69 expression and the release of TNFα, IFNγ and IL-2 (Extended Data Fig. 1j–m). Next, we selected multiple target cell lines for testing cytotoxicity: the lymphoblast K562 cell line ectopically expressing high or low HER2 (Fig. 3c), the ovarian cancer cell line SKOV3 expressing high HER2 and the colon cancer cell line HT29 expressing low HER2 (Fig. 3d). These target cells were cocultured with the control or IDR CAR-Ts. We found that FUS and TAF15 CAR-Ts displayed higher cytotoxicity toward all four cell lines tested, as compared to the control CAR-T (Fig. 3e–h and Extended Data Fig. 6a,b). Consistent with that, FUS and TAF15 CAR-Ts secreted a higher level of cytotoxic factors, including IFNγ, perforin and FasL, than the control CAR-T (Fig. 3i–m and Extended Data Fig. 6c). When comparing HER2 CAR with and without superfolder GFP, we did not find a significant difference in CD69 expression, TNFα production or cytotoxicity (Extended Data Fig. 6d–j), suggesting that the inclusion of superfolder GFP did not significantly affect HER2 CAR-T function. Together, these data suggest that FUS and TAF15 enhanced the cytotoxicity of both HER2high and HER2low CAR-Ts in vitro.

Fig. 3: IDRs enhance the cytotoxicity of HER2 CAR-T in vitro and in vivo.

a, Schematics of the HER2 CAR used in this study. The scFv targeting HER2 is H3B1. b, Expression of the control or IDR CAR targeting HER2 in human primary T cells by flow cytometry. WT, wild type. c,d, Quantification of the HER2 level in K562, SKOV3 and HT29 cells. e–h, Cytotoxicity of HER2 CAR-T in vitro targeting HER2high or HER2low K562 cells, SKOV3 and HT29 cells (n = 3 technical replicates). Shown are the means ± s.d. (P values at various E:T ratios determined using an unpaired two-sided Student’s t-test). i–m, Production of cytotoxic factors by CAR-T cocultured with HT29 cells for 1 day at an E:T of 3:1 (n = 3 technical replicates). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). n, Timeline of examination of the antitumor effect of HER2 IDR CAR-Ts in subcutaneously engrafted HT29 tumor model. Low doses of CAR-T were infused intravenously. o, Quantification of tumor progression in vivo by measuring the tumor size using an electronic digital caliper (n = 5 mice). Shown are the means ± s.d. (P values determined using a two-way ANOVA). p, HT29 tumor was treated with high doses of CAR-T injected intravenously. The CAR-T cells were generated from a donor different than the one in n,o. q, Quantification of tumor progression (n = 6 mice). Shown are the means ± s.d. (P values determined using a two-way ANOVA). r–t, Quantification of total T cell and CD8+ T cell abundance and granzyme B expression in blood on day 22 after CAR-T infusion by flow cytometry (n = 6 mice). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). u–w, Quantification of tumor-infiltrated T cells, CD8 percentage and CD69 expression in tumor-infiltrated T cells on day 27 after CAR-T infusion by flow cytometry (n = 3 mice). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). x,y, Quantification of expression of granzyme B and exhaustion markers including LAG3, PD1 and TIM3 in bone marrow T cells on day 27 after CAR-T infusion (n = 3 mice). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test).

Source data

To determine the antitumor effect of IDR CAR-Ts toward HER2low cells in vivo, HT29 cells were injected into the immune-deficient NSG mice subcutaneously. Eight days later, the control or IDR CAR-T cells were infused through the tail vein. A second dose was administered 5 days later. The tumor progression was monitored by an electronic caliper over 6 weeks (Fig. 3n). We found that FUS and TAF15 CAR-T inhibited tumor growth better than the control CAR-T (Fig. 3o and Extended Data Fig. 6k), which is consistent with the in vitro killing results. To characterize T cell phenotypes, we repeated the experiment with two modifications: (1) by increasing the infused CAR-T cell number so that sufficient tumor-infiltrating T cells could be isolated for analysis and (2) by terminating the experiment at an earlier time point when tumors did not undergo necrosis and T cells were not entirely exhausted (Fig. 3p). During the new trial, we observed that FUS and TAF15 slowed tumor progression (Fig. 3q and Extended Data Fig. 6l), which is consistent with the result from previous experiments. When analyzing circulating T cells, we observed a higher percentage of CD8+ T cells in the FUS group (Fig. 3r,s and Extended Data Fig. 7a–c) and a higher expression of granzyme B in both the FUS and TAF15 group (Fig. 3t and Extended Data Fig. 7d). By isolating tumor-infiltrating T cells, we found that the FUS and TAF15 group showed an increased number of tumor-infiltrating T cells (Fig. 3u and Extended Data Fig. 7e,f), although the T cell numbers in blood were comparable across the control, FUS and TAF15 groups (Fig. 3r and Extended Data Fig. 7a,b). An increase in CD8+ population and CD69 expression was also observed, albeit not statistically significant, which is potentially because of the large variance in samples (Fig. 3v,w and Extended Data Fig. 7g,h). For the bone marrow T cells, we also observed a higher granzyme B expression in the FUS and TAF15 group (Fig. 3x and Extended Data Fig. 7i). No significantly lower expression was found in the exhaustion makers (Fig. 3y and Extended Data Fig. 7j–l). Together, these data suggest that FUS and TAF15 promoted a CD8/cytotoxic signature in vivo, which is consistent with the in vitro data showing that they displayed a higher cytotoxicity.

IDRs from FUS and EWS enhance cytotoxicity of CD22 CAR-T

In addition to the CD19 and HER2 CAR, we also tested how IDRs affect the cytotoxicity of an CD22 CAR (RFB4), which showed very low signaling efficiency because it targets the membrane-distal epitope position on CD22 (refs. 40,41). Using a similar design strategy to the CD19 CAR, we constructed the IDR CARs targeting CD22 by fusing the IDR on the C terminus (Fig. 4a). Human primary T cells were infected with lentivirus encoding the control, FUS, EWS or TAF15 CAR (Fig. 4b). The wild-type Raji B or Nalm6 cells, which express a medium level of CD22 (Fig. 4c), were cocultured with CAR-Ts. We found that FUS and EWS CAR-T displayed a higher cytotoxicity as compared to the control CAR-T when cocultured with either Raji B or Nalm6 cells (Fig. 4d,e and Extended Data Fig. 8a–c). We also measured the release of cytotoxic factors and found that FUS and EWS CAR-Ts released slightly higher granzymes A and B and FasL than the control CAR-T, albeit not statistically significant (Fig. 4f–j and Extended Data Fig. 8d). Together, these data show that FUS and EWS enhanced the cytotoxicity of a low-signaling CD22 CAR-T.

Fig. 4: IDRs enhance the cytotoxicity of CD22 CAR-T in vitro and in vivo.

a, Schematics of the CD22 CAR used in this study. The scFv targeting CD22 is RFB4. b, Expression of the control or IDR CAR targeting CD22 in human primary T cells. CAR expression was detected by recombinant CD22 proteins and measured by flow cytometry. c, Quantification of the CD22 level in the wild-type Raji B and Nalm6 by flow cytometry. d,e, Cytotoxicity of CD22 CAR-T in vitro targeting Raji B or Nalm6 cells cocultured for 3 days (d) or 1 day (e) (n = 3 technical replicates). Shown are the means ± s.d. (P values at various E:T ratios determined using an unpaired two-sided Student’s t-test). f–j, Production of cytotoxic factors by CAR-T cocultured with the wild-type Nalm6 cells for 1 day at an E:T of 1:1 measured by flow cytometry (n = 3 technical replicates). Shown are the mean ± s.d. (P values determined using an unpaired two-sided Student’s t-test). k, Timeline of the antitumor effect of CD22 IDR CAR-Ts in wild-type Nalm6-derived tumor xenograft model. l,m, Tumor progression (rainbow color) quantified by bioluminescent imaging (n = 5 mice). Shown are the means ± s.d. (P values determined using a two-way ANOVA). n, Quantification of T cells in the blood on day 10 after CAR-T infusion by flow cytometry (n = 5 mice). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). o, Expression of T cell exhaustion markers including TIM3, LAG3 and PD1 during cancer progression on day 19 after CAR-T infusion by flow cytometry (n = 5 mice). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test).

Source data

To determine whether IDRs improve the antitumor efficacy of the above low-signaling CD22 CAR in vivo, wild-type Nalm6 cells were injected into NSG mice intravenously. Three days later, the control, FUS or EWS CAR-T cells were infused through the tail vein. The cancer progression was monitored by bioluminescence imaging (Fig. 4k). Similar to the in vitro killing result, FUS and EWS CAR-T inhibited tumor growth better than the control CAR-T (Fig. 4l, m). The circulating T cell number was higher in the FUS and EWS group than the control group (Fig. 4n and Extended Data Fig. 8e). The expression of markers for T cell differentiation (Extended Data Fig. 8f,g) or exhaustion (Fig. 4o and Extended Data Fig. 8h–j) was not significantly different between the control and IDR CAR-Ts. Together, these data suggest that IDRs from FUS and EWS promoted the antitumor effect of a low-signaling CD22 CAR.

IDRs promote the formation and signaling of CAR-T synapse

Next, we investigated the molecular mechanism by which CAR condensation increases T cell activation. CAR condensation is expected to promote multivalent interactions with antigens, thereby increasing the binding between CAR-T cells and cancer cells. Indeed, all three IDR CAR-Ts formed a higher percentage of cell–cell conjugation with CD19+ Nalm6 cells as compared to the control CAR-T (Fig. 5a). To probe the mechanical strength of the cell–cell conjugates, we exploited the z-Movi ‘cell avidity’ instrument by Lumicks42,43. Nalm6 cells were preseeded on the chip, followed by CAR-T addition to form the cell–cell conjugates. An increasing acoustic force was applied and the detachment of CAR-Ts from Nalm6 was monitored over time. We found that the force to disassemble the synapse was higher in IDR CAR-T as compared to the control CAR-T (Fig. 5b–e), suggesting that the cell–cell interaction was stronger in the IDR CAR-T group as compared to the control CAR-T group. Next, we investigated the localization of a key signaling protein, CD45, in the synapse. After cell–cell conjugates formed, CD45, a phosphatase that dephosphorylates CAR, was excluded from the synapse because of its large extracellular domain44. Our previous work showed that CD45 exclusion is mediated by CAR–antigen interaction and the level of CD45 exclusion influences CAR signaling40. Therefore, we examined CD45 exclusion in the synapse and found that CD45 was excluded to a higher level in the IDR CAR group as compared to the control CAR group (Fig. 5f). Together, these data suggest that IDRs promoted the formation, mechanical strength and CD45 exclusion of the immunological synapse formed between CAR-T and cancer cells.

Fig. 5: IDRs promote the formation, mechanical strength and signaling of the CAR-T synapse.

a, Cell–cell conjugation percentage of CD19 control and IDR CAR-Ts with Nalm6 CD19high or CD19low cells (n = 3 technical replicates). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). b, Rupture forces to detach CD19 control and IDR CAR-T from Nalm6 CD19low cells by z-Movi from Lumicks. CAR-T cells were generated from n = 3 donors. Shown are the means. The mean ranges (mean ± s.d.) are shown in c–e (P values determined using a two-way ANOVA). f, CD45 exclusion in the synapse formed between CD19 CAR-T with Raji B CD19low cells was imaged by confocal microscopy. CD19 CAR, n = 39 cells; FUS CAR, n = 41 cells; EWS CAR, n = 43 cells; TAF15 CAR, n = 47 cells. CAR-T cells were generated from n = 3 donors. Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). g,h, Phosphorylation kinetics of CD3ζ (pY142) and LAT (pY171). Displayed are traces for three independent experiments using CAR-T cells generated from n = 3 donors. i,j, Phosphorylation of CD3ζ at the CAR-T synapse imaged by confocal microscopy and quantification of pCD3ζ at the CAR-T synapse by ImageJ. The cyan and magenta colors represent CAR-T and tumor cells; the yellow color represents pCD3ζ. CD19 CAR, n = 77 conjugates; FUS CAR, n = 100 conjugates; EWS and TAF15 CAR, n = 107 conjugates (P values determined using an unpaired two-sided Mann–Whitney U-test). k,l, Phosphorylation of LAT at the CAR-T synapse imaged by confocal microscopy and quantification of pLAT at the CAR-T synapse by ImageJ. The cyan and magenta colors represent CAR-T and tumor cells; the yellow color represents pLAT. CD19 CAR, n = 79 conjugates; FUS CAR, n = 81 conjugates; EWS CAR, n = 105 conjugates; TAF15 CAR, n = 86 conjugates (P values determined using an unpaired two-sided Mann–Whitney U-test).

Source data

Next, we monitored membrane-proximal signaling in CAR-Ts upon engaging with cancer cells. Following a CAR–antigen engagement, the CD3ζ domain on CAR is phosphorylated, recruiting kinase ZAP70 which further phosphorylates LAT, a key adaptor protein nucleating multiple cytosolic effectors. We cocultured CAR-Ts with CD19low Raji B cells and monitored the antigen-dependent signaling kinetics by flow cytometry. With comparable phosphorylation of CD3ζ and LAT in a resting state (Extended Data Fig. 9a), we found that the FUS, EWS and TAF15 IDRs enhanced the phosphorylation of CD3ζ and LAT following CAR-T’s binding to Raji B (Fig. 5g, h and Extended Data Fig. 9b, c). Because CD3ζ and LAT are two transmembrane proteins enriched in the CAR-T synapse, we examined their phosphorylation level in the synapse by confocal microscopy. Consistent with the flow cytometry measurement, all three IDRs enhanced the phosphorylation of CD3ζ and LAT in the synapse (Fig. 5i–l and Extended Data Fig. 9d, e). It should be noted that the above assays were based on microscopy and flow cytometry, enabling us to score CAR-positive cells only to avoid the influence from variance in CAR expression on the signaling readout. However, in the cytotoxicity and cytokine production assay (Fig. 2c–i), the readouts were contributed from all cells; therefore, the variance in CAR expression had a significant influence on the readouts and needs to be taken into account when interpreting the results. In addition to CD3ζ and LAT, we compared the activation of other membrane-proximal signaling proteins between the control and FUS CAR-Ts using western blot. We found a higher phosphorylation (>1.5-fold) in ZAP70, PLCγ1, SLP76 and ERK (Extended Data Fig. 10a). Consistently, we also observed a higher calcium influx in the FUS CAR-T as compared to the control CAR-T (Extended Data Fig. 10b). Together, these data suggest that IDRs promote CAR-T activation by promoting membrane-proximal signaling pathways.

Coiled-coil-mediated oligomerization reduces CAR-T activation

In addition to IDR-mediated protein condensation, the coiled-coil domain is a commonly used tool to induce oligomerization of protein of interest. To test whether coiled-coil domains could promote CAR-T activation, we fused a coiled-coil domain that mediate dimerization, tetramerization or hexamerization45,46 to the C terminus of a CD19 CAR (Fig. 6a) and introduced these coiled-coil CARs into human primary T cells. Interestingly, whereas the coiled-coil dimer maintained a similar cell surface expression level as compared to the control CAR, the coiled-coil tetramer and hexamer CARs showed a dramatic reduction in the cell surface localization (Fig. 6b, c). Consequently, the cell–cell conjugation percentage between CAR-T and Nalm6 cells was significantly reduced in the coiled-coil tetramer and hexamer as compared to the control CAR (Fig. 6d). Consistent with this, CAR-T activation, as evaluated by CD69 expression (Fig. 6e and Extended Data Fig. 10c) and IFNγ secretion (Fig. 6f), was significantly reduced in the coiled-coil tetramer or hexamer. The reductions in cell–cell conjugation and CAR-T activation were recapitulated when using Raji B as a target cell (Extended Data Fig. 10d–f). Similar to CAR, the oligomerization-induced receptor internalization was frequently observed in other transmembrane receptors including the epidermal and fibroblast growth factor receptors47,48. The fact that IDRs did not affect the cell surface expression of CARs suggests that IDRs present a unique advantage to promote CAR clustering without causing enhanced receptor internalization.

Fig. 6: Oligomerization by coiled-coil domain reduced CAR surface localization and CAR-T activation.

a, Schematics of the coiled-coil CD19 CAR. It contained an scFv targeting CD19 (FMC63), a CD8 hinge, a LAT transmembrane domain, an intracellular signaling domain composed of CD28, 41BB and CD3ζ and a coiled-coil domain. b, Expression of coiled-coil CARs detected by flow cytometry using an anti-FMC63 antibody. A GFP tag was fused on the C terminus of each CAR to monitor the total expression of CAR. The red dashed line indicates the slope of the total versus surface CAR expression of the control CAR. The dark-blue and light-blue lines indicate the slope for the expression of the coiled-coil tetramer and coiled-coil hexamer CAR, respectively. c, Confocal microscopy revealing the cellular localization of coiled-coil CAR shown as a cyan color. The plasma membrane is labeled by a CellMask deep red dye shown as magenta in the image. d, Cell–cell conjugation between CAR-T and Nalm6 cells with E:T = 1:1 imaged by confocal microscopy (n = 3 technical replicates). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test). e,f, Activation of coiled-coil CAR-Ts assessed by CD69 expression and IFNγ release cocultured with Nalm6 for 1 day at E:T = 1:1 (n = 3 technical replicates). Shown are the means ± s.d. (P values determined using an unpaired two-sided Student’s t-test).

Source data