Targeting Immune-Related Molecules in Cancer Therapy: A Comprehensive In Vitro Analysis on Patient-Derived Tumor Models

This study investigated the impact of immune-related pathway inhibition, among them indolamine 2,3-dioxygenase (IDO), alone and together with immune cells on growth and viability of colorectal cancer (CRC) cells. A panel of patient-derived CRC cell lines with different molecular characteristics (CpG island methylator phenotype, chromosomal, and microsatellite instability) was included. Initial phenotyping of CRC cell lines (n=17) revealed high abundance of immunosuppressive checkpoint-molecules in general, but an individual profile for IDO. Presence of immune-related molecules was independent of the molecular subtype. Selective treatment of CRC cell lines showing high or low IDO expression (n=2 cell lines each) was performed with single agents and combinations of Indoximod, Curcumin, and Gemcitabine with and without the addition of peripheral blood lymphocytes (PBL) in an allogeneic setting. All substances affected CRC cell growth in a cell line specific manner. The combination of Curcumin and Gemcitabine proved to be most effective in tumor cell elimination. Functional read-out analyses identified cellular senescence, after both single and combined treatment. Curcumin alone exerted strong cytotoxic effects by inducing early and late apoptosis. Necrosis was not detectable at all. Addition of lymphocytes generally boosted antitumoral effects of all IDO-inhibitors, with up to 80 % cytotoxicity for the Curcumin treatment. Here, no obvious differences became apparent between individual cell lines. Combined application of Curcumin and low-dose chemotherapy is a promising strategy to kill tumor target cells and to stimulate antitumoral immune responses.


Introduction
Immune-checkpoint inhibitors constitute one of the most promising novel therapeutic approaches for cancer [1]. These molecules reconstitute the hosts' antitumoral immune response by interrupting tumor-induced tolerance and are now at the forefront of immunotherapy development. Unlike great advances in some tumor types including melanoma and non-small cell lung cancer, immunotherapy of colorectal cancer (CRC) remains challenging due to the broad clinicopathological and molecular heterogeneity [2]. Three molecular pathways have been implicated in colorectal tumorigenesis: chromosomal instability (CIN, ∼60 %), CpG island methylator phenotype (CIMP, ∼30 %), and microsatellite instability (MSI, ∼15 %). This latter subgroup is more likely to respond to immunotherapy [3]. An ultrahigh mutational load due to accumulating insertions/deletions in short repetitive sequences (=microsatellites) constitutes the underlying molecular mechanism and Vice versa; their high immunogenicity forces MSI + tumor cells to escape the otherwise effective immune attack by creating an immunosuppressive microenvironment. Additionally to downregulate major histocompatibility complex (MHC) class I on the tumor cells' surface and inactivation of the antigen-processing machinery, upregulation of immune-checkpoint-molecules (such as

Material and Methods
. . Tumor Cell Lines and Treatment. Patient-derived HROC cell lines (nomenclature: HRO, Hansestadt Rostock; C, Colon; T, transfer; M, mouse) were established and characterized in our lab [18, Table 1 and unpublished]. Authenticity for these lines was verified by means of short tandem repeat fingerprinting (comparison of cell lines at different passages, matched normal tissue, as well as corresponding B cells) from genomic DNA according to [21]. For analyses, cells in passages <40 were maintained in complete medium: DMEM/F12 supplemented with 10 % fetal calf serum, glutamine (2 mmol/l) and antibiotics (PAN-Biotech GmbH, Aidenbach, Germany). Cells were seeded at the appropriate density for each cell line and were incubated 24 h prior to treatment. For all in vitro experiments, the following substances and their combinations were used in these concentrations: 11.5 휇M Indoximod, 1 휇M Gemcitabine, and 20 휇M Curcumin. Antitumoral effects were examined after 72 h of incubation.
. . Functional Analysis of Substance-Mediated Growth Inhibition via Crystal Violet Staining. All experiments were performed in 96-well plates in triplicate and replicated at least three times. Treated cells were stained with crystal violet (0.2 %, 10 min, RT). After adding sodium dodecyl sulfate (1 %) the absorbance was measured at 570 nm (Tecan Trading AG, Männedorf, Switzerland). For sequential treatment, cells were incubated for 72 hours with Indoximod. After removal of media, the cells were incubated for another 72 hours with the test substances and combinations prior to crystal violet staining as described before. Finally, drug effects from triplicate wells were determined in comparison to untreated controls (=set to be 100 %), measured at 570 nm (reference wavelength: 620 nm).
. . RNA Isolation, cDNA Synthesis, and Quantitative Real-Time PCR. RNA from treated cells was isolated with "Gene Matrix Universal RNA Purification Kit" from EURX5 (Gdansk, Poland) according to the manufacturers' instructions. RNA was reverse-transcribed into cDNA from 1 휇g RNA using 1 휇l dNTP mix, 1 휇l oligo (dT)15 Primer, 1 휇l reverse transcriptase, and 4 휇l 5x RT buffer complete (all purchased from Bioron GmbH, Ludwigshafen, Germany). A volume of 20 휇l was filled with RNAse free water. cDNA synthesis conditions were as follows: 25 ∘ C for 10 min, 37 ∘ C for 120 min, and 85 ∘ C for 5 min. Target cDNA levels of human cell lines were analyzed by quantitative real-time PCR using TaqMan Universal PCR Master Mix and predesigned TaqMan gene expression assays for serine/threonine-specific protein kinase Ataxia Telangiectasia Mutated (ATM), CCNE (encoding the cyclin E protein), cyclin-dependent kinase inhibitor 2A (CDKN A), Murine double minute 2 (MDM ), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; housekeeping gene as control) in the light cycler Viia7 (Applied Biosystems, Foster City, USA). PCR conditions were as follows: 95 ∘ C for 10 min, 45 cycles of 15 s at 95 ∘ C, and 1 min at 60 ∘ C. Reactions were performed in triplicate. Expression levels of the gene of interest were calculated in relation to the housekeeping gene (ΔCT = CT target -CT GAPDH ). Relative gene expression values are expressed as 2-(ΔΔCT), resulting from the difference between ΔCT target -ΔCT Calibrator . DMSOtreated cells were used as calibrator.
. . Analysis of Senescence via Light Microscopy. Experiments were performed in 48-well plates replicated three times using the senescence 훽-galactosidase staining kit from Cell Signaling Technology, Inc. (Beverly, USA) according to the . . Detection of Autophagy. Tumor cells were cultured on cover slips in 24-well plates (2.5 x 10 4 cells/well). Upon treatment with the indicated substances/substance combinations, residual cells were stained with a mixture of acridine orange (4 mg/mL; Applichem, Darmstadt, Germany) and calcein AM (1.6 휇M; AAT Bioquest, CA, USA) for 10 min at RT according to the protocol described in [22]. Slides were analyzed on a laser scanning microscope (Zeiss) using 20x objectives.
. . Cell Cycle Analysis via Flow Cytometry. Treated cells and cell culture supernatants were harvested, collected in FACS tubes, and washed in 1x PBS (8 min, 200 x g). Then, supernatant was discarded and cells were resuspended in icecold ethanol (70 %), followed by incubation at -20 ∘ C for at least 12 hours. Afterwards, the cells were pelleted again by centrifugation (8 min, 200 x g), washed with 1x PBS and the cell pellet was resuspended in 1x PBS/0.1 % Tween 20 (Sigma-Aldrich Chemie GmbH, München, Germany)/RNase (1 mg/mL). Cells were incubated for at least one hour at -4 ∘ C. Following the addition of 50 휇g propidium iodide (PI, 0.1 mg/mL) (AppliChem GmbH, Darmstadt, Germany) cells were subjected to cytofluorometric analysis. Cell cycle analysis was performed on a BD FACS Calibur (BD) using the software Cell Quest Pro (BD). 10.000 events were measured for each sample. Data quantification was done by applying the BD ModFit LT Software for cell cycle analysis. Experiments were replicated three times.

. . Discrimination between Apoptosis and Necrosis via Flow
Cytometry. Treated cells and cell culture supernatants were stained with YO-PRO-1 (0.2 휇M, 20 min, RT) and washed. PI (0.1 mg/mL) was added before performing the analysis on a BD FACS Calibur (BD). Cell staining and analysis was done as described [23].
. . Coculture with Immune Cells. Peripheral blood lymphocytes (PBLs) were obtained from healthy volunteers following pancoll (PAN-Biotech GmbH) density-gradient centrifugation. Tumor cells seeded in 24-well plates were treated with the test substances in the presence or absence of PBLs (ratio tumor cell to PBL 1:6). After 72 h of incubation, PBLs were removed by aspirating supernatant and remaining tumor cells were mixed with fluorescent beads (7x10 3 beads/sample) (Fluoresbrite6Plain Beads, Polysciences Inc., Warrington, USA). PI (0.1 mg/mL) was added before performing the flow cytometric analysis on a FACSCalibur Cytometer (BD Pharmingen). Therefore, cells per 5,000 beads (= gate 1) were counted in the FSC/SSC dot plot. Cells cocultured with PBLs but without addition of test substances were used as control. Data analysis was performed using BD CellQuest6 Pro software.
. . Statistical Analysis. All values are expressed as mean ± SE. After proving the assumption of normality, differences between controls and experimental samples (=treated cells) were determined by using the unpaired Student's ttest (SigmaPlot 12.5). The criterion for significance was set to p < 0.05. Correlations were determined using Pearson correlation coefficient (r), with a range between -1 and +1. A value of 0 indicates that there is no association between the two variables. A value greater than 0 indicates a positive association (+1 = strong correlation).

Results
. . Phenotyping of Immune-Checkpoint-Molecules. The immune-phenotype of 17 low-passage CRC cell lines, covering the three main molecular subtypes, was characterized. All cell lines displayed high amounts of immunosuppressive molecules, such as PD-L1/2 ( Figure 1(a)). Similarly, CTLA-4 and B7-H3 were detected in high abundance (always >90 %). IDO expression was heterogeneous among cell lines, ranging from < 5 % (HROC278 T0 M1) to > 60 % (HROC59 T1 M1). Contrary to that, costimulatory molecules were not detectable at all (i.e., CD27L, CD80, and CD86) with the exception of CD40L, which was heterogeneous. HROC40, HROC60, HROC183, and HROC113 T0 M1 cells were found to be MHC class I-positive and HROC257 T0 M1 showed positive staining for MHC class II. The remaining cell lines did not express MHC molecules. A comprehensive expression profile is given in Figure 1(a).
Indoximod alone did not impair cell viability significantly. Cytotoxic effects, if any, were detectable at high doses (> 20 휇M), but this did not reach statistical significance (data not shown). Of note, there was an inverse correlation between IDO abundance and response to high dose treatment with Indoximod (≥ 50 휇M). HROC50 T1 M5 responded best but showed lowest IDO expression in flow cytometry (35 % biomass reduction versus 5 % IDO1 expression, data not shown); conversely, HROC257 T0 M1 had high amounts of IDO1 (65 %) but did not respond to Indoximod-mediated growth inhibition (5 % biomass reduction versus control, data not shown). Curcumin had dose dependent cytotoxic activity (up to 80 % at 50 휇M). Gemcitabine was applied at subtherapeutic doses as determined in previous experiments [  because of its promising antitumor activity when given in conjunction with other therapeutics. Substance combinations enhanced antitumoral effects in almost all cases (Figure 1(b)). Best killing activity was obtained after treatment with Curcumin plus Gemcitabine with up to 70 % biomass reduction in MSI + cell lines. CIMPassociated cell lines' response was weaker.
Sequential treatment (preincubation with Indoximod for 72 h) did not boost effects of the monotherapies. Thus, the direct cytotoxic potential of Indoximod on CRC cells was rather low. Even interferon-훾 pretreatment, described to induce IDO expression and rendering cells more vulnerable to cytolysis [27], did not increase Indoximod-mediated growth inhibition (data not shown).
. . Gene Expression Analysis. Next, therapy-induced gene expression changes were analyzed. Indoximod monotherapy at a dose of 11.5 휇M led to a heterogeneous expression profile, with an increased expression of DNA-damage response gene ATM in MSI + cell lines HROC257 T0 M1 and HROC50 T1 M5. Expression of CDKN A and CCNE , both being involved in cell cycle regulation, was largely unchanged (Figure 2(a)). Lower doses had no influence on gene expression at all (Figure 2(a)). Then, the gene expression profile of HROC50 T1 M5 cells after various treatments (i.e., Indoximod, Curcumin, Gemcitabine, Indoximod/Curcumin, Indoximod/Gemcitabine, and Curcumin/ Gemcitabine) was investigated to see if Indoximod-induced alterations are preserved or changed in the combination. Most significant changes were identified upon combined Curcumin/Gemcitabine treatment, resulting in a 5-fold induction of ATM and CDKN A (p < 0.05 versus control). CCNE and MDM were also upregulated in this combination (Figure 2(b)).
. . Induction of Senescence by Test Substances. We next aimed to unravel the underlying mechanisms of biomass reduction under treatment. Autophagy induction was studied by immunofluorescence based on staining of acidic vesicles with acridine orange. Frequency of autophagy was quite low (Figure 2(c)). By contrast, all substances and combinations caused senescence in HROC tumor cells with a tendency towards higher 훽-galactosidase activity after single treatment ( Figure 3). Most pronounced effects were evident in HROC183 T0 M2 (IDO low ) and HROC257 T0 M1 (IDO high ) cells. . . Altered Cell Cycle upon Treatment. Since senescence is associated with cell cycle arrest, flow cytometric Nicoletti staining was performed next. Almost all experimental conditions readily revealed an altered cell cycle accompanied by an increased amount of dead cells. This was most pronounced after Curcumin treatment, of note, both alone and in combination (Figure 4). Numbers of cells in G0/G1 phase decreased (HROC257 T0 M1 > HROC50 T1 M5 > HROC60 > HROC183 T0 M2; a representative example is given in Figure 4(a) showing Curcumin/Gemcitabine treated HROC257 T0 M1 (IDO high ) cells; sub-G1 defines dead cells). The best antiproliferative effect was inducible in HROC183 T0 M2 (IDO low ) cells after Indoximod/Curcumin treatment (Figure 4(b)). Percentages of cells in S-phase were 17 %± 9 % versus 33 %± 11 % (control).
. . Discrimination between Apoptosis and Necrosis. The above described findings identify cell death as main treatment effect. To test whether apoptosis, necrosis, or a mixture of both is responsible, Yo-Pro-1/PI staining was applied ( Figure 5(a)). In all cases, apoptosis was the underlying mode of cell death (Figure 5(a)). Curcumin was most effective in inducing cellular apoptosis (at least 2-fold increase in all cell lines). In HROC257 T0 M1 (IDO high ) cells, apoptosis was two to three times higher than in controls (early apoptosis: 36 ± 20 % versus 11 %± 2 % control and late apoptosis: 26 %± 10 % versus 15 %± 8 % control). Adding Gemcitabine to Curcumin further enhanced cytotoxic effects ( Figure 5(a)).

. . Coculture with Immune Cells Boosts Antitumoral Effects.
IDO has differential effects on tumor and immune cells and thus can impair T-cell mediated tumor killing. In an in vitro coculture system, consisting of immune effector and tumor target cells, the potential of the different therapeutics to block IDO-induced negative immune effects was subsequently analyzed (Figure 5(b)).
All substances reduced tumor cell numbers in this test system. The best cytotoxic effect could be induced by Curcumin, resulting in a massive tumor cell reduction in all four cell lines (> 80 % versus control). Of note, combining Curcumin either with Indoximod or Gemcitabine even enhanced this toxic effect with nearly complete elimination of tumor cells (Figure 5(b)). Best tumor cell responder was the HROC60 (IDO high ) cell line. However, even with 72-hour incubation time, a specific antigenic activation is unlikely to occur and thus the observed effects are most likely due to a more unspecific stimulation of lymphocytes by the tested drugs.

Discussion
In this study, we describe (I) the expression profile of immune-modulating molecules on a panel of molecularly well-characterized patient-derived CRC cell lines and (II) the effects of IDO-inhibition directly on CRC cells as well as in coculture experiments with immune cells.
Phenotyping analyses revealed regularly high expression of immune-checkpoint-molecules PD-L1/2, B7-H3, B7-H4, and CD152 confirming the tumors' natural immunosuppressive character. IDO abundance was quite heterogeneous between cell lines and independent of their molecular phenotype. Studies on human CRC material similarly describe varying IDO expression in vivo [28]. Of note, abundance correlates inversely with the number of tumor-infiltrating CD3 + lymphocytes and clinical patients' outcome [29,30]. In vitro, IDO decreases tumor cell proliferation and mediates mediating drug resistance [31,32]. While an association of cell growth with IDO expression was not obvious in our CRC lines (data not shown), we indeed found some overlap between IDO level and sensitivity towards clinically used drugs, such as 5-fluorouracil (5-FU; Pearsons r = 0.47, moderate positive relationship, Table 2). Hence, IDO plays a role in several pathways, making this molecule an interesting target for in vitro evaluation of a multimodal treatment strategy. Specific IDO blockade with Indoximod identified an inverse correlation between biomass reduction and IDO1 abundance in our patient-derived cell lines. Of note, IDO low expressing cells responded better towards this treatment, supporting recent findings that Indoximod likely targets additional pathways. The precise mechanism of action is not completely known. Indoximod was described to interfere with IDO translation and transcription as well as to inhibit the tryptophan-transporter in the plasma membrane [33]. Additionally, blocking IDO may sensitize tumor cells to certain chemotherapeutics and radiation by altering cell cycle (G1 㨀→ G2/M shift) [32]. However, in the present study, no chemosensitizing effects of Indoximod-"preconditioning" were detectable. Cytotoxic effects of Irinotecan, 5-FU, and Gemcitabine were also not boosted (data not shown).
The missing direct cytotoxic potential of Indoximod monotherapy in vitro encouraged us to use combinations with other IDO-blocking substances. Curcumin and Gemcitabine were selected based on the following criteria: (I) interference with IDO; (II) direct cytotoxic activity against tumor cells (via induction of immunogenic cell death) in (pre-)clinical studies, (III) safe application with minimal toxic side effects (at least in low-doses) in patients, and (IV) enhanced antitumoral effects when given in conjunction with other drugs. These experiments identified Curcumin as very promising candidate for IDO-blocking cancer therapy. Biomass of tumor cells significantly reduced and oncolytic effects even increased when Curcumin was combined with Gemcitabine, a quite surprising finding. In functional analyses, senescence and apoptosis could be identified as underlying molecular mechanisms. Effects were cell line and treatment specific, but again largely independent from IDOexpression status. Cytotoxicity of Curcumin is mainly based on apoptosis induction, without significant impairment of the cell cycle. This was somewhat unexpected, since shorttime treatment with Curcumin has been reported to decrease cyclin D1 expression and to induce G1-/S-phase arrest [34,35].
Finally, a coculture system mimicking tumor-immune cell interactions was applied to test the impact of IDOinhibition for reconstitution of immune function. In vivo, IDO blockade by Indoximod breaks immune tolerance by increasing the level of inflammatory molecules, such as C-reactive protein and IL6 [10]. We could confirm such immune stimulating effects of Indoximod in 2/4 cell lines, but this was again rather independent of IDO-expression levels. Addition of Curcumin efficiently boosted the immunemediated tumor cell elimination. This antitumoral effect is most likely attributable to unspecific immune stimulation by Curcumin and thus supports recent results of others [16,36]. Gemcitabine, similar to Indoximod, also stimulated immunemediated killing. Here again, NK cells are very likely the main effector cells [37]. However, effects were considerably lower than those of Curcumin. With the known toxic side effects of Gemcitabine, our findings argue in favor of Curcumin for further development of combinatorial treatment strategies. Interestingly, there was a trend towards highest susceptibility of HROC60 cells, established from a 71-year-old male patient with a low-grade CIMP tumor in the colon ascendens. HROC60 cells show high abundance of immunosuppressive molecules, i.e., IDO, PD1/PD-L1. It is therefore of particular interest whether such Curcumin-based combinations provide a real opportunity for selected CRC patients in the future.
However, it is worth mentioning that coculture experiments were performed in an allogeneic setting using naïve and thus "unprimed" immune cells from healthy volunteers, without tumor-induced T-cell dysfunction. Prospective studies will have to test if comparable strong oncolytic effects can be achieved in an autologous setting with partially exhausted lymphocytes from cancer patients.
Although we were unsuccessful to directly link IDO-level with treatment response, we would like to bring forward the argument that this molecule plays a central role in tumor-induced immunosuppression. To this end, not only