Leukemic B Cell CTLA-4 Suppresses Costimulation of T Cells

This article is featured in In This Issue, p.2515

By acting as a dimmer switch to balance T cell activation and inhibition, CTL Ag 4 (CTLA-4) critically mediates T cell responses (1–3). Clinically targeted through blockade of CTLA-4 as an immunostimulatory treatment for cancer (ipilimumab) and administration of soluble CTLA-4 (abatacept) to enhance immunosuppression as a treatment for autoimmunity, this protein’s biology has become a major area of investigation in broadly applicable settings (4–6). The last two decades have shown intense study of CTLA-4 as a T cell protein despite multiple reports that CTLA-4 is expressed by both normal B cells and select tumor types (7–9). In fact, expression of CTLA-4 on T cells is not currently used as a prognostic indicator for response to ipilimumab treatment, likely owing to the complex biology of CTLA-4 but additionally to unknown contributions of CTLA-4 expressed by the tumor or alternative immune effector cells. The consequences of non–T cell CTLA-4 are poorly defined.

Genetically engineered CTLA-4 loss in mouse models has demonstrated its extreme potency as a negative regulator of T cell immunity (10, 11). Loss of CTLA-4 leads to lymphoproliferative-like T cell disease, immune-mediated destruction of multiple organs, and premature death of the mice at 3–4 weeks old. Conditional knockout (KO) of CTLA-4 on T regulatory cells (Tregs) is sufficient to mimic lymphoproliferative-like T cell disease, immune-mediated organ destruction, and early death of mice, although delayed to 8 weeks old (12). In humans, Tregs comprise a rare portion of CD4⁺ T lymphocytes (∼1–2%), CD4⁺ T cells account for ∼50% of lymphocytes, and lymphocytes account for ∼23–33% of leukocytes (13, 14). The potency of CTLA-4 on this rare subset of T cells points to the importance of studying the function of CTLA-4 on tumor cells, which would largely outnumber the levels of Tregs.

Although no direct signaling pathway for CTLA-4 exists, it is heavily supported that CTLA-4 antagonizes costimulation of naive T cells to effector T cells by competing with the CD28 costimulatory molecule on T cells (15–17). This function occurs via downmodulation of the shared T cell costimulatory ligands CD80 and CD86. Downmodulation is regulated via CTLA-4 binding and/or removal from the cell surface through trans-endocytosis. Lack of CD80/CD86 renders APCs less capable of activating the CD28 T cell costimulatory pathway. Recently, tumor-expressed CTLA-4 was reported to associate with worse prognosis in nasopharyngeal carcinoma and shorter overall survival esophageal carcinoma (18, 19). In chronic lymphocytic leukemia (CLL), CTLA-4 is expressed by tumor cells and is positively correlated with stages indicative of progressive disease (20). A comparison of autologous peripheral blood, lymph nodes, and bone marrow identified CTLA-4 (in tumor B cells) as part of a gene signature that was associated with disease progression (21). CTLA-4 was upregulated in the lymph nodes, a site of leukemic B cell proliferation. Because of CTLA-4’s capacity as an extrinsic regulator and potent influence on skewing immune homeostasis, we hypothesized that tumor cell–expressed CTLA-4 is capable of potentiating immunosuppression. Indeed, immunosuppression is a major contributor to the morbidity and mortality of CLL.

It has yet to be rigorously demonstrated that CTLA-4 on tumor cells has relevance to the disease, immunosuppression, or perhaps to therapy. Using CLL as a model disease, we show that leukemic B cell–expressed CTLA-4 is able to decrease CD80 from neighboring cells, leading to suboptimal costimulation and reduced IL-2 production in vitro. Blockade of tumor CTLA-4 in vivo is capable of reducing leukemic burden.

Patient blood was obtained in ACD tubes at The Ohio State University with consent and in accordance with the Declaration of Helsinki. B and T cells were negatively selected using RosetteSep (STEMCELL Technologies) and Ficoll. The Mec1 cell line was obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen and the OSU-CLL cell line from The Ohio State University (22, 23). Except for where directly indicated that cells had been frozen, all cells used were freshly isolated. Normal donor cells were collected using the same methods as patient cells from fresh blood (volunteers or American Red Cross). Mec1 and OSU-CLL were maintained in RPMI 1640 (10% FBS + 56 U/ml penicillin + 56 μg/ml streptomycin + 2 mM l-glutamine). HEK293 (American Type Culture Collection) and Phoenix-AMPHO (Orbigen) cells were maintained in DMEM (10% FBS + 56 U/ml penicillin + 56 μg/ml streptomycin + 2 mM l-glutamine).

RNA was isolated using Trizol (Invitrogen), alcohol precipitation, and column purification (Qiagen). cDNA was prepared using random hexamers and MMLV reverse transcriptase (Invitrogen). Taqman assays were used for real-time quantitative PCR (RT-qPCR) (Applied Biosystems).

The pRetro-tight-pur system was used to produce dox-inducible CTLA-4 or empty vector B cell lines (Clontech Laboratories). Full-length CTLA-4 cDNA (sequence NM_005214.3) was obtained from OriGene, restriction digested with NotI, and ligated into pRetro. The CTLA-4pRetro or empty vector retroviral plasmids were packaged by Phoenix cells, supernatant collected, and 0.45 μm filtered. Tet⁺ Mec1 and OSU-CLL cell lines were infected with CTLA-4pRetro or empty vector virus and selected using 1 μg/ml puromycin + 500 μg/ml G418. CD80-GFP and CD86-GFP plasmids were obtained from OriGene and stably transfected into HEK293 cells using calcium phosphate (Promega), and selected with 500 μg/ml G418. Full-length CTLA-4, CD80, and CD86 sequence inserts were all validated by Sanger Sequencing at The Ohio State University Nucleic Acid Shared Resource Core facility. Primers for sequencing were as follows: VP1.5 forward: 5′-GGACTTTCCAAAATGTCG-3′, XL39 reverse: 5′-ATTAGGACAAGGCTGGTGGG-3′, Retro forward: 5′-ATTAGGACAAGGCTGGTGGG-3′, 5′-ATCTGAGGCCCTTTCGTCTTCACTC-3′, Retro reverse: 5′-TGTGTGCGAGGCCAGAGGCCACTT-3′, Nested CTLA-4 forward: 5′-GACCTGAACACCGCTCCCATAAAGC-3′, Nested CD86GFP forward: 5′-GCCTCCCCCAGACCACAT-3′, and Nested CD86GFP reverse: 5-′GGTGCTCTTCATCTT GTTGGTCAT-3′.

Anti-human Abs CTLA-4 (clone BNI3; PE, allophycocyanin, or BV421), CD80 (clone L307.4; FITC, PE, or V450), CD86 (clone 2331/FUN-1; PE or PerCP-Cy5.5), CD69 (clone FN50-V450; TP1.55.3-PE), CD19 (clone HIB19; FITC or Alexa Fluor 647), CD5 (clone UCHT2; allophycocyanin), CD3 (clone UCHT1; ECD or Alexa Fluor 700), and isotype controls (PE or allophycocyanin) were obtained from BD Biosciences, BioLegend, and Beckman Coulter. Violet and Near-IR LIVE/DEAD stains (Life Technologies) and claret membrane dye (Sigma-Aldrich) were used for flow cytometry. Anti–murine CTLA-4 (mCTLA-4; clone UC10-4F10-11-PE), CD19 (clone 1D3-Alexa Fluor 647), and CD5 (clone 53-7.3; FITC or BV421) and human or murine Fc block were purchased from BD Biosciences. Cells were surface stained in flow buffer (5% FBS, 0.1% NaN₃) and fixed and permeabilized for intracellular staining using BD Cytofix/Cytoperm. Intracellular stains were in BD Perm/Wash buffer. T cells were stimulated with 10 μg/ml plate-bound anti-CD3 (eBioscience) ± 1 μg/ml soluble anti-CD28 (eBioscience) or 1:1 beads:T cells anti-CD3/CD28 dynabeads (Life Technologies). Ipilimumab was obtained from the Ohio State University Pharmacy.

Cells were analyzed on an FC500 (Beckman Coulter), Gallios (Beckman Coulter), or LSRFortessa (BD Biosciences). Adherent cells were removed from the plate using Accutase (Life Technologies). Dynabeads were removed using a dynabead magnet and washed 1× with PBS. Briefly, cells were surfaced stained for 15–20 min at room temperature or on ice, respectively, in either PBS or flow buffer (5% FBS + 0.1% NaN₃) depending on the stains used. When applicable, surface staining was followed by 20 min fixation and permeabilization (BD Cytofix/Cytoperm) on ice and 30 min intracellular staining in BD Perm/Wash buffer. Mouse peripheral blood analysis was performed by whole blood staining for 15 min at 4°C, RBCs lysed (eBioscience), no wash, and CountBrite beads (Life Technologies) added prior to obtaining absolute lymphocyte counts.

Cells were plated at a 1:1 ratio of B:T cells (except in autologous experiments, 1:1–2.5:1 B:T) and at 3–5 ×10⁶ cells/ml. Surface CTLA-4 expression was determined by flow cytometry at 48 h. For Mec1/T cell cocultures, Mec1 cells were treated ± doxycycline and ± 10 μg/ml ipilimumab for 24 h and washed 2× prior to coculture at a 1:1 ratio of Mec1:T cells. Cells were gated on live/lymphocytes by foward scatter/side scatter/CD19⁺/CD5⁺ or Singlets/live/Claret⁺ for CLL B cells. Where cytokines were assessed, supernatant was collected at 48 h and frozen at −80°C. Supernatants were assessed for cytokines using the Th1/Th2/Th17 human cytokine bead array (CBA; BD Biosciences) and analyzed using a five-parameter logistic curve using FCAP Array software (Soft Flow). Purity after cell isolation was assessed via flow cytometry prior to all cocultures and staining B cells with Claret membrane dye (Sigma-Aldrich).

For cell lines, 1 × 10⁶ cells/ml Mec1 were treated ±1 μg/ml Dox and ± ipilimumab for 24 h and washed, stained for surface CTLA-4 expression, and measured by flow cytometry, and resuspended at 1 × 10⁶ cells/ml prior to 4 h coculture with CD80-GFP HEK293 or CD86-GFP HEK293 stably transfected cell lines. HEK293 cells were plated overnight in 100-mm culture dishes at 50% confluency before coculture. Mec1 and HEK293 were cocultured for 4 h and stained for flow cytometry. For primary cells, previously RosetteSep-isolated and cryo-preserved CLL B cells were cocultured with allogeneic T cells plus anti-CD3/anti-CD28 activating dynabeads for 48 h. Following coculture with GFP⁺ HEK293 cell lines, lymphocytes were stained and GFP transfer measured by flow cytometry. Primary CLL B cells were assessed for surface CTLA-4 expression by flow cytometry and resuspended at 3 × 10⁶ cells/ml prior to 18 h coculture with CD80-GFP HEK293 or CD86-GFP HEK293 stably transfected cell lines. Surface expression of CD80 and CD86 and coexpression of GFP by HEK293 cells were measured as controls for each assay.

Protein binding aldehyde/sulfate latex beads (Life Technologies) were coated with 1 μg murine recombinant his-tagged CTLA-4 or 1 μg human recombinant his-tagged CTLA-4 (Life Technologies). Protein was detected using commercial flow Abs compared with isotype controls for murine or human CTLA-4 (hCTLA-4; BD Biosciences). In vivo mAb to mCTLA-4 (clone 9d9; Bio X Cell) compared with isotype control (MPC-11; Bio X Cell) was detected for specificity to mCTLA-4 and no cross-reactivity to hCTLA-4 using a secondary anti-mouse IgG2b APC Ab (Abcam) and analysis by flow cytometry.

All animal studies were approved by the Institutional Animal Use and Care Committee at the Ohio State University. Eμ-TCL1 mice were received from Dr. C. Croce and backcrossed to C57BL/6 (24, 25). C57BL/6 hCTLA-4 knock-in mice were received from Y. Liu (Children’s Hospital, Washington, DC) (26). FcRγ KO mice were received from Jeanette Leusen (University Medical Center, the Netherlands) (27). C57BL/6N hCTLA-4^+/+ mice were bred with C57BL/6N FcRγ^−/− mice to produce hCTLA-4^+/− FcRγ^+/− mice. Heterozygous mice were crossed to produce CTLA-4^+/+ FcRγ^−/− mice. A total of 1 × 10⁷ TCL1 cells was engrafted by tail vein injection. Anti–CTLA-4 (9D9; Bio X Cell) or isotype control (MPC-11; Bio X Cell) were administered via i.p. injection at 100 μg/dose on days 0, 3, 6, 9, and 12 post-leukemia diagnosis (previously defined as ≥20% CD19⁺ CD5⁺ of CD45⁺ cells) in a blinded and randomized manner (28). Peripheral blood was collected by submandibular bleed on a weekly basis. Mice were euthanized based on standard early removal criteria (respiratory distress, rough coat, weight loss, lethargy, etc.) as stated in an approved protocol by The Ohio State University Institutional Animal Care and Use Committee. Genotypes of all mice were confirmed by PCR using the following primers 5′ > 3′: hCTLA-4 forward (5′-CAC CAA TGT TGG GGA GTA G-3′), mCTLA-4 forward (5′-CTT GTC CCT TTG ATG GCA CT-3′), common CTLA-4 reverse (5′-GGT TCT GGA TCT GCA ACA GAA-3′), Neo forward (5′-AAG ATG GAT TGC ACG CAG GT-3′), Neo reverse (5′-TCG ATG CGA TGT TTC GCT TG-3′), mFcRγ forward (5′-GCC CTT CCC TTC CCT CTA CA-3′), mFcRγ reverse (5′-CCT TCA GAC CAT GGG GAA CC-3′), hFcRγ forward (5′-CCA GTT CCA GAG ACC TGA GC-3′), and hFcRγ reverse (5′-CAC CAA CAC ACA CAC ACC AA-3′).

For in vitro studies, data were analyzed by mixed-effect model, incorporating repeated measures for each experiment. For in vivo lymphocyte count studies, data were analyzed by mixed-effect model and repeated measures were incorporated for each subject, followed by comparisons between groups at each time point. For in vivo survival, Kaplan–Meier estimates of the survival function were generated for both groups, and the log-rank test was used to compare the curves.

CLL has profound immunosuppression as a central component of its pathogenesis. We initially noted by microarray that CTLA-4 expression in CLL B cells (pool of n = 5) as compared with normal B cells (pool of n = 6) was the fifth most differentially expressed gene, with average 19-fold change over normal B cells (data not shown) (29). We validated this observation by examining transcript expression in CLL and normal B cell samples by RT-qPCR and the comparative C_T method (Fig. 1A) (30). CTLA-4 was abnormally, constitutively expressed in CLL B cells, whereas normal B cells did not express transcript (−ΔCt −2.774 CLL versus −12.03 NB, p < 0.0001). We examined protein expression by flow cytometry and identified 17/28 (61%; ≥10% CTLA-4⁺) patients to have constitutive intracellular expression of CTLA-4 ranging from 0.15 to 96.34% positive cells compared with isotype control and significantly upregulated on average compared with normal B cells (p < 0.0001). All patients (n = 27) were negative for surface expression of CTLA-4 (Fig. 1B, 1C). For 14 patients, mRNA and intracellular protein expression were available on the same patient. For these patient samples, all patients who tested positive for CTLA-4 protein by flow cytometry also had detetable CTLA-4 transcript by RT-qPCR, although no correlation between mRNA and positive intracellular protein expression of CTLA-4 was observed (Spearman correlation coefficient of −0.06, p = 0.8398; data not shown). It is likely that multiple isoforms of CTLA-4 are detected transcriptionally that may cause discordance with protein expression. Both soluble and full-length CTLA-4 transcripts are detectable with the primer/probe set used. Whereas both isoforms are of interest and potential functional consequence, the focus of this study is on full-length and not soluble CTLA-4. Confocal microscopy was performed to assess the cellular location of CTLA-4, and we observed CTLA-4 to be expressed in a punctate pattern within the cytoplasm, but absent from the nucleus (Fig. 1D). In agreement with the protein expression by flow cytometry, CTLA-4 was heterogeneously expressed and mimicked known expression patterns in T cells, in which activated T cells express variable levels of CTLA-4 as opposed to uniform expression. The Jurkat T cell line was negative for CTLA-4, consistent with known literature, and was used as a control (31, 32). Similar to human CLL, we examined primary CLL tumors from the TCL1 transgenic mouse. Interestingly, we detected surface CTLA-4 protein expression on CD19⁺ CD5⁺ cells compared with both normal CD19⁺ CD5⁻ B cells from the same mice and compared with an isotype control (n = 5, Supplemental Fig. 1), consistent with supplementary findings in TCL1 control mice on B220^low CD19⁺ cells in prior literature (33). Contrasting with both primary human and murine CLL, EBV-transformed cell lines (Mec1, OSU-CLL, OSU-NB, Ramos, Raji, and 697) lacked expression of CTLA-4 at both mRNA and protein levels (data not shown).

There is currently no known role for intracellular CTLA-4 in any cell type, including in CLL or other cancer cells, prompting us to hypothesize that this protein could be expressed on the cell surface. We used multiple B cell–activating factors (CpG, PMA/ionomycin, anti-IgM, IL-4, CD40L, IL-4+CD40L, LPS, IL-4+LPS) to mimic activation-induced expression of CTLA-4 on T cells, but found that activating CLL B cells did not result in surface localization (data not shown). In contrast, cultures with activated T cells or activated T cell membranes have been published to result in detection of surface CTLA-4 on normal B cells, and we found that this was also true with CLL B cells (7, 9). Coculture with activated T cells (anti-CD3/CD28 stimulated) for 48 h induced surface expression of CTLA-4 on tumor CLL cells ≥5% CTLA-4⁺ in nine patient samples and <5% CTLA-4⁺ in three samples in which CLL only versus CLL+T were not significantly different in CTLA-4 positivity, p = 0.5664, and CLL+T without stimulation versus CLL+T+anti-CD3/anti-CD28 stimulation was significantly different at p < 0.0001 in either allogeneic or autologous coculture conditions (Fig. 2A, 2B). Additionally, this expression required cell–cell contact when CTLA-4 was detectable with activated T cell coculture (n = 6, p = 0.0233), as determined by trans-well experiments (Fig. 2C). Whereas induced surface CTLA-4 expression appears relatively modest for most samples tested, comparatively, surface CTLA-4 on T cells is known to be limited and highly regulated in comparison with intracellular stores, for example, with reported surface expression on murine T cells after activation approximated at 9–22% positivity (34–38). With a large ratio of leukemic cells to T cells, this suggests a non-negligible effect of leukemic cell CTLA-4, even at low levels, if tumor cell CTLA-4 has functional capacity.

To determine if tumor cell expressed CTLA-4 is functional, we established two CLL-derived leukemic B cell lines (Mec1 and OSU-CLL) to express doxycycline-inducible CTLA-4 and empty vector controls to further our studies (Fig. 3A, data not shown). Mec1 and OSU-CLL are negative for CTLA-4 and express high levels of the ligands for CTLA-4, CD80, and CD86, making them appropriate models for studying CTLA-4 function in the context of cognate ligand. Known leakiness of the pRetro system resulted in detectable transcript levels of CTLA-4 without doxycycline induction of the CTLA-4pRetro Mec1 cells; however, this baseline level of transcript led to minimal intracellular protein and no detectable surface CTLA-4 protein. Upon doxycycline induction of CTLA-4, CTLA-4 is surface expressed on these cell lines without the help of activated T cell coculture. Both of these CTLA-4–expressing cell lines are able to bind the soluble cognate ligand, CD80, when CTLA-4 is expressed on the surface (data not shown). We found that in the condition in which surface CTLA-4 was expressed compared with the doxycycline-uninduced cell line or the empty vector control± doxycycline, CD80 was consistently downmodulated on the Mec1 cell line by an average change in mean fluorescence intensity of 10.5, p < 0.0001, and that this phenotype was reproducible in the OSU-CLL cell lines, as well (Fig. 3B, Supplemental Fig. 2). Blockade of CTLA-4 on the Mec1 cell line with therapeutic reagent, ipilimumab, restored CD80 expression with a significant increase in average mean fluorescence intensity from 6.873 up to 13.987, p < 0.0001 (Fig. 3B). We used RT-qPCR to assess if transcriptional changes in CD80 were responsible for changes in protein expression, but there were no significant differences between CTLA-4⁺ Mec1 versus any of the controls (Fig. 3C). This finding suggests that CTLA-4 may have function on tumor cells similarly to known blockade and/or costimulatory protein downmodulatory function on T cells.

We next checked the reported mechanism by which CTLA-4 on T cells is able to downmodulate CD80 and/or CD86: trans-endocytosis (17). We stably transfected HEK293 cells to express C-terminal tagged CD80-GFP or CD86-GFP and cocultured these cells with the CTLA-4⁺ doxycycline-inducible Mec1 cell line and measured uptake of GFP into the CTLA-4⁺ cell line (Fig. 4A, 4B, data not shown). Using the Mec1 CTLA-4⁺ or empty vector cell lines ± dox, we determined that only the surface-expressing CTLA-4⁺ cell line was able to receive CD80-GFP from the HEK293 cells after 4 h of coculture. We extended these studies to CLL cells cocultured with the CD80-GFP or CD86-GFP HEK293 cells and again observed significant transfer of CD80-GFP by flow (average uptake of 27.78% GFP⁺ compared with cocultured cells in the absence of CD80-GFP HEK293 cells) into primary CLL cells (n = 10, p < 0.0001) and T cells shown as an assay control. Although not statistically significant (p = 0.2158), the transfer of CD80-GFP tended to decrease (6.06% drop in GFP⁺ cells) in a CTLA-4–dependent manner as evidenced when cocultured in the presence of a CTLA-4–blocking Ab, ipilimumab (Fig. 4C–E). Within the same assay, this result was comparable to blockade of CTLA-4 on T cells and subsequent transfer of GFP (Fig. 4E).

Because CD80 and CD86 are necessary for T cell costimulation, we subsequently studied the consequences of CD80 downmodulation by tumor-expressed CTLA-4 in an MLR. We cocultured CTLA-4⁺ Mec1 cells and allogeneic primary T cells from normal donors. Normal donor T cells were activated in a coculture with Mec1 cells as detected by upregulation of CD69 at 48 h (data not shown). Supernatant was collected at 48 h of coculture and IL-2 measured by flow cytometry using a CBA. IL-2, likely produced by the allogeneic T cells (n = 3), was significantly decreased with tumor CTLA-4 expression at p = 0.0172 (Fig. 5A). We noted that the MLR with the Mec1 CTLA-4 vector led to more IL-2 production than the empty vector Mec1 cell line. This effect is likely due to differences in the resulting clones postselection, and as a result, we compared the most relevant conditions (±dox within one cell line) and include the empty vector line as a control to rule out the effect of doxycycline. Anti–CTLA-4 pretreatment of the Mec1 cell line concurrently with doxycycline induction partially rescues the downmodulation of CD80 and, subsequently, downmodulation of IL-2 in the coculture system (n = 3, IL-2 is significantly downmodulated at p < 0.05 and is not significantly downmodulated with ipilimumab treatment, p = 0.3908) (Fig. 5B). Pretreatment with anti–CTLA-4 Ab was washed out of culture prior to coculture with T cells to negate direct effects of CTLA-4 blockade on T cells.

Based on our initial detection of constitutive surface CTLA-4 expression on leukemic cells in the Eμ-TCL1 transgenic model of CLL, we produced a murine model to study whether blockade of tumor-expressed CTLA-4 (without blocking T cell CTLA-4) could impact leukemic progression in vivo (24, 25). To isolate the effect of targeting only tumor-expressed CTLA-4, we adoptively transferred mCTLA-4–positive leukemic B cells into humanized CTLA-4 knock-in mice, which instead express hCTLA-4 on their T cells. This permitted the study of tumor-expressed CTLA-4 without compromising the effect of T cell–expressed CTLA-4. We confirmed that a reported mouse anti–mCTLA-4 clone, 9d9, versus its isotype control was reactive to mCTLA-4 and lacked cross-reactivity to hCTLA-4 using recombinant protein bound to beads (Supplemental Fig. 3A). In addition, we acquired a hCTLA-4 knock-in mouse (referred to as hCTLA-4) and crossed it with FcRγ-chain KO mouse (referred to as FcRγ KO) until we reached homozygosity of both genes (Supplemental Fig. 3B) (26, 39). The FcRγ KO was used to prevent depletion of engrafted cells via known Fc-mediated mechanisms upon blockade with an mAb (40). We confirmed that these mice were engraftable with leukemic B cells isolated from the Eμ-TCL1 mouse (data not shown). After screening a series of five TCL1 mice, we identified variability in CTLA-4 expression on the TCL1 B cells (Supplemental Fig. 1). Using an identified mCTLA-4⁺ TCL1 donor (B4000), we engrafted mCTLA-4⁺ TCL1 B6 leukemic cells into hCTLA-4^+/+ mFcRγ^−/− B6, immune competent mice (Fig. 6A). At onset of disease (previously defined as ≥20% CD19⁺ CD5⁺ of CD45⁺ cells), mice were randomized into two blinded treatment groups receiving either anti–CTLA-4 blocking Ab (9D9) or isotype control (MPC-11) at 100-μg doses every 3 d for five total injections (Fig. 6B) (28). The amount administered and treatment regimen were based on recent success with this anti–CTLA-4 Ab in a solid tumor model (41). Leukemic progression was assessed by expansion of leukemic cells in the blood, an established measurement for tumor burden in this disease model (25, 28, 42). We measured absolute counts of leukemic CD19⁺ CD5⁺ cells in the peripheral blood on a weekly basis. We observed that mice treated with anti–CTLA-4 blocking Ab had significantly reduced leukemic burden at day 34, and continued through day 41 after the start of treatment compared with control animals (Fig. 6C; p = 0.03 at day 34 and p = 0.006 day 41). The difference between blocking tumor-expressed CTLA-4 compared with the isotype control on overall survival did not reach statistical significance (Supplemental Fig. 4; p = 0.1442); however, decrease in tumor burden alone suggests CTLA-4 on tumor cells to have a non-negligible effect in understanding response to anti–CTLA-4 therapy (Fig. 7).

It is well established that CTLA-4 is a potent T cell inhibitory protein, but it has not been demonstrated that CTLA-4 must be expressed by T cells to execute negative immune regulation. Our observation that CTLA-4 was in the top five most differentially expressed genes by microarray of normal B versus CLL B cells led us to hypothesize that this highly expressed immune regulator could be co-opted to support leukemic progression. Because we found that surface expression of CTLA-4 occurs in the specific context of activated T cells, we further hypothesized that tumor cell CTLA-4 may have an immunosuppressive role directed at T cells. We discovered that, similar to Tregs, tumor cell CTLA-4 was able to directly modulate costimulatory proteins CD80/CD86. This downmodulation, which consistently favored CD80, resulted in decreased T cell–produced IL-2 in vitro. In vivo blockade of CTLA-4 on tumor cells (while sparing T cell–expressed CTLA-4) reduced leukemic burden in a murine model of CLL. Future studies connecting the immune regulation seen in vitro with the decrease in leukemic burden in vivo by assessing antitumor cytotoxicity of T cells from anti–CTLA-4 treated versus control mice will be highly interesting.

With no true signaling signature to date, one of the most well-defined and accepted mechanisms of action for CTLA-4 has been its cell-extrinsic role in decreasing costimulatory proteins (15, 16, 43). The primary mechanism by which decrease in costimulatory proteins, CD80/CD86, is thought to occur is via trans-endocytosis (17). Intriguingly to us, CTLA-4 has been reported numerous times to be expressed on primary tumor cells of multiple cancer types and studied as a potential prognostic factor, but its function on tumor cells has been a relatively untouched subject (8, 18, 19, 44). In CLL, no consensus has been reached regarding the effect of tumor-expressed CTLA-4. Reports indicate correlation with progressive disease and, contradictorily, negative correlation with CD38 expression, a prognostic factor for aggressive disease (20, 45). However, CLL patients are known to be severely immunocompromised, with infections being the leading cause of morbidity and mortality (46). The observation that matched comparisons of tumor cells in the blood versus the lymph nodes, a site of CLL cell proliferation, show an increase in CTLA-4 expression suggests that a potent immune inhibitory protein that is highly expressed in this cancer may have immunosuppressive function. Similar to loss of CD80 or CD86 from the surface of APCs mediated by T cell–expressed CTLA-4, we expect that tumor cell–expressed CTLA-4 shares this function. Considering the clinical impact of CTLA-4 blockade in multiple cancer types, this finding is highly relevant to understanding the biological mechanism of this treatment. Future studies will need to assess loss of surface costimulatory proteins from APCs by tumor cell–expressed CTLA-4 in the absence of T cell effects.

We demonstrated in vitro the potential functional role of tumor-expressed CTLA-4 to mimic known mechanisms on T cells. To study the potential functional role of leukemic cell CTLA-4 in vivo, we developed, to our knowledge, a novel murine model. By using a species-specific Ab to block mCTLA-4 on adoptively transferred tumor cells and spare hCTLA-4 expressed on host T cells (in the hCTLA-4 knock-in mouse), we observed the functional consequences of tumor CTLA-4. We additionally crossed the hCTLA-4 knock-in mouse to the FcRγ-chain KO mouse to remove Fc-mediated depletion of our CTLA-4⁺ engrafted tumor cells. The resulting host mouse (hCTLA-4^+/+mFcRγ^−/−) can be used to study other engraftable syngeneic tumors in the context of an intact immune system sans phagocytic function.

Using this model, our findings suggest tumor CTLA-4 promotes an increase in leukemic burden rather than proapoptotic intrinsic signaling effects on the CTLA-4–expressing cell in CLL, as suggested by in vitro work of others (45). Although the overall survival was not significantly extended with CTLA-4 blockade, there are multiple factors that should be considered prior to concluding a lack of effect on overall survival. First, CTLA-4 blockade in this model was done using an mAb. In CLL, the lymph nodes and bone marrow are notable areas of CLL cell activation and proliferation (47, 48). Potentially incomplete penetration of these sites with anti–CTLA-4 Ab could lead to less potent reduction in leukemic cells than otherwise noted. This idea is circumstantially supported by similarly observed lymphadenopathy in both CTLA-4– and control-treated mice in our studies (data not shown). Future studies of tumor CTLA-4 contribution to overall survival may be better studied with complete inhibition of B cell–specific CTLA-4 using in vivo CRISPR/Cas9. Secondly, the relative contribution in these studies of T cell–expressed CTLA-4 were not quantified. Anti–CTLA-4 treatment (either globally or T cell specific) has not been tested in the TCL1 murine model of CLL. In this model, it is possible that CTLA-4 blockade does not extend overall survival. Follow-up studies looking at the contribution of tumor CTLA-4 versus T cell CTLA-4 using murine- or human-specific Abs are achievable in our model and would clarify the currently observed phenotype. Additionally, other tumor types should be assessed that have known effects from CTLA-4 blockade.

Our results imply that non–T cell–expressed CTLA-4 may have a broader role outside of CLL. Not only do other diverse tumor types express CTLA-4 (such as melanoma, carcinoma, and others), but the literature indicates that CTLA-4 is also expressed by normal B cells, and at least two studies have peripherally identified CTLA-4 expression in B regulatory cells (7, 9, 18, 19, 49, 50). Because CTLA-4 expression on T cells and, it appears, even more so on non–T cells is exquisitely regulated, it becomes an extremely difficult task to study the function of this protein on diverse cell types. As our results serve as a prelude to the broader context of CTLA-4 irrespective of cell type, it will become essential to conduct cell-specific and anatomical site–specific studies of CTLA-4 in vivo. Not just expression but location and context are important for understanding complex immune regulation, as demonstrated in studies of tumor-associated lymph nodes (51).

Overall, our findings extend and enhance our understanding of the general biology of CTLA-4 from an immunosuppressive T cell protein, to an immunosuppressive protein, irrespective of cell type (summarized in Fig. 7). These results provide a critical contribution to a more complete understanding of the CTLA-4 pathway as it applies to immunotherapy.

We thank Emilia Mahoney for assistance with confocal microscopy, Virginia Goettl for technical assistance with animal studies, Jeanette Leusen for providing the FcγR KO mouse, and Jason Dubovsky for guidance on the project.

The authors have no financial conflicts of interest.