T cell profiling reveals high CD4⁺CTLA-4⁺ T cell frequency as dominant predictor for survival after Prostate GVAX/ipilimumab treatment

Prostate GVAX and ipilimumab

The Prostate GVAX vaccine is a cellular vaccine consisting of two prostate cancer cell lines, LNCaP (CG8711) and PC-3 (CG1940), which have been transduced with an adeno-associated viral vector to secrete GM-CSF. These cell lines were propagated, frozen, and irradiated to arrest further cell division [21, 22]. The product was stored and shipped in gaseous nitrogen phase and administered within 60 min after thawing. All manufacturing was conducted according to good manufacturing practice. Ipilimumab (MDX-010), a fully human IgG1κ monoclonal antibody directed against CTLA-4, was provided by Medarex/Bristol-Myers Squibb (Plainsboro, NJ, USA).

Study population and sampling of peripheral blood

Twenty-eight chemonaive patients with asymptomatic mCRPC received 13 biweekly vaccinations of the Prostate GVAX vaccine (as described [21, 22]) and 6 four-weekly infusions of ipilimumab from the time of prime vaccination. In the first 12 patients, ipilimumab was administered at escalating doses of 0.3, 1, 3, and 5 mg/kg (3 patients each). In the expansion phase, 16 additional patients were included at 3 mg/kg ipilimumab [20].

Responses to treatment were defined as described [23]. In brief, PSA partial response (PR) was defined as >50 % PSA decline from baseline, which was confirmed by a second PSA test 3 or more weeks later. PSA progressive disease (PD) was defined as >25 % PSA increase and an absolute increase of 2 ng/ml or more from baseline, whereas stable disease (SD) was defined as no PR and no PD on treatment.

For immunological monitoring, blood samples were taken from the patients before start of therapy and every 4 weeks until 4 weeks after the last treatment (i.e., follow-up (fu)). Peripheral blood was used for whole-blood analysis by flow cytometry, or for peripheral blood mononuclear cell (PBMC) isolation by density centrifugation (Nycomed AS, Oslo, Norway). PBMC was either immediately stained and assessed by flow cytometry, or cryopreserved for later analysis. Absolute lymphocyte counts (ALC) were determined by routine differentiated white blood cell counts (WBC). For CD4⁺CTLA-4 frequency analysis, we also collected blood from eight sex- and age-matched healthy volunteers.

Antibodies and 4-color flow cytometry

Peripheral blood-circulating lymphocyte frequencies and activation status were flow cytometrically assessed before, during, and after treatment by routine whole-blood or PBMC staining. Whole-blood analysis was performed with lysing solution (BD Biosciences, Mountain view, CA) as described [24]. Cell surface antibody staining of PBMC was performed in FACS buffer for 30 min at 4 °C. Intracellular FoxP3 and CTLA-4 staining was conducted with the antihuman FoxP3 staining kit (eBioscience, San Diego, CA) according to manufacturers’ protocol. The following antibodies were used: fluorescein isothiocyanate (FITC)-, phycoerythrin (PE)-, peridinin chlorophyll protein-Cy5.5 (PerCP)-, or allophycocyanin (APC)-labeled antibodies directed against human CD3, CD4, CD8, CD14, CD16, CD19, CD25, CD27, CD45, CD45RO, CD45RA, CTLA4, HLA-DR, PD-1 (all BD Bioscience), CD56 (IQ products, Groningen, the Netherlands) FoxP3 and ICOS-biotin (eBioscience, San Diego, CA), APC-conjugated streptavidin (BD Biosciences), and matching isotype control antibodies. Stained cells were analyzed on a FACScalibur (BD Biosciences) using Cell Quest software. Events collected were 100,000–150,000 per sample.

T cell subset and differentiation state definitions

Naive CD4⁺ T cells were defined as CD4⁺CD45RO⁻ cells and effector/effector memory CD4⁺ T cells as CD4⁺CD45RO⁺ cells (also referred to as non-naive CD4⁺ T cells). For CD8⁺ T cells, naive T cells were defined as CD27⁺CD45RA⁺, effector cells as CD27⁻CD45RA⁺, effector memory cells as CD27⁻CD45RA⁻, and central memory cells as CD27⁺CD45RA⁻ [25]. The non-naive CD8⁺ T cell population is the sum of effector, effector memory, and central memory T cells. Tregs were defined as CD3⁺CD4⁺CD25^hi and FoxP3⁺. As FoxP3 has also been described to be transiently up-regulated on dividing (activated) effector T cells [26–28], we also analyzed FoxP3 expression within these activated (effector-like) T cells, which we defined as CD4⁺CD25^intermediate (CD4⁺CD25^int) cells. Conventional CD4⁺ T cells were defined as CD4⁺CD25^−/int. As CTLA-4 expression was determined within the total T cell population, the percentage of CTLA-4-expressing conventional T cells was determined by subtracting the percentage of CD4⁺ Tregs (which all express CTLA-4 [29]) from the total percentage of CD4⁺CTLA-4⁺ T cells.

Statistical analyses

Differences between immune parameters before treatment (w0v1; w = week; v = visit; i.e., baseline levels) and during and/or after treatment (w4v3, w8v5, w12v7, w16v9, w20v11, w24v13, and follow-up [fu, i.e., 1 month after last Prostate GVAX and 2 months after last ipilimumab administration]) were analyzed with the repeated measures ANOVA with a post hoc Dunnett’s multiple comparisons test. To determine whether the identified immune parameters were indicative of response to treatment or useful for survival prediction, optimal cut-off points were determined by Cox regression analysis, according to which patients were subsequently divided into two groups. OS for the two groups was plotted using the Kaplan–Meier method, and statistical significance of the survival distribution was analyzed by log-rank testing. In case changes occurred after the first event, patients were excluded from the analysis (this proved only necessary for 1 patient and only in the survival analysis according to absolute lymphocyte counts). To analyze whether prognosis impacted the value of the identified response/survival parameters, the median HPS was also determined for both groups [30]. Differences in HPS between groups of patients and in parameters between prostate cancer patients and age- and sex-matched healthy volunteers were analyzed with the two-sample Mann–Whitney U test. Above-listed statistical analyses were performed with either GraphPad or SPSS software. Differences were considered significant when p < 0.05. To identify clusters of correlated markers, hierarchical cluster analysis using TIGR software was performed and complete linkage analysis was done by Pearson correlation analysis. For this purpose, the values of the treatment-induced and pre-treatment parameters were taken for each patient and divided by the cut-off value, after which the resulting ratios were log-transformed (base 2). Three patients were excluded from this analysis since <70 % of the analyzed biomarkers were available for these patients due to withdrawal from the study or sampling failure.

Clinical results

In this study 28 CRPC patients received 13 biweekly injections of the Prostate GVAX vaccine and 6 four-weekly infusions of ipilimumab. Based on serum PSA levels, five patients experienced PR with PSA declines of more than 50 % and 12 demonstrated disease stabilization (SD) during treatment; PR/SD was significantly correlated with prolonged overall survival (p = 0.0034; [20]). Moreover, the median overall survival (OS) was 29 months, whereas median survival of patients with similar disease characteristics is generally around 16–21 months [3–5]. IRAE were observed in nine patients. Seven patients, five of which with a PR, developed grade 2–3 endocrinopathies consistent with hypophysitis; two of these patients also developed grade 4 alveolitis or grade 1–2 colitis [20].

Absolute lymphocyte count (ALC) and T cell differentiation in relation to survival

As shown in Fig. 1a, absolute lymphocyte counts (ALC) significantly increased upon prostate GVAX/ipilimumab treatment, and increases of more than 25 % were associated with prolonged OS (median survival 41 vs. 20.5 months, HR = 0.181, 95 % CI = 0.046–0.720, p = 0.015; Fig. 1b). To assess the frequency and activation state of peripheral blood-circulating lymphocytes before, during, and after treatment, routine whole-blood or PBMC samples were analyzed by flow cytometry. Despite the observed ALC rises, no differences were observed in overall frequencies of circulating CD3⁺, CD4⁺, and CD8⁺ T lymphocytes and of CD3⁻CD56⁺ NK cells following treatment (Fig. 1c). Prostate GVAX/ipilimumab therapy induced T cell differentiation, as reflected by increased percentages of CD4⁺CD45RO⁺ memory T cells and of effector and effector/central memory CD8⁺ T cells (defined by CD45RA and CD27 expression) (see Fig S1A and C, and Table 1). Treatment-induced increases of >30 % of non-naive (memory) CD4⁺ T cells, but not of non-naive (effector and central/effector memory) CD8⁺ T cells, were associated with significantly prolonged OS (median survival, 41 vs. 20 months, HR = 0.402, 95 % CI = 0.147–0.940, p = 0.036; Fig. S1B and D, and Table 1).

T cell activation in relation to treatment response and survival

To assess treatment-induced T cell activation, circulating CD4⁺ and CD8⁺ T cells were analyzed for the expression of HLA-DR, ICOS, FoxP3, CTLA-4, and PD-1 over the course of treatment. As shown in Figure S2 and Table 1, Prostate GVAX/ipilimumab treatment resulted in generalized CD4⁺ and CD8⁺ T cell activation. Expression of the activation markers increased with escalating doses of ipilimumab (data not shown). As we recently reported, early up-regulation of HLA-DR surface expression during treatment (at week 4, i.e., after one ipilimumab infusion) was more pronounced in patients with PR/SD versus patients with PD [20]. Yet, this early up-regulation did not correlate with survival (Table 1 and [20]). In fact, of all the aforementioned activation markers, only up-regulation of FoxP3 in activated, non-regulatory CD4⁺ T cells (defined by intermediate CD25 expression levels) was related to improved OS (median survival, 41 vs. 21 months, HR = 0.37, 95 % CI = 0.13–0.90, p = 0.03; see also Table 1).

Treatment-induced increase in Treg frequency is negatively associated with OS

Treg frequencies gradually increased over the course of treatment in patients with SD or PD, but not in patients with PR (Fig. 2a). Increases of more than 50 % were associated with a shorter OS following treatment (median survival, 21 vs. 37 months, HR = 2.63, 95 % CI = 1.03–19.5, p = 0.045; see also Table 1 and Fig. 2b). Previous data from (pre)clinical studies on CTLA-4 blockade have shown that increases in the Teff/Treg ratio in both blood and tumors were associated with clinical responses [31, 32]. As shown in Table 1 and Fig 2c, we also observed increases in the CD4CD45RO Teff/Treg ratio, and increases of more than 50 % were associated with prolonged OS following treatment (median survival, 57 vs. 21 months, HR = 0.294, 95 % CI = 0.109–0.794, p = 0.0157).

Preexisting T cell characteristics in relation to survival

In order to identify biomarkers for patient selection, we correlated pre-treatment frequencies of non-naive CD4⁺ and CD8⁺ cells, and of ICOS⁺, HLA-DR⁺, FoxP3⁺, PD-1⁺, or CTLA-4⁺ T cells and Tregs with OS after Prostate GVAX/ipilimumab treatment. An overview of significant and near-significant associations is listed in Table 2.

High pre-treatment frequencies of non-naive CD8⁺ cells or CD4⁺PD-1⁺ cells were associated with a significantly prolonged OS (median survival, not reached vs. 20.5 months, HR = 0.334, 95 % CI = 0.135–0.892, p = 0.028 for non-naïve CD8⁺ cells; and median survival, 41 vs. 18 months, HR = 0.336, 95 % CI = 0.103–0.772, p = 0.014 for CD4⁺PD-1⁺ cells; see also Fig. 3a, b). In contrast, high pre-treatment frequencies of non-naive CD4⁺ cells or Tregs were associated with shorter OS (median survival, 19 vs. 41 months, HR = 2.66, 95 % CI = 1.21–9.09, p = 0.020 for non-naive CD4⁺ cells; Fig. 3c and d).

Significantly higher levels of intracellular CTLA-4⁺ were detected in CD4⁺ T cells of prostate cancer patients compared with age- and sex-matched healthy volunteers (Fig. 3e). Interestingly, CTLA-4 could be detected in both total CD4⁺ and conventional CD4⁺ T cells (i.e., total CD4⁺ T cells minus Tregs) of prostate cancer patients, whereas CTLA-4 was virtually undetectable in conventional CD4⁺ T cells of healthy donors (Fig. 3e). Patients who displayed these cancer-related high levels of conventional CD4⁺CTLA-4⁺ cells before treatment had a significantly longer OS than patients who did not (median survival, 52 vs. 20.5 months, HR = 0.343, 95 % CI = 0.083–0.725, p = 0.011; Fig. 3f). Of note, this positive correlation with survival was also observed for total CD4⁺CTLA-4⁺ T cells (see Fig. S3B).

CD4⁺CTLA-4⁺ T cell frequency as a dominant predictor of OS

To assess whether clinical prognosis impacted the predictive value for treatment outcome of any of the identified response/survival parameters, the median Halabi predicted survival (HPS) was determined for the patient groups above and below the designated cut-offs [30]. No significant differences were observed, indicating that better prognosis before treatment was not the determining factor for any of these parameters (data not shown).

Unsupervised clustering of all pre-treatment and treatment-induced T cell biomarkers with predictive value for survival on treatment revealed two major clusters of patients, designated group 1 and group 2 (Fig. 4a). There was a clear trend for higher OS of patients in group 1 (Fig. 4b). Interestingly, all except two patients from group 1 exhibited the cancer-related expression of CTLA-4⁺ in CD4⁺ T cells, and these patients (designated clustered group 3, highlighted in Fig. 4a) experienced significantly prolonged OS after therapy (median survival, 46.5 vs. 21 months, HR = 0.271, 95 % CI = 0.0791–0.931, p = 0.036; Fig. 4c). Thus, unsupervised clustering analysis demonstrated the cancer-related CTLA-4 expression in CD4⁺ T cells to be a dominant positive predictor for survival after Prostate GVAX/ipilimumab therapy.