Lack of ex vivo peripheral and intrahepatic α‐fetoprotein‐specific CD4+ responses in hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is one of the most common malignancies with a poor prognosis and limited therapeutic options that is often characterized by the expression of the tumor‐associated antigen α‐fetoprotein (AFP). CD4+ helper T cells are important in generating potent anticancer immunity as they prime and expand CD8+ T‐cell memory and may also have direct antitumor activity. However, very little information is currently available about the relative frequency, immunodominance and peripheral versus intratumoral distribution of AFP‐specific CD4+ T‐cell responses in patients with HCC. We, therefore, analyzed AFP‐specific CD4+ responses in blood and tumor tissue of patients with HCC by using overlapping peptides spanning the entire AFP protein and novel sensitive approaches such as antigen‐specific upregulation of CD154. We found that AFP‐specific CD4+ T‐cell responses were not detectable in the peripheral blood ex vivo. However, after in vitro stimulation, AFP‐specific CD4+ T‐cell responses were detectable in a large fraction of patients targeting different previously unreported epitopes with no clear immunodominance. These results indicate that AFP‐specific CD4+ T‐cell responses are not completely deleted but only present at very low frequencies. Importantly, AFP‐specific CD4+ T‐cell responses were also rarely detectable in tumor tissue, suggesting that the relative absence of these cells in the circulation ex vivo is not due to a rapid accumulation to the tumor side. Taken together, these results suggest that the lack of sufficient CD4+ T‐cell help, especially within the tumor tissue, may be one central mechanism responsible for the failure of AFP‐specific immune responses to control HCC progression.

Hepatocellular carcinoma (HCC) is the fifth most common malignancy worldwide with a poor prognosis and limited therapeutic options. Therefore, the development of novel therapeutic strategies is of high priority. Immunotherapy represents a promising potential option for several reasons: First, a correlation has been reported between high numbers of tumor-infiltrating T cells in HCC tissue and the prognosis of disease. 1,2 Second, adoptive immunotherapy with anti-CD3and interleukin-2-stimulated autologous lymphocytes lowers postsurgical HCC recurrence rates in humans. 3 Finally, the induction of anti-a-fetoprotein (AFP) cell-mediated immune responses can control tumor growth in the mouse model. 4 AFP is a serum marker for HCC that is elevated in 50-80% of patients with HCC. Physiologically, AFP is highly expressed in fetal liver, gastrointestinal tract and yolk sac but is transcriptionally downregulated after birth. Importantly, after birth, AFP can be elevated in patients with HCC or testicular cancers. 5 Of note, this can be associated with the emergence of AFP-specific immune responses. For example, several human leukocyte antigen (HLA) class I-restricted AFP-specific epitopes have been identified by different approaches and shown to be present in HCC patients. [6][7][8][9] Indeed, in a recent comprehensive analysis using overlapping AFP peptides, we have shown that the majority of patients with HCC showed AFP-specific CD8þ T-cell responses directed against previously unreported epitopes and that these responses were primarily detectable in the tumor tissue. 5 Importantly, however, we did not find an association between the presence of CD8þ T-cell responses to AFP and the progression or prognosis of HCC, 5 suggesting a failure of the tumor-specific CD8þ T cells to control tumor growth. The mechanisms responsible for the failure of the AFP-specific CD8þ T-cell response are not well understood and may include the action of regulatory T cells, inhibitory receptors or immunosuppressive cytokines. [10][11][12][13] It is also possible that the CD8þ T-cell failure may be due to the lack of AFP-specific CD4þ T-cell help. Indeed, although CD4þ helper T cells are important in generating potent anticancer immunity as they prime and expand CD8þ T-cell memory 14 and may also have direct antitumor activity, [15][16][17] very little information is currently available about the AFP-specific CD4þ T-cell response. Studies have shown that CD4þ T-cell responses to immunodominant AFP-derived peptides are mainly detectable in HCC patients with low serum AFP concentrations and at an early stage of disease. 18,19 However, in these studies, only a limited set of peptides was studied; the analyses were based on prediction programs, were limited to the peripheral blood and only used the measurements of functional effector functions as a readout. Thus, the important question as to whether AFP-specific CD4þ T-cell responses are present ex vivo but exhausted and functionally impaired in their effector functions could not be addressed. To overcome these limitations, we used the novel CD154 assay that is robust, independent from the HLA type as well as from the knowledge of previously identified epitopes. 20 Of note, by using this approach, we could recently show that antigen-specific CD4þ T cells are not deleted but fully exhausted in chronically HCV-infected patients. 21 In our study, we set out to analyze the actual frequency, function and immunodominance of AFP-specific CD4þ T-cell responses as well as their peripheral versus intrahepatic distribution by using these novel assays.

Subjects and study samples
Thirty patients with HCC and ten healthy subjects were enrolled in the study (Table 1). Fifty milliliters of EDTA-anticoagulated blood was obtained after informed consent and in agreement with federal guidelines and after approval by the local ethics committee. Peripheral blood mononuclear cells (PBMCs) were isolated by Pancoll density gradient (PAN Laboratories, Eidenbach, Germany) and washed three times with phosphate-buffered saline (PBS) (Gibco, Karlsruhe, Germany). Twelve of 30 patients with HCC underwent diagnostic liver biopsy or potentially curative liver resection. Part of the liver biopsy or resection was put into RPMI 1640 medium (Gibco) containing 10% fetal bovine serum and processed as described below.

Synthetic peptides and recombinant proteins
Overlapping peptides spanning the entire AFP protein (18 amino acids each, overlapping by ten amino acids) were obtained from JPT Peptide Technologies GmbH and used in a final concentration of 10 lg/mL. AFP protein (BioProcessing, Portland, ME) was used in a final concentration of 10 lg/mL. A lysate of CMV-infected cells was obtained by Virusys (North Berwick, ME; final concentration 4 lg/mL). Staphylococcus enterotoxin B (SEB final concentration 0.1 lg/mL) and phorbol 12-myristate 13-acetate (PMA, 10 ng/mL; Sigma-Aldrich, Taufkirchen, Germany) were used as positive controls as indicated. Tetanus toxoid (TT; Sanofi pasteur MSD, Leimen, Germany; final concentration 4 lg/mL) was used only in selected cases as additional control antigen.

Antigen-specific expansion of PBMC
A total of 10 Â 10 6 PBMC were resuspended in 10 mL RPMI 1640 medium containing 10% fetal calf serum, 1% streptomycin/penicillin, 1.5% 1 M Hepes and 100 U/mL IL-2 (Hoffmann-La Roche) and were stimulated with AFP protein. No additional antigen-presenting cells (APCs) were added as PBMCs already contain a sufficient amount of the latter. Recombinant IL-2 was added on Day 5 of culture. After a total of 10-12 days of culture, the cells were assayed for IFN-c secretion as described below in response to a 5-hr stimulation with pools of overlapping AFP peptides. If responses were detectable, peptide pools were deconvoluted, and PBMCs were then retested and stimulated with individual peptides. Background levels (no peptide or irrelevant peptide) were between 0.01 and 0.08%. The average of all AFP-pool responses in healthy donors was <0.015% after subtraction of the background, and none of these responses was >0.03% after subtraction of the background. Positive responses were, therefore, defined as >0.03% peptide-specific IFN-c secretion after subtraction of the background. This sensitive threshold was validated by the successful deconvolution of peptide pool-specific responses to the single peptide level (representative stainings are shown in Supporting Information Fig. S2) and is in line with previous observations of our group for both, HCV-specific CD4þ and CD8þ T-cell responses. 21,22 Assays were performed on fresh PBMC and for some selected samples on cryopreserved cells.

CD4 selection and polyclonal expansion of PBMC
A total of 4 Â 10 6 PBMCs were resuspended in 5 mL PBS and incubated with magnetic beads coupled to anti-CD4 antibodies (Dynabeads; Dynla, Oslo, Norway) for 20 min at 4 C. Bound CD4þ T cells were isolated by a particle magnetic concentrator. The purity of CD4þ T cells was confirmed to be >95% by fluorescence-activated cell sorting (FACS) analysis. FACS analysis was performed on a BD FACSCanto II flow cytometer. CD4þ T cells were plated into one well of a 24-well plate (Corning) in 1 mL complete medium containing 100 U/mL IL-2 (Hoffmann-La Roche), 0.04 lg/mL anti-human CD3 monoclonal antibody (Immunotech, Marseilles, France) and 2 Â 10 6 irradiated autologous PBMC as feeder cells. Twice a week, 1 mL of medium was exchanged and 100 U/mL IL-2 was added. After 2 weeks, the expanded CD4þ PBMC and intratumoral lymphocytes were tested for AFP-specific T-cell responses by intracellular IFN-c staining, first on pool level and, if tested positively, on a single peptide level.

Isolation and expansion of intratumoral CD41 T cells
Isolation of tumor-infiltrating T cells was performed as described. 23 Briefly, liver biopsy specimens were homogenized using a 70-lm Dounce tissue grinder (BD Biosciences, Heidelberg, Germany). Cell suspensions were incubated with magnetic beads coupled to anti-CD4 antibodies (Dynabeads) for 20 min at 4 C. Bound CD4þ T cells were isolated using a particle magnetic concentrator. The purity of CD4þ T cells was >95% by FACS analysis. The intratumoral CD4þ T cells were then expanded and stained for IFN-c production as described for PBMC (see above). Background levels of nonspecific IFN-c production (no peptide or irrelevant peptide) for intratumoral lymphocytes and PBMC after nonspecific expansion were between 0 and 0.05%. Importantly, the expansion of peripheral and intratumoral CD4þ T cells from a given patient was always performed in parallel and for the same time before analysis to obtain comparable results. The expansion is needed to obtain a sufficient number of intratumoral T cells for analysis, and the same approach has been successfully used in HCV immunobiology. 22,24,25 All nonspecifically expanded PBMC and intratumoral lymphocytes were used fresh.

Intracellular IFN-c staining and cytokine assays
The intracellular IFN-c staining was performed essentially as described. 26 Briefly, PBMC ex vivo or after antigen-specific expansion and nonspecifically expanded PBMC and intratumoral T cells were stimulated with peptides (10 lg/mL), 1 lL/mL Brefeldin A (BD Pharmingen) and IL-2 (50 U/mL). After incubation for 5 hr (37 C, 5% CO 2 ), cells from each well were blocked with immunoglobulin G1 antibodies and stained with antibodies against CD4 and with Via-Probe to exclude dead cells. After permeabilization with Cytofix/Cytoperm (BD Pharmingen), cells were stained with antibodies against IFN-c (BD Pharmingen) and fixed in 100 lL 2% paraformaldehyde/PBS per well before FACS analysis. The frequency of cytokine-positive T cells was defined as the difference between the frequency detected in peptide-stimulated and unstimulated cells with a minimum of 0.03%. CD107a staining was conducted as described. 27 Staining for IL-2, TNF-a and IFN-c for characterization of peptide-specific CD4þ T-cell responses was performed as described for the intracellular IFN-c staining, except for Brefeldin A which was added 3 hr before staining. All CMV samples were stimulated for 16 hr. Statistical analysis was performed using Graph-PadPrism version 4 with the statistical tests as indicated.

Ex vivo CD154 staining of PBMC
The procedure was performed as described previously. 21 A total of 1 Â 10 6 fresh PBMCs were stimulated with pools of 11 overlapping 18-mer peptides for 5 hr in the presence of IL-2 (50 U/mL). Corresponding samples were stimulated with CMV-lysate or TT as control antigens for 16 hr and cocultured with a CD154-PE antibody as described previously. 20 The cells were blocked with immunoglobulin G1 and then stained with Via-Probe (to exclude dead cells, which may bind nonspecifically to the PE magnetic beads) and anti-CD4-APC. Cells were then fixed in 100 lL 2% paraformaldehyde/PBS per well and analyzed for CD154 expression by FACS analysis. FACS analysis was performed on a BD FACS-Canto II flow cytometer. Ex vivo assays for CD154 expression or IFN-c secretion were performed on either fresh or cryopreserved PBMC, and experiments performed side by side revealed comparable results.

Patient cohort
The AFP-specific CD4þ T-cell response was tested in 30 patients with HCC. The characteristics of the patients are summarized in Table 1. The diagnosis of HCC was based on the American Association for the Study of Liver Diseases guidelines. The most frequent causes for HCC development in our cohort were chronic viral infection (n ¼ 8) and alcohol (n ¼ 11) or both of these two risk factors (n ¼ 3). One patient had hemochromatosis, whereas in seven patients the cause of HCC development remained unknown. AFP was elevated (AFP > 7 ng/mL) in 18 patients, ranging from 7.4 to >60.500 ng/mL. Nineteen of all patients enrolled in the study were therapy naive. The most frequent therapy applied was transarterial chemoembolization (n ¼ 9). As a control group, ten healthy subjects were also tested for AFP-specific CD4þ T-cell responses.

Absence of AFP-specific IFN-c-producing CD41 T cells in the peripheral blood
In a first set of experiments, we analyzed the AFP-specific CD4þ T-cell response in 22 patients with HCC and compared it to the CMV-specific CD4þ T-cell response in the same cohort of patients. For these experiments, PBMCs from HCC patients were stimulated with overlapping peptides spanning the entire AFP protein for 5 hr before IFN-c staining. As shown in Figure 1a, we did not detect any AFP-specific IFN-c production ex vivo. In contrast, 13 of these patients (59%) displayed a CMV-specific CD4þ T-cell response that is in a similar range as previously described by us and other groups. 21,28,29 Original dot blots from these data are shown in Figure 1c. Results were similar in healthy donors: three of the eight healthy donors (38%) displayed a CD4þ T-cell response against CMV, whereas we were unable to detect specific CD4þ T-cell responses directed against AFP in these subjects (Fig. 1b). Thus, the results clearly indicate that HCC patients do not have a general impairment of their CD4þ T-cell response but rather a specific lack of AFPspecific CD4þ T-cell responses.

Detection of AFP-specific CD41 T-cell responses after antigen-specific expansion
Next, we set out to determine whether AFP-specific CD4þ T-cell responses can be detected after antigen-specific expansion. For these experiments, PBMCs derived from HCC patients and healthy donors were stimulated with the whole AFP protein for 10-12 days and subsequently tested for AFP-specific responses to overlapping peptides covering the entire AFP protein. Importantly, by using this approach, we were unable to detect AFP-specific CD4þ T-cell responses in healthy donors (Fig. 2a). In contrast, AFP-specific CD4þ

Tumor Immunology
Int. J. Cancer: 129, 2171-2182 (2011) V C 2010 UICC T-cell responses were readily detectable in 15 of 29 (52%) HCC patients tested (Fig. 2a). In these 15 patients, a median of 2 [range: 1-8] epitopes was targeted, and these responses were found in a frequency between 0.035 and 3.43%. Original dot blots from these results are shown in Figure 2b. All positive peptide-specific CD4þ T-cell responses are listed in Table 2.
It is important to note that the AFP-specific CD4þ T-cell responses were heterogeneous and spread over the entire AFP protein with no consistently recognizable immunodominant epitopes (Fig. 2c).
In 13 of 18 patients (72%) with elevated AFP level (>7 ng/mL) AFP-specific CD4þ T-cell responses were detectable. In two remaining patients showing T-cell responses, serum AFP level was not determined. Patients with normal serum AFP level (n ¼ 10) did not show responses. Statistical analysis revealed that an elevated serum AFP level correlated significantly with the presence of AFP-specific CD4þ T-cell responses in the peripheral blood in our study (p ¼ 0.0003, obtained by Fisher's exact test). The serum AFP level of responding patients was predominantly mildly elevated and significantly higher compared to that of nonresponding patients (p ¼ 0.024, Fig. 2d). This is in agreement with the study by Behboudi et al. who could detect CD4þ T-cell responses to AFP mainly in patients with mildly elevated serum AFP level. 19 However, in contrast to our findings, responding patients in that study showed significant lower serum AFP levels compared to nonresponding patients.
Further analysis of the effector functions of 12 AFPspecific IFN-c-producing CD4þ T-cell responses revealed that some of these cells were also able to produce TNF-a, whereas only one response was able to degranulate (CD107aþ 1/12) and two produced IL-2 (2/12) (Supporting Information Fig. S1a). Interestingly, CMV-specific CD4þ T cells showed a similar functional profile, although with a higher tendency to degranulate (3/5) (Supporting Information Fig. S1a). Original dot blots from these data are shown in Supporting Information Figure S1b.

Analysis of de novo AFP-specific CD154 (CD40 ligand) expression
The fact that IFN-c-producing AFP-specific CD4þ T-cell responses are readily detectable after antigen-specific expan-sion but not ex vivo raises the important question of whether AFP-specific CD4þ T cells might indeed be present ex vivo but primarily impaired in specific effector functions, such as IFN-c production. To address this issue, we determined de novo CD154 expression in response to AFP antigens in selected patients who showed AFP-specific IFN-c production after antigen-specific expansion. De novo CD154 expression is a sensitive ex vivo assay that has the advantage of being   robust and independent from the HLA type and previous knowledge of AFP epitopes. We have recently used this approach to show that HCV-specific CD4þ T cells are indeed present ex vivo but are unable to perform effector functions, such as IFN-c production. 21 In case of HCC, however, as shown in Figure 3a, even by using this sensitive approach, we were unable to detect AFP-specific T-cell responses ex vivo, whereas CMV-and TT-specific responses were readily detectable. These results clearly indicate that AFP-specific CD4þ T-cell responses are not detectable in the peripheral blood of patients with HCC, at least when using the most sensitive methods currently available. However, they can be expanded by antigen-specific stimulation indicating that they are not completely deleted from the peripheral pool albeit present at a very low frequency.

Analysis of intratumoral AFP-specific CD41 T-cell responses
The relative rarity of circulating AFP-specific CD4þ T cells in the peripheral blood could be explained by a rapid accumulation of these cells within the tumor tissue. To address this question, the intratumoral AFP-specific CD4þ T-cell response was analyzed in a total of 12 patients. Liverderived lymphocytes were either isolated from diagnostic liver biopsies (n ¼ 7) or from potentially curative liver resections (n ¼ 5) and expanded nonspecifically as described in the Material and Methods section. The antigen nonspecific expansion is necessary to obtain a sufficient amount of lymphocytes required for this comprehensive analysis. Lymphocytes derived from the peripheral blood were expanded in the exact same manner to allow a strict com-parison of the strength and hierarchy of the immune response between the two compartments. In addition, the AFP-specific CD4þ T-cell response was also analyzed after AFP-specific expansion in 11 of these patients (Figs. 4a and  4b). Importantly, as shown in Figure 4, AFP-specific CD4þ T-cell responses, each targeting a single epitope, were detectable in the tumor-derived lymphocytes of only 2 of 12 (17%) patients. Similarly, AFP-specific CD4þ T-cell responses were detectable in the blood of only 1 of 12 (8%) patients analyzed, although this patient (HCC3) did mount a response to eight different epitopes (Figs. 4a and 4b, Table  2). This patient had a mildly elevated serum AFP level and was therapy naive. Figure 4c shows original dot blots of an intratumoral CD4þ T-cell response to AFP. As expected, antigen-specific expansion of PBMC led to the detection of AFP-specific responses in 7 of 11 (64%) patients, again suggesting that these cells are present at least at low frequencies (Figs. 4a and 4b).

Discussion
Our study was performed to analyze the peripheral and intratumoral AFP-specific CD4þ T-cell responses in patients with HCC. The first important finding of our study is that these responses are almost completely absent ex vivo in this patient cohort. Indeed, by using sensitive approaches, such as ICS, we failed to detect AFP-specific CD4þ T-cell responses ex vivo in the peripheral blood. Importantly, however, by using control antigens such as CMV or TT, we were easily able to detect antigen-specific immune responses, clearly indicating the specific absence of AFP-specific CD4þ T-cell responses ex vivo and not a general state of tumor-induced T-cell suppression. Moreover, in agreement with a previous study, 19 we were able to detect AFP-specific immune responses after antigen-specific expansion, strongly suggesting that these cells are not completely deleted from the T-cell pool but probably just present at very low frequencies. This is also supported by a study by Alisa et al. who have shown that AFP-specific CD4þ T-cell responses are detectable, but primarily in patients with an early stage of HCC (Okuda I or II) or with AFP levels <1,000 ng/mL, and only after antigen-specific expansion, like in our study. 18 In addition, Evdokimova et al.
reported the absence of AFP-specific CD4þ T-cell responses ex vivo but could detect such responses by ELISpot after activation with AFP protein-fed and AdVhAFP-engineered dendritic cells. 30 Thus, these combined results clearly indicate the absence of AFP-specific CD4þ T-cell responses ex vivo in the peripheral blood of patients with HCC.
Here, by using the highly sensitive and specific CD154 assay, we further extend these findings by showing that AFPspecific CD4þ T cells are indeed almost completely absent from the circulation and not just present in a dysfunctional state. In contrast, by using the same approach, we could recently show that HCV-specific CD4þ T cells are readily detectable in chronic HCV infection although they were also undetectable by functional assays and thus in a dysfunctional state. 21 These results indicate that two different major mechanisms contribute to CD4þ T-cell failure in HCC and HCV: almost complete absence of T cells in HCC and T-cell dysfunction in HCV. This is further supported by our finding that AFP-specific CD4þ T-cell responses were also virtually absent in the tumor tissue. These results clearly indicate that the relative absence of these cells in the blood is not due to a rapid accumulation to the tumor site. This is an important finding because it suggests that the absence of sufficient CD4þ help may be one central mechanism responsible for the failure of AFP-specific immune responses to HCC. Indeed, CD4þ T cells have been shown to be key regulators of the adaptive immune responses in tumor models and chronic viral infections, e.g., HBV and HCV, 14,16,17,[31][32][33] where the absence of CD4þ T-cell help causes tumor progression or viral persistence.

Tumor Immunology
The mechanisms responsible for the weak or absent AFPspecific CD4þ T-cell responses in HCC are currently unclear. Possible mechanisms include an AFP-induced impairment of APCs, 34 dysfunctional dendritic cells, 35 the induction of AFPspecific TGF-b-producing CD4þ T cells 36 or the action of CD4þ CD25þ regulatory T cells. [10][11][12] Indeed, high serum AFP levels >2,500 ng/mL have been shown to impair APC function. 34 However, in our study, patients with a serum level up to >5,000 ng/mL showed AFP-specific responses; thus, no evidence of a marked immune-suppressive effect of serum AFP up to this level could be observed although only a few patients with high serum levels were enrolled in our study. It is also possible that AFP-specific CD4þ T cells undergo deletion during thymic migration, as AFP is a tumor-associated, but not completely tumor-specific antigen. The readily detection of AFP-specific CD8þ T-cell responses, however, argues against the thymic deletion of AFP-specific T cells as a major mechanism. 5 Our results may have important implications for vaccine design in HCC. Indeed, to develop a successful vaccine, it will be important to induce both arms of the adaptive immune responses, CD8þ and CD4þ T cells. Studies in other tumors have clearly demonstrated that the combination of CD4þ and CD8þ T-cell epitopes provides a more powerful and long-lasting immunity than CD8þ epitopes alone. 37 In addition, autologous DC vaccination using a tumor cell line lysate that allows presentation of CD4þ and CD8þ T-cell epitopes has been shown to be safe and to have antitumor Tumor Immunology efficacy even with generation of antigen-specific immune responses in the serum of patients with HCC. 38 However, for the successful development of peptide-based immune therapies, it will be important to identify dominant AFP-specific CD4þ T-cell epitopes, as until now only a few such epitopes have been identified. 18,39 In our study, we were unable to detect a clear immunodominance of the CD4þ T-cell responses that were spread across the entire AFP protein. This is similar to our findings regarding AFP-specific CD8þ T-cell responses in patients with HCC. 5 These combined results suggest that AFP-based immunotherapeutic strategies should include the whole AFP, allowing the endogenous processing and presentation of several different AFP-specific CD4þ and CD8þ T-cell epitopes, which may have better clinical efficacy than vaccination with a limited set of AFP peptides.
In sum, the results of our study demonstrate the almost complete absence of AFP-specific CD4þ T-cell help in the peripheral blood ex vivo and in tumor tissues. This is in contrast to the easy detection of AFP-specific CD8þ T-cell responses in the same compartments, as previously described. 5 These results clearly suggest that AFP-specific CD4þ and CD8þ T-cell responses are differentially regulated and that the absence of sufficient help is a hallmark of HCCspecific immune responses, probably explaining the failure of these responses to control HCC progression.