Over the past two decades, cell therapies have been approved for cartilage repair, tissue engineering of skin, and most recently for the personalized, cellular immunotherapy of cancer (1). Another personalized procedure that has been in development is adoptive cell therapy using tumor-infiltrating lymphocytes (TIL). In this setting, tumors are surgically removed from the patient, processed into a single cell suspensions, and TIL are expanded with interleukin-2 (IL-2) (2, 3). When adherent melanoma cells are no longer detected by visual inspection, TIL are then rapidly expanded and given back to patients who have been preconditioned with non-myeloablative chemotherapy or myeloablative total body irradiation and chemotherapy (2, 4−6). Using these procedures, clinical response rates of 50–70% in patients with metastatic melanoma have been reported (2, 4−6).
To treat a patient, cellular products must be tested for microbials, identity, and purity (7). One of the challenges with providing personalized medicine, such as TIL adoptive therapy, is meeting the demands of product characterization and purity when the starting cell population from which the biological product is derived is highly variable in gene expression and cellular composition. TIL may be distinguished from tumor cells using morphology, size, and antigen expression. Flow cytometry is a common method of characterizing cellular therapeutics (1). Besser et al. (2) recently showed that adoptive transfer of TIL with high percentages of CD8+ T cells and overall high numbers of CD8+ T cells correlated with clinical responses in patients with melanoma. Dudley et al. (8) observed similar results when purified CD8+ T cells were adoptively transferred into patients with melanoma. Additional characterization of TIL by flow cytometry also suggest that CD27+, CD28+, and telomere length correlates with clinical responses (9, 10).
Flow cytometry has been utilized to characterize TIL, but it has not been used to evaluate the presence of tumor cell contamination of TIL cultures. TIL products for adoptive transfer have been tested for purity by immunohistochemistry (IHC), which is a sensitive assay that can distinguish TIL from tumors by analyzing cell morphology, size, and antigen expression. However, there are several disadvantages to IHC. The assay is unable to detect viable cells, it is subjective in interpretation, and it semiquantitative at best (11, 12). Advantages of flow cytometry include its ability to acquire multiple parameters and large numbers of data points in a short period of time which make it an ideal technology to evaluate cellular purity in a qualitative and quantitative manner.
The objective of this study was to determine the utility of flow cytometry in the evaluation of the purity of TIL cultures expanded from surgically-resected tumors. To detect melanoma cells, antibodies to the intracellular tumor antigens gp100, Mart1, tyrosinase and S100 were combined with an antibody to the surface receptor melanoma chondroitin sulfate proteoglycan (MCSP). The antibody mixture for staining tumor-antigens was combined with anti-CD3 and a viability dye to determine if flow cytometry may be used to detect viable tumor cells in TIL cultures. We demonstrated that this cocktail of antibodies could detect melanoma cell lines spiked into T cell cultures. Furthermore, we showed that flow cytometry was able to detect tumor cells, in primary tumor-cell suspensions and that TIL expanded from these suspensions were tumor cell free. This data supports the use of flow cytometry as an alternative method of evaluating TIL purity for clinical purposes.
MATERIALS AND METHODS
The protocol for obtaining metastatic melanoma tumors was approved by the Aurora Health Care Institution Review Board and informed consent was obtained. Additional tumor samples were obtained from the Cooperative Human Tissue Network (CHTN http://www.chtn.nci.nih.gov/). Primary tumor samples were from patients with metastatic melanoma. Acquisition of primary tumor samples occurred between March 2009 and February 2011.
Cell Culture and Cell Lines
Complete medium (CM) was composed of RPMI 1640 with L-glutamine (Lonza, Walkersville, MD), 10% fresh human serum (Key Biologicals, Memphis, TN), 2 mM L-glutamine (Lonza), 25 mM HEPES (Lonza), 100 U/ml penicillin (Lonza), 100 μg/ml streptomycin (Lonza), 10 μg/ml gentamicin (Lonza) and 5.5 × 10−5 M 2-mercaptoethanol (Invitrogen, Carlsbad, CA). Cell lines and primary tumor lines were cultured in T2 media composed of RPMI1640 with L-glutamine, 10% FBS (Hyclone, Logan, UT), 2 mM L-glutamine, 25 mM HEPES, 100 U/ml penicillin, 100 μg/ml streptomycin, and 10 μg/ml gentamicin. ASLMC-024, ASLMC-048, and ASLMC-049 are primary melanoma cell lines developed in our laboratories; SK-Mel-5 and HT-144 were purchased from ATCC (Manassas, VA); and 624, 2514, and 888mel were generous gifts from Dr. Mark Dudley at the National Cancer Institute (NCI).
Generation of Activated T cells
Activated T cells (ATC) were generated from peripheral blood mononuclear cells (PBMC) obtained from healthy subjects cultured with 20 ng/ml anti-CD3 (Clone OKT3, Orthoclone, Centocor Ortho Biotech, Philadelphia, PA) and 100 IU/ml of IL-2 (Proleukin (aldesleukin), Novartis, East Hanover, NJ) in RPMI 1640 media containing L-glutamine and 2% human AB serum (Gemini Bio-products, West Sacramento, CA). Starting on day 5 and every 2–3 days thereafter, ATC were counted and maintained at 1 × 106 cells/ml by adding fresh media containing 100 IU/ml IL-2. On day 14, ATC were counted and cryopreserved. Thawed ATC were used as a negative control for tumor antigen detection or as a positive control when spiked with a melanoma cell line.
Generation of TIL
TIL were generated as described in detail by Dudley et al. (3) with minor modifications. Briefly, tumor fragments were enzymatically digested overnight in media containing collagenase (Sigma-Aldrich, St. Louis, MO), hyaluronidase (Sigma-Aldrich), and DNAse (Pulmozyme [Dornase Alpha], Genentech, San Francisco, CA) to form a single cell suspension. To generate TIL, viable cells were cultured at 5 × 105/ml in 2 ml of complete media (CM) containing IL-2 (6000 IU/ml). After 5 days of culture, half of the CM was removed and replaced with fresh CM containing IL-2. Every 2–3 days thereafter, the process was repeated until a cell concentration of 1 × 106 cells/ml was obtained. At that time, cells were split into daughter wells, at 5 × 105 cells/ml and cultures continued or cells were stimulated in a rapid expansion protocol (REP). The rapid expansion of TIL was performed by culturing TIL with a 200-fold excess of irradiated (50 Gy) autologous or pooled allogeneic PBMC with 30 ng/ml anti-CD3 (OKT3) and 6000 IU/ml IL-2. Culture media was composed of 50% CM containing 10% human AB serum instead of fresh human serum, 50% AIM-V (Invitrogen), 10 μg/ml ciprofloxacin IV (Sicor Pharmaceuticals, Irvine, CA), and 1.25 μg/ml amphotericin B (X-gen Pharmaceuticals, Big Flats, NY). Half of the media was exchanged on day 5 with fresh media and 6000 IU/ml IL-2. When cell concentration reached 1 × 106 cells/ml, additional AIM-V containing 5% human AB serum, 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 10 μg/ml ciprofloxacin IV, 1.25 μg/ml amphotericin B, and 6000 IU/ml IL-2 was added to reduce the cell concentration to 5 × 105cells/ml. Additional AIM-V media was added as required until cell harvest on day 14.
IHC was performed on cytospin preparations of tumor cells. The antibodies used in featured studies were purchased from DAKO and included mouse anti-human melanosome (Clone HMB45; Code: N1545), mouse anti-human Melan-A (Clone A103; Code: N1622), mouse anti-human tyrosinase (Clone T311; Code: N1634), and polyclonal rabbit anti-S100 (Code: N1573). Cytospin slides were prepared using 50,000 tumor cells/slide on a Cytotech centrifuge. Slides were dried for at least 1 h after which cells were fixed with ice cold 100% acetone for 10 min. After evaporation of the acetone, cells were stained using the EnVision G/2 System from DAKO according to the manufacturer's instructions with a few modifications. The modifications included a 5% FBS in PBS blocking step prior to the addition of primary antibodies, decreasing the time of washes from 5 min to 15 s, and decreasing the incubation times from 30 min to 10 min.
To screen tumor antigens, tumor cells were stained with antibodies against intracellular and surface tumor antigens. Intracellular flow cytometry was performed on 2.5–5 × 105 tumor cells using the Fixation and Permeabilization Solution Kit with BD GolgiStop™ (Cat. No.554715, BD Bioscience, San Jose, CA). Indirect staining of intracellular tumor-antigens gp100 (Melanoma Marker, Clone HMB45, 1:50 dilution, Cat. No. sc-59305; Santa Cruz Biotechnology, Santa Cruz, CA), Mart-1 (Melan-A, Clone (A103, 1:50 dilution, Cat. No. sc-20032; Santa Cruz Biotechnology), tyrosinase (Clone T311, 1:50 dilution, Cat. No. 35-6000, Invitrogen), and S100 (Clone 4C4.9, 1: 50 dilution, Cat. No. NB600-563, Novus Biologicals, Littleton, CO). Anti-mouse F(ab)'2 PE was used to detect the intracellular tumor antigens. On the same number of cells a direct surface stain was used to evaluate MCSP expression on tumor cells.
To develop the assay for tumor detection within a TIL culture, ATC were spiked with known numbers (1% of ATC) of melanoma-tumor cells. To evaluate the number of tumor cells in TIL cultures, various points across the TIL generation process were tested including enzyme digest, pre-REP and REP. To access the number of tumor cells in ATC or TIL cultures, 1 × 106 cells were stained with the viability dye Live/Dead Fixable Green Dead Cell Stain Kit (Cat. No L23101. Invitrogen) followed by fixation and permeabilization. Unconjugated antibodies against gp100 (Melanoma Marker, Clone HMB45, 1:400 dilution, Santa Cruz Biotechnology) and Mart1 (Melan-A, Clone A103, 1:200 dilution, Santa Cruz Biotechnology), tyrosinase (Clone T311, 1:200 dilution, Invitrogen), and anti-S100 (Clone 4C4.9, 1: 200 dilution, Novus Biologicals, Littleton, CO) were combined for tumor cell staining after permeabilization. A secondary anti-mouse F(ab)'2 PE (1:100 dilution, Jackson ImmunoResearch, West Grove, PA) was then used to detect intracellular tumor antigens (ICTA). Surface antigens that identify tumor and T cells were stained with anti-MCSP APC (Clone LMH-2, 1:10 dilution, Cat. No. MAB2585A, R&D Systems Minneapolis, MN), and anti-CD3 PerCP (Clone SK7, 20 μl/100μl, Cat. No. 347344, BD Biosciences), respectively. Surface antigens were stained after the intracellular antigens to prevent the anti-mouse F(ab)'2 from binding to antibodies bound to surface antigens. Approximately 100,000 viable cells were acquired on a FACS Calibur from BD Bioscience.
Tumor Detection and Statistics
The percentage of tumor in the TIL product was calculated from the following equation:
% Viable Tumor Cells = (Test Tube(Q1+Q2+Q4)/Total Viable Cells) × 100, where Q1, Q2, and Q4 are ICTAs+MCSP−, ICTAs+MCSP+, and ICTAs−MCSP+ quadrants of a 2D plot, respectively.
Because of nonnormal distribution of results, data were analyzed by the nonparametric Kruskal–Wallis One Way Analysis of Variance on Ranks followed by Dunn's pairwise multiple comparison procedure (Sigmaplot 11.0).
To develop a flow cytometry assay to test TIL purity, we first evaluated anti-melanoma antibodies by immunohistochemistry to demonstrate melanoma-tumor antigen expression. Figure 1 shows expression of the intracellular antigens gp100, Mart-1, S100, and tyrosinase in two melanoma cell lines 2514 and HT144 by immunohistochemical staining. In the 2514 cell line, gp100 and Mart-1 brightly stained the cytoplasm, while tyrosinase and S100 had a weak cytoplasmic stain and no stain, respectively. In contrast, the HT-144 cell line was brightly stained with S100 and to a lesser extent with gp100, while Mart-1 and tyrosinase weakly stained the cells. We then tested the antibodies for the intracellular tumor antigens by flow cytometry. In Figure 2a, we depict the gating strategy to evaluate tumor-antigen expression. Tumor cells were gated to remove debris followed by an exclusion of dead cells using Live/Dead far Red. Live/Dead far Red dim viable tumor cells indirectly stained for intracellular tumor-antigens are overlayed against the F(ab')2-PE control. In Figure 2b, we show that flow cytometry was able to detect all the intracellular antigens. Consistent with IHC, the pattern of staining between 2514 to HT-144 demonstrated a decrease in Mart-1 and increase in S100 staining while the other two tumor-antigens tyrosinase and gp100 staining remained approximately the same. In addition, we evaluated the staining of anti-MCSP, a known melanoma surface molecule. In Figure 2C, we show that anti-MCSP staining was bright on both cell lines. These data show that flow cytometry maybe used to detect tumor antigens.
To obtain a better idea of melanoma tumor antigen expression, we evaluated cell antigen expression on several other tumor cell lines. The results of staining eight different melanoma tumor cell lines are shown in Table 1. One observation was that individual antibodies were heterogeneous in tumor cell staining. For instance, in eight-melanoma cell lines, anti-Mart-1 stained from 39 to 99% of the cells depending on the cell line tested. To reduce the possibility of a false negative due to variation in antigen expression, we combined antibodies that bound intracellular gp100, Mart-1, S100 and tyrosinase, and the surface antigen MCSP. This strategy proved beneficial when staining the primary melanoma cell line ASLMC-049. The best staining was found to be MCSP that stained 76% of the cells; however, when combine with all the other antibodies, 84% of the cells were detected. The combination of antibodies against melanoma tumor antigens appeared to improve the detection of the antigens in all the cell lines tested, even though some were close to 100% with a single antibody.
Each percentage is the result of a single staining experiment. Gating was performed as described in Figure 3 with a histogram gate being set at 1% of the negative control.
Combination of antibodies against gp100, Mart-1, S100, Tyrosinase, and MCSP
To determine if viable tumor cells could be detected within a T-cell population, a melanoma cell line was mixed with OKT3-activated T cells (ATC). The cell mixture was stained and a gating strategy was developed to evaluate expression of tumor antigens on viable cells by flow cytometry (Fig. 3). In plot one (FSC vs. SSC), a gate (R1) was used to select the lymphocyte and tumor populations and to exclude debris. In plot two (live/dead green vs. SSC), a gate (R2) was used to detect viable cells by selecting the Live/Dead Green dim population. A gate joining R1 and R2 was used to select the viable target population composed of lymphocytes and tumor cells. Given that most TIL cultures have high numbers of T cells, a third plot (CD3 vs. SSC) was generated to exclude T cells by gating on the CD3− population. The exclusion of CD3+ cells was done to minimize the number of cells that would be evaluated in the CD3− population to make identification of tumors more feasible. The CD3− population was then evaluated in plot 4 (ICTAs vs. MCSP) for tumor antigen expression by staining with the combination of intracellular tumor antigens (ICTAs) and the surface antigen MCSP. Table2 is a summary of the calculations of events obtained from ATC spiked with various melanoma cell lines. This assay could readily detect melanoma cells spiked into a T cell population.
Table 2. Identification of melanoma tumor cells in ATC samples
To determine if viable tumor cells could be detected within TIL populations, samples were taken throughout culture (enzyme digests, pre-REP, and REP TIL) and stained according to the method described earlier. Figure 4a shows flow cytometry staining from a TIL culture in stepwise progression from the enzyme digest through REP. Samples were gated to exclude debris and dead cells. From the earliest culture (enzyme digest) to the final TIL preparation (REP), there was an increase in the CD3+ population, a decrease in the CD3− population and a decrease in tumor cell detection. Figure 4B summarizes flow cytometry data from the cultures evaluated for the presence of tumor cells in TIL cultures. In the enzyme digests, the median (range) of tumor cell detection was 46.42% (4.96–84.69%). The detection of tumor cells decreased significantly from the enzyme digest to pre-rep (P < 0.05). The median (range) of tumor cell detection in pre-REP samples was 0.03% (0.01–1.45%). There were two samples in which tumor were definitely detected one was at 1.45% and the other was 0.87%. Overall, there was no difference between the pre-REP, REP and ATC biological control as the median (range) of tumor cell detection in REP and ATC samples were 0.04% (0.01–0.13%) and 0.07% (0.01–0.14%), respectively.
The advent of IHC initiated a renewed effort to detect minimal residual disease in patients with cancer (12), and it has become the gold standard for evaluating minimal residual disease (13). IHC has also been the method of choice to assess the presence of tumor cells within TIL cultures for adoptive transfer therapy. However, IHC is labor intensive, requires large number of slides, does not allow an assessment of viability, is technically challenging, is semi-quantitative and requires a trained eye and/or pathologist (11–13). The amount of time involved in performing IHC has decreased by utilizing automated microscopy; however, manual review is still required to confirm positive cells (12). This allows for other technologies such as immunomagnetic enrichment, PCR, and flow cytometry to evaluate rare cells in a cellular population. We now report a multiparameter flow cytometry assay to evaluate the presence of viable melanoma cells in TIL cultures.
Although flow cytometry may be technically challenging, there are several advantages of multiparameter flow cytometry over other technologies that are used to identify rare cells. First, immunomagnetic separation either negatively or positively selects the population of interest, thereby enriching the population. After enrichment, the selected population stillrequires a detection method such as IHC, PCR, or flow cytometry for positive identification of tumor cells. Multiparameter flow cytometry performs the same function by generating a dump gate, where irrelevant cell populations maybe excluded (14). This technique also has the advantage over preselection as it preserves the quantitative accuracy of the analysis (14). Second, IHC and PCR are unable to distinguish viable cells from dead cells, where as impermeant fluorescent dyes such as 7-AAD or the amine-reactive dyes have increased fluorescent signals in dead cells that are easily distinguished by flow cytometry. Third, flow cytometry may be superior to IHC in the detection of weakly expressed antigens (11). Finally, IHC is semi-quantitative (11) where an estimate is given of the percent positive. In flow cytometry, data for thousands to millions of cells maybe collected and each event is accounted for in the analysis allowing quantitative as well as qualitative analysis at the single cell level in a relatively short period of time.
One critical task in flow cytometry that detects rare events is to enhance the signal to noise ratio (14). To maximize the signal, we needed to account for tumor-antigen expression that, in melanoma, varies according to stage and type of antigen (15). To improve the chances of tumor cell detection, antibodies against intracellular melanoma tumor antigens, gp100, Mart-1, S100 and tyrosinase were pooled and detected by an indirect PE stain and combined with an APC conjugated antibody against the surface antigen MCSP. Pooling of antibodies and detecting them with multiple PMTs is consistent with a study in breast cancer, where antibodies against EpCAM and cytokeratins were labeled with PE and FITC, respectively, to detect breast cancer in blood samples (16). The pooling was performed to decrease the possibility of false-negatives (16), and we saw similar data indicating improved detection of tumors when antibodies were pooled. Instead of FITC and PE, we used APC and PE to detect tumor antigens as APC and PE each have a very high signal to noise ratio where FITC has a lower signal to noise ratio (17). To allow for detection of cell viability, amine-reactive viability dyes were required to stain TIL populations as other viability dyes fail to discriminate viable cells in fixed samples (18). By excluding dead cells in our analysis (first dump gate), a major contribution to noise or high background staining was eliminated (14). A second dump gate to exclude CD3+ cells was used to eliminate potential nonspecific binding of antibodies to T cells. The exclusion of debris, dead cells and T cells by R3 often resulted in <1% of the TIL culture being analyzed for tumor cell detection. To decrease noise further, antibody titrations (19) were performed to minimize the background associated with nonspecific CD3− staining by intracellular tumor-antigens. In addition to optimizing the staining and gating strategy, we also ensured that carryover from previous tubes and instrument background were minimized (20). All of these techniques were used to maximize the signal to noise ratio.
Tumor cells were detected at culture initiation in the enzyme digest samples but not at culture termination in REP samples. Therefore, the final TIL population was tumor-cell free. Evidence to support the absence of tumor cells in REP populations, is built into the TIL culture itself, because manual inspections by light microscopy that count the number of TIL and tumor cells by virtue of their difference in size and shape were conducted throughout culture (data not shown). The lack of tumor cells by visual inspection defines the transition between pre-REP and REP (2, 3). A biological comparison control is potentially the most relevant control as it accounts for all spillover and background events that may come with the test sample (21). We used ATC as a biological comparison control, as it is similar to TIL in that ATC is composed mainly of T cells and NK cells. The difference is that ATC have had no previous exposure to tumor cells. However, the background fluorescent properties of ATC were not always the same as that found in TIL, possibly associated with different concentrations of IL-2 found in the two cultures. Regardless of the variations in culture, the background established by ATC was no different than those found in TIL giving further evidence that no tumor cells were found in TIL cultures. We also performed IHC on TIL cultures. In the seven REP samples tested, there was no detection of tumors (Data not shown). However, debris in the digested tissue tumor samples increased background staining which complicated our ability to identify tumor cells. Overall, the identification of tumor cells generated using flow cytometry was consistent with data generated from cell counts, biological controls and IHC demonstrating that no tumor cells were present in the final TIL culture (REP).
The development of this flow cytometric assay is a starting point for the assessment of TIL for contaminating melanoma cells. Although CD3+ T cells encompass the majority of cells in TIL cultures, CD56+CD3− NK cells were detected in several cultures. One improvement to the assay maybe the use of CD45+ cells instead of CD3+ cells to exclude all leukocytes. Gross et al. demonstrated that flow cytometry may reach the sensitivity of one per million cells (20) which is comparable to the best detection performed with immunohistochemistry (22). However, both of these were performed by spiking tumors into PBMC or bone marrow. Evaluating primary cells is more challenging as heterogeneous antigen expression is often not as robust as that found in melanoma cell lines. In our experience we were able to detect primary tumors in all enzyme digests tested and tumor cells were absent in REP cultures. These results suggest that flow cytometry maybe an alternative method used to evaluate TIL purity for use in adoptive transfer therapy.
The authors thank Anjali Choithani, Kate Dennert, Lynn Erickson, and Ruth Siebenlist for technical assistance.