Cancer Cell-targeted and Activatable Photoimmunotherapy Spares T Cells in a 3D Co-culture Model

Abstract

Photodynamic therapy (PDT) is an established therapeutic modality that uses non-ionizing near infrared light to activate photocytotoxicity of endogenous or exogenous photosensitizers (PSs). An ongoing avenue of cancer research involves leveraging PDT to stimulate anti-tumor immune responses; however, these effects appear to be best elicited in low-dose regimens that do not provide significant tumor reduction using conventional, non-specific PSs. The loss of immune enhancement at higher PDT doses may arise in part from indiscriminate damage to local immune cell populations, including tumor-infiltrating T cells. We previously introduced “tumor-targeted, activatable photoimmunotherapy” (taPIT) using molecular-targeted and cell-activatable antibody–PS conjugates to realize precision tumor photodamage with microscale fidelity. Here, we investigate the immune cell sparing effect provided by taPIT in a 3D model of the tumor immune microenvironment. We report that high-dose taPIT spares 25% of the local immune cell population, 5 times more than the conventional PDT regimen, in a 3D co-culture model incorporating epithelial ovarian cancer cells and T cells. These findings suggest that the enhanced selectivity of taPIT may be utilized to achieve local tumor reduction with sparing of intratumor effector immune cells that would otherwise be lost if treated with conventional PDT.

GRAPHICAL ABSTRACT

A targeted, activatable photosensitizer immunoconjugate was synthesized and compared to its untargeted analogue. We hypothesized that cancer cells that overexpress epidermal growth factor receptor preferentially take up and activate the immunoconjugate compared to surrounding cell types, whereas the untargeted photosensitizer is indiscriminately distributed. This was tested in a 3-dimensional co-culture model of human ovarian cancer cells and murine T cells. We showed that T cells are spared after photodynamic treatment with the targeted immunoconjugate compared to the conventional therapy.

INTRODUCTION

Photodynamic therapy (PDT) using conventional photosensitizers (PSs) has been shown to stimulate acute inflammation, tumor-antigen exposure and anti-tumor immunity in a number of preclinical studies (1). PDT induces the release of pro-inflammatory cytokines (1, 2) as well as the exposure of tumor-antigens to the host immune system (1, 3). However, high-dose PDT regimens that elicit maximal local tumor reduction interfere with anti-tumor immunity (4) possibly due to indiscriminate damage to tumor-infiltrating effector immune cell populations as well as tumor microvasculature shutdown that could impede immune cell trafficking in and out of the tumor. Gollnick and colleagues introduced a two-step PDT regimen where an initial low-dose promotes immune-enhancing effects and a follow-up high-dose PDT regimen to reduce the tumor volume once anti-tumor immunity is activated (4). High fidelity effector immune cell sparing and rescue (5) is further motivated by the correlation between the presence of tumor-infiltrating T cells and improved patient outcomes for several malignancies (6–10).

Photoimmunotherapy (PIT) using antibody (Ab)–PS conjugates is a targeted form of PDT (11–22) that is in clinical trials for cancer therapy (e.g., NCT02422979). PIT has traditionally been applied using clinical tumor biomarker targeted antibodies, such as cell-surface epidermal growth factor receptor (EGFR) targeting using cetuximab (23) or panitumumab (17). Recently, PIT has been applied to target tumor-associated regulatory T cells (24) and carcinoma-associated fibroblasts (25) in preclinical models with promising results suggesting enhancements of anti-tumor immunity by disrupting mechanisms of tumor immunosuppression. However, it is not clear if PIT preserves off-target, tumor-infiltrating effector immune cells or not. The data presented by Sato et al. (24) suggests sparing of tumor-infiltrating T cells and natural killer cells, but the analysis was performed 30 minutes after PIT such that many direct and indirect cell death pathways induced by PIT (photodynamic action) (26–28), or any mode of therapy, would not have sufficient time to take their course. It is well-known that molecular and cellular signatures of apoptosis, as well as tumor volume reduction, do not appear for several hours after therapeutic stress in monolayer and 3D cell cultures (29–31) as well as in vivo (32, 33). In addition, a control group using conventional, unconjugated PS was not performed to help interpret the results (34). Therefore, exploration of preserving tumor-infiltrating effector immune cells is still needed in model systems that facilitate probing cell death quantitatively.

A limitation of conventional PDT and PIT is that the PS, or the antibody–PS conjugate, accumulates non-specifically in off-target tissues where the PS remains “always-on” and imparts off-target photodamage when the illumination is not confined to the tumor site. This leads to the risk of unwanted damage to the surrounding tissues, especially when diffuse light is used to treat disease with metastatic spread. We recently introduced molecular-targeted, cell-activatable antibody–PS conjugates that remain quenched (“off”) in off-target sites to protect vital organs, and which are activated (turned “on”) upon cancer cell binding, internalization and lysosomal catabolism (35). These photoimmunotherapeutic constructs (PICs) overcome bowel phototoxicity to enable high-dose PIT in the peritoneal cavity of a micrometastatic mouse model of ovarian cancer (with the photodynamic dose estimated as the product of the PS and light administered: 2 mg·kg⁻¹ PS×50–100 J·cm⁻¹ per quadrant) (36) whereas “always-on” antibody–PS conjugates have an order-of-magnitude reduced maximum tolerated dose (14, 18, 37). This new therapy is termed tumor-targeted, activatable PIT (taPIT) with enhanced selectivity and safety for treating microscopic disease compared to conventional PIT.

Here, we examined the tumor cell-type fidelity of taPIT in a 3D co-culture model consisting of human epithelial ovarian cancer cells (EOCCs) and T cells isolated from mouse splenocytes. Although unconventional, a non-syngeneic model was chosen to reduce confounding mechanisms of cell death via intercellular interactions. Specifically, the selected murine model incorporates transgenic T cell receptors designed to recognize ovalbumin, which is not known to be expressed by the EOCC cell line used in the co-culture. Thus, direct T cell cytotoxicity to tumor cells is not likely and tumor reduction may be attributed to light-activation. Each cell type was genetically engineered to express fluorescent protein tracers to enable quantitative analysis of surviving cell populations following treatment. We compare taPIT with its unconjugated “always-on” control to elucidate treatment effects on each cell type and allow sufficient time to elapse to permit complete cell death processes to occur. This opens the door to implementing a single, high-dose taPIT regimen that facilitates both local tumor reduction and preservation of local effector immune cells in place of the two-step regimen required for conventional PDT.

MATERIALS AND METHODS

All methods involving cell culture were performed in a sterile biosafety cabinet using aseptic technique. Cell growth and incubation occurred in a humidified incubator at 37°C and 5% CO₂.

PIC synthesis.

Cetuximab (Cet, Erbitux) was conjugated to benzoporphyrin derivative monoacid A (BPD, verteporfin) at a 1:7 Ab-PS ratio (Cet-BPD) according to previous protocols (21, 22, 35). Briefly, the N-hydroxysuccinimide ester of BPD was reacted with Cet coated in methoxypolyethylene glycol (mPEG, Sigma, 85976) in dimethyl sulfoxide (DMSO, Sigma, D2650) at a molar ratio of 9:1 NHS–BPD:mPEG–Cet. The product was purified using a 7kDa molecular weight cutoff (MWCO) Zeba spin desalting column (Thermo Scientific, 89893), and again with a 30 kDa MWCO centrifugal filter (Amicon, UFC903024) in a final solution of 5% DMSO in phosphate buffered saline (Gibco, 20012027). This protocol yields highly pure PIC concentrate with no more than 5% residual free photosensitizer (22).

PS characterization.

Final concentrations of Cet and BPD within the PIC solution were determined independently to determine the PS loading ratio. Final Cet concentration was measured by BCA Protein Assay (Pierce). BPD concentration was assessed by absorption spectrophotometry using a UV-Visible spectrophotometer (Thermo Scientific Evolution 300) and the known molar extinction coefficient of BPD at 687 nm ε₆₈₇ = 34,859 M⁻¹ cm⁻¹. Similarly, the stock concentration of unconjugated BPD (Visudyne®, liposomal BPD, Bausch + Lomb) was 3.14 mM. The results of these measurements are summarized in Table 1. PIC and BPD solutions were stored protected from light at 4°C until use.

Table 1.

PIC characteristics.

Target (Antibody)	EGFR (Cetuximab)
Photosensitizer (PS)	BPD (Verteporfin)
Photodynamic Excitation (nm)	690
Optimal PS:Ab Loading Ratio (21)	7–10 : 1
Ab Concentration (μM)	17.5 ± 2.9
PS Concentration (μM)	109.0
Measured PS:Ab Loading Ratio	6.2 ± 1.0 : 1

Cancer cell model.

EOCC line NIH:Ovcar5 (Ovcar5) of human origin was obtained from Fox Chase Cancer Center (Philadelphia, PA, USA) under a Material Transfer Agreement (MTA). A derivative of Ovcar5 cells, EGFP–Ovcar5, that stably expresses EGFP (enhanced green fluorescent protein), was created in this study to monitor 3D spheroid growth and PDT treatment response following a previously published protocol (38). Briefly, the wild-type Ovcar5 cells were infected with lentiviruses encoding an EGFP gene. To generate the infective lentiviral particles, a third generation EGFP transfer plasmid pHAGE-CMV-EGFP-W (Harvard Plasmid Repository, EvNO00061634) was co-transfected along with a viral packaging plasmid psPAX2 (Addgene, 12260) and a viral envelope plasmid pMD2.G (Addgene, 12259) into packaging cell line Lenti–X™ 293T at a molar ratio of 2:1:1 using X–fect transfection reagent (TaKaRa, 631317). Viral particles were released into the supernatant 48 hours after the transfection, harvested by pooling the supernatant, and filtered through a 0.22 μm polyethersulfone filter (GE Healthcare, 67802504) to remove cellular debris. Ovcar5 cells in 2D monolayer were transduced with the viral suspension at low multiplicity of infection in the presence of 6 μg/ml polybrene (hexadimethrine bromide, Sigma H9268). Transduced cells that express EGFP were sorted using a FACS Aria cell sorter (BD Biosciences) after two days of culturing following the transduction. The sorted cells were expanded and cryopreserved in fetal bovine serum (FBS, Gibco™, 10437028) with 8% DMSO. All cell lines used in this study tested negative for Mycoplasma contamination via MycoAlert™ PLUS Kit (Lonza, LT07–710).

2D cell culture.

Monolayer Ovcar5 and EGFP–Ovcar5 cells were cultured in 75 cm² flasks (Corning, 353136) in RPMI 1640 medium (Corning Cellgro, 10–040-cv) with 10% (v/v) heat inactivated FBS (Hyclone, SH30071.03HI) and 1% (v/v) 1X penicillin–streptomycin (Gibco™ 15140122) (SuppRPMI). At 90% confluency, cells were lifted with Trypsin–EDTA (Corning, 25–053-CI), washed, and re-plated at a 1:10 dilution. Cultures were discarded after 28 passages and new vials were thawed as needed.

T cell culture and priming.

DsRed fluorescent protein-expressing OT–1 mice (C57BL/6–TgTcra/Tcrb homozygous, 1100 Mjb/J) were a generous gift from Professor Mei Wu (Wellman Center for Photomedicine, Massachusetts General Hospital and Harvard Medical School, Boston). Mice 10 weeks of age or older were euthanized by CO₂ asphyxiation following an approved IACUC protocol. Spleens were collected aseptically immediately after euthanasia and T cells were primed following a published protocol (39) with minor modifications. Briefly, red blood cells in the splenic suspension were depleted by incubating with ammonium–chloride–potassium lysing buffer (Gibco, A1049201) for 4 minutes. Three days before addition to 3D cultures, T cell blasts were generated in 6–well plates (Corning, 353046) by inoculating 0.75 μg/mL of SIINFEKL peptide (ovalbumin sequence 257–264, New England Peptide, BP10–915) into 1×10⁶ splenocytes/mL for 3 days in the presence of 10 U/ml recombinant murine interleukin–2 (IL–2) cytokine (Peprotech Inc., 212–12). This protocol yields a >90% pure solution of CD8+ T cells (39, 40) and was not purified further. Cells were washed twice in RPMI 1640 media before addition to 3D cultures. In all cases, T cells were maintained in SuppRPMI medium further supplemented with 50 μM β–mercaptoethanol (Sigma, M3148) and 10 U/mL recombinant murine IL–2 (T cell RPMI).

3D cell culture.

3D cultures of EGFP–Ovcar5 nodules were developed with guidance from a previous protocol (41) on beds of Growth Factor Reduced Matrigel® (GFRM) Matrix Basement Membrane (Corning, 354230, lot# 8143001). GFRM was thawed overnight at 4°C on ice. 250 μL of 4°C GFRM was added to each well of a black-walled, glass-bottomed 24-well plate (Greiner Bio-One, 662892). The pre-chilled plate was kept on a flat ice-pack to keep GFRM cold during plating, pipetting slowly to avoid bubble formation and swirling gently to ensure homogenous well coverage. The plate was then incubated for 30 minutes to polymerize the GFRM, after which 500 μL of a 4% GFRM in SuppRPMI solution was added to each well to prevent further drying of the gel. During GFRM incubation, EGFP–Ovcar5 cells were harvested, washed, and re-suspended at 20,000 cells per mL in SuppRPMI. 500 μL of cell solution was added to each well, for a final concentration of 10,000 cells in 1 mL of 2% GFRM–SuppRPMI. Cells were incubated for 7 days to allow nodule formation. Media was changed on day 4 by carefully aspirating the media above the GFRM bed and replacing with 1 mL of 2% GFRM–SuppRPMI.

One day before light activation, T cells were harvested, washed, and re-suspended in T cell RPMI at 80,000 cells per mL. Media was carefully aspirated from each well and replaced with 1 mL of the T cell suspension. This represents a moderate effector-to-target (E:T) ratio of 8:1 informed by prior in vitro experiments (42, 43).

Light activation.

3D co-cultures were incubated with T cell RPMI containing 1 μM BPD–equivalent Cet–BPD for 24 hours (for taPIT) or 1 μM unconjugated BPD(BPD–PDT) for 90 minutes prior to light activation. In both cases, media was refreshed with PS–free T cell RPMI after incubation. Following previously described methods (41, 44), 690 nm light was administered vertically through the plate bottom via a diode laser source at 150 mW/cm² irradiance. The total energy deposit (light dose) was given at 0, 1, 3, 10, 30, and 60 J/cm² (three biological replicates per group including two co-cultures and one mono-culture as a control for cancer–immune cell signaling cross-talk) by varying the exposure time using custom code to control a programmable TTL shutter. Plates were incubated post light administration for three days to allow complete development of cell death processes.

Imaging.

Plates were imaged with an Operetta CLS High–Content Analysis System (Perkin Elmer, LIVE configuration) with the incubation chamber set to 37°C and 5% CO₂. EGFP (ex/em: 489/510 nm) and dsRed (ex/em: 545/572 nm) fluorescence was collected via LED excitation and a 5×, 0.16 NA air objective. 3D volumes were imaged in a z–stack format with four fields per image plane over 14 planes with 70 μm separation, collecting a total volume of 38 mm³. Two fluorescence plus one brightfield z–stack were collected for each well using Harmony 4.6 (Perkin Elmer), an automated plate reading software.

Image analysis.

Images were analyzed via custom MATLAB (MathWorks) script that applies the following analysis scheme to each well from each fluorescence channel. First, each image field was reassembled by concatenating four raw images in each plane. The full z–stack was reduced to a 2D image via maximum intensity projection (MIP) to analyze in–focus light, from which a fluorescence histogram was calculated. MIPs were dominated by background signal due to the wide field of view, so the mean background signal from the fluorescence control wells (in which no fluorescent cells were plated) was subtracted from the histogram of each well to center the background distribution about zero. Finally, the mean fluorescent protein fluorescence of the corrected histograms was computed and used as a measure of cell viability.

Statistics.

Unnormalized treatment response was fit to the four-parameter Hill equation to assess the viability of each cell type in response to taPIT vs. conventional PDT. Data and fits were then normalized using the mean no-treatment control group fluorescence (100% viable) and empty well controls (0% viable). Dose-response fits were compared with an extra sum-of-squares F test to determine if the EC₅₀ differed between data sets. Individual treatment groups were compared via unpaired, two-tailed t-test. For both tests, the significance level was set to α = 0.05. Sparing of T cells vs. induction of tumor nodule death was quantified by dividing the T cell response by the tumor response (both data points and fits), which we interpret as the average number of remaining viable T cells per viable tumor cell. All calculations and error propagation were carried out using Prism 8 (GraphPad) statistical analysis software. Data are presented as mean ± S.E.M unless otherwise stated.

RESULTS AND DISCUSSION

We demonstrate for the first time, in an intuitive 3D co-culture model of the tumor immune microenvironment, that taPIT spares T cells at light doses that effectively eliminate the disease, whereas the equivalent, untargeted therapy actually destroys more T cells than cancer cells. The cell-type response was evaluated using total fluorescence intensity as a proxy for cell viability.

Maximum intensity projections of the 3D volumes (Fig. 1) reveal in general that EOCC nodules were more sensitive to taPIT, whereas, T cells were more sensitive to BPD–PDT. To quantify this response, mean fluorescence intensity was fit to a standard four–parameter dose response model and analyzed (Fig. 2). For both cell types, an extra sum-of-squares F test determined that the EC₅₀ did not differ between therapies, suggesting similar toxicity profiles and cell death mechanisms among BPD formulations at low to moderate light doses. Specifically, the EOCC nodules responded equivalently to each treatment modality (for both data sets, EC₅₀ = 18.7 ± 2.2 J/cm²). However, at the highest light dose, 60 J/cm², nodules treated with taPIT were significantly less viable (8.1 ± 1.2%) than those that received free BPD (14.4 ± 1.8%) (Fig. 2a).

Figure 1.

3D EOCC-T cell co-culture response to targeted vs. untargeted PDT. Maximum intensity projections of 3D tumor nodules (green) and T cell (magenta) co-cultures 72 hours after treatment with taPIT (Cet–BPD, left column) or untargeted PDT (BPD, right column). All images are set to the same intensity scale; arrows indicate dim tumor nodules. Scale bar, 200 μm.

Figure 2.

TaPIT spares T cells compared to untargeted PDT. (a) The EOCC dose response to taPIT (Cet–BPD) was not significantly different than the BPD–PDT response at moderate light doses. Significantly more tumor nodule death is achieved by taPIT compared to BPD–PDT at 60 J/cm². (b) Off-target T cell dose response was also not significantly different in response to taPIT vs. BPD–PDT at moderate light doses. However, significantly more T cells were preserved by taPIT treatment than by BPD–PDT at 60 J/cm². (c) The ratio of T cell-to-EOCC relative viability illustrates the relative difference in tumor cell-type response to each therapy. The results at 60 J/cm² indicate that taPIT is 3.4 ± 0.8 times more selective for cancer cells than T cells, whereas BPD–PDT results in more efficient destruction of T cells than cancer cells (only 0.4 ± 0.1 T cells survive per cancer cell). In all cases, significance was tested by an unpaired, two-tailed t test (*P<0.05).

T cells responded in similar fashion and exhibited slightly higher sensitivity to the treatment (for both data sets, EC₅₀ = 12.1 ± 4.4 J/cm²) than EOCC nodules. Notably, T cells treated with taPIT were significantly more viable (27.7 ± 6.5%) than those treated with unconjugated BPD (5.2 ± 1.3%) at the highest light dose, 60 J/cm² (Fig. 2b).

A T cell:EOCC viability ratio was computed to quantify the number of surviving immune cells per EOCC cell after treatment (Fig. 2c). Small values of this ratio (<1) indicate more immune cells are eliminated than EOCC cells in the same culture, whereas large values (>1) signify the opposite, that is, more T cells survive than cancer cells. We observe low doses of PDT (≤10 J/cm²) elicit equal responses from each cell type (ratio ≈1). However, at 60 J/cm², taPIT spared 3.4 ± 0.8 T cells per cancer cell, whereas conventional PDT eradicated more T cells than cancer cells (0.36 ± 0.09). Furthermore, curve fitting suggests that the taPIT ratio trend results in an immune cell sparing “sweet-spot” near 50 J/cm², where the immune cell sparing effect is maximized. In contrast to taPIT, higher doses of conventional PDT deplete the local T cell population in agreement with prior in vivo studies (4).

Although encouraging, T cell sparing was not ubiquitous, suggesting imperfect selectivity and remaining off-target cytotoxicity. This may conceivably be due in part to bystander effects that propagate cell death signaling pathways to neighboring cells (45, 46) or cytotoxic cytokines (e.g., interferon–γ and tumor necrosis factor–α) released by dying T cells after high-dose PDT. It is also possible, but much less likely, that surrounding T cells take up EOCC–activated BPD after its intracellular transport and extracellular release from EOCCs during the 24-hour PIC incubation. Finally, it has been shown that Foxp3+ regulatory T cells express EGFR under inflammatory conditions (47), which may account for partial T cell reduction in this model, although a complete T cell phenotypic profile was not verified for this study. Further investigation is warranted to elucidate the mechanisms of PIC-mediated T cell death and to optimize the dosimetry in order to maximize the observed off-target cell sparing phenomenon. Another question yet to be answered is if T cell function is altered after high-dose PDT, an important feature for future taPIT strategies designed to reverse anti-tumor immunity through simultaneous EOCC and regulatory T cell targeting (47, 48).

Design of new therapies that integrate priming and immune checkpoint inhibition (ICI) are emerging as more is understood about the complex tumor immune microenvironment (TIME). As ICIs have not generally been successful on their own due to a number of immunological and physiological challenges, combination treatments that perturb the TIME from cold to hot are appealing (49). The advantage of taPIT over conventional modalities is the potential to more precisely modulate the TIME through specific targeting of cancer and immunosuppressive tumor cell subtypes including tumor associated macrophages (50, 51), fibroblasts (25, 52), and regulatory T cells (24, 53). Although the in vitro model in this work overestimates clinical measurements of E:T ratios in ovarian cancer (6), sparing local cytotoxic T cells may prove critical in future taPIT–ICI treatment strategies that facilitate a robust immune response after photodynamic destruction.

In summary, the results demonstrate that taPIT can be applied at doses lethal to cancer cells while preserving neighboring immune cells. In contrast, conventional PDT using unformulated PSs results in indiscriminate destruction of EOCCs and T cells. The enhanced selectivity of taPIT paired with the use of 3D immune-oncology models opens the door to developing and optimizing selective destruction of specific tumor cell subtypes beyond epithelial cancer cells, including immunosuppressive tumor cell types, with preservation of effector immune cells to potentially induce local and systemic anti-tumor immune responses following a single cycle of phototherapy. Quantitative optimization of these effects in 3D biological models will pave the way towards investigating these new concepts in mouse models and ultimately in humans.