Co-recruitment analysis of the CBL and CBLB signalosomes in primary T cells identifies CD5 as a key regulator of TCR-induced ubiquitylation

Abstract

T-cell receptor (TCR) signaling is essential for the function of T cells and negatively regulated by the E3 ubiquitin-protein ligases CBL and CBLB Here, we combined mouse genetics and affinity purification coupled to quantitative mass spectrometry to monitor the dynamics of the CBL and CBLB signaling complexes that assemble in normal T cells over 600 seconds of TCR stimulation. We identify most previously known CBL and CBLB interacting partners, as well as a majority of proteins that have not yet been implicated in those signaling complexes. We exploit correlations in protein association with CBL and CBLB as a function of time of TCR stimulation for predicting the occurrence of direct physical association between them. By combining co-recruitment analysis with biochemical analysis, we demonstrated that the CD5 transmembrane receptor constitutes a key scaffold for CBL- and CBLB-mediated ubiquitylation following TCR engagement. Our results offer an integrated view of the CBL and CBLB signaling complexes induced by TCR stimulation and provide a molecular basis for their negative regulatory function in normal T cells.

Keywords: CBL; CBLB; CD5; ubiquitylation.

Figure EV1. T cells from CBL^OST and CBLB^OST mice develop and function normally

1. Structure of the 3′ end of the wild‐type Cbl allele and of the targeted Cbl ^OST allele following homologous recombination and CRE‐mediated excision of the loxP‐neo^r‐loxP cassette. Exons are shown as filled black boxes and numbered. In the Cbl ^OST allele, the One‐STrEP‐tag (OST) is shown in red and the remaining loxP site in orange.

2. Structure of the 3′ end of the wild‐type Cblb allele and of the targeted Cblb ^OST allele following homologous recombination and FLP‐mediated excision of the frt‐neo^r‐frt cassette. Exons are shown as filled black boxes and numbered. In the Cblb ^OST allele, the One‐STrEP‐tag (OST) is shown in red, the IRES‐mTFP1 cassette in green, the two loxP sites in orange, and the remaining frt site in blue.

3. Flow cytometry analysis of thymus and spleen from wild‐type (WT) and CBL^OST mice for expression of CD4 versus CD8 and CD5 versus CD45R. Numbers adjacent to outlined areas indicate percentage of cells.

4. Flow cytometry analysis of thymus and spleen from wild‐type (WT) and CBLB^OST mice for expression of CD4 versus CD8 and CD5 versus CD45R. Numbers adjacent to outlined areas indicate percentage of cells.

5. Cellularity of thymus, spleen, and pooled mesenteric lymph nodes (Mln) from wild‐type (WT) and CBL^OST mice. Data are expressed as mean value ± SEM.

6. Cellularity of thymus, spleen, and pooled mesenteric lymph nodes (Mln) from wild‐type (WT) and CBLB^OST mice. Data are expressed as mean value ± SEM.

7. ATP content of CD4⁺ T cells purified from WT and CBL^OST mice and activated for 48 h with PMA and ionomycin (PI) or with plate‐bound anti‐CD3 (0.3 μg/ml) in the presence or absence of soluble anti‐CD28 (1 μg/ml). ATP content is directly proportional to the numbers of proliferating cells in the well and assessed by luminescence. Data are expressed as mean value ± SEM.

8. ATP content of CD4⁺ T cells purified from WT and CBLB^OST mice and activated for 48 h with PMA and ionomycin (PI) or with plate‐bound anti‐CD3 (0.3 μg/ml) in the presence or absence of soluble anti‐CD28 (1 μg/ml), assessed by luminescence. Data are expressed as mean value ± SEM.

9. IL‐2 in supernatants of WT and CBL^OST CD4⁺ T cells activated as in (G). Data are expressed as mean value ± SEM.

10. IL‐2 in supernatants of WT and CBLB^OST CD4⁺ T cells activated as in (H). Data are expressed as mean value ± SEM.

Data information: Data in (C‐J) are representative of at least three experiments with at least two mice per genotype.

Figure 1. Purification of CBL‐OST and CBLB‐OST proteins and quantification of their relative abundance

1. CD4⁺ T cells from wild‐type (WT), CBL^OST, and CBLB^OST mice were left unstimulated (−) or stimulated for 2 min with anti‐CD3 and anti‐CD4 antibodies (+). Equal amounts of cell lysates were subjected to affinity purification on Strep‐Tactin Sepharose beads, followed by elution of proteins with D‐biotin. Eluted proteins were analyzed by immunoblot with antibodies specific for CBLB and CBL. Equal amounts of proteins from whole‐cell lysates (WCLs) were also probed for CBLB and CBL. Note that in the case of CBLB, two bands are detected and likely corresponded to the presence of two isoforms differing at their N‐terminus (http://www.uniprot.org/uniprot/Q3TTA7). Also shown are the loading control corresponding to the same immunoblot probed with anti‐CD5, and molecular masses (kDa). Data are representative of at least three experiments.

2. CD4⁺CD8⁺ thymocytes and mature CD4⁺ T cells isolated from mice with the specified genotypes were permeabilized and stained with saturating amounts of Strep‐Tactin APC and analyzed by flow cytometry. Background staining was deduced using cells from WT mouse. Data are representative of two experiments.

Figure EV2. Assessment of biological and technical variability across samples

1. A, B

   For CBL‐OST (A) and CBLB‐OST (B) samples, the variability between samples corresponding to a given condition of activation was estimated by computing the Pearson correlation coefficient from log‐transformed intensities—normalized using indexed Retention Time peptide intensities (iRT Kit; Biognosys)—of all detected proteins for all pairs of technical and biological replicates (denoted as Rep. Tech. and Rep. Bio., respectively). For each condition of activation (i.e. prior to t = 0 s or following TCR engagement for t = 30 s, t = 120 s, t = 300 s, and t = 600 s), scatter plots of log‐transformed and iRT‐normalized intensities for all pairs of technical and biological replicates are represented along with the corresponding Pearson correlation coefficients (R). The average Pearson correlation coefficient across all pairs of technical and biological replicates (denoted as Average R) is also indicated for each time point. Strong correlations exist between technical replicates corresponding to the same biological replicate, thereby illustrating the reproducibility of the LC‐MS measurement.

2. C, D

   The iRT‐normalized intensity of the CBL (C) or CBLB (D) bait proteins is represented for all the LC‐MS runs corresponding to CBL‐OST, CBLB‐OST, and wild‐type (WT) backgrounds. In each sample, a strong enrichment of the bait protein is observed in AP‐MS purifications performed in the CBL‐OST or CBLB‐OST backgrounds as compared to control purifications performed in the WT background. In the case of control purifications from WT backgrounds, missing values are not represented.

Figure 2. Detection of specific interacting partners of CBL and CBLB in peripheral CD4⁺ T cells prior to and following TCR stimulation

CD4⁺ T cells from wild‐type (WT), CBL^OST, and CBLB^OST mice were left unstimulated (unstim) or stimulated for 30, 120, 300, or 600 s with anti‐CD3 and anti‐CD4 antibodies. Equal amounts of cell lysates were subjected to affinity purification using Strep‐Tactin Sepharose beads and then to MS analysis as described in Materials and Methods.

1. A

   Histogram comparing the normalized CBL protein intensities resulting from AP‐MS analysis of CD4⁺ T cells from WT and CBL^OST mice prior to and after the specified stimulation times. For each time point, the distributions of the normalized intensities of the nine data points (three biological replicates × three technical replicates) corresponding to CD4⁺ T cells from wild‐type (WT) and CBL^OST mice were compared using the adjusted P‐value from a one‐way ANOVA test (denoted as P(t)) and the value r(t) corresponding to the enrichment observed in the CBL^OST samples as compared to WT samples. Different colors represent distinct biological replicates.

2. B

   Proteins are classified as CBL interactors according to their position in a volcano plot in which the mean P‐value <P(t)> is plotted against the corresponding mean enrichment <r(t)> for CBL^OST versus WT samples. The time point shown corresponds to the unstimulated condition (t = 0 s). Mean was calculated using bootstrap resampling (see Results and Materials and Methods). Proteins that displayed an enrichment r(t) greater than twofold and a corresponding P‐value P(t) lower than a set threshold (see Materials and Methods) in more than 90% of the bootstrap iterations were selected as specific partners (indicated in red). Names of four of the most significantly enriched proteins are indicated. Dashed lines represent thresholds on the P‐value P* = 10⁻³ and the enrichment r* = 2 used to identify specific interactors of CBL.

3. C, D

   List of specific interacting partners of CBLB (C) and CBL (D) ranked according to their maximum mean enrichment across all time points. Heat maps show, for each time point, the fraction of bootstrap iterations for which the corresponding proteins were detected as specific partners (denoted as “P(Detection)”). Red dots indicate the proteins that have been previously reported as CBLB (C) or CBL (D) interactors in the BioGRID database. Ubiquitin (UB) is highlighted with a black arrow.

4. E

   Histograms showing the numbers of specific interactors binding to the CBL‐OST or CBLB‐OST proteins prior to and after TCR stimulation for different times and for which P(Detection) > 0.9.

5. F

   Pie charts showing the percentages of CBL and CBLB interactors detected in both the present study and the BioGRID public database. The size of the pie charts is proportional to the numbers of interactors per bait.

Figure 3. Comparison of the CBL and CBLB signalosomes of peripheral CD4⁺ T cells

Each protein interacting with either or both the CBL‐OST and CBLB‐OST baits is represented as a node and is linked by an edge to the corresponding bait. Proteins are identified by their BioGRID designations and color‐coded according to their function or protein family; see key (right). As specified, a set of proteins interacts with both CBL and CBLB, and CBLB is found in the CBL signalosome.

Figure EV3. GO term enrichment analysis of CBL and CBLB signalosomes

1. A, B

   Organization of the CBL (A) and CBLB (B) signalosomes according to GO annotations. Enriched GO terms were extracted from the list of interacting partners identified in this study using the ClueGO plugin (http://apps.cytoscape.org/apps/cluego) on Cytoscape.

Figure 4. Analysis of the correlations existing in the assembly with and disassembly from the CBL‐OST or CBLB‐OST baits for all pairs of identified preys

1. A

   Workflow used for the generation of the co‐recruitment correlation network (CN). Left: Normalized intensities of recruitment of two identified partners (denoted X _i and X _j) to the considered bait as a function of time. Center left: Normalized intensities of recruitment for preys X _i and X _j are plotted against each other (R _ij: Pearson correlation value; P _ij corresponding P‐value). Center right: Correlation matrix (R _ij) between all pairs of identified interactors partitioned into different clusters. Right: Portion of the obtained correlation network after filtering according to P _ij values.

2. B, C

   Representation of the CBL (B) and CBLB (C) CNs. Nodes represent specific interacting partners and edges connect pairs of nodes for which the P‐value associated with the Pearson correlation coefficient exceeded a set threshold. Edge width is coded according to Pearson correlation coefficient (R < 0.8 or R > 0.8). The identified edges are depicted in red if they have been already reported as direct interaction partners in the BioGRID database.

3. D, E

   Venn diagrams showing overlaps between protein–protein interactions deduced from the CBL (D) or CBLB (E) CNs and those reported in the BioGRID database for Homo sapiens and Mus musculus. Two sets of analysis are shown according to Pearson correlation coefficient values (R < 0.8 or R > 0.8). Histograms below the Venn diagrams compare the probabilities to predict existing interactions with a random network (Random), the complete CN, or the high‐confidence CN (R > 0.8).

Figure EV4. K‐means clustering analysis of the CBL and CBLB signalosomes

1. A, B

   Representation of the CBLB (A) and CBL (B) correlation matrix (R _ij) partitioned into different clusters using a K‐means clustering algorithm. The normalized recruitment intensity to the bait as a function of time is represented for the different interactors grouped into corresponding clusters. Within each cluster, interactors are listed in alphabetical order. Proteins from the 14‐3‐3 family and PI3K subunits are highlighted within the CBL signalosome.

Figure 5. Co‐recruitment network predicts the occurrence of physical association between pairs of recruited proteins

1. A

   CBLB first‐neighbors network within the CBL CN. Edge width is coded according to Pearson correlation coefficient (R < 0.8 or R > 0.8). The identified edges are depicted in red if they have been already reported as direct interaction partners in the BioGRID database. Proteins detected in the CBLB signalosome shown in Fig 3 are colored in red.

2. B

   Histograms showing the number and corresponding percentage of CBL interactors also present in the CBLB signalosome according to their distance from CBLB in the CBL CN. The two depicted analyses correspond to the complete and to the high‐confidence (R > 0.8) CNs.

3. C

   Zoom on the high‐confidence (R > 0.8) CRKL and CSK first‐neighbors subnetwork found within the CBL CN. The identified edges are depicted in red if they have been already reported as direct interaction in the BioGRID database. CSK and CRKL are highlighted in yellow.

4. D–G

   Experimental validation of the PPIs predicted on the basis of the high‐confidence CN shown in (C). CD4⁺ T cells were left unstimulated (−) or stimulated (+) with anti‐CD3 and anti‐CD4 antibodies and subsequently lysed. Equal amounts of cell lysates were incubated with isotype control (Iso Ctrl) or the specified antibodies and the resulting immunoprecipitates (IP) analyzed by immunoblot with the antibodies specified in the left margin. The validated PPIs predicted in (C) correspond to CSK‐CBLB and CSK‐CD5 (D), CRKL‐PI3Kp85β and CSK‐PI3Kp85β (E), PI3Kp85β‐PI3Kp110α and CSK‐PI3Kp110α (F), and CRKL‐PI3Kp110α (G). Molecular masses are shown (kDa). Data are representative of at least two experiments.

5. H

   Summary of the PPIs found in the high‐confidence CSK and CRKL first‐neighbors subnetwork (see C) that were validated by coimmunoprecipitation (see D–G). Interactions validated and not validated by coimmunoprecipitation are shown using red and blue edges, respectively. Interactions untractable by coimmunoprecipitation analysis due to the lack of specific antibodies are shown as gray edges.

Data information: In (C) and (H), PI3Kp85β and PI3Kp110α are denoted as PIK3R2 and PIK3CA, respectively.

Figure 6. Asymmetric functional relationships between CBL and CBLB

1. CD4⁺ T cells from wild‐type (WT), CBL^OST, and CBLB^OST mice were left unstimulated (−) or stimulated for 1 min with anti‐CD3 and anti‐CD4 antibodies (+). Equal amounts of cell lysates were subjected to affinity purification on Strep‐Tactin Sepharose beads, and the resulting protein eluates were analyzed by immunoblotting with anti‐ubiquitin (UB), anti‐CBLB, or anti‐CBL antibodies. Also shown are a loading control corresponding to equal amounts of proteins from whole‐cell lysates (WCLs) probed with anti‐CD5, and molecular masses (kDa). TCR stimulation resulted in some CBLB degradation and in the detection of a higher molecular weight CBLB form, likely due to its ubiquitylation (marked with an asterisk). In contrast, no degradation of CBL was detectable upon TCR stimulation and its molecular weight remained unchanged. Data are representative of at least two experiments.

2. Representation of the ubiquitin (UB) first‐neighbors subnetwork within the CBL and CBLB CN. Edge width is coded according to Pearson correlation coefficient (R < 0.8 or R > 0.8). Red edges depict direct interactions reported in the BioGRID database. Ubiquitin (UB) is highlighted in yellow.

3. Immunoprecipitation of CBL from cell lysates of CD4⁺ T cells isolated from WT or CBLB‐deficient mouse (Cblb ^−/−) and kept unstimulated (−) or stimulated as in (A). Immunoprecipitates and whole‐cell lysates (WCLs) were analyzed by immunoblot with antibodies specific for the proteins specified in the left margin. Molecular masses are shown (kDa). Data are representative of at least two experiments.

4. Immunoprecipitation of CBLB from cell lysates of CD4⁺ T cells isolated from WT or CBL‐deficient mouse (Cbl ^−/−) and kept unstimulated (−) or stimulated as in (A). Immunoprecipitates and whole‐cell lysates (WCLs) were analyzed by immunoblot with antibodies specific for the proteins specified in the left margin. Molecular masses are shown (kDa). Data are representative of at least three experiments.

Figure 7. CD5 regulates CBL‐mediated ubiquitylation following TCR stimulation in mature CD4⁺ T cells

1. A, B

   CD4⁺ T cells from wild‐type, Cblb ^−/−, and Cd5 ^−/− mice were stimulated for 1 min with anti‐CD3 and anti‐CD4 antibodies (+). CBL (A) and PI3KR1 (PI3Kp85α) (B) were immunoprecipitated from equal amounts of protein lysates. Level of ubiquitylation and amount of CD5 associated proteins were evaluated by immunoblot with antibodies specific for the proteins specified in the left margin. Whole‐cell lysates (WCLs) were also analyzed with the specified antibodies. Molecular masses are shown (kDa). Data are representative of at least two experiments.

Figure EV5. CD5 contributes to control the phosphorylation of the negative‐regulatory tyrosine found at position 505 of LCK via CSK

1. Expression of CD5 on short‐term expanded CD4⁺ T cells stimulated by cross‐linking biotinylated anti‐CD3 and anti‐CD4 antibodies with streptavidin for 1 min at 37°C. The depicted gates define three populations of T cells with different expression of CD5 at their surface (CD5^lo, CD5^med and CD5^hi). Numbers adjacent to outlined areas indicate percentage of cells.

2. In addition to having been stained with anti‐CD5, the cells described in (A) were permeabilized and stained with an anti‐pLCK(Y505). The histogram represents the levels of phospho‐LCK(Y505) found in the three populations defined in (A) on the basis of CD5 levels.

3. The upper panel represents the mean (± SD) of the log fluorescence intensity of phospho‐LCK(Y505) as a function of the geometric mean fluorescence intensity of CD5. Mean and SD were computed from populations of cells defined using a regular binning of the CD5 expression histogram (lower panel).

4. Effect of CD5 cross‐linking on the phosphorylation of Y505 of LCK. Short‐term expanded CD4⁺ T cells were stimulated with 2 μg biotinylated anti‐CD3 plus 2 μg biotinylated anti‐CD4 (as in A) in the presence or absence of 2 μg biotinylated anti‐CD5 (clone 53‐7.3). The intensity of phospho‐LCK(Y505) is represented as percent of phospho‐LCK(Y505) intensity in the unstimulated condition.