Research ArticleCell Biology

ESCRT proteins restrict constitutive NF-κB signaling by trafficking cytokine receptors

See allHide authors and affiliations

Science Signaling  19 Jan 2016:
Vol. 9, Issue 411, pp. ra8
DOI: 10.1126/scisignal.aad0848

“ESCRT”ing receptors out of endosomes

Members of the nuclear factor κB (NF-κB) family of transcription factors are critical for cell survival, differentiation, inflammation, and immune responses. Many receptors that activate NF-κB are internalized and degraded, and receptor degradation depends on ESCRT (endosomal sorting complexes required for transport) complexes. By knocking down individual ESCRT components in cells, Mamińska et al. found that lack of ESCRT function enhanced NF-κB signaling as a result of increased clustering of ligand-free cytokine receptors at endosomes. Clustering at the endosomes activated these receptors, resulting in NF-κB–dependent gene expression. Thus, the ESCRT complex constitutively directs cytokine receptor trafficking to prevent endosomal accumulation and block spurious NF-κB activation.


Because signaling mediated by the transcription factor nuclear factor κB (NF-κB) is initiated by ligands and receptors that can undergo internalization, we investigated how endocytic trafficking regulated this key physiological pathway. We depleted all of the ESCRT (endosomal sorting complexes required for transport) subunits, which mediate receptor trafficking and degradation, and found that the components Tsg101, Vps28, UBAP1, and CHMP4B were essential to restrict constitutive NF-κB signaling in human embryonic kidney 293 cells. In the absence of exogenous cytokines, depletion of these proteins led to the activation of both canonical and noncanonical NF-κB signaling, as well as the induction of NF-κB–dependent transcriptional responses in cultured human cells, zebrafish embryos, and fat bodies in flies. These effects depended on cytokine receptors, such as the lymphotoxin β receptor (LTβR) and tumor necrosis factor receptor 1 (TNFR1). Upon depletion of ESCRT subunits, both receptors became concentrated on and signaled from endosomes. Endosomal accumulation of LTβR induced its ligand-independent oligomerization and signaling through the adaptors TNFR-associated factor 2 (TRAF2) and TRAF3. These data suggest that ESCRTs constitutively control the distribution of cytokine receptors in their ligand-free state to restrict their signaling, which may represent a general mechanism to prevent spurious activation of NF-κB.


Signaling mediated by the transcription factor nuclear factor κB (NF-κB) between cells in a tissue and between organs in an organism ultimately controls cell survival and differentiation, stress responses, inflammation, and immunity (1). Without stimulation, this ubiquitous pathway is latent but ready for rapid activation (2). In an inactive state, NF-κB transcription factors are complexed in the cytoplasm with members of the inhibitor of κB (IκB) family (in the canonical pathway) or with the NF-κB precursor p100 (in the noncanonical pathway). Signaling, which is initiated typically by cytokines, leads to inhibitor degradation or p100 processing and the subsequent nuclear translocation of active NF-κB dimers (3). Because cytokine receptors lack intrinsic enzymatic activity, ligand-induced signaling occurs through receptor oligomerization and the recruitment of adaptors, E3 ubiquitin ligases, and kinases to their cytoplasmic tails (4). For canonical ligands, such as tumor necrosis factor–α (TNF-α), signaling proceeds through activation of the IκB kinase 2 (IKK2) complex, degradation of IκBα, and nuclear accumulation of the NF-κB subunit RelA. For noncanonical ligands, such as lymphotoxin or CD40L, induction of IKK1 activity leads to the proteolytic processing of p100 to generate its transcriptionally active form p52 (5). Although these pathways are stimulated by extracellular ligands that stimulate receptor internalization, little is known about the role of endocytosis in NF-κB signaling.

Endocytic trafficking modulates the downstream signaling of internalized ligand-receptor complexes by targeting them for degradation or recycling (68). Certain receptors continue to signal during endosomal transport until they are incorporated into the intraluminal vesicles (ILVs) of endosomes, which separates them from their cytoplasmic effectors before they are degraded (9). This step is controlled by four ESCRT (endosomal sorting complexes required for transport) complexes (10), which couple cargo recognition to ILV budding by binding to cargo, to endosomal membranes, and sequentially to each other (11). The early-acting ESCRT-0 and ESCRT-I complexes recognize and cluster ubiquitylated receptors, whereas ESCRT-II recruits ESCRT-III filaments, which are crucial for membrane deformation and scission of the resulting vesicle. Cargo-loaded ILVs are then degraded in lysosomes. ESCRTs are thus key gatekeepers that control cargo flow toward degradation. In addition, some ESCRT components mediate other membrane-remodeling events during cytokinesis, retroviral budding, plasma membrane repair, and nuclear envelope sealing (1215).

ESCRT dysfunction is expected not only to inhibit receptor degradation but also to elicit pleiotropic effects, which result from their multiple roles. Indeed, depletion of selected ESCRT subunits causes the accumulation of ligand-activated epidermal growth factor receptor (EGFR) and increased EGF-dependent signaling by mitogen-activated protein kinases (MAPKs) (16, 17). Unexpectedly, a genome-wide mRNA analysis did not reveal substantial changes in the overall transcriptional response to EGF under these conditions (18). Instead, NF-κB target genes were identified as the only group of genes whose expression was specifically induced upon depletion of Hrs or Tsg101, which are components of ESCRT-0 and ESCRT-I, respectively. This result suggested the specific action of ESCRTs in the NF-κB pathway, but the underlying mechanism remained unexplored. Here, we show that ESCRT components restrict constitutive NF-κB signaling by preventing the accumulation and activation of NF-κB–inducing receptors.


ESCRT components are negative regulators of constitutive, but not TNF-α–induced, NF-κB activity

To investigate the role of membrane trafficking in the regulation of NF-κB signaling, we systematically depleted 26 subunits of the ESCRT complexes in human embryonic kidney (HEK) 293 cells with two RNA interference (RNAi) libraries. The effect of knockdown on NF-κB signaling was measured by luciferase reporter assay under basal conditions (serum-grown cells) or after stimulation with TNF-α. For both RNAi libraries, silencing of VPS28 or TSG101 activated constitutive NF-κB signaling to a greater extent than did depletion of the IκBα inhibitor (NFKBIA; Fig. 1, A and B). Silencing of VPS37A or CHMP4B by RNAi in either one of the libraries also resulted in activation of the pathway (Fig. 1, A and B). However, in TNF-α–stimulated cells, depletion of none of the ESCRT components resulted in an increase in pathway activity that was more than twofold (fig. S1). Therefore, we focused on the role of potential hits in basal, constitutive NF-κB signaling, that is, in the absence of exogenous cytokines.

Fig. 1 RNAi screen identifies ESCRT components that negatively regulate the NF-κB pathway.

(A and B) Luciferase reporter assays in unstimulated HEK 293 cells upon depletion of 26 ESCRT components with siRNAs from endoribonuclease-prepared siRNA (esiRNA) (A) or SMARTpools (B) libraries. Cells depleted of the established pathway components IκBα (NFKBIA), IKKγ (IKBKG), p105 (NFKB1), and RelA served as controls. (C to F) The screen results were validated with single siRNA duplexes (siRNAs #1 to #6), esiRNA targeting another gene region (esiRNA #2), or both. In addition, the effects of UBAP1 knockdown were analyzed (F). The efficiency of knockdown was verified by Western blotting analysis (bottom). Data are means ± SEM of three to six (A and D to F) or two (B and C) independent experiments. Data were analyzed by Student’s t test (C, D, and F) or Mann-Whitney U test (E). ns (not significant), P > 0.05; *P ≤ 0.05; **P ≤ 0.01.

The activation of NF-κB signaling by the silencing of Tsg101, Vps28, or CHMP4B was validated with additional short interfering RNA (siRNA) reagents (Fig. 1, C to E, and fig. S2A). The increase in reporter activity observed for all five reagents targeting Vps28 and for most of the siRNAs against Tsg101 and CHMP4B correlated with the knockdown efficiency of the given gene. We could not unequivocally confirm the effects of VPS37A silencing. This may reflect partly redundant functions between multiple Vps37 isoforms (Vps37A through Vps37D) (19, 20).

Vps28, Tsg101, and Vps37A are subunits of the ESCRT-I complex (21). We additionally verified that the knockdown of UBAP1, a recently identified component of ESCRT-I (22, 23), increased NF-κB reporter activity (Fig. 1F and fig. S2A). We thus concluded that the ESCRT-I complex, containing Vps28, Tsg101, and UBAP1, inhibits constitutive NF-κB activity. Of these three nonredundant ESCRT-I subunits (which are encoded by single genes in humans), UBAP1 acts specifically in the sorting of cargo into ILVs, but not in cytokinesis or viral budding (23), which suggests an endosomal role for ESCRT-I in NF-κB signaling. CHMP4B is the most abundant ESCRT-III subunit, which builds membrane-constricting filaments (24). Whereas the depletion of ESCRT-0 or ESCRT-II did not potentiate NF-κB signaling (Fig. 1, A and B), knockdown of CHMP4B enhanced pathway activity similarly to the silencing of ESCRT-I (Fig. 1E). This may be explained by CHMP4B playing a role in the early steps of cargo sorting into ILVs together with ESCRT-0 and ESCRT-I (25).

Depletion of Tsg101, Vps28, UBAP1, or CHMP4B induces the expression of human NF-κB target genes

To measure signaling output upon knockdown of ESCRT components, we profiled the expression of 84 NF-κB target genes upon knockdown of Tsg101 or Vps28 and identified overlapping sets of genes whose mRNA abundance was greatly increased (Fig. 2, A to C). We verified the increased expression of six genes upon depletion of Tsg101, Vps28, UBAP1, or CHMP4B with SMARTpools (Fig. 2D) or single siRNA duplexes (figs. S2B and S3A). The knockdown of Vps28 had the greatest effects (for example, a >50-fold induction in the expression of IL8 and a >10-fold induction in the expression of TNF, ICAM1, and RELB), whereas the knockdown of Tsg101, UBAP1, or CHMP4B resulted in 2- to 10-fold increases in gene expression. In agreement with the results of the screen (Fig. 1, A and B), these effects were specific for the four selected ESCRT members: depletion of Hrs, EAP30, or CHMP3 (subunits of ESCRT-0, ESCRT-II, and ESCRT-III, respectively) did not induce the expression of NF-κB target genes (fig. S3B). Furthermore, depletion of Tsg101, Vps28, UBAP1, or CHMP4B resulted in the secretion of interleukin-8 (IL-8) and TNF-α proteins (Fig. 2, E and F). The amounts of IL-8 produced by the knockdown cells were comparable or greater than those generated upon stimulation of cells with TNF-α (Fig. 2E). Together, these results suggest that Tsg101, Vps28, UBAP1, and CHMP4B inhibit the expression of NF-κB target genes in serum-grown cells without exogenous stimulation.

Fig. 2 Depletion of Tsg101, Vps28, UBAP1, or CHMP4B induces the expression of NF-κB target genes.

(A to C) Depletion of Tsg101 (A) or Vps28 (B) activates the expression of several genes of the Human NF-κB Signaling Pathway PCR Array. The black line indicates no change (fold change = 1); the magenta lines mark a threefold change, whereas genes whose expression was induced ≥ 3-fold are shown in red. (C) Genes whose expression was increased at least threefold by both the SMARTpool (SP) and single siRNAs specific for Tsg101 or Vps28. A single experiment for each siRNA reagent was performed. (D) The expression of six NF-κB target genes (indicated above the graphs) was measured by quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis upon the silencing of four ESCRT subunits with single Dharmacon SMARTpool siRNAs. Data are means ± SEM of at least three independent experiments. (E and F) The concentrations of secreted IL-8 (E) and TNF-α (F) were measured by enzyme-linked immunosorbent assay (ELISA)–based analysis of culture medium collected upon the silencing of four ESCRT subunits in HEK 293 cells with Ambion siRNA duplexes or upon treatment with TNF-α (2 ng/ml) or AgoLTβR for 72 hours. Data are means ± SEM of at least three independent experiments. Data in (D) to (F) were analyzed by Mann-Whitney U test. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01.

Tsg101, Vps28, UBAP1, and CHMP4B decrease the expression of NF-κB target genes during zebrafish development and the immune response in flies

We extended our cell culture findings to the development of a vertebrate and the immune response of an invertebrate model organism. We analyzed the expression of NF-κB target genes in zebrafish embryos injected with morpholinos targeting fish orthologs of Tsg101, Vps28, UBAP1, and CHMP4B (which is encoded by two genes: chmp4ba and chmp4bb). As a control, we used previously validated morpholinos against IKK1, depletion of which induces the expression in fish of NF-κB target genes, such as nfkbiaa, il1b, and nfkb2 (26, 27). We measured their expression by RT-PCR assays in whole-embryo lysates (Fig. 3A). We detected a substantial increase of nfkb2 expression in vps28 morphants, as well as of il1b expression in ubap1 and chmp4ba/chmp4bb morphants, which in all cases exceeded the corresponding values in ikk1 morphants. Consistently, we found the increased abundance of nfkbiaa mRNA by ISH in vps28, ubap1, and chmp4ba/chmp4bb morphants (Fig. 3, B to D). In all cases, increased nfkbiaa expression was detectable in the same tissues, with enrichment in the pronephros, which suggests that these phenotypes were specific. Together, these analyses of target gene expression in fish embryos suggest that the ESCRT components inhibit constitutive NF-κB signaling in lower vertebrates.

Fig. 3 Depletion of ESCRTs induces the expression of NF-κB target genes in zebrafish embryos and in adult fly fat bodies.

(A) The expression of nfkb2, il1b, and nfkbiaa was analyzed by qRT-PCR analysis of 48-hour-old fish embryos depleted of Tsg101, Vps28, UBAP1, CHMP4BA/BB, or IKK1 and compared to wild-type (WT) or control morpholino (Ctrl MO)–treated embryos. Data are means ± SEM of three to five experiments and were analyzed by Student’s t test. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01. Right: Depletion of Tsg101, Vps28, and UBAP1 was determined by Western blotting analysis with antibodies against the indicated human proteins. The available CHMP4B-specific antibodies did not recognize fish proteins. (B to D) In situ hybridization (ISH) for nfkbiaa (IκBα) mRNA in fish embryos upon depletion of Vps28 (B), UBAP1 (C), or CHMP4BA/BB (D). Lateral view of the tail of 28-hour-old embryos. Arrows mark the posterior extremity of the pronephric ducts. Representative examples of nfkbiaa expression in two noninjected controls [left in (C) and (D)] or an embryo injected with a control morpholino [left in (B)]. Middle and right pictures in (B) to (D) depict two different embryos with altered nfkbiaa expression after the knockdown of different ESCRT components. At least three independent experiments were performed for each gene. (E) The expression of the NF-κB target genes Diptericin and Drosomycin was measured by qRT-PCR assays in adult Drosophila fat bodies after inducing the expression of RNAi against Tsg101 and Vps28 specifically in this immune organ. (F) Measurement of the expression of the EGFR target vein was performed as described in (E). Data in (E) and (F) are means ± SEM of five to nine independent experiments (offspring from different crosses). Data were analyzed by Kruskal-Wallis test with Dunnett’s post test. ns, P > 0.05; **P ≤ 0.01.

In Drosophila melanogaster, the NF-κB–like transcription factors Relish and Dif/Dorsal drive the expression of genes encoding antimicrobial peptides downstream of the Imd and Toll pathways, respectively (28). We used RNAi in adult flies to assess the effect of ESCRT-I knockdown on NF-κB target gene expression in the fat body, the main immune-responsive tissue in flies. As a control, we overexpressed the activating receptors peptidoglycan recognition protein LCx (PGRP-LCx, for Imd) and Toll10b (for Toll). In the absence of any infection, RNAi against Vps28 induced Relish-dependent, but not Dif/Dorsal-dependent, gene transcription in the fat body (Fig. 3E), suggesting that in flies, the regulation of NF-κB signaling by ESCRT proteins is specific to the Imd pathway. Expression of vein, a target gene of the EGFR pathway, was not altered by the knockdown of ESCRT components (Fig. 3F).

Depletion of Tsg101, Vps28, UBAP1, or CHMP4B activates both canonical and noncanonical NF-κB signaling

We observed the hallmarks of canonical NF-κB activation upon depletion of ESCRT components. Knockdown of Tsg101, Vps28, UBAP1, or CHMP4B in HEK 293 cells induced the nuclear accumulation of RelA (Fig. 4, A and B) and the activating phosphorylation of RelA on Ser536 (29), as well as the phosphorylation of IκBα and, to various extents, its degradation (Fig. 4C). We then verified that noncanonical NF-κB signaling also functioned in HEK 293 cells. Agonistic antibody to the lymphotoxin β receptor (AgoLTβR) (30) induced the processing of p100 protein to the transcriptionally active p52, whereas TNF-α exerted a weaker effect (Fig. 4D, top). Knockdown of Tsg101, Vps28, UBAP1, or CHMP4B resulted in an accumulation of the p100 precursor protein (consistent with the increased expression of NFKB2; Fig. 2D) and in the processing of p100 to p52 (Fig. 4D, bottom). These data suggest that there was a marked induction in RelA-dependent canonical and p52-dependent noncanonical NF-κB signaling upon depletion of ESCRT-I or CHMP4B.

Fig. 4 ESCRT silencing independently activates canonical and noncanonical NF-κB signaling.

(A) Immunofluorescence analysis of RelA in HEK 293 cells upon silencing of the indicated ESCRT subunits or stimulation of control siRNA–treated cells with AgoLTβR or TNF-α (10 ng/ml), both of which induced the nuclear translocation of RelA. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). (B) RelA fluorescence in nuclei of cells from three experiments performed as described in (A) was quantified. Data are means ± SEM of three independent experiments, each quantifying ~200 cells. *P ≤ 0.05, **P ≤ 0.01 by Student’s t test. (C) Lysates of unstimulated, ESCRT-depleted cells were analyzed by Western blotting to detect the degradation and phosphorylation of IκBα and the phosphorylation of RelA. Cells transfected with control siRNA (Ctrl) were also treated with TNF-α (10 ng/ml) for 30 min, which induces canonical NF-κB signaling. (D) HEK 293 cells treated for the indicated times with TNF-α (2 ng/ml) or AgoLTβR (top) and unstimulated cells depleted of ESCRT subunits (bottom) were lysed and analyzed by Western blotting to examine the abundance and processing of p100. (E) HEK 293 cells were transfected individually with 30 nM SMARTpool (SP) siRNAs targeting Vps28, RelA, or p100, or were cotransfected with 30 nM Vps28 SP in combination with 30 nM siRNA specific for p100 or RelA. The siRNA amounts in each transfection were equalized by control SMARTpool. Cells transfected with control SMARTpool (Ctrl) were also treated with TNF-α (10 ng/ml) for 30 min. Samples were analyzed by Western blotting as described for (C) and (D). Western blots in (C) to (E) are representative of two to four experiments.

We examined whether both branches of NF-κB signaling were activated independently of one another or whether one of the pathways induced the other in cells in which ESCRT proteins were knocked down. We performed double knockdown experiments, targeting Vps28 together with either RelA or p100 (Fig. 4E). In Vps28 knockdown cells, the additional depletion of RelA did not block the processing of p100 to p52, whereas the additional depletion of p100 did not prevent the phosphorylation of RelA. These findings suggest that the canonical and noncanonical NF-κB pathways are activated independently upon depletion of ESCRT subunits.

ESCRT subunits act upstream of the IKK complexes

To identify at which level of the NF-κB pathway the ESCRTs acted, we inhibited NF-κB signaling by ectopic expression of the IκBα-S32/36A super-repressor (31, 32) or the IKK2-K44M dominant-negative mutant (33). Either mutant abolished the reporter activity induced by depletion of Tsg101 or Vps28 (fig. S4, A and B). As a control, ESCRT silencing did not affect the activation of the NF-κB pathway caused by RelA overexpression (fig. S4C). This finding suggests that Tsg101 and Vps28 act upstream of IKK2 and IκBα, indicating that ESCRTs may control the trafficking of ligands and receptors that activate the NF-κB pathway.

The activation of NF-κB upon ESCRT depletion depends on LTβR and TNF receptor 1

We excluded the possibility that the action of ESCRT proteins requires serum-derived ligands, because the NF-κB activation observed in cells depleted of Tsg101 or Vps28 was preserved in serum-starved cells (Fig. 5A). Because ESCRTs sort membrane receptors toward degradation, we investigated the amounts of such receptors that activate NF-κB signaling and are present in HEK 293 cells: TNF receptor 1 (TNFR1), LTβR, and the IL-1 receptor 1 (IL1R1). Knockdown of ESCRT subunits did not change the abundances of TNFR1 or IL1R1, whereas the observed increases in the amounts of LTβR in the knockdown cells were not statistically significant (Fig. 5B). We confirmed that the abundance of LTBR mRNA was not altered under these conditions (fig. S5A). In cells depleted of ESCRT components, there was also no detectable increase in LTβR abundance at the plasma membrane as verified by immunofluorescence analysis of nonpermeabilized cells (fig. S5B) and by flow cytometric analysis (fig. S5, C and D). This finding suggests that the lack of ESCRT-I does not increase the recycling of LTβR, in contrast to the reported recycling of EGFR that occurs upon depletion of the early-acting ESCRT subunits (17).

Fig. 5 The effects of ESCRT depletion depend on receptors, but not on serum-derived ligands.

(A) HEK 293 cells transfected with the indicated SMARTpool (SP) siRNAs and kept in standard medium (10% FBS) or deprived of serum for the indicated times (Starv.) were lysed and subjected to luciferase assays to assess NF-κB pathway activity. (B) Top: Lysates of HEK 293 cells depleted of the indicated ESCRT subunits were analyzed by Western blotting with antibodies against LTβR, TNFR1, or IL1R1, with GAPDH, vinculin, and tubulin used as loading controls. Bottom: Quantification of LTβR protein abundance upon ESCRT knockdown. (C) HEK 293 cells transfected with siRNAs targeting ESCRT components alone or in combination with siRNA specific for LTβR or TNFR1 were lysed and subjected to luciferase assays to assess NF-κB pathway activity. Data in (A) to (C) are means ± SEM of three to four experiments. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 by Student’s t test. (D) The expression of NF-κB target genes (indicated above the graphs) was measured by qRT-PCR analysis of HEK 293 cells in which ESCRT subunits were knocked down alone or in combination with depletion of LTβR or TNFR1 with Ambion single siRNA duplexes. Additionally, control siRNA–transfected cells (Ctrl) were treated with TNF-α (2 ng/ml) or a 1:1000 dilution of AgoLTβR. Data are means ± SEM of four independent experiments. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 by Student’s t test or Mann-Whitney U test when compared to knockdown of individual ESCRT subunits. ns, P > 0.05; #P ≤ 0.05, ##P ≤ 0.01, ###P ≤ 0.001 by Student’s t test when compared to control normalized to 1 (cells transfected with control siRNA and left unstimulated).

To test whether NF-κB–inducing receptors were required for the pathway activation that occurred after knockdown of ESCRT proteins, we analyzed the effects of co-depletion of Tsg101, Vps28, UBAP1, or CHMP4B with either LTβR or TNFR1, the two prototypical receptors that stimulate the noncanonical and canonical NF-κB signaling pathways, respectively. Silencing of LTβR or TNFR1 prevented the activation of the NF-κB luciferase reporter in cells depleted of individual ESCRT subunits (Fig. 5C). Moreover, the expression of endogenous NF-κB target genes as a result of ESCRT protein silencing in HEK 293 cells was in most cases reduced when either LTβR or TNFR1 was also depleted (Fig. 5D and fig. S6A). We repeated these experiments in A549 cells, which also express both receptors, and we observed the same tendency: knockdown of ESCRT subunits activated transcription of NF-κB target genes, but this was reduced upon co-depletion of LTβR or TNFR1 (fig. S6, B and C). Together, these results suggest that the activation of NF-κB that occurs upon depletion of ESCRT components requires LTβR and TNFR1.

We verified that signaling by agonist-stimulated LTβR did not require the presence of TNFR1 and vice versa, as evidenced by the unchanged transcriptional activation of NF-κB target genes upon depletion of the other receptor (fig. S7). Therefore, both receptors are likely to contribute to NF-κB activation upon ESCRT silencing independently of each other and in an additive manner. Further supporting this notion, depletion of one of these receptors only reduced the induction of target gene transcription upon ESCRT silencing but did not abolish it (Fig. 5D and fig. S6).

LTβR and TNFR1 localize on enlarged endosomes upon ESCRT depletion

We tested whether LTβR or TNFR1 was enriched on intracellular compartments upon depletion of Tsg101, Vps28, UBAP1, or CHMP4B. ESCRT dysfunction induces the formation of oversized endosomes (34). Indeed, LTβR and TNFR1 were detected and enriched on enlarged vesicles (Fig. 6A and fig. S8A), as measured by increased fluorescence intensities per vesicle for both receptors, with higher values for LTβR (Fig. 6B). As a control, we depleted components of ESCRT-0, ESCRT-II, and ESCRT-III, but we observed no substantial increases in the fluorescence intensity of LTβR or TNFR1 per vesicle (fig. S8, B and C), consistent with the data that were suggestive of specific roles of ESCRT-I components and CHMP4B in the regulation of NF-κB (fig. S3B).

Fig. 6 ESCRT-dependent trafficking of LTβR and TNFR1.

(A and B) LTβR and TNFR1 accumulate on vesicles upon ESCRT depletion. (A) Immunofluorescence analysis of LTβR and TNFR1 in HEK 293 cells upon silencing of the indicated ESCRT components (individual channels shown in fig. S8A). Insets: Magnified views of boxed regions in the main images. Scale bars, 20 μm. (B) Fluorescence of LTβR- and TNFR1-positive vesicles was quantified by MotionTracking software. Integral vesicle intensities were divided by the number of vesicles, indicating the intensity of LTβR or TNFR1 staining per vesicle. Data are means ± SEM of three independent experiments, each quantifying 150 to 200 cells per condition. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01 by Mann-Whitney U test and Student’s t test. (C) The effects of ESCRT depletion do not depend on autocrine TNF-α. NF-κB pathway activity was measured by luciferase assays in lysates of cells transfected with the indicated SMARTpool (SP) siRNA and incubated with (+αTNF) or without (−αTNF) TNF-α–blocking antibody. As a control for the activity of the TNF-α–blocking antibody, cells were stimulated with TNF-α. Data are means ± SEM of three independent experiments. ns, P > 0.05; *P ≤ 0.05 by Student’s t test. (D and E) Colocalization of LTβR with UBAP1 or CHMP4B (D) and with TRAF2 or ubiquitin (Ub) (E) in A549 cells upon stimulation with AgoLTβR for the indicated times. Individual channels are shown in fig. S11 (A and B). Insets: Magnified views of boxed regions in the main images. Scale bars, 20 μm.

Ligands of LTβR and TNFR1 are absent or dispensable for signaling induced by ESCRT deficiency

Vesicular accumulation of receptors and their indispensability for NF-κB activation upon ESCRT depletion suggested that intracellular receptors might be actively signaling. We thus tested the involvement of their ligands under these conditions. TNF-α–blocking antibodies did not affect the activation of NF-κB upon ESCRT depletion, whereas they blocked the stimulation of cells with TNF-α (Fig. 6C). Therefore, although TNF gene expression and TNF-α production were increased in our system (Fig. 2, D and F), we could exclude the involvement of a stimulatory TNF-α autocrine loop, which was reported in various models (3537). Moreover, we verified that HEK 293 cells do not express detectable amounts of the LTβR ligands lymphotoxin α and β, which are encoded by LTA and LTB, respectively. In qRT-PCR analysis with validated probes, the Ct values for LTA and LTB exceeded 35, as compared to 27 for LTBR (LTβR) and TNFRSF1A (TNFR1) (table S1). Together, these data suggest that ligands of cytokine receptors may be dispensable for NF-κB signaling under ESCRT-depleted conditions.

LTβR signaling contributes to NF-κB activation upon loss of ESCRT components

We set out to delineate the mechanism by which an unbound receptor could contribute to NF-κB activation upon ESCRT deficiency. We focused on LTβR because its ligands were lacking in HEK 293 cells and because of its greater contribution to NF-κB activation (Fig. 5, C and D) and its increased intracellular accumulation (Fig. 6, A and B) compared to TNFR1. Moreover, in addition to stimulating noncanonical signaling, LTβR can also stimulate the canonical NF-κB pathway (Fig. 4, A and B) (38), and its overexpression potently activates NF-κB signaling without exogenous ligands (39). Indeed, we confirmed that the overexpression of LTβR induced expression of the same NF-κB target genes as did ESCRT depletion in HEK 293 cells (fig. S9).

To further corroborate the involvement of LTβR in NF-κB signaling upon ESCRT deficiency, we used its previously described deletion mutants, Δ389 and Δ359 (40, 41). Both are signaling-deficient, with Δ359 reported as being dominant-negative. We verified that in contrast to overexpression of the wild-type receptor, overexpression of either mutant receptor did not stimulate the expression of NF-κB target genes, luciferase reporter activity, or p100-p52 processing (fig. S10, A to C). When expressed in ESCRT-silenced cells, both mutants reduced the extent of activation of NF-κB; however, residual activation above baseline in control nonsilenced cells was still visible in most cases (fig. S10D), consistent with the proposed parallel involvement of other receptors. As a control, overexpression of wild-type LTβR further increased NF-κB activity beyond that induced by knockdown of the ESCRT subunits (fig. S10D). Consistently, AgoLTβR also potentiated the NF-κB response in ESCRT-depleted cells (fig. S10E). Together, these data suggest that LTβR signaling contributes to NF-κB activation upon depletion of ESCRT proteins.

Accumulated LTβR oligomerizes and signals from early endosomes upon depletion of ESCRT proteins

To understand the relevance of the accumulation of LTβR on vesicles upon loss of ESCRT proteins (Fig. 6, A and B), we first investigated whether its normal agonist-induced trafficking proceeds through ESCRT-positive endosomes. To visualize endogenous ESCRT subunits, we established a protocol to permeabilize cells before fixation, which reduces the cytoplasmic background. With this procedure, we observed costaining of LTβR with UBAP1 and CHMP4B on vesicles in cells stimulated with AgoLTβR for 15 to 60 min (Fig. 6D and fig. S11A). We could not detect specific signals for available antibodies against Tsg101 or Vps28; however, LTβR-positive endosomes colocalized with ubiquitin and, to a lesser degree, with the signaling adaptor TNFR-associated factor 2 (TRAF2) (Fig. 6E and fig. S11B), as expected for an active receptor. These data suggest that agonist stimulates the internalization of LTβR and its ESCRT-mediated endosomal transport.

To characterize receptor trafficking under ESCRT-depleted conditions, we measured its colocalization with endosomal markers. ESCRT depletion led to increased colocalization of LTβR with EEA1, which marks early endosomes (Fig. 7, A and B, fig. S12A), but not with the late endosomal marker LAMP1 (fig. S12B). Quantitative analysis of LTβR fluorescence on vesicles showed marked increases in total intensity (Fig. 7C) and mean intensity per unit of vesicle area (Fig. 7D). This latter parameter denotes the increased concentration of LTβR on endosomes when ESCRT proteins were depleted. Furthermore, intracellular cross-linking after knockdown of ESCRT components revealed a greater extent of LTβR oligomerization, which is indicative of receptor activation (Fig. 7E). When stimulated, LTβR recruits the adaptor proteins TRAF3 and TRAF2 (5), the latter also being a signaling effector of TNFR1 (1). The active signaling status of LTβR was confirmed by its coimmunoprecipitation with TRAF2 and TRAF3 in cells depleted of Tsg101, Vps28, or UBAP1 (Fig. 7F). In the same cell lysates, interactions between TNFR1 and TRAF2 were undetectable, possibly because they were less stable or more transient. No TRAF2 or TRAF3 was associated with LTβR in control cells, suggesting that ESCRT deficiency caused LTβR activation. Consistently, TRAF2 was detected on LTβR-positive endosomes by microscopy, similar to ubiquitin (Fig. 7G and fig. S12C). Together, these results suggest that the accumulation of LTβR stimulates NF-κB signaling from endosomes.

Fig. 7 LTβR accumulates at and signals from early endosomes upon ESCRT depletion.

(A) Immunofluorescence of LTβR in HEK 293 cells upon silencing of the indicated ESCRT components or in control siRNA–transfected cells stimulated with AgoLTβR. Individual channels are shown in fig. S12A. Insets: Magnified views of boxed regions in the main images. Scale bars, 20 μm. (B to D) LTβR-positive vesicles were analyzed by MotionTracking software. Control siRNA–transfected cells (Ctrl) were treated with TNF-α (10 ng/ml) or a 1:1000 dilution of AgoLTβR for 30 min. (B) Colocalization of LTβR with EEA1 was quantified. Integral (C) and mean (D) vesicle intensities were measured. Data in (B) to (D) are means ± SEM of three independent experiments, each quantifying 150 to 200 cells per condition. Data in (B) and (D) were analyzed by Student’s t test, whereas data in (C) were analyzed by Mann-Whitney U test. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01. (E) Oligomerized LTβR was detected by intracellular cross-linking (+DSP) and Western blotting analysis under nonreducing conditions. (F) Immunoprecipitation (IP) of LTβR or TNFR1 upon ESCRT depletion. TRAF2, TRAF3, and the indicated receptors were detected in immunoprecipitates by Western blotting analysis. The asterisk indicates the LTβR-specific signal that remains after reprobing of the blot. Western blots are representative of three experiments. (G) Colocalization of LTβR with TRAF2 or ubiquitin (Ub) in A549 cells depleted of the indicated ESCRT components was determined by immunofluorescence analysis. Individual channels are shown in fig. S12C. Insets: Magnified views of boxed regions in the main images. Scale bars, 20 μm.


Our data identify ESCRT-I and CHMP4B as inhibitors of constitutive NF-κB activity. ESCRT complexes sort ligand-induced receptors of various classes toward degradation. Here, we showed that ESCRTs limited basal NF-κB signaling by regulating the distribution of cytokine receptors. Specifically, we propose that constitutive ESCRT-dependent trafficking of ligand-free cytokine receptors prevents their accumulation and activation. In the absence of ESCRT-I function, receptors such as LTβR or TNFR1 became concentrated on and signaled from enlarged endosomes. In the case of LTβR, endosomal accumulation led to its activation by oligomerization and the recruitment of the signaling adaptors TRAF2 and TRAF3 in the absence of ligands. Such endosomal signaling of LTβR, TNFR1, and possibly of other cytokine receptors resulted cumulatively in the NF-κB transcriptional response. This is in contrast to EGFR, which also accumulates on endosomes upon ESCRT deficiency; however, the expression of EGF target genes is not induced (18). Our study further establishes that cytokine receptors are also signaling-competent when localized to endosomes.

Oligomerization-induced receptor activation represents a key principle of cytokine signaling. It is thus possible that such receptors, devoid of enzymatic activity, may be especially prone to spurious activation due to their increased local concentration even in the absence of ligands. It is plausible that active, oligomerized receptors are preferentially recognized by ESCRTs, which target them for degradation, preventing overstimulation of the pathway. Mechanistic details of ESCRT-dependent trafficking may vary for different cytokine receptors, possibly depending on the availability of ligands. In case of TNFR1, TNF-α was produced by HEK 293 cells upon knockdown of ESCRT components (although it seemed to be dispensable for signaling; Fig. 6C). In contrast, LTβR ligands were absent in HEK 293 cells; however, upon depletion of ESCRT components, the receptor was constitutively active. Such ligand-independent receptor signaling also operates in other model systems. For example, Drosophila PGRP-LC activates NF-κB signaling in a ligand-independent manner through the clustering of overexpressed receptors (42), similarly to what we observed for overexpressed LTβR.

The repertoire of ESCRT subunits involved suggests that the early steps of endosomal sorting of cytokine receptors are crucial to inhibit their signaling. Three of the identified regulators are components of ESCRT-I, whereas CHMP4B (ESCRT-III) cooperates with ESCRT-0 and ESCRT-I in the early steps of cargo sorting (25). Although the depletion of Hrs (ESCRT-0) did not induce NF-κB signaling in our system, its function may be redundant with that of other adaptor proteins (43, 44). Depletion of ESCRT-I or CHMP4B resulted in the capture of ligand-free LTβR at the limiting membrane of endosomes where it could still signal. This mechanism may be also responsible for the ligand-induced internalization of LTβR because the silencing of ESCRT components and treatment of cells with the LTβR agonistic antibody AgoLTβR potentiated the NF-κB response (fig. S10E). Consistent with this, components of the early-acting ESCRT-0 and ESCRT-I were reported to inhibit ligand-induced NF-κB signaling downstream of Toll-like receptor 4 (45) and nucleotide-binding oligomerization domain-containing protein 2 (NOD2) (46).

Our results offer a fresh perspective on the role of intracellular trafficking in constitutive NF-κB signaling. Under basal conditions, endocytic defects could mimic cytokine stimulation and induce a pronounced proinflammatory response through receptor activation, also in a ligand-independent manner. When degradation of cytokine receptors is perturbed, their endosomal concentration and oligomerization appear to be sufficient for their signaling to activate transcription factors. Although ESCRT deficiency leads to endosomal accumulation of multiple receptor types in an active form, including EGFR, transcriptional changes are restricted to the NF-κB pathway (Fig. 3F) (18). In general, the role of ESCRTs in preventing excessive constitutive NF-κB signaling underscores the function of endosomal sorting in maintaining cellular homeostasis.


Cell culture and stimulation

HEK 293 cells and A549 (human lung carcinoma) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM, high glucose, Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS) and 2 mM l-glutamine (Sigma-Aldrich). In cell stimulation experiments, TNF-α (1 to 10 ng/ml) or AgoLTβR (1:1000 dilution) was diluted in starvation medium (DMEM, 0.2% bovine serum albumin), except for the TNF-α–blocking experiment.

TNF-α blocking

HEK 293 cells were transfected with the appropriate siRNAs and luciferase reporter plasmids with Attractene (QIAGEN). After 16 hours, the culture medium was exchanged to standard DMEM containing 10% FBS in the absence or presence of anti–TNF-α antibody (5 μg/ml). As a control, cells were stimulated with TNF-α (2 ng/ml) in the presence or absence of anti–TNF-α antibody (5 μg/ml). Cells were lysed 72 hours after transfection and were used for luciferase assays.

FBS starvation of cells

HEK 293 cells were transfected with 50 nM SMARTpool siRNA and luciferase reporter plasmids with Attractene. Either 24 or 48 hours before lysis, the culture medium was exchanged for starvation medium. The cells were lysed 72 hours after transfection and were used for luciferase assays.

Luciferase assays

HEK 293 cells were reverse-transfected with SureFECT (SABiosciences) or Attractene. The cells were cotransfected with the reporter plasmids pNF-κB-luc (encoding firefly luciferase, reporting pathway activity) and pRL-TK-luc (encoding Renilla luciferase, reporting transfection efficiency) and with esiRNA, SMARTpool siRNA, or single siRNA duplexes. Seventy-two hours after transfection, cells were lysed in passive lysis buffer (Promega), and either luciferase assays or Western blotting analysis (Fig. 1) was performed. If needed, cells were stimulated with TNF-α (2 ng/ml; 1 ng/ml for the RNAi screen) or a 1:1000 dilution of AgoLTβR 5 hours before cell lysis. In cell lysates, the luciferase activity was measured with a microplate luminometer (BMG LUMIstar Galaxy or Berthold Technologies Centro XS3 LB 960). The firefly luciferase signal was normalized to its respective Renilla luciferase count. Measurements are presented as the fold change relative to the basal activity of the pathway in cells transfected with control vector or nontargeting siRNA. Assay details are provided in the Supplementary Materials.

qRT-PCR assays for HEK 293 and A549 cells

Standard RNA extraction and cDNA synthesis protocols were used. The expression of 84 genes of interest in HEK 293 cells was measured with a 96-well PCR array (Human NF-κB Signaling Pathway PCR Array, PAHS-025, SABiosciences). The Ct-based fold change calculations were performed with RT2 PCR Data Analysis software (SABiosciences). The expression of six selected NF-κB target genes was verified with the pairs of primers listed in table S5. Relative quantification method and DataAssist Software (Applied Biosystems) were used to estimate the fold change in gene expression. For the profiling of LTBR, TNFRSF1A, LTA, and LTB expression, TaqMan (Life Technologies) products were used. Details of the qRT-PCR methods used are provided in the Supplementary Materials.

Enzyme-linked immunosorbent assay

Concentrations of IL-8 or TNF-α in HEK 293 cell culture medium were measured with the quantitative sandwich enzyme immunoassay technique (Quantikine ELISA, R&D Systems). Culture medium was collected 72 hours after the cells were transfected with 10 nM siRNA or treated with TNF-α (2 ng/ml) or a 1:1000 dilution of AgoLTβR. Assays were performed according to the manufacturer’s instructions. Data were analyzed with a four-parameter logistic curve fit.

Immunofluorescence and microscopy

Immunofluorescence staining protocols for permeabilized and nonpermeabilized cells are provided in the Supplementary Materials. Quantification of RelA-positive nuclei was performed with ImageJ software, whereas analyses of LTβR or TNFR1 fluorescence were performed with MotionTracking (

Intracellular cross-linking

To analyze the oligomerization of LTβR in HEK 293 cells, cross-linking with membrane-permeable DSP [dithiobis(succinimidyl propionate); Thermo Scientific] was performed 72 hours after transfection. Cells were washed twice with phosphate-buffered saline (PBS) and then incubated for 30 min at room temperature with 0.5 mM DSP in PBS. After cross-linking, the cells were washed twice with PBS and incubated for 15 min at room temperature in 20 mM tris-HCl-PBS (pH 7.4) to stop the reaction. The cells were lysed in RIPA (radioimmunoprecipitation assay) buffer [1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM tris (pH 8.0), 150 mM NaCl, 0.5 mM EDTA] and analyzed by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) under nonreducing conditions.


HEK 293 cells were transfected with 30 nM siRNA with HiPerFect. After 72 hours, the cells were washed once with PBS and lysed in RIPA buffer at 4°C. Lysates were clarified by centrifugation, and in addition, DNA was sheared with QIAshredder columns (QIAGEN). After preclearance for 3 hours with goat immunoglobulin G (Sigma-Aldrich), nonspecifically bound proteins were depleted with Protein G agarose (Roche) for 1 hour. Cell lysates were then incubated overnight at 4°C with the appropriate antibodies coupled to Protein G agarose. The agarose beads were washed four times with IP buffer (50 mM Hepes, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 10% glycerol) before elution with 0.1 M glycine (pH 2.5). Samples were then resolved by SDS-PAGE.

Developmental staging and maintenance of zebrafish embryos

General maintenance, collection, and staging of the zebrafish were performed according to The Zebrafish Book: A Guide for the Laboratory Use of Zebrafish (Danio rerio) (47). RT-PCR experiments were performed in wild-type zebrafish (AB strain), whereas ISH was performed in wild-type AB/TÜ hybrids. Embryos were maintained in Danieau zebrafish medium and grown at 28°C. The approximate developmental stages are given in hours postfertilization (hpf) at 28°C, as described previously (48).

Morpholino injections

Translation-blocking morpholino (MO) antisense oligonucleotides (Gene Tools) were designed for tsg101a, vps28, ubap1, chmp4ba, and chmp4bb. The IKK1 MO was previously described (26). To determine the effects of gene silencing on NF-κB–dependent genes, 48-hpf MO-injected embryos were manually dechorionated, frozen in liquid nitrogen, and stored at −80°C for RNA extraction. Morpholino sequences and doses as well as details of the RT-PCR and ISH procedures are provided in the Supplementary Materials.

Drosophila husbandry, RNAi, and qRT-PCR

Interfering RNA was expressed in a tissue-specific manner using the UAS-Gal4 system (49). Driver females (w;c564-Gal4;tub-Gal80ts) were crossed to control males (w or w;;UAS-PGRP-LCx or w;UAS-Toll10b) or to males carrying RNAi constructs. Inducible RNAi lines were obtained from the Vienna Drosophila RNAi Centre (transformant ID in brackets): tsg101 (23944, GD library); vps28 (31894, GD library; 105124, KK library). These lines have been validated in independent RNAi screens (50, 51). Crosses were raised at the restrictive temperature of 18°C to prevent developmental effects. Female progeny were shifted to 29°C for 3 days to induce RNAi expression. Fat bodies (10 to 15) were dissected for mRNA extraction and RT-PCR with the Roche SYBR Green kit compatible with a 96-well plate LightCycler. Primers used for target genes are listed in table S7.

Statistical analysis

Statistical analyses were performed with STATISTICA 8 software (StatSoft), Microsoft Excel, or GraphPad Prism 5.0 (GraphPad Software Inc.). Data were tested for normality with the Shapiro-Wilk normality test. Statistical analyses of nonnormally distributed data were performed with the Mann-Whitney U test. Data with normal distribution were analyzed with one sample or two independent sample t test (referred to as Student’s t test). Kruskal-Wallis test with Dunnett’s posttest was performed for fly data (Fig. 3E). Data points were marked according to the P value, where P > 0.05 is marked “ns” or left unmarked, *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001.


Materials and Methods

Fig. S1. RNAi screen does not identify inhibitors of the NF-κB pathway upon TNF-α stimulation.

Fig. S2. Depletion of Tsg101, Vps28, UBAP1, or CHMP4B activates the NF-κB pathway.

Fig. S3. Depletion of Tsg101, Vps28, UBAP1, or CHMP4B activates the expression of NF-κB target genes.

Fig. S4. ESCRTs act in the NF-κB pathway upstream of the IKK complexes.

Fig. S5. Depletion of ESCRT components does not affect the expression of LTBR or the abundance of LTβR on the plasma membrane.

Fig. S6. Knockdown of TNFR1 or LTβR reduces the expression of NF-κB target genes induced by depletion of ESCRT components.

Fig. S7. LTβR and TNFR1 signal independently of each other.

Fig. S8. LTβR and TNFR1 are enriched on vesicles upon depletion of Tsg101, Vps28, UBAP1, or CHMP4B, but not Hrs, EAP30, or CHMP3.

Fig. S9. Overexpressed LTβR activates the expression of NF-κB target genes.

Fig. S10. Expression of LTβR mutants inhibits the NF-κB signaling induced by depletion of ESCRT components.

Fig. S11. Agonist-stimulated LTβR colocalizes with ESCRT subunits, TRAF2, and ubiquitin.

Fig. S12. LTβR accumulates on and signals from early endosomes upon depletion of ESCRT components.

Table S1. The average qRT-PCR threshold cycle values for genes in HEK 293 cells.

Table S2. Sequences of oligonucleotides used for esiRNA production.

Table S3. Sequences and catalog numbers of Dharmacon siRNAs and Stealth siRNAs.

Table S4. Sequences and catalog numbers of Ambion siRNAs.

Table S5. Sequences of oligonucleotides used for qRT-PCR analysis of HEK 293 and A549 cells.

Table S6. Sequences of oligonucleotides used for qRT-PCR analysis of zebrafish.

Table S7. Sequences of oligonucleotides used for qRT-PCR analysis of Drosophila.

References (5259)


Acknowledgments: We thank A. Mieżaniec for technical assistance and J. Cendrowski and N. Budick-Harmelin for critical reading of the manuscript. We thank A. Rao for the pcDNA3-IKK2-K44M construct, W. Green for the pCMV4-3 HA/IκBα (SS32,36AA) construct, and T. Lipniacki for the pcDNA-3Flag-Strawberry-RelA construct. Funding: This work was supported by a grant from Switzerland through the Swiss Contribution to the enlarged European Union (EU) (Polish-Swiss Research Programme project PSPB-094/2010) to M.M. and M.G.-G. A.B. and L.W.-N. were supported by the EU FP7 grant FishMed GA No. 316125; M.B.O. and E.S. by the Parent-Bridge program of the Foundation for Polish Science, co-financed from the EU under the European Regional Development Fund (POMOST/2010-1/1 and /2011-3/11); and K.J. by the Foundation for Polish Science within the International PhD Project “Studies of nucleic acids and proteins—From basic to applied research,” co-financed by the European Union Regional Development Fund. M.P. benefited from an ARC postdoctoral fellowship. Work in the laboratory of M.F. was supported by CNRS/INSERM ATIP/Avenir, Human Frontier Science Program, a Project ARC grant, and the Labex SIGNALIFE. Author contributions: A.M. conceived, designed, performed, and analyzed the experiments in Figs. 1, 4, 5 (A and C), 6 (A to C), and 7 (A to D) and figs. S1, S4, S5 (A to C), S8, S9, and S12 (A and B), and wrote the manuscript; A.B. designed and performed the experiments in Figs. 2 (E and F), 5D, and 7 (E and F) and figs. S2, S5D, S6, and S10; M.B.-O. contributed data to Figs. 2D, 3A, and 5D and figs. S3 and S6, and performed statistical analyses; I.P. performed the experiments in Fig. 2 (A to D) and fig. S3; K.J. performed the experiments in Figs. 6 (D and E) and 7G and figs. S11 and S12C; D.Z.-B. performed the experiments in fig. S7 and contributed data to Figs. 2 (E and F), 4D, and 5B and fig. S6 (A and C); I.C. performed morpholino injections for qRT-PCR analysis in Fig. 3A; M.P. and M.F. performed the experiments presented in Fig. 3 (B to D); C.N. performed the experiments in Fig. 3 (E and F); L.W.-N. contributed data to Fig. 3A; A.T. and E.S. designed and prepared the esiRNA library; A.K. performed the flow cytometric analysis in fig. S5 (C and D); K.P. helped in flow cytometry design and analysis; A.S., H.S., M.F., and M.G.-G. helped in conceiving or designing the experiments and provided reagents; and M.M. conceived, designed, and analyzed the experiments and wrote the manuscript. Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Signaling

Navigate This Article