Research ArticleCAR T CELLS

Phosphoproteomic analysis of chimeric antigen receptor signaling reveals kinetic and quantitative differences that affect cell function

See allHide authors and affiliations

Science Signaling  21 Aug 2018:
Vol. 11, Issue 544, eaat6753
DOI: 10.1126/scisignal.aat6753

Strength limits potency

T cells engineered to express chimeric antigen receptors (CARs) may represent a curative treatment for B cell malignancies. To understand how to optimize these therapies, Salter et al. used phosphoproteomics to compare the activities of CARs encoding CD28 or 4-1BB costimulatory domains in primary human T cells. Although CARs with CD28 domains provoked more robust signaling than did CARs with 4-1BB domains (due to constitutive Lck association), both CARs activated overlapping T cell signaling pathways. However, CARs that initiated stronger signals also exhibited increased T cell dysfunction, which reduced their potency in a mouse model of lymphoma. These data indicate that the CAR costimulatory domain does not predict the signaling pathways activated in CAR T cells but instead suggest that reducing the strength of signaling may counterintuitively enhance CAR T cell therapeutic efficacy.

Abstract

Chimeric antigen receptors (CARs) link an antigen recognition domain to intracellular signaling domains to redirect T cell specificity and function. T cells expressing CARs with CD28/CD3ζ or 4-1BB/CD3ζ signaling domains are effective at treating refractory B cell malignancies but exhibit differences in effector function, clinical efficacy, and toxicity that are assumed to result from the activation of divergent signaling cascades. We analyzed stimulation-induced phosphorylation events in primary human CD8+ CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells by mass spectrometry and found that both CAR constructs activated similar signaling intermediates. Stimulation of CD28/CD3ζ CARs activated faster and larger-magnitude changes in protein phosphorylation, which correlated with an effector T cell–like phenotype and function. In contrast, 4-1BB/CD3ζ CAR T cells preferentially expressed T cell memory–associated genes and exhibited sustained antitumor activity against established tumors in vivo. Mutagenesis of the CAR CD28 signaling domain demonstrated that the increased CD28/CD3ζ CAR signal intensity was partly related to constitutive association of Lck with this domain in CAR complexes. Our data show that CAR signaling pathways cannot be predicted solely by the domains used to construct the receptor and that signal strength is a key determinant of T cell fate. Thus, tailoring CAR design based on signal strength may lead to improved clinical efficacy and reduced toxicity.

INTRODUCTION

Synthetic receptors that mimic natural T cell signaling cascades are being developed as immunotherapeutic reagents for cancer, autoimmunity, and infections. Chimeric antigen receptors (CARs), chimeric costimulatory receptors, and engineered T cell receptors (TCRs) can be introduced into T cells by gene transfer to redirect specificity and promote signaling pathways that initiate effector T cell functions (1). The most successful of these novel therapeutics to date are CARs, which are composed of an extracellular antigen-specific single-chain variable immunoglobulin fragment (scFv) fused to intracellular signaling domains that activate T cells upon ligand binding (2). Although treatment with CAR T cells can result in the complete remission of relapsed or refractory B cell malignancies, it can also cause life-threatening toxicities including cytokine release syndrome (CRS) and neurotoxicity (312). Both efficacy and toxicity result from activation of intracellular signaling pathways mediated by CAR engagement.

T cell activation occurs after TCR engagement with antigen-specific peptide bound within the major histocompatibility complex (MHC). TCR binding stimulates intracellular phosphorylation of immunoreceptor tyrosine-based activating motifs (ITAMs) on the CD3δ, ε, γ, and ζ chains of the TCR (13). Combined with protein phosphorylation signals delivered in trans from costimulatory molecules and cytokines, these events alter T cell transcriptional programs, induce proliferation, promote cytotoxic functions, and stimulate cytokine release. Most CARs use a simplified format to recapitulate the signals necessary for T cell effector function and proliferation. CAR constructs combine the CD3ζ endodomain in a single chain with a costimulatory domain from CD28 or 4-1BB. Both CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells are effective for treating patients with B cell malignancies but may behave differently in vivo. CD28/CD3ζ CAR T cells generally undergo intense proliferation within 7 days of adoptive transfer and seldom persist more than 60 days after infusion (7, 14). In contrast, 4-1BB/CD3ζ CAR T cells reach peak number by 7 to 14 days after transfer and can persist for several months (8, 15). However, variability in patient cohorts, CAR structures, and clinical trial designs impede robust comparisons across trials. Studies comparing CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells in vitro and in animal models partially explain these clinical differences by demonstrating that 4-1BB/CD3ζ CAR T cells have greater mitochondrial mass and a more memory T cell–like surface phenotype, and better retain effector functions in settings of chronic antigen stimulation (16, 17).

It is widely assumed that the differences in phenotype and function of CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells are conferred by activation of divergent signaling pathways through the distinct CAR costimulatory molecule domains. Analyses of signaling pathways stimulated by 4-1BB/CD3ζ or third-generation CD28/4-1BB/CD3ζ CARs performed using phospho–flow cytometry, Western blot, or blot array do not identify differences between CD28/CD3ζ and 4-1BB/CD3ζ CAR signaling modules and only profile a small number of known signaling events to which there are experimentally validated antibodies (18, 19). A more comprehensive, unbiased and quantitative approach to examine CAR signaling would be to use liquid chromatography–tandem mass spectrometry (LC-MS/MS). When applied to study TCR signaling, LC-MS/MS uncovered hundreds of novel phosphoprotein signaling events that were missed by antibody-based techniques (2024).

Here, we used LC-MS/MS to analyze CAR stimulation–induced signaling events in primary human CD8+ T cells that express clinically relevant CD28/CD3ζ or 4-1BB/CD3ζ CARs specific for CD19 or ROR1. We found that stimulation through CD28/CD3ζ and 4-1BB/CD3ζ CARs produced nearly identical protein phosphorylation events. Instead, stimulation of CD28/CD3ζ CAR T cells prompted more rapid and intense phosphorylation of signaling intermediates and a more effector cell–like phenotype than stimulation of 4-1BB/CD3ζ CAR T cells. CD28/CD3ζ CAR T cells were less potent at eradicating disseminated lymphoma in a xenograft mouse model than 4-1BB/CD3ζ CAR T cells. Increased basal phosphorylation of the CAR CD3ζ chain and CAR-associated Lck contributed to the rapid kinetics and stronger signal strength of CD28/CD3ζ CARs. Thus, the major distinction between CD28/CD3ζ and 4-1BB/CD3ζ CARs related not to divergent phosphoprotein signaling pathways but rather signaling strength and kinetics, which, in turn, affect T cell function and fate. These results may inform the design of new therapeutic receptors.

RESULTS

CAR designs containing a Strep-tag II sequence enable selective activation of CAR signaling in primary T cells

TCR signaling has been studied using LC-MS/MS analysis of transformed Jurkat T cells stimulated with anti-CD3 monoclonal antibodies (mAbs) (20, 21, 24). Jurkat cells were selected for signaling studies because of the ease at which they can be grown and manipulated with common molecular biology techniques. Primary T cells from human subjects are more appropriate for studying CAR signaling, and we previously developed a method to activate CAR signaling in T cells without the need for ligand-expressing stimulator cells (25). We modified two lentiviral vectors encoding CD19- and ROR1-specific 4-1BB/CD3ζ CARs that are currently being tested in clinical trials by adding a nine–amino acid Strep-tag II (STII) sequence to the extracellular CAR hinge. For comparison, we cloned structurally identical CD19- and ROR1-specific CD28/CD3ζ CARs containing the STII tag (Fig. 1A). All constructs contained a truncated epidermal growth factor receptor (EGFRt) marker downstream of a T2A ribosomal skip element for purification of CAR-expressing T cells (26). Inclusion of the STII sequence does not interfere with CAR T cell recognition or function, and STII CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells are efficiently activated and expanded in vitro by simulation with STII microbeads (Fig. 1B) (25). Primary CD8+ T cells were transduced with each lentiviral vector, sorted for EGFRt expression, and expanded to >1.6 × 108 cells with a single cycle of stimulation before subsequent analysis (Fig. 1C). The abundance of CD28/CD3ζ and 4-1BB/CD3ζ CARs of each scFv specificity was similar on the cell surface as measured by staining with STII mAb (Fig. 1D). CAR T cells also expressed similar amounts of CD45RO, CD62L, CD27, and CD28 (Fig. 1E), indicating that CAR T cells retained markers associated with memory and proliferative potential. We only detected small frequencies of PD-1– or Tim-3–positive CAR T cells, suggesting that the cells were not activated or exhausted after cell culture (Fig. 1E). Accordingly, >85% of CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells were in the G0-G1 cell cycle phase, indicating that the cells were resting (Fig. 1F).

Fig. 1 Both CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells can be activated through an engineered STII hinge.

(A) Schematic of CARs incorporating an STII sequence in the extracellular hinge. CARs contained either the CD19-specific FMC63 scFv or the ROR1-specific R12 scFv. (B) Schematic of CAR T cell activation through the STII hinge using magnetic beads coated with antibody against STII. (C) Flow cytometry analysis of CD8 and EGFRt staining on singlet CD19 CAR T cells after expansion. Dot plots are representative of three independent experiments. The frequency values of positive cells are means from all experiments. (D and E) Flow cytometry analysis of STII staining of cell surface CAR (D) or CD45RO, CD62L, CD27, CD28, PD-1, and Tim-3 phenotypic marker staining (E) on sort-purified CD19- or ROR1-specific singlet CD8+ CAR T cells after expansion. Histogram plots of CD28/CD3ζ CAR T cells (red), 4-1BB/CD3ζ CAR T cells (blue), or isotype control staining (gray) are representative of four independent experiments. The frequency values of positive cells are means from all experiments. (F) Flow cytometry analysis of the DNA content of CAR T cells after expansion. Histograms are representative of four independent experiments. The frequency values of cells in G0-G1 gate are means from all experiments. (G) Western blot analysis for CD3ζ, CD3ζ pTyr142, SLP-76, and SLP-76 pSer376 in lysates of ROR1 4-1BB/CD3ζ CAR T cells after 45 min of coculture with varying quantities of STII microbeads, K562 cells, or K562/ROR1 cells. Blots and fold change (log2FC) of normalized band intensity values are representative of two independent experiments. The indicated P values were calculated by paired two-tailed t test (E).

We evaluated canonical T cell signaling events induced by STII ligation and found that cell-free STII microbead stimulation of CARs was similar to CAR antigen–expressing tumor cells. ROR1-specific 4-1BB/CD3ζ CAR T cells were incubated for 45 min with increasing amounts of STII microbeads or with ROR1-transduced K562 (K562/ROR1) tumor cells, and CD3ζ Tyr142 and SLP-76 Ser376 were measured by Western blot. At the highest bead to CAR T cell ratio, the phosphorylation of CD3ζ and SLP-76 was grossly similar to that observed in lysates from CAR T cells stimulated with K562/ROR1 cells (Fig. 1G and fig. S1A). This bead-to-cell ratio was used for all subsequent experiments. Thus, STII microbead stimulation provided a precise method to selectively activate CAR signaling in primary T cells.

MS identifies common protein phosphorylation events after stimulation of CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells

We performed LC-MS/MS analysis to interrogate the signaling pathways activated in CD28/CD3ζ or 4-1BB/CD3ζ CARs in an unbiased manner. Human T cells expressing CD28/CD3ζ or 4-1BB/CD3ζ CARs were incubated with STII or uncoated (control) microbeads for 10 or 45 min (Fig. 2A). CD19-specific CAR T cells generated from two different donors were used in two independent experiments, and a third independent experiment used ROR1-specific CAR T cells derived from one of the two donors (Fig. 2B). Given that ROR1- and CD19-specific CARs displayed similar phenotypes across the two donors and were stimulated identically in a ligand-independent manner, the corresponding measurements were considered biological replicates. To provide relative quantitation of phosphopeptides within each experiment, we labeled each trypsin-digested lysate with a unique isobaric tandem mass tag (TMT) and globally enriched for phosphopeptides by both phosphorylated tyrosine (pTyr) immunoprecipitation and immobilized metal affinity chromatography (IMAC) (27) (fig. S1B). We identified a total of 26,804 phosphorylation sites across the three experiments corresponding to 4849 proteins. Among phosphorylation sites, 571 (2.13%) were pTyr, 4647 (17.33%) were phosphorylated threonine (pThr), and 21,586 (80.53%) were phosphorylated serine (pSer). (Fig. 2C). Considering the stochastic limitations inherent in data-dependent acquisition shotgun proteomics (28), we found considerable overlap in the captured phosphoproteome between replicate experiments (Fig. 2D). As previously described in murine T cells, 99% of phosphorylation sites detected in each experiment were present in both unstimulated and stimulated T cell lysates, enabling quantitation of changes induced by CAR activation (22, 29).

Fig. 2 CAR T cells signal through endogenous T cell signaling proteins.

(A and B) Human CAR T cell treatment conditions and experimental groups. (C to E) MS/MS analysis of phosphorylated peptides from lysates of CAR T cells stimulated as in (A). The total number of pSer, pThr, and pTyr peptides identified (C) and the Venn diagram of the overlap among phosphorylation sites (D) are pooled from three independent experiments. The fold change (log2FC) data in phosphorylation at the indicated times at sites involved in canonical TCR signaling (E) are means ± range from two or three independent experiments. (F) Western blot analysis for CD3ζ, CD3ζ pTyr142, ZAP-70 pTyr319, and PLC-γ1 pTyr783 in lysates from CD19-specific CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells at the indicated times after stimulation. Blots are representative of three independent experiments. Fold change (log2FC) data of normalized band intensity are means ± SD from all experiments. The indicated P values were calculated by repeated-measures one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test comparing CD28/CD3ζ and 4-1BB/CD3ζ CAR samples (F).

We analyzed well-described TCR phosphorylation events within the MS data set to identify sites that were modified by CAR engagement (13). We calculated the log2 of the fold-change (log2FC) value for all phosphorylation sites within each experiment by comparing stimulated samples to appropriate controls (that is, CD28/CD3ζ CAR 10-min stimulation versus CD28/CD3ζ CAR 10-min control). At 10 min, phosphorylation of CD3ζ at Tyr83, Tyr111, and Tyr142, was increased in stimulated CAR T cells (Fig. 2E). All CAR T cells displayed increased phosphorylation of CD28 at Tyr209 after 10 min, which demonstrated that 4-1BB/CD3ζ CAR activation stimulates phosphorylation of endogenous CD28. After 45 min of stimulation, phosphoprotein signaling spread to downstream TCR nodes, and we measured increased phosphorylation of phospholipase C–γ1 (PLC-γ1) at Ser1248 and BCL10 at Ser138 (Fig. 2E) (30, 31). We did not observe reproducible changes at either time point in the phosphorylation of CD3δ, ε, or γ ITAMs. Because these sites are phosphorylated after TCR stimulation (13), our data indicated that CAR signaling only partially mimics endogenous TCR activation.

The MS data suggested that CD28/CD3ζ CAR stimulation may initiate a greater magnitude log2FC than 4-1BB/CD3ζ CARs at both 10 and 45 min, and we confirmed this finding by Western blot. Evaluation of cell lysates for CD3ζ pTyr142, ZAP-70 pTyr319, and PLC-γ1 pTyr783 demonstrated that both CD28/CD3ζ and 4-1BB/CD3ζ CAR stimulation increased phosphorylation of these sites, but we observed more intense ZAP-70 and CAR CD3ζ phosphorylation in stimulated CD28/CD3ζ CAR T cells compared to 4-1BB/CD3ζ CAR T cells (Fig. 2F). We also detected a low level of basal CAR CD3ζ phosphorylation in unstimulated CD28/CD3ζ CD19 and ROR1 CAR T cells that was not present in 4-1BB/CD3ζ CAR T cells. Western blot analysis of lysates from primary CD4+ T cells transduced with CD19 CARs and stimulated with STII microbeads also demonstrated similar patterns of basal CAR CD3ζ phosphorylation as well as more rapid and robust phosphorylation of CD3ζ and SLP-76 within CD28/CD3ζ CAR T cells (fig. S2A). Constitutive phosphorylation of the CAR CD3ζ domain or tonic signaling has been shown to occur with some CARs, including a CD19-specific CD28/CD3ζ CAR that is being used in the clinic (32). Strong tonic signaling has been linked to sequences in the scFv that promote clustering of CAR molecules at the cell surface and results in the up-regulation of T cell exhaustion markers (33, 34). Because we did not observe differences in PD-1 or Tim-3 expression, or evidence of CAR clustering in unstimulated CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells (Fig. 1, E and F, and fig. S2B), the low level of basal CAR CD3ζ domain phosphorylation detected here may be distinct from more extreme tonic signaling observed in some CARs with different scFv specificities.

Phosphorylation events mediated by CD28/CD3ζ or 4-1BB/CD3ζ CARs differ in kinetics and magnitude

An advantage of shotgun MS is that it can quantitatively and temporally measure thousands of phosphorylation events outside of the canonical TCR signaling pathway to which there are no experimentally validated antibodies. We leveraged the limma statistical framework and associated R package to identify phosphorylation sites that were increased or decreased in abundance after CD28/CD3ζ and 4-1BB/CD3ζ CAR ligation (35). We assigned a phosphorylation site to be CAR stimulation–responsive if it was detected in at least two of the three experiments, displayed an average |log2FC| ≥ 0.7 between stimulated and unstimulated conditions at 10 or 45 min, and met a 5% false discovery rate (FDR) cutoff. A log2FC cutoff of 0.7 was chosen because this represents approximately 2 SDs of the distribution of log2FC values (fig. S3).

Using these cutoffs, 26 phosphorylation sites were identified as stimulation-responsive at 10 min after activation of CD28/CD3ζ CAR T cells. These sites were enriched for proteins in the Kyoto Encyclopedia of Genes and Genomes (KEGG) TCR signaling pathway including increased phosphorylation of p21-activated kinase 2 (PAK2) at Ser64, CD8α (CD8A) at Ser231, protein kinase C θ (PKCT) at Ser370, and proto-oncogene vav (VAV1) at Ser748 and Thr749 (Fig. 3A and table S1). We also detected increased phosphorylation of neuroblast differentiation-associated protein (AHNAK) at Ser5857, which activates PLC-γ1 and is required for T cell calcium (Ca2+) mobilization and effector functions (36, 37). In contrast, no sites met log2FC and FDR criteria at 10 min after stimulation of 4-1BB/CD3ζ CAR T cells (Fig. 3A and table S1). The lack of robust alterations in early protein phosphorylation in 4-1BB/CD3ζ CAR T cells was consistent with Western blot data showing a very low level of phosphorylation of CAR CD3ζ, ZAP-70, and PLC-γ1 at 10 min after stimulation (Fig. 2F).

Fig. 3 The kinetics and strength of signaling vary after stimulation of CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells.

(A) Volcano plots of fold change (log2FC) and FDR for phosphorylation sites identified by MS/MS in Fig. 2. Green dots indicate sites with increased phosphorylation, and red dots indicate sites with decreased phosphorylation after CAR stimulation in at least two experiments. (B) Comparison of stimulation-responsive phosphorylation sites identified by MS/MS at 45 min after activation in either CD28/CD3ζ or 4-1BB/CD3ζ CAR samples. Green dots specify sites that exhibited opposite responses after CD28/CD3ζ or 4-1BB/CD3ζ CAR activation, whereas red dots indicate sites phosphorylated to a greater extent after stimulation of 4-1BB/CD3ζ CAR T cells in at least two MS/MS experiments in Fig. 2. (C) Fold change in phosphorylation sites on known CD28 and 4-1BB signaling pathway members at 45 min after CAR T cell stimulation. Data are means ± range from two or three MS/MS experiments in Fig. 2. (D) Fold change in the 20 most phosphorylated sites identified by MS/MS in Fig. 2. Data are means from all experiments. (E) Absolute fold change of phosphorylation sites on known KEGG TCR signaling pathway proteins identified by MS/MS in Fig. 2. Data are means from all experiments. (F) Western blot analysis for CD3ζ, DAPP1, DAPP1 pTyr139, SLP-76, SLP-76 pSer376, PLC-γ1, and PLC-γ1 pTyr783 in lysates from ROR1 CAR T cells stimulated with STII microbeads for the indicated times. Blots are representative of three independent experiments. Fold change data of normalized band intensity are means ± SD from all experiments. The indicated P values were calculated by unpaired two-tailed t test (D and E).

After 45 min of stimulation, more intense changes in protein phosphorylation had occurred, and we detected phosphorylation at 1289 sites that met the log2FC and FDR cutoffs in either CD28/CD3ζ or 4-1BB/CD3ζ CAR T cell samples. These included 1279 phosphorylation sites from 743 gene products in CD28/CD3ζ CAR T cells and 522 sites from 346 gene products in 4-1BB/CD3ζ CAR T cells (Fig. 3A and tables S2 and S3). These data indicated that stimulation of CD28/CD3ζ CAR T cells increased phosphorylation at a greater number of sites than did stimulation of 4-1BB/CD3ζ CAR T cells. However, both CD28/CD3ζ and 4-1BB/CD3ζ CAR activation stimulated highly similar changes in the phosphorylation of activation-responsive sites (Fig. 3B). Only 12 (0.93%) of the 1289 phosphorylation sites that met the cutoffs in CD28/CD3ζ CAR samples exhibited an opposite response after 4-1BB/CD3ζ CAR stimulation, and only 43 (3.3%) sites exhibited a greater magnitude log2FC after 4-1BB/CD3ζ CAR stimulation. These observations were consistent with our earlier data suggesting that signaling downstream of CD28/CD3ζ CAR activation was more intense than downstream of 4-1BB/CD3ζ CAR activation. Furthermore, these subsets of 12 and 43 phosphorylation sites did not map to currently defined 4-1BB signaling networks (38).

Finding relatively few differences in the phosphorylation events stimulated after activation of CD28/CD3ζ and 4-1BB/CD3ζ CARs was unexpected. When we questioned how proteins involved in the canonical T cell costimulatory signaling pathways were affected, we found that the phosphorylation of CD28 signaling intermediates VAV1, PIK3C2A, and PKCT was also increased after stimulation of both CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells (Fig. 3C) (39). This result appeared to agree with our earlier observations that endogenous CD28 was phosphorylated after activation of 4-1BB/CD3ζ CAR T cells (Fig. 2E). Within the 4-1BB signaling pathway, phosphorylation of lymphocyte-specific protein 1 (LSP1), a direct target of 4-1BB and TRAF2 signaling, was altered by stimulation of both CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells (40). Thus, rather than activating divergent costimulatory pathways, as might be predicted from the distinct costimulatory domains encoded by the receptors, activation of CD28/CD3ζ and 4-1BB/CD3ζ CARs induced highly similar changes in intracellular protein phosphorylation that encompassed both canonical CD28 and 4-1BB signaling intermediates.

A map illustrating the major pathways and individual protein phosphorylation events affected by CD28/CD3ζ and 4-1BB/CD3ζ CAR stimulation includes proteins involved in canonical TCR signaling and mitogen-activated protein kinase (MAPK) signaling pathways (Fig. 4 and table S3). In addition, stimulation of either CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells affected phosphorylation of actin-cytoskeletal regulatory proteins, RNA processing proteins, nuclear pore proteins, and mitochondrial fission regulators such as DRP1 (41, 42). A list of CAR stimulation–responsive phosphorylation sites is provided in tables S1 and S2. Many of these proteins have not been previously associated with TCR or CAR signaling.

Fig. 4 Stimulation of CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells alters protein phosphorylation across similar signaling pathways and cellular compartments.

Map of select proteins differentially phosphorylated after 45 min of CAR T cell stimulation from analysis of all MS/MS experiments in Fig. 2.

Differences in the magnitude of CD28/CD3ζ and 4-1BB/CD3ζ CAR signaling persist across time

Phosphorylation events activated by CD28/CD3ζ and 4-1BB/CD3ζ CAR T cell stimulation were highly similar but differed in intensity at the vast majority of sites. Signaling strength during T cell activation, measured as a composite of TCR affinity, costimulation, and cytokine signals, dictates clonal expansion and the effector capacities of a T cell response (4345). To provide a holistic measure of CAR signal strength, we sorted stimulation-responsive phosphorylation sites at 45 min from CD28/CD3ζ and 4-1BB/CD3ζ CAR samples by decreasing log2FC. In line with results showing that stimulation of both CARs altered similar phosphorylation sites, 15 of the 20 most phosphorylated sites after stimulation were shared between CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells (table S4). However, phosphorylation of the top 20 sites increased from control samples by 11.15-fold on average in CD28/CD3ζ CAR samples but only by 5.8-fold on average in 4-1BB/CD3ζ CAR samples (Fig. 3D). Stratifying CAR stimulation–responsive phosphorylation sites by signaling pathway further indicated that the average phosphorylation site within the KEGG TCR signaling pathway was modulated by 2.52-fold in CD28/CD3ζ CAR samples but only by 1.69-fold in 4-1BB/CD3ζ CAR samples (Fig. 3E). To determine whether 4-1BB/CD3ζ CAR signals reached a similar intensity to those of CD28/CD3ζ CARs at later times, we stimulated identically prepared CD8+ CAR T cells for 60, 120, or 180 min and measured phosphorylation of canonical and newly identified signaling intermediates (table S4). Phosphorylation of PLC-γ1 Tyr783 and DAPP1 Tyr139 was apparent after stimulation of CD28/CD3ζ but not 4-1BB/CD3ζ CAR T cells (Fig. 3F). These data suggest that 4-1BB/CD3ζ CAR stimulation may not achieve the same signal intensity as CD28/CD3ζ CAR T cells during this time frame.

Increased CAR signal intensity is associated with an effector cell–like phenotype and reduced in vivo antitumor activity

The strength of T cell activation and signal transduction influences transcriptional programs that regulate effector cell differentiation and memory formation (44). We used RNA sequencing (RNA-seq) to analyze changes in transcription within CD28/CD3ζ or 4-1BB/CD3ζ CD8+ CAR T cells at 6 hours after STII microbead stimulation. Consistent with the faster and more intense phosphoprotein signal, CD28/CD3ζ CAR stimulation initiated more marked transcriptional changes. A comparison of stimulated to unstimulated CD28/CD3ζ CAR T cells identified 4894 differentially expressed genes at 6 hours, whereas 4-1BB/CD3ζ CAR stimulation resulted in 197 differentially expressed genes. CD28/CD3ζ or 4-1BB/CD3ζ CAR stimulation increased expression of the canonical T cell activation marker CD69 to a similar degree (Fig. 5A), but greater fold increases in expression of the effector molecules granzyme B (GZMB), interferon-γ (IFNG), interleukin-2 (IL2), tumor necrosis factor–α (TNF), macrophage inflammatory protein 1α (CCL3), and macrophage inflammatory protein 1β (CCL4) were observed in CD28/CD3ζ CAR T cells (Fig. 5B). Direct comparison of CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells identified 1673 differentially expressed genes after stimulation (table S5). Of these, Krüppel-like factor 2 (KLF2), IL-7 receptor (IL7R), and Rho family–interacting cell polarization regulator 2 (RIPOR2; previously known as FAM65B) expression was decreased in CD28/CD3ζ CAR T cells when compared to 4-1BB/CD3ζ CAR T cells (Fig. 5C). KLF2 and IL7R are associated with memory T cell formation and are targets of the FOXO family of transcription factors (4648). Consistent with this observation, FOXO4 expression was reduced in stimulated CD28/CD3ζ CAR T cells when compared to 4-1BB/CD3ζ CAR T cells. Quantitative polymerase chain reaction (qPCR) confirmed that some of these T cell memory–associated genes were only differentially expressed in stimulated and not unstimulated CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells (Fig. 5D).

Fig. 5 Increased CD28/CD3ζ CAR signal intensity is associated with an effector cell–like phenotype and reduced in vivo antitumor activity.

(A to C) RNA-seq analysis of total RNA expression in CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells with and without stimulation. The fold change values of the indicated transcripts (A and B) are means ± SD from three independent experiments. Transcripts in (B) met an FDR of 1% for differential expression between CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells. Volcano plot analysis (C) indicates genes with increased expression in CD28/CD3ζ CAR T cells (green) or increased expression in 4-1BB/CD3ζ CAR T cells (red). (D) Quantitative PCR analysis of IL7R, KLF2, and FOXO4 expression in CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells. Fold change data are means ± SD from three biological replicates. (E) Enzyme-linked immunosorbent assay (ELISA) analysis of cytokine production 24 hours after coculture of ROR1-specific CAR T cells with K562/ROR1 cells. Data are means ± SD of three to four independent experiments. (F) Flow cytometry analysis of T cell proliferation as measured by carboxyfluorescein diacetate succinimidyl ester (CFSE) dye dilution at 72 hours after STII microbead stimulation. Histogram plot of unstimulated CD28/CD3ζ CAR T cells (gray), unstimulated 4-1BB/CD3ζ CAR T cells (black), stimulated CD28/CD3ζ CAR T cells (red), and stimulated 4-1BB/CD3ζ CAR T cells (blue). The proliferation index values of cells are means from five independent experiments (P = 0.0747). (G to J) At 7 days after Raji/ffluc engraftment, NSG mice were treated with a single infusion of the indicated dose of CAR T cells. Survival analyses (G) of 6, 9, or 15 mice per group are pooled from two to three independent experiments. Bioluminescence images of Raji/ffluc tumor burden in mice at the indicated time points (H) are representative of all experiments. PBS, phosphate-buffered saline. (I and J) Flow cytometry analysis of CAR T cell frequency in bone marrow or peripheral blood (I) or abundance of PD-1, Lag-3, or Tim-3 on CAR T cells in the bone marrow (J) on day 20. Frequency and mean fluorescence intensity (MFI) data are means ± SD of 10 mice per group from two independent experiments. The indicated P values were calculated by one-sample t test with null hypothesis H0 = 0 (D), paired two-tailed t test (E and F), log-rank (Mantel-Cox) test (G), or unpaired two-tailed t test (I and J).

Because differences in TCR signal quantity affect T cell functions (44, 45), we measured CD28/CD3ζ and 4-1BB/CD3ζ CAR T cell effector functions in vitro. We activated CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells with ROR1- and CD19-expressing K562 cells or STII microbeads and measured cytokine production and proliferation. After 24 hours of coculture, CD28/CD3ζ CAR T cells secreted markedly more IFN-γ, IL-2, and TNF-α than 4-1BB/CD3ζ CAR T cells (Fig. 5E). By 72 hours, both CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells proliferated (Fig. 5F). These data suggested that stronger signaling by CD28/CD3ζ CAR T cells correlated with increased short-term effector responses.

We also investigated the function of CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells in vivo. For these studies, we used a defined 1:1 CD4/CD8 ratio CAR T cell product that displays superior control of xenograft Raji lymphoma in NOD/SCID/γc−/− (NSG) mice (49). Adoptive transfer of 3 × 106 CD28/CD3ζ or 4-1BB/CD3ζ CAR T cells into tumor-bearing mice mediated complete tumor regression (Fig. 5G). However, when fewer (7.5–8 × 105) cells were transferred, CD28/CD3ζ CAR T cells were much less potent at eliminating Raji lymphoma cells than 4-1BB/CD3ζ CAR T cells, and all CD28/CD3ζ CAR T cell–treated mice died of tumor progression within 40 days (Fig. 5, G and H). Tumor progression occurred in mice treated at the lower CD28/CD3ζ CAR T cell dose despite accumulation of CAR T cells to higher frequencies in tumor-involved bone marrow (Fig. 5I). CD28/CD3ζ CAR T cells in the bone marrow expressed higher levels of PD-1, Lag-3, and Tim-3 (Fig. 5J), consistent with the acquisition of an exhausted phenotype. In summary, the rapid and intense signaling mediated by CD28/CD3ζ CAR activation correlated with an exhausted phenotype in CD8+ and CD4+ CAR T cells and reduced antitumor activity.

CD28/CD3ζ and 4-1BB/CD3ζ CARs differentially associate with endogenous CD28 and Lck

To interrogate possible causes of increased CD28/CD3ζ CAR signaling kinetics and strength, we immunoprecipitated CAR complexes from unstimulated CD8+ T cells and probed for differences among associated T cell signaling proteins in the basal state. Western blot confirmed efficient CAR pull-down and showed association of endogenous CD28 and Lck with the CD28/CD3ζ CAR but only minimal CD28 and Lck association with the 4-1BB/CD3ζ CAR (Fig. 6A). Because basal CAR phosphorylation was conferred by the presence of the CD28 costimulatory domain and Tyr206, Tyr209, and Tyr218 were intensely phosphorylated after CAR stimulation, we constructed CD28/CD3ζ CARs with tyrosine to phenylalanine mutations at these residues (Fig. 6B). CD19 and ROR1 CARs with mutations of Y218F (Y1) or all three tyrosines (Y3) were efficiently expressed in T cells and led to vigorous IFN-γ production and proliferation in response to coculture with ROR1+ or CD19+ tumor cells (Fig. 6, C and D). We observed partial (Y1) or complete (Y3) abrogation of basal CAR CD3ζ phosphorylation in CARs containing Y to F mutations as compared to CD28/CD3ζ CARs; however, Y1 and Y3 CARs still phosphorylated SLP-76 and PLC-γ1 with similar kinetics and intensity as the wild-type CD28/CD3ζ CAR after STII microbead stimulation (Fig. 6E). Immunoprecipitation of each CAR indicated that Lck association was not abrogated by the Y1 and Y3 mutations, although endogenous CD28 did not associate with the Y3 CAR (Fig. 6F). Together, these results suggested that neither basal CAR CD3ζ phosphorylation nor endogenous CD28 association was responsible for the increased signal kinetics and strength of CD28/CD3ζ CARs.

Fig. 6 CD28/CD3ζ and 4-1BB/CD3ζ CARs differentially associate with endogenous Lck and CD28.

(A) Western blot analysis for Lck, CD28, and CD3ζ in whole-cell lysates (L) and STII immunoprecipitated fractions (IP) from unstimulated ROR1-specific CAR T cells. Blots are representative of three to four independent experiments. (B) Schematic of mutations made to the CAR CD28 signaling domain. (C) Flow cytometry analysis of ROR1-specific CAR T cell proliferation as measured by CFSE dye dilution at 72 hours after coculture with K562/ROR1 cells. Histograms of untransduced T cells (gray) or 4-1BB/CD3ζ (blue), CD28/CD3ζ (red), Y1 (green), and Y3 CAR T cells (purple) after stimulation are representative of three independent experiments. (D) ELISA analysis of IFN-γ production by ROR1-specific 4-1BB/CD3ζ (blue), CD28/CD3ζ (red), Y1 (green), or Y3 CAR T cells (purple) after coculture with K562/ROR1 cells for 24 hours. Fold change data are means ± SD of three independent experiments. (E) Western blot analysis for CD3ζ, CD3ζ pTyr142, SLP-76, SLP-76 pSer376, PLC-γ1, and PLC-γ1 pTyr783 in lysates from ROR1-specific CAR T cells stimulated for the indicated times with STII microbeads. Blots are representative of three independent experiments. Fold change data of normalized band intensity are means ± SD from all experiments. (F) Western blot analysis for Lck, CD28, and CD3ζ in whole-cell lysates (L) and STII immunoprecipitated fractions (IP) from resting CAR T cells. Blots are representative of at least three independent experiments. Fold change data of normalized band intensity are means ± SD from three (CD28) or four (Lck) independent experiments. The indicated P values were calculated by repeated-measures one-way ANOVA with Tukey’s multiple comparisons test comparing samples at equivalent time points (E and F).

Constitutive Lck activation promotes T cell effector functions (50, 51). Therefore, we tested whether association of Lck with CD28/CD3ζ CARs mediated rapid and robust phosphorylation of signaling intermediates after CAR activation. We generated proline to alanine mutations at the Lck binding site of CD28 (Fig. 7A) (52, 53). Immunoprecipitation of the wild-type and mutant CD28/CD3ζ CARs showed that Lck association was absent in CARs with proline to alanine mutations (Fig. 7 and fig. S4). Signaling analyses of these constructs demonstrated that mutation of proline residues alone in CD28P CARs partially abrogated basal CAR CD3ζ phosphorylation but did not reduce signal intensity (Fig. 7C). However, simultaneous mutation of proline and tyrosine residues in Y3P CARs fully abrogated basal CAR phosphorylation and reduced the magnitude of SLP-76 and PLC-γ1 phosphorylation after stimulation (Fig. 7C). Thus, differences in CAR signal intensity between CD28/CD3ζ CARs and 4-1BB/CD3ζ CARs were, in part, related to greater Lck association with CD28/CD3ζ CARs. These data suggest that CAR signal intensity could be altered by mutating tyrosine and proline residues in the CD28 signaling domain.

Fig. 7 Mutations that reduce Lck binding diminish CD28/CD3ζ CAR signal intensity.

(A) Schematic of mutations made to the CAR CD28 signaling domain. (B) Western blot analysis for Lck and CD3ζ within whole-cell lysates (L) and STII immunoprecipitated fractions (IP) from resting CAR T cells. Blots and fold change in the normalized band intensity values are representative of two independent experiments. (C) Western blot analysis for CD3ζ, CD3ζ pTyr142, SLP-76, SLP-76 pSer376, PLC-γ1, and PLC-γ1 pTyr783 within lysates from CAR T cells stimulated for the indicated times with STII microbeads. Blots are representative of three independent experiments. Fold change data in the normalized band intensity values are means ± SD from all experiments. The indicated P values were calculated by repeated-measures one-way ANOVA with Tukey’s multiple comparisons test comparing samples at equivalent time points (C).

DISCUSSION

The adoptive transfer of CD28/CD3ζ or 4-1BB/CD3ζ CD19 CAR T cells is an effective therapy for patients with certain B cell malignancies but can cause serious toxicities that originate from CAR signaling after tumor recognition (1). However, CD28/CD3ζ or 4-1BB/CD3ζ CAR signaling modules confer differences in effector functions and metabolic profiles, and both types of CAR T cells can cause serious toxicities that originate from CAR signaling after T cell activation (16, 17). Crafting safer and more effective CAR T cells might be accomplished by modifying CAR design, but optimization is hindered by the rudimentary understanding of how signaling by synthetic CARs directs T cell functional outputs and cell fate decisions.

We used MS to study changes in the phosphoproteome of T cells after CAR ligation and captured 26,804 unique phosphorylation sites that describe changes induced by CD28/CD3ζ and 4-1BB/CD3ζ CARs in detail. Our experiments used structurally identical CARs specific for two target antigens being pursued in clinical trials. In contrast to the current perception of how costimulatory domains function in CARs, our data show that, in the space of this data set, CD28/CD3ζ and 4-1BB/CD3ζ CAR activation initiated nearly identical directional changes in protein phosphorylation. Notably, both CARs altered phosphorylation of proteins involved in canonical TCR, CD28, and 4-1BB signaling pathways. Only 12 of 1289 (0.93%) phosphorylation sites exhibited opposite responses to stimulation when comparing CAR constructs at 45 min after activation, and these few sites are not directly linked to either CD28 or 4-1BB signaling pathways. Therefore, our data refute the notion that CD28/CD3ζ and 4-1BB/CD3ζ CARs activate divergent signaling cascades and indicate that one cannot predict the signaling cascades initiated by synthetic CARs by focusing solely on the domains encoded by the receptor polypeptide. Rather, synthetic CARs initiate a plethora of signals that both encompass and surpass those of the natural molecules on which the synthetic receptors are based.

Instead, our data demonstrate that a major difference between CD28/CD3ζ and 4-1BB/CD3ζ CARs is that CD28/CD3ζ receptors signal with markedly increased kinetics and intensity. Within 45 min, CD28/CD3ζ CAR stimulation altered the phosphorylation state of hundreds of proteins involved in TCR signaling, MAPK signaling, actin-cytoskeleton rearrangement, and RNA processing pathways. Alterations in protein phosphorylation after 4-1BB/CD3ζ CAR stimulation affected identical proteins as CD28/CD3ζ CAR stimulation but proceeded more slowly and with reduced intensity. Subsequent analysis of CAR binding partners demonstrated more Lck in immunoprecipitates from unstimulated CD28/CD3ζ CAR T cells than from 4-1BB/CD3ζ CAR T cells. Mutational analysis of the CAR CD28 signaling domain identified residues responsible for basal CAR CD3ζ phosphorylation as well as Lck association, and showed that abrogation of these features reduced CAR signal intensity. Thus, future analyses of novel CAR designs should consider how interactions with potential binding partners might affect signaling.

Our data also showed that the more rapid and intense phosphoprotein signaling in CD28/CD3ζ CAR T cells induced an effector cell–like transcriptional profile and more robust T cell cytokine production early after CAR ligation, but less potent antitumor function in an in vivo tumor model of disseminated lymphoma. These data are consistent with previous findings that intense TCR signaling and IL-2 production promote differentiation of short-lived effector CD8+ T cells (54, 55), and also support a model by which TCR and costimulatory signals sum linearly to affect cell fate (45). Some amount of CD28 signaling is beneficial for T cell memory formation (56), and we found that the 4-1BB/CD3ζ CARs used in this study phosphorylated endogenous CD28 and activated the CD28 signaling pathway. However, encoding a fully functional CD28 signaling domain on a CAR polypeptide chain may yield excessive stimulation that increases the incidence of CRS, promotes T cell exhaustion, and reduces persistence. Indeed, analyses of acute lymphoblastic leukemia patients who were in morphologic relapse at the time of treatment with CD28/CD3ζ CAR T cells showed poor long-term CAR T cell persistence and low overall survival due to frequent relapse with CD19+ disease (57).

CARs are often currently selected to maximize measures of in vitro T cell effector functions and proliferation, and our findings indicate that these selection criteria may not be optimal for predicting in vivo efficacy. Future optimizations to CAR design should consider both the signaling domains contained within the receptor, interactions with potential binding partners, and how modifications of costimulatory domains might tailor overall signal strength for downstream T cell functional outputs. Mutating the CAR CD28 signaling domain is a logical approach for reducing excessive signal strength, and further studies of mutated CAR signaling domains are warranted. However, variability in CAR structural features outside of the costimulatory CD28 or 4-1BB and CD3ζ signaling domains may limit the universality of this approach. Most CARs have a unique combination of scFv, hinge, and transmembrane domain, and this structural variation can affect CAR signaling and may alter association with endogenous T cell signaling molecules. For instance, the CAR CD28 transmembrane domain used in this study may have influenced association with endogenous CD28 and Lck. Further immunoprecipitation analysis of CARs with CD8α transmembrane domains and other structural features could answer this question (3).

Our approach evaluating CD28/CD3ζ and 4-1BB/CD3ζ CAR signaling using a cell-free stimulation method has the advantage of identifying phosphorylation events induced by the synthetic receptors in isolation. Although this approach did not integrate the multitude of other events that may occur when CAR T cells encounter tumor cells in vivo (58), it could be extended to interrogate cross-talk between CAR signaling and costimulatory or inhibitory pathways by immobilizing various ligands onto the magnetic beads. Our approach was also not designed to compare TCR and CAR signaling. A direct comparison of TCR and CAR ligation-induced phosphoprotein signaling is likely to be valuable and is a focus of ongoing research. Thus, MS analysis can provide important insights into CAR signaling that will affect next-generation CAR designs.

MATERIALS AND METHODS

Acquisition of peripheral blood T cells from healthy donors

Healthy adults (>18 years old) were enrolled in an institutional review board–approved study for peripheral blood collection. Informed consent was obtained from all enrollees. Researchers were blinded to all personally identifiable information about study participants and were provided only donor age and a nondescript donor ID number. Peripheral blood (400 cm3) was collected by venipuncture, and mononuclear cells [peripheral blood mononuclear cells (PBMCs)] were isolated by density gradient using Lymphocyte Separation Media (Corning). CD4+ and CD8+ T cells were isolated using the EasySep Human CD4+ and CD8+ T Cell Isolation Kits (STEMCELL Technologies). For the three shotgun MS experiments, CD8+CD62L+ T cells were further enriched by staining with CD62L-PE (DREG-56, Thermo Fisher Scientific) followed by the EasySep Human PE Selection Kit (STEMCELL Technologies). Isolations were performed in accordance with the manufacturer’s instructions.

Cell culture

293T Lenti-X cells (Clontech) were cultured in Dulbecco’s modified Eagle’s medium (Gibco) supplemented with 10% fetal bovine serum, 1 mM l-glutamine (Gibco), 25 mM Hepes (Gibco), and penicillin/streptomycin (100 U/ml; Gibco). K562 (CCL-243) and Raji (CCL-86) cells were obtained from the American Type Culture Collection and cultured in RPMI 1640 (Gibco) supplemented with 5% fetal bovine serum, 1 mM l-glutamine, 25 mM Hepes, and penicillin/streptomycin (100 U/ml). Primary human T cells were cultured in CTL medium consisting of RPMI 1640 supplemented with 10% human serum, 2 mM l-glutamine, 25 mM Hepes, penicillin/streptomycin (100 U/ml), 50 μM β-mercaptoethanol (Sigma), and human IL-2 (50 U/ml; Prometheus). All cells were cultured at 37°C and 5% CO2 and tested bimonthly for the absence of mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza).

Generation of CARs and recombinant lentiviral vectors

CD19- and ROR1-specific CAR constructs have been previously described (49, 59). For work in this study, a single STII sequence and two G4S linkers were inserted between the FMC63 or R12 scFv and immunoglobulin G4 (IgG4) hinge (25). These were linked to the 27–amino acid transmembrane domain of human CD28 (UniProt: P10747) and to a signaling module comprising either (i) the 41–amino acid cytoplasmic domain of human CD28 with an LL→GG substitution located at positions 186 to 187 of the native CD28 protein (60) or (ii) the 42–amino acid cytoplasmic domain of human 4-1BB (UniProt: Q07011), each of which was linked to the 112–amino acid cytoplasmic domain of isoform 3 of human CD3ζ (UniProt: P20963-3). Mutant CD28/CD3ζ CARs with tyrosine to phenylalanine substitutions at CD28 UniProt positions 206, 209, and 218 and/or with proline to alanine substitutions at CD28 UniProt positions 208 and 211 were generated by site-directed mutagenesis. All CAR constructs were linked by T2A sequence to EGFRt, codon-optimized, and cloned into an HIV7 lentiviral vector. For fluorescence microscopy, the CD3ζ endodomain was directly fused to enhanced green fluorescent protein (eGFP). To make CAR antigen–expressing K562 cells, amino acids 1 to 325 of human CD19 (UniProt: P15391) were cloned into an HIV7 lentiviral vector, and amino acids 1 to 937 of human ROR1 (UniProt: Q01973) were cloned into an mp71 retroviral vector, which was a gift of W. Uckert (Max Delbruck Center for Molecular Medicine). All cloning was performed by PCR, enzyme digest, and/or Gibson assembly. Plasmids were verified by capillary sequencing and restriction digest.

Lentivirus preparation and transduction

To prepare CAR T cells, Lenti-X cells were transiently transfected with the HIV7 CAR vector, as well as psPAX2 (Addgene plasmid no. 12260) and pMD2.G (Addgene plasmid no. 12259) packaging plasmids. One day later (day 1), primary T cells were activated using Dynabeads Human T-Activator CD3/CD28 (Thermo Fisher Scientific) and cultured in CTL supplemented with IL-2 (50 U/ml). On the next day (day 2), lentiviral supernatant was harvested from Lenti-X cells, filtered using 0.45-μm polyethersulfone (PES) syringe filters (Millipore), and added to activated T cells. Polybrene (Millipore) was added to reach a final concentration of 4.4 μg/ml, and cells were spinoculated at 800g and 32°C for 90 min. Viral supernatant was replaced 8 hours later with fresh CTL supplemented with IL-2 (50 IU/ml). Half-media changes were then performed every 48 hours using CTL supplemented with IL-2 (50 U/ml). Dynabeads were removed on day 6; CD8+EGFRt+-transduced T cells were FACS (fluorescence-activated cell sorting)–purified on a FACSAriaII (BD Biosciences) on day 9.

To prepare K562/CD19 cells, Lenti-X cells were transiently transfected with psPAX2, pMD2.G, and an HIV7 lentiviral vector encoding CD19. To prepare K562/ROR1 cells, Lenti-X cells were transiently transfected with MLV g/p, 10A1, and an mp71 retroviral vector encoding ROR1. To prepare Raji/ffluc cells, Lenti-X cells were transiently transfected with psPAX2, pMD2.G, and an HIV7 lentiviral vector encoding GFP and firefly luciferase. Two days later, viral supernatant was filtered using a 0.45-μm PES syringe filter and added to K562 or Raji cells. Five days later, transduced cells were stained with mAbs specific for CD19 (HIB19, BioLegend) or ROR1 (2A2, Miltenyi Biotec) and FACS-purified on a FACSAria II to greater than 97% purity.

T cell expansion for MS and functional analyses

FACS-purified CD8+EGFRt+ cells were expanded over a single stimulation cycle before MS and/or functional analyses. CD19-specific CAR T cells were expanded by coculture with irradiated CD19+ lymphoblastoid cell lines (LCLs) in a 1:7 (T cell/LCL) ratio and assayed 8 days after stimulation. ROR1-specific CAR T cells were expanded using a rapid expansion protocol containing purified OKT3, irradiated LCL, and irradiated PBMCs and were assayed 11 days after stimulation. During expansion, cultures were fed with fresh CTL media containing IL-2 (50 IU/ml) every 2 to 3 days.

Flow cytometry and cell phenotyping

T cells were stained with a 1:100 dilution of fluorophore-conjugated mAbs specific for human CD4 (RPA-T4), CD8 (SK1), CD27 (M-T271), CD28 (CD28.2), CD45 (HI30), CD45RO (UCHL1), CD62L (DREG-56), CD223 (3DS223H), CD279 (eBioJ105), or CD366 (F38-2E2) purchased from BD Biosciences, Thermo Fisher Scientific, or BioLegend. T cells were also stained with isotype control fluorophore-conjugated antibodies when appropriate. Cetuximab (anti-EGFR, Bristol-Myers Squibb) and 3E8 [anti-STII, Fred Hutchinson Cancer Research Center (FHCRC)] were biotinylated using the EZ-Link Sulfo-NHS-Biotin Kit (Thermo Fisher Scientific) followed by cleanup with the Zeba Spin Desalting Column (Thermo Fisher Scientific) and used to stain T cells in conjunction with streptavidin-allophycocyanin (Thermo Fisher Scientific). DNA content staining was performed by fixing T cells with 70% ice-cold ethanol, permeabilizing cells with 1% Triton X-100 (Sigma), degrading RNA with ribonuclease A (100 μg/ml; Thermo Fisher Scientific), and staining DNA with propidium iodide (20 μg/ml; Thermo Fisher Scientific). All data were collected on a FACSCanto II or FACSAria II (BD Biosciences).

STII and control microbead preparation

Streptavidin-coated magnetic particles (1 ml; Spherotech) were washed once in excess 1× PBS supplemented with penicillin/streptomycin (100 U/ml) (PBS + P/S) using a benchtop magnet. STII microbeads were prepared by resuspending beads in PBS + P/S (1 ml) and then slowly adding 16.67 μg of STII biotin mAb (GenScript) while vortexing. Beads were incubated overnight at 4°C on a three-dimensional orbital shaker, washed three times with excess PBS + P/S using a benchtop magnet, and resuspended in PBS + P/S (4 ml). To make control beads, streptavidin-coated magnetic particles (1 ml) were washed once using a benchtop magnet and resuspended in PBS + P/S (4 ml). Beads were stored at 4°C.

Cell stimulations, protein lysates, and RNA isolation

CAR T cells were washed and resuspended in warm CTL medium at a concentration of 2 × 107 cells/ml. Cells were then incubated with STII or control microbeads at a ratio of 30 μl of beads per 1 × 106 cells in a 37°C water bath. After the allotted time, cells were quickly washed twice using ice-cold PBS and lysed in a 6 M urea, 25 mM tris (pH 8.0), 1 mM EDTA, 1 mM EGTA solution supplemented with protease (Sigma) and phosphatase inhibitors (Sigma) at a 1:100 dilution, hereon referred to as lysis buffer. Lysates were sonicated for 15 s before centrifuging at 10,000g and 4°C for 10 min. Beads were removed during lysate clearing. After 6 hours of stimulation, RNA isolations were performed using a NucleoSpin RNA kit (Macherey-Nagel) according to the manufacturer’s instructions. Beads were removed using a benchtop magnet before cell lysis and RNA extraction.

Protein digestion, TMT labeling, and pTyr peptide immunoprecipitation

Protein was quantified in lysates by Micro BCA Assay (Thermo Fisher Scientific), and lysates were diluted to 2 mg/ml using lysis buffer. Lysates were reduced in 24 mM TCEP [tris(2-carboxyethyl)phosphine hydrochloride] (Thermo Fisher Scientific) for 30 min at 37°C with shaking, followed by alkylation with 48 mM iodoacetamide (Sigma) in the dark at room temperature for 30 min. Lysates were then diluted with 200 mM tris (pH 8.0) to a urea concentration of 2 M. Lys-C (Wako) was dissolved in 25 mM tris (pH 8.0) at 200 μg/ml and added to lysates at 1:100 (enzyme/protein) ratio by mass and incubated for 2 hours at 37°C with shaking. Samples were further diluted with 200 mM tris (pH 8.0) to a urea concentration of 1 M before adding trypsin at a 1:50 trypsin/protein ratio. After 2 hours, a second trypsin aliquot was added at a 1:100 trypsin/protein ratio. Digestion was carried out overnight at 37°C with shaking. After 16 hours, the reaction was quenched with formic acid (FA) to a final concentration of 1% by volume. Samples were desalted using Oasis HLB 96-well plates (Waters) and a positive pressure manifold (Waters). The plate wells were washed with 3 × 400 μl of 50% MeCN/0.1% FA and then equilibrated with 4 × 400 μl of 0.1% FA. The digests were applied to the wells and then washed with 4 × 400 μl of 0.1% FA before being eluted drop by drop with 3 × 400 μl of 50% MeCN/0.1% FA. The eluates were lyophilized, followed by storage at −80°C until use. For TMT labeling (Thermo Fisher Scientific), desalted peptides were resuspended in 50 mM Hepes at 1 mg/ml based on starting protein mass. TMT reagents were resuspended in 257 μl of MeCN and transferred to the peptide sample. Samples were incubated at room temperature for 1 hour with mixing. Labeling reactions were quenched by the addition of 50 μl of 5% hydroxyl amine (Sigma) and incubated for 15 min at room temperature with mixing. The independent labeling reactions were then pooled together and lyophilized. The labeled peptides were desalted as above and then lyophilized and stored at −80°C. Immunoprecipitation of pTyr peptides was performed using the PTMScan P-Tyr-1000 Kit (Cell Signaling Technology). The enriched pTyr peptide fraction was purified using a C18 Spin Tip (Thermo Fisher Scientific), lyophilized, and stored at −80°C until analysis. The flow-through fraction was desalted, lyophilized, and stored at −80°C.

Basic (high-pH) reversed-phase LC

The desalted and pTyr peptide–depleted flow-through was fractionated by high-pH reversed-phase (RP) LC. Protein digest (4 mg) was loaded onto an LC system consisting of an Agilent 1200 high-performance LC with mobile phases of 5 mM NH4HCO3 (pH 10) (A) and 5 mM NH4HCO3 in 90% MeCN (pH 10) (B). The peptides were separated by a 4.6 mm × 250 mm Zorbax Extend-C18, 3.5 μm, column (Agilent) over 96 min at a flow rate of 1.0 ml/min by the following timetable: hold 0% B for 9 min, gradient from 0 to 10% B for 4 min, 10 to 28.5% B for 50 min, 28.5 to 34% B for 5.5 min, 34 to 60% B for 13 min, hold at 60% B for 8.5 min, 60 to 0% B for 1 min, reequilibrate at 0% B for 5 min. Fractions were collected at 1-min intervals from 0 to 96 min by the shortest path by row in a 1-ml deep well plate (Thermo Fisher Scientific). The high-pH RP fractions were concatenated into 24 samples by every other plate column starting at 15 min (for example, sample 1 contained fractions from wells B10, D10, F10, etc.). The remaining fractions were combined such that fractions from 12 to 14 min were added to sample 1, all fractions after 86 min were added to sample 24, and all fractions from 0 to 11 min were combined into sample “A.” Ninety-five percent of every 12th fraction of the 24 samples was combined (1,13; 2,14;…) to generate 12 more samples, which were dried down and stored at −80°C before phosphopeptide enrichment by IMAC.

Immobilized metal affinity chromatography

IMAC enrichment was performed using Ni-NTA agarose beads (Qiagen) stripped with EDTA and incubated in a 10 mM FeCl3 solution to prepare Fe3+-NTA agarose beads. Fractionated lysate was reconstituted in 200 μl of 0.1% trifluoroacetic acid (TFA) in 80% MeCN and incubated for 30 min with 100 μl of the 5% bead suspension while mixing at room temperature. After incubation, beads were washed three times with 300 μl of 0.1% TFA in 80% MeCN. Phosphorylated peptides were eluted from the beads using 200 μl of 70% acetonitrile, 1% ammonium hydroxide for 1 min with agitation at room temperature. Samples were transferred into a fresh tube containing 60 μl of 10% FA, dried down, and resuspended in 0.1% FA and 3% MeCN. Samples were frozen at −80°C until analysis.

Nano–LC-MS/MS

Phosphopeptide-enriched samples were analyzed by LC-MS/MS on an Easy-nLC 1000 (Thermo Fisher Scientific) coupled to an LTQ-Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific) operated in positive ion mode. The LC system, configured in a vented format, consisted of a fused-silica nanospray needle [PicoTip emitter, 50 μm inside diameter (ID) × 20 cm, New Objective] packed in-house with ReproSil-Pur C18-AQ, 3 μm, and a trap (IntegraFrit Capillary, 100 μm ID × 2 cm, New Objective) containing the same resin as in the analytical column with mobile phases of 0.1% FA in water (A) and 0.1% FA in MeCN (B). The peptide sample was diluted in 20 μl of 0.1% FA, 3% MeCN, and 8.5 μl was loaded onto the column and separated over 210 min at a flow rate of 300 nl/min with a gradient from 5 to 7% B for 2 min, 7 to 35% B for 150 min, and 35 to 50% B for 1 min, hold 50% B for 9 min, 50 to 95% B for 2 min, hold 95% B for 7 min, 95 to 5% B for 1 min, reequilibrate at 5% B for 38 min. A spray voltage of 2000 V was applied to the nanospray tip. MS/MS analysis occurred over a 3-s cycle time consisting of one full-scan MS from 350 to 1500 mass/charge ratio (m/z) at a resolution of 120,000, followed by data-dependent MS/MS scans using high-energy collision dissociation activation with 27% normalized collision energy of the most abundant ions. Selected ions were dynamically excluded for 45 s after a repeat count of 1.

Immunoprecipitation

Protein G Dynabeads (Thermo Fisher Scientific) were incubated with anti-STII antibody (GenScript) for 60 min, cross-linked for 30 min using 20 mM dimethyl pimelimidate (Thermo Fisher Scientific) diluted in 200 mM triethanolamine (Thermo Fisher Scientific), quenched with 150 mM monoethanolamine (Thermo Fisher Scientific), and washed three times with 1× PBS. T cells were lysed in NP40 Cell Lysis Buffer (Thermo Fisher Scientific) supplemented with protease and phosphatase inhibitors. Lysates were incubated on ice for 15 min and then centrifuged at 10,000g and 4°C for 10 min. Immunoprecipitations were performed according to the manufacturer’s instructions, where Dynabeads were incubated with equal masses of cleared lysates for 90 min at room temperature.

Western blotting

Equal masses of protein lysate or equal volumes of immunoprecipitation eluents were loaded into NuPAGE gels (Thermo Fisher Scientific). After protein transfer onto nitrocellulose membranes (Thermo Fisher Scientific), membranes were blocked with Western Blocking Reagent (Sigma). Membranes were stained with primary and secondary antibodies diluted in SuperBlock (Thermo Fisher Scientific) supplemented with 0.1% Tween. The following antibodies were used: anti-human CD247 (8D3, BD Biosciences), anti-human CD247 pTyr142 (K25-407.69, BD Biosciences), anti–ZAP-70 pTyr319 (65E4, Cell Signaling Technology), anti–SLP-76 (polyclonal, Cell Signaling Technology), anti–SLP-76 pSer376 (D9D6E, Cell Signaling Technology), anti–PLC-γ1 (D9H10, Cell Signaling Technology), anti–PLC-γ1 pTyr783 (D6M9S, Cell Signaling Technology), anti-DAPP1 (D9K4O, Cell Signaling Technology), anti-DAPP1 pTyr139 (D7G4G, Cell Signaling Technology), anti-Lck (D88, Cell Signaling Technology), anti-CD28 (D2Z4E, Cell Signaling Technology), anti-mouse horseradish peroxidase (HRP) (polyclonal, Cell Signaling Technology), and anti-rabbit HRP (polyclonal, Cell Signaling Technology). Typical antibody dilutions ranged from 1:10,000 to 1:2500. Band intensities were quantified using ImageJ [National Institutes of Health (NIH)]; normalized to total protein, loading control, or immunoprecipitation input; and then renormalized to a control sample.

RNA sequencing

RNA was extracted from 24 samples from three donors. Total RNA integrity was checked using an Agilent 4200 TapeStation (Agilent Technologies) and quantified using a Trinean DropSense96 spectrophotometer (Caliper Life Sciences). RNA-seq libraries were prepared from total RNA using the TruSeq RNA Sample Prep Kit v2 (Illumina) and a Sciclone NGSx Workstation (PerkinElmer). Library size distributions were validated using an Agilent 4200 TapeStation. Additional library quality control, blending of pooled indexed libraries, and cluster optimization were performed using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific). RNA-seq libraries were pooled (6- to 8-plex) and clustered onto a flow cell lane. Sequencing was performed using an Illumina HiSeq 2500 in rapid mode using a paired-end, 50-base read length sequencing strategy.

Quantitative polymerase chain reaction

RNA was extracted, and total RNA integrity was verified using an Agilent 4200 TapeStation (Agilent Technologies). RNA (500 ng) was used to prepare complementary DNA (cDNA) with SuperScript III (Thermo Fisher Scientific). cDNA (15 ng) was added to a reaction with TaqMan Universal Master Mix II with uracil N-glycosylase (Thermo Fisher Scientific) and one of the following Thermo Fisher Scientific TaqMan assay probes: Hs00172973_m1 (FOXO4), Hs00902234_m1 (IL7R), Hs00360439_g1 (KLF2), or Hs99999907_m1 (B2M). Reactions were run in duplicate or triplicate on one 384-well plate. ΔCt values were calculated by dividing mean Ct of technical replicates from FOXO4, IL7R, and KLF2 probes by the mean Ct of the housekeeping gene β2 microglobulin (B2M). CD28/CD3ζ versus 4-1BB/CD3ζ ratios for each donor were calculated and subjected to a log2 transformation.

In vitro functional assays

CAR T cells were cocultured with K562, K562/CD19, or K562/ROR1 cells at a T cell–to–tumor cell ratio of 2:1. In some experiments, CAR T cells were also incubated with control or STII microbeads at a ratio of 30 μl of beads per 1 × 106 cells. Cytokine concentrations in cellular supernatant were quantified by ELISA (Thermo Fisher Scientific) 24 hours after stimulation. T cell proliferation was quantified by staining CAR T cells with a 0.2 μM solution of CFSE dye (Thermo Fisher Scientific) and incubating CAR T cells with K562/CD19 cells, K562/ROR1 cells, control beads, or STII microbeads for 72 hours.

Transfer of T cells in NSG mice

Six- to 8-week-old male or female NSG mice were obtained from the Jackson Laboratory or bred in-house. Mice were engrafted via tail vein with 5 × 105 CD19+ Raji/ffluc cells and, 7 days later, injected intravenously with PBS or a defined product of purified CD8+ and CD4+ CD19-specific CAR T cells mixed together in a 1:1 ratio. Bioluminescence imaging was performed as described (49). Mice were either followed for survival or sacrificed on day 20 for analysis of T cell frequencies and phenotypes by flow cytometry. Peripheral blood was extracted, red blood cells were lysed using ACK Lysing Buffer (Thermo Fisher Scientific), and remaining cells were stained with fluorochrome-labeled mAbs. Bone marrow was isolated from hindlimbs by mechanical disruption, followed by red blood cell lysis, and staining with fluorochrome-labeled mAbs. Mice handlers were blinded to group allocation. The FHCRC Institutional Animal Care and Use Committee approved all experimental procedures.

Fluorescence microscopy

CD8+ T cells from healthy donors were transduced as previously described. Instead of FACS purification on day 9, cells were imaged on a DeltaVision Elite microscope (GE Healthcare). At least eight cells were visualized per condition. Raw images were subjected to a linear adjustment of brightness and contrast using ImageJ (NIH).

Shotgun MS data analysis

Raw MS/MS spectra from each replicate experiment were searched together against the reviewed Human Universal Protein Resource (UniProt) sequence database (release 2016_01) with common laboratory contaminants using the MaxQuant/Andromeda search engine version 1.6.0.1 (61). The search was performed with a tryptic enzyme constraint for up to two missed cleavages. Variable modifications were oxidized methionine, pSer, pThr, and pTyr. Carbamidomethylated cysteine was set as a static modification. Peptide MH+ mass tolerances were set at 20 parts per million (ppm). The overall FDR was set at ≤1% using a reverse database target decoy approach.

For the three TMT experiments, phosphopeptide site localization was determined by MaxQuant and converted to phosphorylation sites using Perseus version 1.6.0.7 (62). At this step, reverse hits and potential contaminants were excluded from further analysis. Data normalization was performed by scaling each TMT channel to the channel median, followed by a log2 transformation. Stimulation versus control ratios were calculated by subtracting the appropriate control channels from stimulated channels. Because of incomplete MS sampling, some phosphorylation sites (features) were only found in one or two replicate experiments, and a much smaller minority (<1%) of sites were not found in every TMT channel.

Differential expression analyses over phosphorylation sites were performed using the limma statistical framework and associated R package (35, 63). For these analyses, we chose to keep only features that had values in at least two experiments and all TMT channels, leaving us with 14,490 quantified phosphorylation sites. A linear model was fitted to each phosphorylation site, and empirical Bayes moderated t statistics were used to assess differences in expression/abundance. Contrasts comparing stimulation versus control treatments were tested. Intraclass correlations were estimated using the duplicate correlation function of the limma package to account for measures originating from the same patients and the same antigens (64). An absolute log2FC cutoff (stimulation versus control) of 0.7 and an FDR cutoff of 5% were used to determine differentially expressed phosphorylation sites. Analyses of signaling networks and KEGG pathways were performed using StringDB.

RNA-seq data analysis

Image analysis and base calling were performed using Illumina’s Real Time Analysis v1.18 software, followed by “demultiplexing” of indexed reads and generation of FASTQ files, using Illumina’s bcl2fastq Conversion Software v1.8.4. The RNA-seq data were aligned to the human genome (University of California, Santa Cruz Human Genome Assembly GRCh38 reference) using STAR (Spliced Transcripts Alignment to a Reference), and gene quantification was performed using RSEM (RNA-Seq by Expectation Maximization) (65, 66). Genes with less than 10 nonzero read counts (taking into account technical replicates) were discarded, leaving 18,498 expressed genes. All libraries passed the quality control criteria (libraries with more than 200,000 reads, 12,000 detected genes, and an exon range > 60%). Raw count data were imported into R. edgeR was used to calculate the normalization factors to scale the raw library sizes, followed by a voom transformation from the limma Bioconductor package (67, 68). It transforms count data to log2 counts per million and estimates the mean-variance relationship to compute appropriate observation-level weights. Linear models with subject random effects were again used for differential gene expression analysis as described in the “Shotgun MS data analysis” section. Contrasts comparing treatments (control versus stimulation) or CARs (CD28/CD3ζ versus 4-1BB/CD3ζ) were tested. An absolute log2FC cutoff of 1 and an FDR cutoff of 1% were used to determine differentially expressed genes.

Analysis of T cell phenotype, function, and in vivo experiments

FlowJo version 9 (Tree Star) was used to analyze flow cytometry files and calculate proliferation indices. Prism version 7 (GraphPad Software) was used to plot data and calculate statistics. P values meeting an α = 0.05 level of statistical significance are indicated in the figures. The precise statistical tests used are indicated in the figure legends.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/544/eaat6753/DC1

Fig. S1. CAR designs containing an STII sequence enable selective activation of CAR signaling in primary T cells.

Fig. S2. CD19- and ROR1-specific CD28/CD3ζ and 4-1BB/CD3ζ CARs signal similarly in CD4+ T cells and are uniformly expressed on the cell surface.

Fig. S3. Log2FC summary statistics.

Fig. S4. Mutations to the CAR CD28 domain abrogate Lck binding.

Table S1. CAR stimulation–responsive phosphorylation sites after 10 min.

Table S2. CAR stimulation–responsive phosphorylation sites after 45 min.

Table S3. KEGG pathways identified by StringDB after 45 min of CAR stimulation.

Table S4. Phosphorylation sites increased by CAR stimulation at 45 min.

Table S5. Differentially expressed genes between stimulated CD28/CD3ζ and 4-1BB/CD3ζ CAR T cells.

REFERENCES AND NOTES

Acknowledgments: We thank the M.J. Murdock Charitable Trust and members of the FHCRC Proteomics Shared Resource for help in data acquisition. We also thank D. E. S. Parrilla and M. Jess for performing mouse husbandry and tumor xenograft experiments. Funding: This research was supported by NIH (R01 #CA136551 and R01 #CA114536 to S.R.R., U01 #CA214114 to A.G.P., and R50 #CA211499 to J.R.W.) and FHCRC (Bezos Immunotherapy Pilot Award to A.G.P. and S.R.R. and Interdisciplinary Training in Cancer Research Training Grant support to A.I.S.). Author contributions: A.I.S., R.G.I., D.S., A.G.P., and S.R.R. designed the experiments. A.I.S., R.G.I., J.J.K., A.R., E.J.A., and U.J.V. conducted the experiments. V.V., C.L., J.R.W., J.J.K., and R.G. processed the data and wrote accompanying software. L.L. developed STII CAR technology and provided reagents. A.I.S. and R.G.I. performed formal data analysis. A.I.S. and S.R.R. wrote the manuscript. A.I.S., R.G.I., V.V., R.G., A.G.P., and S.R.R. edited the manuscript. A.G.P. and S.R.R. supervised all work. Competing interests: S.R.R. is a founder, shareholder, and scientific advisor of Juno Therapeutics. R.G. is a consultant for Juno Therapeutics. STII CAR technology used in this study is licensed to Juno Therapeutics. A.I.S. and S.R.R. have filed a provisional patent application number 62/635,450 covering applications of mutant CD28 CARs for cellular therapy. The other authors declare that they have no competing interests. Data and materials availability: LC-MS/MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD007921. RNA-seq data have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) under accession number 109161. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Navigate This Article