Granzyme B- and Fas ligand-mediated cytotoxic function induced by mitogenic CD28 stimulation of human memory CD4⁺ T cells

Mitogenic CD28 (or Superagonistic CD28) antibodies activate CTL functions by human CD4+ T cells—functions previously unknown to be induced by these stimulatory antibodies.

Abstract

Some human memory CD4⁺ T cells have cytotoxic functions best understood in the context of viral infections; however, their possible role in pathologic processes is understudied. The novel discovery that mitogenic CD28 antibodies induced proliferation and expansion of Tregs offered therapeutic promise for autoimmune disorders. However, the failed TGN1412 trial forced reassessment of this concept. As memory CD4⁺ T cells are known to produce toxic molecules, including granzyme B (GrzB) and FasL, we wondered whether mitogenic CD28 was able to induce these cytotoxic molecules. A commercially available mitogenic human CD28 mAb (clone ANC28.1) was used to determine whether mitogenic CD28 induces cytotoxic function from human memory CD4⁺ T cells. We found that stimulation of memory CD4⁺ T cells by ANC28.1, as well as by conventional costimulation (CD3/CD28 mAb), robustly induced enzymatically active GrzB, along with increased surface expression of FasL. These functional phenotypes were induced in association with increased expression of T cell activation markers CD69 and CD25, and elimination of target cells by ANC28.1-activated memory CD4⁺ T cells involved both GrzB and FasL. Additionally, ANC28.1-activated memory CD4⁺ T cells caused disruption of epithelial cell monolayer integrity, which was partially mediated by GrzB. These findings reveal functions of memory CD4⁺ T cells previously unknown to be induced by mitogenic CD28, and suggest that these pathogenic mechanisms may have been responsible for some of the widespread tissue destruction that occurred in the TGN1412 trial recipients.

Introduction

In vivo, T cells are activated after receiving two signals from an antigen-presenting cell (APC) (for review, see ref. [1]). The first signal comes from their T cell receptor (TCR) upon ligation with MHC class I or II molecules containing TCR-specific peptide, and the second signal comes from a costimulatory molecule, such as CD28, OX40, ICOS, or CD40L. CD28 is the most important and best-characterized costimulatory molecule and interacts with CD80 (B7.1) and CD86 (B7.2) molecules on APCs. Concomitant delivery of these two activating signals to a T cell results in their proliferation, expansion, and initiation of certain effector functions, depending on the particular T cell subset, whereas delivery of one signal in the absence of the other usually results in the T cell becoming nonresponsive or anergic. Recapitulation of T cell activation in vitro can be conducted using stimulatory mAb against CD3 and CD28, immobilized onto plastic or on beads, or by using more broadly activating polyclonal stimulators, such as PHA or phorbol esters (PMA) and ionomycin.

The finding that CD28 agonism (via stimulatory CD28 mAb) without TCR ligation results in full T cell activation and proliferation at levels comparable with activation by conventional T cell stimulation prompted reassessment of the two-signal model of T cell activation [2, 3]. Structure-function studies revealed that these mAb have different epitope specificities for CD28 compared with natural CD28 ligands, such as CD80/CD86 [4–6], and these mAb were designated “mitogenic CD28” or “superagonistic CD28”. Further studies using murine models showed that these stimulatory mAb induced proliferation and expansion of CD4⁺ Tregs [7–9], which elevated interest in these mAb as a potential therapeutic for T lymphopenias and autoimmune diseases. This focus resulted in development of a humanized superagonistic CD28 mAb (clone TGN1412). Following the testing of TGN1412 using murine and non-human primate (NHP), TGN1412 was administered to human volunteers but resulted in adverse effects, including severe inflammation and multiorgan failure (for review, see ref. [10]). Subsequent investigation into the clinical trial revealed a number of possible reasons for the severe responses, including massive production of inflammatory cytokines [11], unexpected redistribution of T cells [12], and species differences with respect to CD28 structure, expression pattern, and signaling between humans and the NHPs used for antibody development and testing [13–15]. Despite these explanations, the mechanisms responsible for the pathological consequences are not completely understood.

The activation and expansion of T cells by various mitogenic CD28 antibodies are well-described using murine, NHP, or human T cells, and many of these CD28 mAb are similar with respect to activation, cytokine production, and proliferation. Studies with human T cells used clones BW 828, 5.11A1, 5D10 (also known as ANC28.1), or TGN1412. Clone BW 828 was the first reported mitogenic CD28 mAb and this clone increased CD25 expression and induced proliferation of naïve CD4⁺CD45RA⁺ and memory CD4⁺CD45RO⁺ T cells [2]. Clone 5.11A1 increased CD25 and CD69 expression and induced IL-2 and IFN-γ, as well as proliferation of naïve and memory CD4⁺ T cells [16]. Clone 5D10 activated T cells similar to that seen with TGN1412, and both mAb cause elevation of intracellular calcium levels for abnormally long time periods in human naïve and memory CD4⁺ T cells, accompanied by proliferation and production of IL-2, IFN-γ, and TNF-α [14]. However, these activation patterns were not observed in rhesus or cynomolgus monkey T cells. Rodent models have used clones such as JJ316, which activates CD4⁺ T cells in rats, as shown by increased expression of CD25 and CD69, and release of inflammatory mediators, IFN-γ, IL-2, IL-17, RANTES, and MCP-1. This activation was also accompanied by unexpected alteration of T cell migratory patterns as a result of aberrant expression of adhesion molecules, such as CD62L and LFA-1, and immediately following this activation period, proliferation of the CD4⁺CD25⁺ Foxp3⁺ Tregs was observed [12]. Another clone used in rodent studies is clone D665, which induced IL-2 production and proliferation from primary peripheral T cells [17].

Previous studies in our lab demonstrated that the commercially available mitogenic human CD28 mAb clone 5D10 (ANC28.1), a clone similar to TGN1412, activated and induced production of inflammatory cytokines from memory CD4⁺ T cells [18]. Memory CD4⁺ T cells are long-lived cells derived from previously activated naïve CD4⁺ T cells following encounter with cognate antigen, and are broadly identified by surface expression of the RO isoform of CD45. A subset within the memory CD4⁺ T cell pool has cytolytic function resembling CD8⁺ T cell and NK cell function in which a target cell is eliminated via effector molecules, such as GrzB or FasL. Additionally, human memory CD4⁺CD45RO⁺ T cells produce substantial amounts of GrzB after conventional activation in vitro using CD3/CD28 mAb [19]. As activation of CD4⁺ T cells by mitogenic CD28 resembles activation by conventional stimulants, we hypothesized that cytotoxic effector functions may also be elicited from circulating memory CD4⁺ T cells after stimulation with ANC28.1 mAb. In these experiments, we show that mitogenic CD28 (ANC28.1) strongly activates purified human memory CD4⁺ T cells and induces them to kill target cells via GrzB and FasL. Furthermore, GrzB from ANC28.1-activated memory CD4⁺ T cells was partially responsible for altering monolayer integrity of Caco-2 epithelial cells. These results demonstrate that a significant fraction of human memory CD4⁺ T cells can become cytotoxic and pathogenic when activated by mitogenic CD28 mAb and may help to explain the multiorgan failure and tissue damage that occurred in the recipients receiving TGN1412 mAb during the clinical trial.

MATERIALS AND METHODS

Cells and stimulations

The majority of experiments was conducted using human memory CD4⁺CD45RO⁺ T cells purified from peripheral blood of individual healthy donors or buffy coat preparations (Gulf Coast Regional Blood Center, Houston, TX, USA), and informed consent was obtained in the case of identified donors. Memory CD4⁺ T cells were negatively selected with magnetic bead-based EasySep kits (Stemcell Technologies, Vancouver, BC, Canada) from PBMCs, which had been isolated from peripheral blood or buffy coats using Histopaque-1077 density-gradient separation (Sigma-Aldrich, St. Louis, MO, USA). Purities of memory CD4⁺CD45RO⁺ T cells were always >95%. Some experiments used CD3⁺CD8⁺ or naïve CD4⁺CD45RA⁺ T cells, which were also purified from PBMCs with EasySep negative selection kits. Cell cultures were conducted at 37°C + 5% CO₂, and media used for experiments were RPMI 1640, supplemented with 10% FBS, 2 mM L-glutamine, 0.1 mM MEM nonessential amino acids, 2 mM sodium pyruvate, 25 mM HEPES, and 1× antibiotic-antimycotic.

The mitogenic human CD28 mAb, clone ANC28.1 (also referred to as clone 5D10, IgG1κ), was purchased from Ancell (Bayport, MN, USA). Most stimulations of memory CD4⁺ T cells were conducted using 2 μg/ml soluble ANC28.1 at cell concentrations of 0.5–1 × 10⁶ cells/ml for a time period of ∼24 h, unless specified otherwise, after which, cells were harvested for appropriate analyses. A conventional CD28 mAb, clone CD28.2 (BD Biosciences, San Jose, CA USA), was used as a CD28 nonstimulatory control. For conventional T cell stimulations, 5 μg/ml immobilized CD3 mAb (clone UCHT1; BD Biosciences) and 2 μg/ml soluble CD28.2 (BD Biosciences) were used.

Cell surface and intracellular flow cytometry

T cell activation was assessed by flow cytometry staining for cell-surface expression of CD69 and CD25. Cell death and apoptosis were assessed by cell-surface staining of inverted phosphatidylserine with fluorochrome-conjugated peptide ligand annexin V. Intracellular flow cytometry staining of GrzB was conducted by fixing and permeabilizing cells with Cytofix/Cytoperm solution (BD Biosciences). The following antibodies were used for flow cytometry staining: CD69-PE (clone FN50; BD Biosciences), CD25-allophycocyanin (clone 2A3; BD Biosciences), annexin V-FITC (BioSource/Invitrogen, Carlsbad, CA, USA), annexin V-allophycocyanin (BD Biosciences), GrzB-FITC (clone GB11; BD Biosciences), and FasL-PE (clone NOK-1; BioLegend, San Diego, CA, USA). Appropriate isotype controls were also used during flow cytometry staining, and compensation settings were established using single-color and fluorescence-minus-one controls. Data were acquired with EPICS XL-MCL or Gallios flow cytometers (Beckman Coulter, Miami, FL, USA), and analyses were performed using FlowJo (Tree Star, Ashland, OR, USA) or Kaluza (Beckman Coulter) software.

Extracellular protein measurements

All extracellular protein amounts were determined from cell-free supernatants. Cytokines were measured using multiplex, flow cytometry-based CBA kits (BD Biosciences). Extracellular GrzB and perforin were measured with ELISA kits (Cell Sciences, Canton, MA, USA).

Flow cytometry bystander cell-killing assay

We devised a simple, two-color flow cytometry assay to measure bystander cell killing by activated memory CD4⁺ T cells in which JCAM1.6 or K562 cell lines were used as target cells. The JCAM1.6 T cell line (ATCC, Manassass, VA, USA) is a Jurkat derivative that does not respond to an ANC28.1 agonism, as it lacks the Lck tyrosine kinase, and the K562 cell line (ATCC) is a myeloid leukemia cell line that does not express Fas (CD95) and is useful for determining Fas-independent cell death [20]. Target cells were prelabeled with a membrane dye, PKH26 (Sigma-Aldrich), which has optimal emission ∼567 nm, and were then mixed with purified memory CD4⁺ T cells, usually at E:T cell ratios of 10:1 (1×10⁶ memory CD4⁺ T cells:1×10⁵ target cells in a culture volume of 1 ml in 48-well plates), in the presence of 2 μg/ml ANC28.1 and various inhibitors or neutralizing antibodies (depending on the experiment) for ∼24 h. Cells were then harvested, and apoptosis of PKH26-positive target cells was assessed by flow cytometry using annexin V staining. The compound, Z-IETD-FMK (MP Biomedicals, Solon, OH, USA), is a caspase-8 and GrzB inhibitor and was used at 10 μM for GrzB inhibition experiments. (Another compound, Z-AAD-CMK, was also tested but was found to be toxic to cells.) For experiments involving neutralization of various TNF family member proteins, the following antibodies were used: 10 μg/ml anti-FasL (clone 2C101; Kamiya Biomedical, Seattle, WA, USA), 10 μg/ml anti-TNF-α (goat polyclonal; R&D Systems, Minneapolis, MN, USA), 10 μg/ml anti-TNF-β (goat polyclonal; R&D Systems), 10 μg/ml anti-TRAIL (goat polyclonal; R&D Systems), and 1 μg/ml anti-OPG/TNFR superfamily 11B (goat polyclonal; R&D Systems).

Assessment of GrzB enzymatic activity

GrzB bioactivity within target cells was examined using the GranToxiLux kit from OncoImmunin (Gaithersburg, MD, USA), a flow cytometry-based assay in which substrate cleavage by GrzB within live cells results in green fluorescence emission, ∼520 nm (GranToxiLux-FITC). Target cells were prelabeled with CellVue Claret dye (Sigma-Aldrich), which fluorescently emits ∼675 nm and then cultured in direct contact with purified memory CD4⁺ T cells and a live cell-permeable GrzB substrate (GranToxiLux) at E:T cell ratios of 10:1 (usually 1×10⁶ memory CD4⁺ T cells:1×10⁵ target cells in a culture volume of 1 ml in 48-well plates) for ∼24 h in the presence of 2 μg/ml ANC28.1, with or without 10 μM caspase-8/GrzB inhibitor (Z-IETD-FMK). Cells were then harvested, washed, and analyzed by flow cytometry. Aliquots were also harvested from each condition for parallel assessment of apoptosis by annexin V measurement.

Epithelial cell monolayer integrity measurements

Monolayer integrity of Caco-2 intestinal epithelial cells after direct contact coculture with memory CD4⁺ T cells was assessed by measuring TER and paracellular translocation of TRITC-labeled dextrans (4 kDa MW; Sigma-Aldrich). For TER measurements, 3 × 10⁵ Caco-2 cells were seeded into collagen-coated Cellagen inserts (MP Biomedicals, Solon, OH, USA) and cultured in DMEM medium at 37°C + 5% CO₂ for 1–2 weeks to establish confluent monolayers and stable baseline resistances. Memory CD4⁺ T cells (3×10⁵) were then seeded onto Caco-2 monolayers and cultured for up to 4 days with 2 μg/ml ANC28.1 or control CD28.2 and also with or without 10 μM caspase-8/GrzB inhibitor (Z-IETD-FMK). TER was measured with an epithelial voltohmmeter (EVOM; World Precision Instruments, Sarasota, FL, USA) every 24 h. For monolayer permeability measurements after appropriate coculture time-points, monolayers were fixed with formaldehyde, and then dextran-TRITC (1 mg/ml) was added to monolayers for 4 h. Aliquots of media were then removed from the lower compartment for dextran-TRITC fluorescence measurements (485 excitation/520 emission), and dextran concentrations were determined by standard curve.

Signaling pathway and inhibitor studies

For experiments examining signaling pathways involved during GrzB production from memory CD4⁺ T cells after ANC28.1 stimulation, cells were cultured with an appropriate inhibitor for 1 h (0.5–1×10⁶ cells/ml in 1 ml media) prior to addition of 2 μg/ml ANC28.1. Cells were then cultured for 24 h. The following compounds and inhibitors were used: 20 nM Rapamycin (Rap) (Sigma-Aldrich), 0.4 μM cyclosporin A (CsA) (Sigma-Aldrich), 1 μM JNK inhibitor (JNK1), and 1 μM JNK inhibitor negative control peptide (JNCP) (Millipore, Billerica, MA, USA); 0.5 μM Erk inhibitor (ErkII) and 0.5 μM Erk inhibitor negative control peptide (ENCP) (Millipore); and 2 μM LY294002/PI3K inhibitor (Millipore).

Statistics

Most data are expressed as the mean ± sd. Logarithm-transferred variables have been used in analysis if outcome measurements are highly skewed. One-way ANOVA with repeated measures and multiple comparison test (Tukey test) were performed for pair-wise comparisons between treatments on each condition. In cases when treatments among conditions were compared, two-way ANOVA with repeated measures was performed. In cases where applicable, a Mann-Whitney two-tailed or Student's t test was performed.

RESULTS

Mitogenic CD28 ANC28.1 activates human memory CD4⁺ T cells

Previous work showed that clone ANC28.1 increased CD25 and CD69 expression, induced production of IL-2, IL-8, IFN-γ, and TNF-α from memory CD4⁺ T cells, and selectively induced proliferation of memory but not naïve CD4⁺ T cells [18]. In the present experiments, memory CD4⁺ CTL effector activities induced by ANC28.1 were examined in association with their activation status. The capacity of ANC28.1 to activate human T cells in a 24-h time period was tested with whole PBMCs and purified subsets of T cells, including CD3⁺CD4⁺, CD3⁺CD8⁺, naïve CD4⁺CD45RA⁺, or memory CD4⁺CD45RO⁺ T cells (Fig. 1). We assessed activation by flow cytometry staining for CD69 and CD25 expression. ANC28.1 activated PBMCs, CD4⁺, and CD8⁺ T cells, as compared with untreated or conventional CD28 (CD28.2)-treated cells after 24 h (Fig. 1A), and CD69 expression was usually higher on CD4⁺ compared with CD8⁺ T cells. Stimulation of purified, naïve CD4⁺CD45RA⁺ and memory CD4⁺CD45RO⁺ T cells by ANC28.1 resulted in robust activation of memory CD4⁺ T cell CD69 expression, up to ∼50%, whereas naïve CD4⁺ T cells were only marginally activated with CD69 expression, typically <5% (Fig. 1B). The activation levels of memory CD4⁺ T cells induced by ANC28.1 were also comparable with those induced by conventional activation using CD3/CD28 mAb. Induction of CD25 expression by ANC28.1 was also observed on memory CD4⁺ T cells (Fig. 1C). A titration experiment with ANC28.1 indicated that ∼2 μg/ml was optimal for activation of ∼0.5–1 × 10⁶ memory CD4⁺ T cells, and this concentration was used for subsequent experiments (Fig. 1D). Additionally, memory CD4⁺ T cells remained highly viable at ANC28.1 concentrations as high as 20 μg/ml, as determined by flow cytometry staining with annexin V (Fig. 1D). Thus, ANC28.1 strongly activates human memory CD4⁺ T cells without inducing apoptosis.

Figure 1. Activation of human memory CD4⁺ T cells by mitogenic CD28 (ANC28.1).

(A) PBMCs, CD3⁺CD4⁺, or CD3⁺CD8⁺ T cells were purified from peripheral blood and treated with control CD28.2, ANC28.1 (2 μg/ml), or conventional costimulation (Costim; CD3/CD28 mAb) for 24 h. Activation by CD69 expression was examined by flow cytometry. Error bars are means ± sd of five to six separate experiments of single donors. UT, Untreated. (B) Purified, naïve (CD45RA⁺) or memory (CD45RO⁺) CD4⁺ T cells were treated with control CD28.2, ANC28.1, or CD3/CD28 costimulation for 24 h and then examined for CD69 by flow cytometry. Error bars are means ± sd of three to five separate experiments of single donors. *P < 0.05; **P < 0.01; ***P < 0.001. (C) Representative flow cytometry histogram overlay of memory CD4⁺ T cell CD69 and CD25 expression after treatment with ANC28.1 for 24 h. APC, Allophycocyanin. (D) Activation (CD69) and viability of memory CD4⁺CD45RO⁺ T cells during dose-dependent stimulation with ANC28.1 for 24 h. Cell death was determined by flow cytometric analysis of annexin V expression. Error bars are means ± sd of three to five separate experiments of single donors.

Mitogenic CD28 ANC28.1 induces memory CD4⁺ and memory CD8⁺ T cells to produce granzyme B

The data in Fig. 1 are, for the most part, consistent with other studies that have examined effects of mitogenic CD28 on T cell activation using a variety of cell lines and primary cells from mice, NHPs, and humans. However, from microarray studies, we found that ANC28.1 also induced substantial up-regulation of the death-inducing effector molecule, GrzB, in Jurkat T cells (data not shown), which suggested that ANC28.1 may be inducing previously unrecognized cytotoxic functions by memory CD4⁺ T cells. We therefore examined intracellular and extracellular GrzB in purified memory CD4⁺ T cells after 24 h treatment with ANC28.1 or control CD28.2, as well as after CD3/CD28 costimulation. For comparison, we also examined GrzB production from purified memory CD8⁺CD45RO⁺ T cells from the same donor. Flow cytometry staining for intracellular GrzB revealed differences between memory CD4⁺ and memory CD8⁺ T cells in which untreated memory CD4⁺ T cells harbored very little, if any, amounts of basal intracellular GrzB. By contrast, untreated memory CD8⁺ T cells usually showed at least 10–20% memory CD8⁺ T cells as expressing GrzB (Fig. 2A). Stimulation with ANC28.1 or CD3/CD28 costimulation increased intracellular GrzB modestly in memory CD4⁺ T cells (5–10% memory CD4⁺ T cells after ANC28.1 or CD3/CD28 costimulation) and more so in memory CD8⁺ T cells (15–25% memory CD8⁺ T cells). Despite less intracellular GrzB in memory CD4⁺ T cells, in comparison with memory CD8⁺ T cells, extracellular GrzB production from ANC28.1- or CD3/CD28-activated memory CD4⁺ T cells was comparable with memory CD8⁺ T cells (Fig. 2B). Extracellular GrzB production was variable between donors, ranging from ∼300 to frequently over 5000 pg/ml; however, concentrations of at least 1000 pg/ml were usually observed from activated CD4⁺ and CD8⁺ T cells. These data indicate that although there is very little stored or presynthesized GrzB in memory CD4⁺ T cells, activation by ANC28.1 or CD3/CD28 costimulation stimulates high levels of GrzB production similar to that of CD8⁺ T cells.

Figure 2. Mitogenic CD28 stimulates memory CD4⁺ and CD8⁺ T cells to produce GrzB.

Comparison of GrzB production from memory CD4⁺ and memory CD8⁺ T cells purified from the same donor after treatment with stimulants for 24 h. (A) Intracellular GrzB (flow cytometry) in memory CD4⁺ and memory CD8⁺ T cells after treatment with control CD28.2, ANC28.1, or CD3/CD28 costimulation for 24 h. Shown are representative flow cytometry histograms depicting intracellular GrzB versus surface CD69 expression from three separate experiments. (B) Extracellular GrzB (ELISA) from memory CD4⁺ and memory CD8⁺ T cells after treatment with control CD28.2, ANC28.1, or CD3/CD28 costimulation for 24 h. Error bars are means ± sd of one experiment (in triplicate) representative of three separate experiments.

Mitogenic CD28 ANC28.1 induces memory CD4⁺ T cells to kill bystander cells

A subset of memory CD4⁺ T cells develops cytotoxic functionality during their differentiation from naïve CD4⁺ T cells following initial antigen-specific activation. We determined whether such functions could be activated by ANC28.1 in a bystander cell-killing assay, as described in Materials and Methods. We first used JCAM1.6 as target cells, a Jurkat-derived T cell line that is Lck-deficient and does not respond to ANC28.1. JCAM1.6 cells were labeled with a PKH dye and cocultured directly with memory CD4⁺ T cells in the presence of ANC28.1 or control CD28.2 for 24 h. When cells were examined for apoptosis by flow cytometry annexin V staining, the PKH-positive JCAM1.6 cells, but not the memory CD4⁺ T cells, stained for annexin V after coculture in the presence of ANC28.1 but not when incubated with the control CD28.2 (Fig. 3A). Levels of target cell annexin V positivity were typically at least 30% when E:T ratios of 10:1 were used, and target cell killing by memory CD4⁺ T cells was directly dependent on the E:T ratio (Fig. 3B). Additionally, target cell death was also observed if JCAM1.6 cells were cocultured with memory CD4⁺ T cells that were preactivated by CD3/CD28 costimulation (Fig. 3A). These data show that bystander cells are killed by memory CD4⁺ T cells after activation with ANC28.1.

Figure 3. Mitogenic CD28 induces bystander cell killing by human memory CD4⁺ T cells.

(A) JCAM1.6 target cells were preloaded with a cytoplasmic labeling dye, PKH26, and cocultured with purified memory CD4⁺ T cells in the presence of ANC28.1 for 24 h. Cell death of the PKH-positive target cells was determined by annexin V staining by flow cytometry. Shown is the gating scheme (upper histograms; untreated condition is shown) and representative flow cytometry histograms (lower histograms) presenting JCAM1.6 target cell annexin V staining after coculture with memory CD4⁺ T cells treated with control CD28.2, ANC28.1, or CD3/CD28 costimulation for 24 h. (B) JCAM1.6 target cell death at different E:T ratios with memory CD4⁺ T cells in the presence of ANC28.1 or control CD28.2. Error bars are means ± sd of one experiment representative of three separate experiments. (C) Cell death of JCAM1.6 target cells after coculture with memory CD4⁺ T cells in the presence of ANC28.1 and caspase-8/GrzB inhibitors Z-AAD-CMK or Z-IETD-FMK for 24 h. Cell death of target cells was determined by annexin V staining by flow cytometry. ***P < 0.001. Error bars are means ± sd of one experiment representative of three separate experiments.

To assess the role of GrzB in cell killing by memory CD4⁺ T cells after activation by ANC28.1, we used the cell-killing assay and measured apoptosis of JCAM1.6 target cells after coculture with memory CD4⁺ T cells in the presence of ANC28.1, with or without the caspase-8/GrzB inhibitor compound, Z-IETD-FMK. Another caspase/GrzB inhibitor, Z-AAD-CMK, was tested but was frequently too toxic in these experimental conditions, and remaining experiments used Z-IETD-FMK. At E:T ratios of 10:1, Z-IETD-FMK at 10 μM reduced JCAM1.6 apoptosis by approximately one-half, and protection of JCAM1.6 cells from GrzB-mediated apoptosis was nearly maximal with Z-IETD-FMK doses of up to 100 μM (Fig. 3C).

Enzymatic activity of GrzB from memory CD4⁺ T cells activated by ANC28.1

The previous data suggest that target cell death is mediated by GrzB, but we wanted to confirm that GrzB was indeed biologically active in target cells. The bioactivity of GrzB within target cells was examined using a flow cytometry-based assay that measures GrzB enzymatic activity via substrate cleavage-dependent fluorescence (GranToxiLux-FITC), as described in Materials and Methods (target cells are also labeled with CellVue Claret dye in this assay). After 24 h coculture of JCAM1.6 cells with memory CD4⁺ T cells in the presence of ANC28.1, up to 30% of JCAM1.6 cells were GranToxiLux-FITC-positive, demonstrating that GrzB was functionally active in target cells (Fig. 4A). Parallel examination of JCAM1.6 apoptosis via annexin V staining also confirmed concomitant cell death. Additionally, the inclusion of 100 μM Z-IETD-FMK (caspase-8/GrzB inhibitor) in this assay mitigated GrzB activity in JCAM1.6 cells. The inclusion of FasL-blocking antibodies did not affect GrzB activity in target cells but did partially prevent apoptosis (Fig. 4A).

Figure 4. Enzymatic activity of GrzB from memory CD4⁺ T cells activated by mitogenic CD28.

(A) Fluorogenic substrate-based detection of GrzB bioactivity in target cells by flow cytometry. JCAM1.6 target cells were labeled with CellVue Claret dye and cocultured with memory CD4⁺ T cells and GrzB substrate in the presence of ANC28.1, with or without caspase-8/GrzB inhibitor (Z-IETD-FMK) or FasL neutralizing antibodies for 24 h. GrzB activity was determined by measuring green fluorescence signal emitted by substrate (GranToxiLux-FITC) following cleavage by GrzB. Shown are representative flow cytometry dot plots (three separate experiments) of GranToxiLux positivity of CellVue Claret-labeled target cells. Annexin V staining on target cells was also examined in parallel with GrzB activity measurements. (B) K562 target cells were preloaded with PKH26 labeling dye and cocultured with purified memory CD4⁺ T cells in the presence of ANC28.1 or control CD28.2 for 24 h. Cell death and intracellular GrzB activity of K562 cells were determined by measuring annexin V and intracellular GrzB substrate cleavage (GranToxiLux assay) by flow cytometry. Shown are representative flow cytometry dot plots (three separate experiments) showing K562 annexin V staining (lower histograms) after coculture with memory CD4⁺ T cells in the presence of ANC28.1 or control CD28.2 after 24 h and GrzB activity in K562 cells via GranToxiLux measurement (upper histograms). FITC-A, FITC-area; APC-A, allophycocyanin-area. (C) Monolayer integrity of Caco-2 epithelial cells during coculture with memory CD4⁺ T cells in the presence of control CD28.2 or ANC28.1 (with or without caspase-8/GrzB inhibitor) after 96 h. After Caco-2 cells were grown to confluence, purified memory CD4⁺ T cells were added directly onto the monolayer with stimulants. After 4 days of culture, monolayer integrity was examined by measuring TER (left graph) and monolayer permeability to dextran-TRITC (right graph). *P < 0.05. Error bars are means ± sd of one experiment representative of two separate experiments.

In similar fashion to JCAM1.6 cells, GrzB activity and apoptosis were examined in K562 target cells after coculture with memory CD4⁺ T cells activated by ANC28.1. K562 cells are a lymphoblastic cell line derived from chronic myelogenous leukemia which does not constitutively express Fas and is thus resistant to Fas-mediated apoptosis, and like JCAM1.6 cells are not responsive to ANC28.1. After 24 h of coculture with memory CD4⁺ T cells in the presence of ANC28.1 (at 10:1 E:T ratios), annexin V levels on K562 cells consistently exceeded 20%, as compared with annexin V levels of <10% after coculture with untreated or control CD28.2-treated memory CD4⁺ T cells (Fig. 4B). As with JCAM1.6 cells, GrzB activity within K562 cells was confirmed by GranToxiLux-FITC positivity. Thus, GrzB-mediated cell death is responsible for bystander cell killing by memory CD4⁺ T cells activated by ANC28.1.

We further tested whether ANC28.1-activated memory CD4⁺ T cells could alter cell permeability on monolayers of Caco-2 epithelial cell lines. These cells are derived from epithelial colorectal adenocarcinoma and when cultured in transwell inserts serve as useful in vitro models of intestinal epithelial cell integrity and permeability [21]. Additionally, these cells do not express Fas [22]. As described in Materials and Methods, monolayer integrity and permeability are assessed by measuring TER and permeability to compounds of various sizes such as TRITC-labeled dextrans. Coculture of memory CD4⁺ T cells with confluent monolayers of Caco-2 cells with ANC28.1 (2 μg/ml) resulted in significant reductions of transepithelial resistance after 3–4 days compared with cells treated with control CD28.2 (Fig. 4C). The inclusion of caspase-8/GrzB inhibitor Z-IETD-FMK (10 μM) significantly mitigated the TER decrease induced by ANC28.1-activated memory CD4⁺ T cells. After 4 days of coculture, Caco-2 monolayer permeability to TRITC-labeled dextrans was also tested. Dextran-TRITC (1 mg/ml) was added to fixed monolayers for 4 h, after which, aliquots of medium from the lower chamber were removed and measured for dextran-TRITC amounts. In corroboration with monolayer TER decreases, dextran-TRITC levels in the lower chamber of Caco-2 transwell cultures contained significantly more dextran, indicating higher paracellular translocation of dextran after coculture with ANC28.1-activated memory CD4⁺ T cells compared with culture with control CD28.2. Additionally, the higher flux of dextran through Caco-2 monolayers cultured with ANC28.1-activated memory CD4⁺ T cells was reduced when the Z-IETD-FMK compound was included in cocultures. Gross morphological examination of monolayers by bright-field microscopy also showed obvious Caco-2 monolayer disruption (not shown). Assessment of GrzB activity within Caco-2 and another epithelial cell line, Madin-Darby canine kidney cells, was also attempted in shorter-term (1-day) coculture experiments using the GranToxiLux method, but results were not clear because of the adherent nature of the Caco-2 cells. These data show that in addition to death induction of lymphoid cells, ANC28.1-activated memory CD4⁺ T cells have detrimental effects on epithelial cell integrity partially mediated by GrzB.

Induction of GrzB from memory CD4⁺ T cells by ANC28.1 signals through PI3K and JNK pathways

The signaling events initiated by mitogenic CD28 mAb in human T cells, with respect to activation and proliferation, have been well-characterized, but we were intrigued by the GrzB production from memory CD4⁺ T cells induced by ANC28.1 and wanted to learn more about the signaling mechanisms involved. The GrzB promoter contains binding sites for a number of transcription factors, including NF-κB, AP-1/core-binding factor, Ikaros, and NF-AT, indicating that multiple signaling pathways are involved in GrzB induction [23–25]. We assessed signaling pathways involved in GrzB production from ANC28.1-activated memory CD4⁺ T cells using various pharmacological inhibitors. Memory CD4⁺ T cells were cultured with inhibitors for 1 h prior to ANC28.1 stimulation. Rapamycin, cyclosporin A, and ErkII inhibitor did not significantly block GrzB production from ANC28.1-stimulated memory CD4⁺ T cells after 24 h, whereas JNK and PI3K inhibitors (LY294002) reduced GrzB production (Fig. 5A), indicating GrzB production involves JNK, and not ERK, signaling activities. Consistent with GrzB production, perforin was also induced from memory CD4⁺ T cells by ANC28.1, and this perforin production was reduced by JNK inhibition, although perforin production was not reduced significantly by PI3K inhibition using LY294002 (Fig. 5A). Inhibition of GrzB production by JNK inhibitor, but not rapamycin or cyclosporin A, was also observed intracellularly by flow cytometry (Fig. 5B), and flow cytometry staining for intracellular GrzB showed that CD25 and CD69 expression was not reduced significantly in conjunction with inhibition of GrzB by JNK inhibitor, indicating divergent pathway events for T cell activation and GrzB induction after ANC28.1 stimulation. In summary, these data suggest that PI3K and JNK signaling is important for GrzB production from ANC28.1-activated memory CD4⁺ T cells.

Figure 5. Production of GrzB in memory CD4⁺ T cells by mitogenic CD28 signals through the PI3K and JNK pathways.

(A) Production of GrzB and perforin (ELISA) by memory CD4⁺ T cells activated with ANC28.1 for 24 h with various pathway inhibitors (LY294002, PI3K inhibitor). Error bars are means ± sd of one experiment representative of four separate experiments. *P < 0.05; **P < 0.01. (B) Representative flow cytometry histograms (four separate experiments) of intracellular GrzB in activated (CD69 and CD25) memory CD4⁺ T cells after treatment with ANC28.1 for 24 h in the presence of JNK inhibitor, rapamycin, or cyclosporin A.

Killing of bystander cells by memory CD4⁺ T cells activated by ANC28.1 is partially mediated by Fas/FasL

The mitogenic CD28 clone ANC28.1 used in this study and the similar clone TGN1412 induce TNF-α from CD4⁺ T cells, emphasizing the proinflammatory nature of ANC28.1 activation [14, 17]. Less explored have been the consequences for this TNF-α release with respect to bystander cell death, and we wondered if other TNF family members, such as FasL, are also involved. We examined cell-surface FasL expression by flow cytometry after stimulation with ANC28.1 or control CD28.2 for 24 h and found significant up-regulation of FasL on memory CD4⁺ T cells compared with control CD28.2 (Fig. 6A). We then cocultured memory CD4⁺ T cells and JCAM1.6 target cells at 10:1 ratios in the presence of ANC28.1 ± FasL-blocking antibodies for 24 h and observed that JCAM1.6 cell death was partially prevented using FasL-blocking antibodies at amounts up to 20 μg/ml (Fig. 6B). To explore the role of other TNF family members, we performed the bystander cell-killing assay using blocking antibodies against TNF family members that mediate cell death, such as TNF-α, TNF-β, or TRAIL. OPG is a TNF family member that acts to block TRAIL-mediated apoptosis of Jurkat T cells [26], and we also included recombinant OPG in the bystander cell-killing assay as another means to block potential TRAIL-induced apoptosis. Consistent with previous reports, TNF-α production was observed from memory CD4⁺ T cells after ANC28.1 stimulation (Fig. 6C). Despite the substantial levels of TNF-α release from memory CD4⁺ T cells, this proinflammatory cytokine was not involved in mediating cell death, as shown by the bystander cell-killing results nor, was cell death mediated by TNF-β or TRAIL (Fig. 6D). These data show that memory CD4⁺ T cells also use Fas/FasL mechanisms for target cell killing after activation by ANC28.1.

Figure 6. Cell death induced by mitogenic CD28-activated memory CD4⁺ T cells is partially mediated by FasL.

(A) Surface expression of FasL on memory CD4⁺ T cells after treatment with ANC28.1 for 24 h. Representative flow cytometry histogram overlay (four separate experiments) showing FasL expression on memory CD4⁺ T cells after treatment with ANC28.1 or control CD28.2. (B) Cell death of JCAM1.6 target cells (measured by annexin V staining) after coculture with memory CD4⁺ T cells in the presence of ANC28.1 and increasing amounts FasL-neutralizing mAb for 24 h. ***P < 0.001. Error bars are means ± sd of one experiment representative of three separate experiments. (C) Production of TNF-α from memory CD4⁺ T cells after 24 h stimulation with ANC28.1. **P < 0.01; ***P < 0.001. Error bars are means ± sd of one experiment representative of three separate experiments. (D) Cell death of JCAM1.6 target cells after coculture with memory CD4⁺ T cells in the presence of ANC28.1 and neutralizing antibodies or inhibitors against TNF family members for 24 h. ***P < 0.001. Error bars are means ± sd of one experiment representative of three separate experiments.

DISCUSSION

The present study shows that mitogenic CD28 mAb induce human memory CD4⁺ T cells to kill bystander cells or disrupt epithelial cell integrity and demonstrates CD4⁺ T cell functions previously unknown to be induced by mitogenic CD28. In these in vitro conditions, significant levels of target cell killing occurred within 24 h and involved the two main mechanisms of target cell elimination used by CTLs, granzyme B and Fas/FasL. Previous reports of mitogenic CD28 have focused mainly on aspects of CD4⁺ T cell activation, proliferation, and Treg functionality that are affected by mitogenic CD28, but the present report shows that CTL functions of CD4⁺ T cells are also elicited by mitogenic CD28. Furthermore, these findings were observed using mitogenic CD28 clone 5D10, a clone very similar to clone TGN1412 in its T cell activation capacity used in the clinical trial. The adverse effects in the trial subjects were thought to be consequences of a proinflammatory “cytokine storm” that ensued immediately following TGN1412 infusion [11], but the present experiments suggest that other detrimental mechanisms could have contributed to the widespread tissue damage. These findings highlight pathogenic functions of activated memory CD4⁺ T cells that should be considered before other such antibodies are tested.

The activation of memory CD4⁺ T cells by ANC28.1 (Fig. 1) shows strong consistency with other reports of T cell activation by mitogenic CD28 mAb but surprising are the potent CTL mechanisms induced by ANC28.1. CD8⁺ CTLs and NK cells use GrzB and FasL to kill target cells, and herein, we show that CD4⁺ T cells also use these mechanisms after activation by mitogenic CD28 mAb. Within 24 h, a significant portion of JCAM1.6 or K562 target cells was apoptotic when directly exposed to ANC28.1-activated memory CD4⁺ T cells (Figs. 3 and 4), and these activities may have been a factor in the severe lymphopenia that occurred during the TGN1412 trial. Also noteworthy is that target cell killing also occurred after conventional costimulation of memory CD4⁺ T cells using CD3/CD28 mAb, indicating that mitogenic CD28 and conventional TCR stimulation can induce pathogenic functions. Cytolytic CD4⁺ T cells (CD4⁺ CTLs) have important roles as direct mediators against infectious disease and for anti-tumor protection and are terminally differentiated effector CD4⁺ T cells that kill target cells via GrzB/perforin and Fas/FasL, closely resembling CD8⁺ CTL and NK cell mechanisms of target cell killing (for review, see ref. [27]). However, CD4⁺ CTLs typically express CD28 at very low levels [28], suggesting that these cells would be unresponsive to mitogenic CD28. We think these CD28-negative CD4⁺ T cells represent a late-stage phenotype and that other CD4⁺CD45RO⁺ T cells that still express CD28 manifest a cytotoxic phenotype in response to mitogenic CD28. By comparison, in our previous study, we showed that the “Tregs” responding to ANC28.1 were CD25-expressing but were not the CD25^bright-expressing Treg cells (which did not respond to ANC28.1) when more stringent purification of CD25^bright Tregs was performed [18].

The present experiments demonstrate that memory CD4⁺ T cell CTL functions are activated by ANC28.1, but the levels of GrzB induced by ANC28.1, as well as by costimulation with CD3/CD28 mAb (Fig. 2), are surprising and have important implications in other physiological contexts where CD4⁺ T cells are pathogenic. Comparison of purified memory CD4⁺ and memory CD8⁺ T cells shows that activated memory CD4⁺ T cells produce GrzB just as robustly as CD8⁺ T cells (Fig. 2B). In addition, the flow cytometry data show basal intracellular levels of GrzB to be very low in memory CD4⁺ T cells compared with CD8⁺ T cells (Fig. 2A; untreated and control CD28.2), implying that GrzB expression and storage differ between CD4⁺ and CD8⁺ T cells. This pattern of intracellular GrzB expression, in light of the extracellular levels of GrzB in which extracellular GrzB production from memory CD4⁺ T cells was comparable with memory CD8⁺ T cells, suggests that activated CD4⁺ T cells may produce more GrzB at a single-cell level compared with CD8⁺ T cells. Additionally, the enzymatic activity and apoptosis-inducing function of GrzB from activated memory CD4⁺ T cells were confirmed in lymphoid target cells, such as JCAM1.6 and K562 (Fig. 4). Activation of subsets of human CD4⁺ T cells, including memory CD4⁺CD45RO⁺ T cells, Th1/Th2, and Tregs, using CD3/CD28 mAb, has been reported to induce GrzB production [19, 29, 30]. We also observed strong activation and high production of GrzB from purified human Th1 cells when stimulated with ANC28.1 or CD3/CD28 costimulation (data not shown). GrzB is important for various cell-killing effector functions of memory CD4⁺ T cells, such as Th1/Th2, Tregs, or Th17 cells. For example, one mechanism used by CD4⁺CD25⁺ Tregs to suppress effector T cells in mice is via GrzB-mediated apoptosis, which is perforin-independent [29]. Interestingly, a more recent study using human cells showed that activated responder (non-Treg) CD4⁺ T cells resist suppression by HLA-DR⁺ Tregs by killing these Tregs via GrzB [30]. Murine Th1 and Th2 cells both express GrzB, but unlike Th1 cells, the Th2 cells are protected from GrzB-mediated activation-induced cell death by enhanced responsiveness to vasoactive intestinal peptide, which results in down-regulation of GrzB expression in Th2 cells [31]. Human Th17 cells mediate neuronal cell death via GrzB following transmigration through the blood-brain barrier [32]. GrzB is also active extracellularly based on histological analyses of biopsies from skin diseases, such as psoriasis and atopic dermatitis, showing GrzB in CD4⁺ and CD8⁺ T cells [33, 34].

Based on our killing assay experiments using inhibitors of TNF family members, such as FasL, TNF-α, TNF-β, or TRAIL, ANC28.1-activated memory CD4⁺ T cells kill bystander cells (JCAM1.6) using FasL, consistent with the increased surface expression of FasL after ANC28.1 activation (Fig. 6A). However, the pathologic effects of ANC28.1-activated memory CD4⁺ T cells were not restricted to just killing of bystander lymphoid cells via GrzB and FasL. As shown in Fig. 4C, disruption of Caco-2 epithelial cell integrity was observed, although these effects usually occurred more slowly (after 1–2 days of coculture with activated memory CD4⁺ T cells). We think this is likely a result of the protease function of GrzB, not the caspase function. Additionally, the partial mitigation of the detrimental effects (decreased TER and increased dextran permeability) induced by ANC28.1-activated memory CD4⁺ T cells with a caspase-8/GrzB inhibitor indicates that other mechanisms are also involved in monolayer disruption. However, as Fas is not expressed on Caco-2 cells [22], disruption of monolayer integrity as a result of Fas/FasL interactions seems unlikely. But consistent with previous reports, ANC28.1 induced high amounts of TNF-α production from memory CD4⁺ T cells (Fig. 6C), and this TNF-α may contribute to monolayer disruption, as TNF-α can disrupt Caco-2 monolayer integrity by reducing protein levels of the occludin tight junction protein [35].

Based on the present experiments using pharmacological inhibitors (Fig. 5), PI3K is an important signaling pathway involved in activation of human T cells by ANC28.1 and is also important for GrzB induction. But also apparently important for GrzB induction in memory CD4⁺ T cells by ANC28.1 is the JNK pathway. T cell activation, as determined by production of IL-2 and CD25 and CD69 expression are well-established features of ANC28.1-stimulated human T cells, and previous work in our lab showed that PI3K is critical for ANC28.1 activation of human T cells, as the PI3K inhibitor, LY294002, abrogated CD69 and CD25 expression in a dose-dependent manner and also prevented Treg expansion [18]. A report by Waibler et al. [14] using this same clone (5D10), as well as clone TGN1412, also implicated PI3K signaling by showing that IL-2 production induced by ANC28.1 in human T cells could be mitigated by LY294002. IL-2 production was also inhibited by other inhibitory compounds, such as Wortmannin, EGTA, cycloheximide, CsA, and Src kinase inhibitor (PP2). Also in line with our JNK inhibitor experiments, the report by Waibler et al. [14] demonstrated strong activation of multiple signaling pathways, including JNK, ErkII, p38, AKT, NF-ATc1, GSK3β-1, and IκBα, and these events led to sustained calcium flux in the activated T cells (although one technical difference that should be noted is that these effects were observed by first cross-linking TGN1412 using goat anti-mouse antibodies, whereas in our experiments, T cells were incubated longer-term with soluble ANC28.1 with no cross-linking). Lastly, in light of our intracellular GrzB flow cytometry data (Fig. 2B), which showed low basal levels of GrzB in memory CD4⁺ T cells compared with CD8⁺ T cells, it would be interesting to know if PI3K and JNK are important specifically for the release of GrzB from cells containing preformed GrzB, such as CD8⁺ T cells or NK cells. Thus, mitogenic CD28 activates many signaling events in T cells, but PI3K and JNK pathways appear to have important roles in GrzB production.

In summary, mitogenic CD28 mAb activates human memory CD4⁺ T cells to kill bystander cells via GrzB and FasL, and additional pathogenic mechanisms as a result of this form of activation may involve TNF-α. Studies of mitogenic CD28 mAb have focused mainly on Tregs and CD4⁺ T cells, but other cells expressing CD28 have the potential to respond to this form of stimulation. We also observed (but not presented in this report) activation, proliferation, and GrzB-mediated target cell killing by purified human CD3⁺CD8⁺ and memory CD8⁺CD45RO⁺ T cells after ANC28.1 activation comparable with levels induced by conventional costimulation by CD3/CD28 mAb, but these findings require further investigation. NK cells also express CD28 and have potential to respond to mitogenic CD28, but we did not test this, although CD28 agonism alone has been shown to activate NK cells to kill target cells [36]. Lastly, human peripheral blood monocytes and granulocytes express CD28 at low levels but are still capable of responding to ANC28.1 stimulation by producing inflammatory cytokines, such as TNF-α [37]. Physiological correlates of mitogenic CD28 stimulation are more difficult to address, but important information with respect to CD28 signaling and T cell activation of memory CD4⁺ T cells is gained from the present study. Effector memory CD4⁺ T cells routinely traffic to multiple tissues, and the nonspecific or uncontrolled activation of these cells suggests that they may be important mediators of tissue damage, as summarized in Fig. 7, but the impact of these mechanisms in the multiorgan failure that occurred in the TGN1412 recipients is not knowable. These data emphasize the diversity of CD4⁺ T cell functions that occur after CD4⁺ T cell activation which could be associated with CD4⁺ T cell-mediated pathology.

Figure 7. Potential multiple mechanisms of tissue destruction mediated by activated memory CD4⁺ T cells.

Tissue-resident memory CD4⁺ T cells activated by mitogenic CD28 may result in local cell death and tissue degradation mediated by GrzB, Fas/FasL, or release of proinflammatory cytokines. Pf, perforin.

AUTHORSHIP

M.A.M. designed and performed the majority of the research and wrote the manuscript. J.C. performed experiments and wrote the manuscript. A.E.M. and M.T. performed experiments. M.L.F., X.Y., and C.A.K. performed statistical analyses. A.F.O. and A.T.H. assisted with data analysis and critiqued the manuscript. J.R.R. assisted with experimental design and data analyses and critiqued the manuscript. T.C.S. performed experiments and critiqued the manuscript. D.E.L. designed research, supervised the project, and reviewed the manuscript.

DISCLOSURES

The authors declare no conflict of interest.