CTLA4 Blockade Induces Frequent Tumor Infiltration by Activated Lymphocytes Regardless of Clinical Responses in Humans Free

Costimulatory and co-inhibitory molecules are key players in the activation step of the adaptive immune system and regulate the expansion and effector functions of antigen-specific T cells (1). CTLA4 has a pivotal role in this interaction, dampening immune responses to self-antigens (2). Ipilimumab, a fully human IgG1 anti-CTLA4 antibody (formerly MDX-010; Bristol-Myers Squibb), has shown improvement in overall survival relative to a peptide vaccine in a phase 3 randomized clinical trial in patients with metastatic melanoma previously treated with standard-of-care therapies (3), presenting the therapeutic activity of this class of antibodies. Despite this success, the clinical experience shows that the objective response rate of patients with metastatic melanoma treated with ipilimumab, or the IgG2 anti-CTLA4 antibody tremelimumab (formerly CP-675,206; Pfizer), is low, in the range of 5% to 15%, and they both have similar rates of inflammatory and autoimmune toxicities (grade ≥3) in approximately 20% of patients in pivotal phase 2 trials in second-line therapy for melanoma (4, 5). However, most patients with objective tumor regression have durable responses, the longest ongoing since 2001 (6). The proof of concept of antitumor activity and patient benefit with CTLA4 blockade have been achieved, but there is a clear need to determine what differentiates patients who respond from those who show tumor progression.

Multiple groups have studied how anti-CTLA4 antibodies impact the human immune system and the mechanisms that determine tumor response or progression. Analysis of the effects of anti-CTLA4 antibodies in patients has been mainly based on the study of peripheral blood samples (7–18). Studying the effects of CTLA4 blocking antibodies in tumor samples allows analysis of the interaction between an activated immune system and its cancer cell targets. Preclinical models suggest a key role for CTLA4 both in the infiltration of T lymphocytes into peripheral tissues including tumors and in the modulation of the duration of the interaction between T cells and cells presenting with cognate antigens (19, 20). These data predict that the use of CTLA4 blocking antibodies should increase intratumoral infiltration (ITI) by lymphocytes and retain tumor antigen–specific T cells within tumors. Clinical data to date show lymphocytic ITI in tumor biopsies of patient responding after the administration of anti-CTLA4 antibodies (16, 17, 21, 22).

In a prior study, we analyzed 15 tumor biopsy samples taken at different time points from 7 patients treated with tremelimumab, with lesions biopsied when there was clinical evidence of either response or progression (22). Clinically responding lesions had diffuse intratumoral infiltrates by CD8⁺ T cells that were markedly increased in cases for whom comparison with a baseline biopsy was available. These T-cell infiltrates were massive at the peak of the response at around 1 to 2 months after the first antibody infusion, occupying much of the biopsied regressing lesions. Interestingly, expression of FOXP3 and indoleamine-2,3-dioxygenase, 2 proteins associated with immunosuppressive cells in the tumor microenvironment [regulatory T cells (Treg) and plasmacytoid dendritic cells, respectively], were actually increased in the regressing lesions, particularly at the sites of immune cell–melanoma cell interaction (22). The retrospective nature of that analysis (22) may have induced bias; patients with responding tumors were prone to be biopsied at one stage of the response whereas those with disease progression were primarily biopsied when the therapy effects may be overwhelmed by melanoma progression.

Therefore, a key question remains whether the presence or degree of ITI of T cells differentiates between patients with and without objective tumor responses in prospectively conducted tumor biopsy samples taken at a defined time point. Therefore, we conducted a clinical trial with paired baseline and postdosing tumor biopsy specimens collected within 1 and 2 months from the first dose of the CTLA4 blocking antibody. Our main finding is a remarkable induction of immune cell infiltrates by CD4⁺ and mostly CD8⁺ T cells after the administration of tremelimumab. This was present both in lesions that went on to objective tumor response and in half of the lesions that progressed.

Thirty-two patients with measurable advanced melanoma (stages IIIc–IV) with metastatic lesions amenable to outpatient biopsies were enrolled in this phase II clinical trial (UCLA IRB# 06-06-093 IND# 100453, clinical trial registration NCT00471887). Patients received single-agent tremelimumab at 15 mg/kg every 3 months with baseline and approximately day 30 to 60 postdosing biopsies. Samples were coded with the study denomination of GA and a patient-specific number. Adverse events attributed to tremelimumab were graded according to the NCI common toxicity criteria version 2.0 (23). Patients who experienced the following adverse events at any time during the previous cycle were considered to have a dose limiting toxicity (DLT) and treatment with tremelimumab was discontinued: grade 4 treatment-related adverse event; grade 3 or higher hypersensitivity reaction; grade 2 or higher colitis; and/or autoimmune reaction in a critical organ (brain, eye, liver, thyroid, hypophysis). Objective clinical responses were recorded following a modified Response Evaluation Criteria in Solid Tumors (RECIST; ref. 24), wherein skin and subcutaneous lesions evaluable only by physical examination were considered measurable if adequately recorded using a photographic camera with a measuring tape or ruler; there was no minimum size restriction for these lesions.

Biopsies samples were formalin-fixed and paraffin-embedded and stained for immunohistochemistry (IHC) as previously described (22) with anti-CD4 (clone 4B12; NeoMarkers) and anti-CD8 (clone C8/144B'; Dako). The Simple-PCI Imaging System (version 5.2.1.1609; Compix Inc. Imaging System) was utilized to quantitatively evaluate T-cell infiltration. The frequency of intratumoral and peritumoral lymphocytes was assessed by analyzing 10 tumor areas from each sample at 200× magnification. The density was compared between pretreatment and posttreatment biopsies. All samples were analyzed without the knowledge of the patients' clinical outcomes.

Postdosing biopsy specimens with significant increase in T-cell infiltrates were stained by immunohistochemical double staining with anti-HLA-DR (clone TAL, 1B5; Dako) and anti-CD45RO (clone UCHL1; Dako), and single staining with Ki67 (clone MIB-1; Dako) or anti-FOXP3 (clone 236A/E7; Abcam).

The statistical design of this clinical trial was based on the assumption of a 20% or higher probability of increased CD8⁺ infiltration in posttreatment biopsies detected by IHC. This assumption was based on the lower boundary of change in CD8⁺ infiltration from our prior studies (22). Two scores using semiquantitative analysis of immunohistochemical data (0 to 2+, 1+ to 3+) were assessed. A binomial test was used at 5% level of significance. If the true probability of infiltration increased by at least 2 score levels in at least 50% of the tremelimumab-treated patients, then 20 to 21 evaluable patients would provide 90% power to reject the null hypothesis. The Mann–Whitney rank-sum test was used to compare values obtained from assessment of the pre- and posttreatment samples. Analyses were done using the SigmaPlot software package, and all tests were 2-sided with the significance level set at P = 0.05.

Thirty-two patients were enrolled (Table 1). The majority of patients had M1c metastatic melanoma (visceral metastasis and/or high lactate dehydrogenase level) and more than half of the patients had received prior systemic therapy, most frequently a chemotherapy-containing regimen. There were 9 patients with clinically relevant toxicities prospectively defined as DLTs that precluded continued dosing with tremelimumab. One patient had immune thrombocytopenia purpura (ITP), which developed within 1 week after the first dose; grade 3 colitis in 5 patients, 2 during the first cycle and the other 3 while on chronic maintenance dosing; 1 patient with a grade 3 skin rash in the first cycle; and 2 with grade 2 hypophysitis, both during the second or later cycles. Three patients had an objective and durable tumor response, all with a complete response of in-transit metastasis (patients GA18, GA29, and GA33). One additional patient had an objective response in supraclavicular and laterocervical lymph nodes meeting partial response criteria, followed by slow disease progression of nodal metastasies 7 months after initiating dosing with tremelimumab (patient GA5). This patient died from an unrelated cause (infectious osteomyelitis after an accidental wound) 20 months after starting tremelimumab, with active nodal metastases of melanoma localized in the supraclavicular area but without widespread systemic metastasis. There are 7 patients alive beyond 2 years, the 3 patients with a complete response (35+, 30+ and 28+ months from study start), and 4 patients who are alive with metastatic melanoma (follow-up between 29+ to 41+ months) despite not having an objective response to tremelimumab (patients GA7, GA19, GA26, and GA32).

Paired tumor biopsy samples before and after the first infusion with tremelimumab were obtained in 21 of the 32 patients enrolled in this clinical trial. Reasons for obtaining only a baseline biopsy were absence of melanoma in the biopsy specimen in 1 case, toxicity within the first cycle resulting in inability to return for the postdosing biopsy in 2 patients (ITP and colitis), and early disease progression in 7 patients who withdrew consent before the proposed postdosing biopsy. The postdosing sample from 1 of the 21 patients with paired biopsies could not be retrieved for analysis (patient GA32). The postdosing biopsy specimen from patient GA25 did not contain melanoma. The presenting characteristics and outcome of the remaining 19 patients with pre- and postdosing biopsies available for analysis are presented in Table 2.

Up to 10 randomly selected fields per sample were analyzed for ITI (when T cells were mixed within the melanoma cells) and peritumoral infiltration (PTI, when T-cell infiltrates are located peripheral to the tumor mass and in collagen bundles that dissected the tumor mass). Overall, there was a marked and highly statistically significant increase in ITI by CD8⁺ cells in biopsy samples taken after tremelimumab treatment (Figs. 1 and 2). The mean pretreatment CD8⁺ cell count was 289 cells/mm² (SEM = 61), and the postdosing density of these cells increased to 955 (SEM = 191, P = 0.005; Fig. 2). The difference in ITI for CD4⁺ cells was also significantly increased but at a lower magnitude (mean predosing 104 ± 32 vs. mean postdosing 428 ± 156, P = 0.018; Fig. 2). Analysis of PTI by lymphocytes was not feasible in 5 cases with metastatic melanoma to the lymph nodes because the great majority of peritumoral cells were nodal lymphocytes. Among the remaining cases, there were no significant changes in PTI by CD8⁺ or CD4⁺ cells (Supplementary Fig. S1), as there was no evidence of tumor adjacent infiltration by T cells in skin biopsies beyond the metastases (data not shown).

Among the 4 patients with objective clinical responses, 2 cases (patients GA5 and GA29) had a marked increase in ITI of CD8⁺ cells (Figs. 1 and 2; Supplementary Fig. S2). Melanoma cells showed degenerative changes, with disassociated connection and abundant lymphocytes between and around the melanoma cells. Pretreatment biopsies showed intact melanoma cells, with little or no intratumoral penetration by T cells. The posttreatment biopsy in patient GA18, who had a durable complete response, showed complete regression of the melanoma, with no residual melanoma cells and abundant infiltrating lymphocytes at the regressed melanoma site (Fig. 1). The postdosing biopsy from the fourth patient with a clinical response (patient GA31) also showed no residual melanoma cells. This completely regressed melanoma included abundant melanin pigment, both free and in macrophages, and extensive scar tissue without lymphocytic infiltration and was therefore interpreted as showing late-stage regression (Supplementary Fig. S2). Many postdosing samples from patients with clinically progressive disease also showed significant lymphocytic ITI (Figs. 1 and 2; Supplementary Fig. S2). In 8 of 16 patients with disease progression, the increase in CD8⁺ cell density was greater than the average increase for the whole series. In addition, 6 cases had increases in CD4⁺ cell density greater than the average overall series increase over the baseline biopsy. There was no correlation between the density of ITI after tremelimumab treatment in responder and nonresponder patients or between patients alive 2 years or more after study initiation and patients who died from melanoma less than 2 years after treatment initiation (Fig. 2).

Because T-cell infiltrates increased in patients with and without a clinical response, we were interested in studying whether the functional characteristics of infiltrating T cells differenced between these samples. HLA-DR is a surface marker of T-cell activation after exposure to CTLA4 blocking antibodies (8, 9, 25), whereas CD45RO is a marker of prior cognate antigen-exposed T cells. Together they mark cells with a surface phenotype of T effector or T effector memory cells (26). The combined analysis of HLA-DR/CD45RO staining showed a marked increase in double-positive cells in postdosing biopsies in all cases, irrespective of whether they had an objective response or not or were alive beyond 2 years from study initiation (Fig. 3 and Table 3). Given the increase in the number of T cells in postdosing biopsies, we stained the samples for the cell proliferation marker Ki67 to determine the extent of cell replication within tumors. There was no change in the frequency of Ki67-positive nuclei among lymphocytic infiltrates when postdosing biopsies were compared with baseline biopsies (Fig. 3 and Table 3). In our prior analysis of lesions regressing after tremelimumab treatment (22), we noted an increase in FOXP3⁺ cells. The current study confirmed this finding. Three postdosing samples from patients who had a durable complete response had a marked increase in FOXP3⁺ cells (Fig. 3). In the 8 nonresponding tumors, there was an increase in FOXP3⁺ cells in 5 tumors, 2 had no apparent change, and 1 had a decrease in FOXP3⁺ cells. When comparing the infiltrates between durably responding and nonresponding patients, the trend was in favor of higher infiltrates of FOXP3⁺ cells in responding lesions (Table 3).

An intratumoral infiltrate with activated T cells has prognostic significance in patients with cancer, in which primary tumors with larger and more diffuse T-cell infiltrates displaying an effector functional phenotype have improved survival (26, 27). The major goal of tumor immunotherapy is to induce such intratumoral infiltrates by using a therapeutic intervention. It should obviously require demonstration that T cells do effectively infiltrate tumor lesions to exert their cytotoxic activity. However, given the practical limitations of obtaining serial biopsy specimens in patients with metastatic cancers, there is a paucity of data for studying immune infiltrates in tumors of patients receiving immunotherapy. This point is particularly relevant for CTLA4 blocking antibodies, as preclinical data suggest that the mechanism of tumor regression should be mediated by the intratumoral accumulation of T cells, with little evidence of changes in the systemic circulation. In the current studies, we analyzed tumor biopsy specimens from patients receiving anti-CTLA4 antibodies for the treatment of advanced melanoma. The main goal was to compare baseline and postdosing samples for the presence and functional characteristics of lymphocytic infiltrates. Contrary to conclusions based on prior experience with biopsies conducted late in the treatment with tremelimumab (22), the current studies with biopsies at 1 to 2 months after the first dose of tremelimumab show sharp increases in tumor-infiltrating lymphocyte (TIL) in half of the patients who showed disease progression. Quantitative analysis of the T-cell infiltrates did not differentiate clinical responders and nonresponders. Additional analyses to determine whether there was a difference in the functionality of these cells by using phenotypic markers also indicate that the cellular infiltrate induced by tremelimumab does not differ significantly between clinical responders and nonclinical responders.

Postdosing intratumoral lymphocyte infiltrates could be as a result of cell mobilization and increased ITI induced by CTLA4 blockade, a possibility supported by some preclinical models (19, 20, 28). Alternatively, the increase may be due to active tumor antigen–specific lymphocyte proliferation with release of the so-called CTLA4 cell-cycle checkpoint with G₁ arrest (29–32). The dominant effect of CTLA4-inhibiting lymphocyte replication is evidenced by studies in CTLA4 knockout mice, which die within days of postnatal antigen exposure due to massive lymphocyte proliferation and peripheral tissue infiltration (33, 34). To study whether active lymphocyte replication within tumors caused the postdosing increase in TILs, we compared pre- and postdosing samples for the nuclear expression of the cell replication marker Ki67. The data showed no such change, suggesting that tumors are not the site of lymphocyte replication after CTLA4 blockade. This information is complemented by our recent experience in using whole body imaging with positron emitting tomography (PET) to study tumor and lymphoid organs for a differential uptake of radiolabeled PET traces in patients treated with tremelimumab (35). After treatment with tremelimumab, there was increased 3′-deoxy-3′-[¹⁸F]fluorothymidine ([¹⁸F]FLT) uptake in the spleen in most patients, which is reflective of cell replication in this large lymphoid organ (35). The PET imaging data, together with the morphologic data from the analysis of tumors presented herein, suggest that tremelimumab induces lymphocyte replication in lymphoid organs that, in turn, leads to increased infiltration of tumors in most patients, whether or not they have a clinical tumor response. Because the changes in intratumoral T-cell infiltrates are well beyond what can be detected in blood or in normal tissues, lymphocyte trafficking changes are the most likely explanation for the observed results in this biopsy series.

Studies analyzing immune parameters after CTLA4 blockade have not yet provided a reproducible explanation of why clinical tumor responses are infrequent despite evidence of immune activation in most patients. Multiple studies reported lymphocyte activation in blood (7–10, 12–16, 18, 36), and our current data confirm immune activation within tumors (17). It is difficult to reconcile the frequent immune responses to CTLA4 blockade with infrequent clinical evidence of tumor regression. Reported mechanisms of tumor escape to tumor immunotherapy include downregulation of MHC and tumor antigen–processing and antigen-presenting machinery (37), or the effects of oncogenes on sensitivity or resistance to apoptosis induced by immune effector cells (38–40).

In conclusion, postdosing melanoma tumor biopsy specimens from more than half of patients treated with the CTLA4 antagonistic antibody tremelimumab show increased T-lymphocyte infiltrates. This increase is pronounced in patients who show an objective tumor response but is indistinguishable quantitatively and phenotypically from infiltrates in half of the patients who showed disease progression. These data indicate that, in most patients, therapeutic CTLA4 blockade induces the desired immune stimulation resulting in T-cell infiltration of tumors. Because only a minority have clinical responses, then differences on how tumors respond to the T cell infiltrates is likely to be a major cause of resistance to anti-CTLA4 antibodies.

J. Gomez-Navarro was an employee of Pfizer Inc. at the time of this work. A. Ribas received honoraria from Pfizer for the participation in advisory boards during the conduct of this study.

We thank the manuscript review by Dr. Margaret Marshall from Pfizer Inc., New London, CT.

This work was funded in part by Pfizer Inc., the Melanoma Research Foundation, the NIH grant 2U54 CA119347, The Fred L. Hartley Family Foundation, the Jonsson Cancer Center Foundation and the Caltech-UCLA Joint Center for Translational Medicine (all to A. Ribas).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.