A PanImmune Feature Matrix of Immune Characteristics as described in (47) was used to obtain lymphocytic infiltrate signature scores (based on the following 18 markers defined in (48): CCL5, CD19, CD37, CD3D, CD3E, CD3G, CD3Z, CD79A, CD79B, CD8A, CD8B1, IGHG3, IGJ, IGLC1, CD14, LCK, LTB, MS4A1) across multiple tumor-samples, containing The Cancer Genome Atlas (TCGA) Participant Barcodes. Batch effect normalized TCGA PAN CANCER CD27 and EGFR expression levels were obtained via Xena Hub (49), which consisted of TCGA data from broad GDAC firehose, that was normalized by RSEM (RNA-seq by Expectation-Maximization) and batch corrected via EB++ (Empirical Bayes++) (synapse ID: syn4976363). Expression levels were matched with lymphocytic infiltrate signature scores based on TCGA Participant Barcodes, and linear regressions were performed in GraphPad Prism 8.0.2. To visualize EGFR expression across multiple tumor-types, violin plots were generated.

The dataset from the Tumor Immune Cell Atlas study (50) consisting of 13 different cancer types, 217 patients, and 526,261 cells was downloaded in the form of a RDS file containing the Seurat object. The data was ingested into Seurat V4 in R language version 4.0.3. The integrated single-cell RNA sequencing (scRNA-seq) data sets were collected as described before by Nieto et al. (50). In brief, after integration, the cells were divided into 25 clusters representing major immune cell types including 12 T cell types. To verify the robustness of the clusters and the associated signatures, a random forest classifier was used to assign cell annotations. A fivefold cross-validation was performed to assess biases and variance (50). The following T cell types were included in our study: Regulatory T cells, T helper cells, Th17 cells, recently activated CD4⁺ T cells, Naïve-memory CD4⁺ T cells, Transitional memory CD4⁺ T cells, Naïve T cells, Proliferative T cells, Pre-exhausted CD8⁺ T cells, Cytotoxic CD8⁺ T cells, Effector memory CD8⁺ T cells, and Terminally exhausted CD8⁺ T cells (For the key markers per subtype, see Supplementary Table 1 ). Differential expression was calculated by using the FindMarkers function from Seurat with MAST as the method of choice (51).

Polyclonal antibody (pAb) Goat anti-human Ig-PE (cat# 2040-09, Southern Biotech, Birmingham, AL, USA), monoclonal antibodies (mAb): anti-CD27-APC (cat# 302810, clone O323, BioLegend, San Diego, CA, USA) (also used as CD27 mAb), anti-EGFR-FITC (cat# sc-120 FITC, clone 528, Santa Cruz Biotechnology, Dallas, TX, USA), anti-CD25-APC (cat# 302610, clone BC96, BioLegend), anti-CD4-FITC (cat# 300506, clone RPA-T4, BioLegend), anti-CD8-Brilliant Violet 421 (cat# 344748, clone SK1, BioLegend), mouse (IgG2A) (mAb 425) (Cat# EWI020, Kerafast, Boston, MA, USA), anti-Myc mAb Alexa Fluor 647 (cat# 2233, clone 9B11, Cell Signaling, Danvers, MA, USA). Atezolizumab was obtained from the pharmacy of the UMCG (Groningen, the Netherlands).

The following wild type (WT) cell lines were obtained from the American Type Culture Collection (ATCC): A431, MDA-MB-231, ES-2, DLD-1, FaDu, and OVCAR-3. OVCAR-3.EGFR knock-out (KO) cells were a kind gift from prof. dr. Helfrich (UMCG/Dept of Surgery, Groningen, the Netherlands). HT1080.CD27 is previously described in (39) and is a kind gift from prof. dr. Harald Wajant (University of Wuerzburg, Wuerzburg, Germany). An overview of the cell lines, including tissue, cell type, cancer type, species, source and transduced genes, can be found in Table 1 . Cells were cultured in RPMI-1640 (cat# 21875034, Gibco, Thermo Fisher Scientific, Waltham, MA, USA) or DMEM (cat# 11965092, Gibco, Thermo Fisher Scientific), supplemented with 10% fetal calf serum (FCS)(cat# F7524, Thermo Fisher Scientific) at 37°C/5% CO₂. The artificial scFvCD3 (UchtV1 anti-CD3 antibody fragment)-presenting cell lines MDA-MB-231^scFvCD3, ES-2^scFvCD3, DLD-1^scFvCD3, FaDu^scFvCD3, OVCAR-3^scFvCD3, and OVCAR-3^scFvCD3EGFR^KO are based on the lentiviral synNotch receptor construct pHR_PGK_antiCD19_ synNotch_Gal4VP64, which was a gift from Wendell Lim (Addgene plasmid #79125; http://n2t.net/addgene:79125; RRID : Addgene_79125) (52). The anti-CD19 scFv was replaced with the scFvCD3 UCHT-1v9 using Gibson cloning (cat# E5510S, New England BioLabs, Ipswich, MA, USA), yielding pHR_PGK_scFvCD3_synNotch_Gal4VP64. Lentivirus was produced by transient transfection of HEK293T cells with transfer vector, psPAX2, and pCMV-VSV-G packaging system using FuGENE (cat# E2312, Promega, Madison, WI, USA) according to manufacturer’s recommendations. Viral supernatant was collected and filtered through a 0.45 μm filter (cat# SLHVR13SL, Millipore, Burlington, MA, USA). Transduction was performed by adding 1.5 mL viral supernatant to 1.5 mL of RPMI containing 2.5 × 10⁵ pre-seeded cells in a 6 well tissue culture plate (cat# 3516, Corning Inc., Corning, NY, USA) in the presence of 4 μg/mL polybrene (cat# TR-1003, Sigma-Aldrich, Saint Louis, MO, USA). Transduced cells were sorted for expression of a Myc-tag (present at the N-terminus of the CD3 scFv) with a cell sorter model SH-800s (Sony Biotechnology, San Jose, CA, USA). Before each experiment, ES-2^scFvCD3, DLD-1^scFvCD3, FaDu^scFvCD3, OVCAR-3^scFvCD3 and OVCAR-3 ^scFvCD3.EGFR^KO cells stably expressing scFvCD3 were characterized for their expression of EGFR, scFvCD3 and CD27, maintaining a consistent fold change from the isotype ( Supplementary Figure 3 ). Cells were also transduced with lentivirus containing vector pLKO.1 mCherry, which was a gift from Oskar Laur (Addgene plasmid #128073; http://n2t.net/addgene:128073; RRID : Addgene_128073), producing the corresponding mCherry-expressing cells lines for visualization in the cytotoxicity assays.

The bsAb CD27xEGFR was constructed in an scFv-scFv-IgG1 format, containing the antigen-binding fragments scFv1F5, targeting CD27, and scFv425, targeting EGFR. These two scFvs were connected by a flexible glycine-serine (GS) linker, consisting of a (GGGS)₃ sequence. The Fc domain of the antibody was designed with LALAPG mutations (L234A, L235A, and P329G) (53) in order to create an effector silent IgG molecule. Another GS linker, with the same (GGGGS)₃ sequence, connects the scFv EGFR to the IgG1 Fc domain. The antibody was produced by Evitria (Schlieren, Switzerland). Supernatant was harvested by centrifugation, filtered (0.2 μm filter), whereupon antibody was purified using MabSelect SuRe (cat# GE17-5438-01, Merck KGaA, Darmstadt, Germany). Purity was evaluated by analytical size exclusion chromatography with an AdvanceBio SEC column (300A 2.7 um 7.8 x 300 mm) (cat# PL1180-5301, Agilent, Santa Clara, CA, USA) and Dulbecco’s phosphate-buffered saline (DPBS) (cat# 14190144, Gibco, Thermo Fisher Scientific) as running buffer at 0.8 mL/min. CD27xEGFR was successfully purified up to 98.6% with only minor amounts of degradation product ( Supplementary Figure 1A ). Endotoxin content was measured with the Charles River Endosafe PTS system (cat# PTS150K, Wilmington, MA, USA) and was < 1 EU/mg.

Binding of His-CD27 or His-EGFR to CD27xEGFR was analyzed using the BLItz system from ForteBio (cat# 45-5000, ForteBio, Menlo Park, CA, USA). His-CD27 (cat# 10039-H08B1, SinoBiological, Beijing, China), produced using the Baculovirus-Insect Cell expression system, encodes for the extracellular domain of human CD27 (Met1-Ile192) and included a C-terminal polyhistidine (His) tag. His-EGFR (cat# Z03194, GenScript, Rijswijk, the Netherlands), generated with the Sf9 insect cell expression system, encodes for the extracellular domain of human EGFR (Leu25-Ser645), and also featured a C-terminal His tag.

Octet protein A biosensors (cat# 18-5010, Satorius, Göttingen, Germany) were wetted for at least 10 min before use in 100 mM Tris-HCl pH 8 and all samples were diluted in the same buffer. In short, a baseline was run for 30 sec, followed by loading of 8 µg/mL of CD27xEGFR for 120 sec, baseline for 30 sec, association of either 125 nM His-EGFR and/or 500 nM His-CD27 for 120 sec, and dissociation for 120 sec. Atezolizumab (50 µg/mL) was used as a control.

The same protocol was used for binding of CD27xEGFR to immobilized His-CD27 and His-EGFR, with the following exceptions: The use of Octet HIS1K biosensors (cat# 18-5120, Sartorius), loading of 500 nM His-EGFR or His-CD27 and association of CD27xEGFR (50 µg/mL). Step corrections were applied to both the start of association and dissociation. Finally, the individual experiments were aligned to the start of association (x=y=0 for t=180 sec).

Buffy coats were purchased from Sanquin (nr. NVT0465), and all donors gave informed consent (Sanquin Blood Supply, Groningen, the Netherlands). Human peripheral blood mononuclear cells (PBMCs) were isolated via density gradient centrifugation using lymphoprep (cat# 07851/07861, STEMCELL Technologies, Vancouver, Canada) and frozen until the day of the assay. T cells were isolated from fresh PBMCs using an autoMACS Pro Separator (Miltenyi Biotec, Bergisch Galdbach, Germany) and a Pan T Cell Isolation Kit (cat# 130-096-535, Miltenyi Biotec) following the manufacturer’s recommendations. After isolation, T cells were frozen until the day of the assay.

Binding of CD27xEGFR to CD27 and EGFR was evaluated using cell lines A431, HT1080.CD27, and primary human T cells. In brief, 5x10⁴ cells were incubated with CD27xEGFR (0.01–10 µg/mL, 45 min at 4°C) in a 96 well plate (cat# 3799, Corning Inc.) washed 3 times with DPBS (cat# 14190144, Gibco, Thermo Fisher Scientific), and then incubated with anti-human-IgG-PE pAb (45 min at 4°C). Following 3 washes with DPBS, cells were evaluated by flow cytometry (Accuri C6 Plus Flow Cytometer, BD, Franklin Lakes, NJ, USA). Binding to primary human T cells was performed analogously, but in the presence of FcR blocking reagent (cat# 130-059-901, Miltenyi Biotec) in all incubation steps. To demonstrate EGFR-specific binding of CD27xEGFR, HT1080.CD27 cells were pre-incubated with a 10-fold molar excess of mAb 425 for 15 min at 4°C. To demonstrate CD27-specific binding of CD27xEGFR, HT1080.CD27 cells were pre-incubated with a 10-fold molar excess of CD27 mAb for 15 min at 4°C. Binding of CD27xEGFR to HT1080.CD27 was blocked by pre-incubation with a 10-fold molar excess of CD27 mAb and mAb 425. The mean fluorescent intensity (MFI) was normalized to the highest obtained MFI (which was set at 100%) with the 0 µg/mL CD27xEGFR condition being set at 0%. Doublet formation between A431 tumor cells and primary human T cells upon addition of CD27xEGFR was analyzed by pre-labeling A431 cancer cells with Vybrant DiD Cell-Labeling Solution (cat# V22887, Thermo Fisher Scientific) and primary human T cells with CellTrace Violet reagent (cat# C34557, Thermo Fisher Scientific) both according to manufacturer’s protocol). Cells were subsequently mixed at a 1:1 ratio (2x10⁵ cells) with or without the addition of 10 μg/mL CD27xEGFR for 45 min at 4°C. Doublet formation was analyzed by flow cytometry (CytoFLEX V5-B5-R3, Beckman Coulter Life Sciences, Indianapolis, IN, USA).

In a 96 well plate (cat# 167008, Thermo Fisher Scientific), 100 μL of media containing 1x10⁴ primary PBMCs were added to 100 μL of media with or without CD27xEGFR (10 µg/mL) containing MDA-MB-231^WT or MDA-MB-231^scFvCD3 cells previously incubated overnight at Effector : Target (E:T) ratios of 1:1, 1:2 and 1:5. After a 24-hour incubation, images were taken at 5x magnification using an EVOS FLoid Imaging System (cat# 4471136, Thermo Fisher Scientific) to visualize T cell clustering. Furthermore, PBMC cells were collected, stained for CD3 and CD25 and the CD25 expression of CD3⁺ cells was measured using flow cytometry (CytoFLEX V5-B5-R3).

In a 96 well plate (cat# 167008, Thermo Fisher Scientific), 100 μL of media containing 4x10⁴ primary human T cells labeled with CellTrace Violet reagent with or without CD27xEGFR (10 µg/mL) were added to 100 μL of media containing 2x10³ ES-2^scFvCD3, DLD-1^scFvCD3 or FaDu^scFvCD3 cells previously incubated overnight. Proliferation was measured using flow cytometry (CytoFLEX V5-B5-R3) on day 5 and quantified using FlowJo Software version 10.8.1.

In a 96 well plate (cat# 167008, Thermo Fisher Scientific), 100 μL of media containing primary human T cells with or without CD27xEGFR (10 µg/mL) were added to 100 μL of media containing 2x10³ ES-2^scFvCD3, DLD-1^scFvCD3 or FaDu^scFvCD3 cells previously incubated overnight in the E:T ratios indicated (5:1, 10:1, 20:1). After 4 days, T cells were collected from the co-culture and stained for CD4, CD8, CD25, and Zombie NIR (cat# 423106, BioLegend) (to distinguish between alive and dead cells) and analyzed by flow cytometry (CytoFLEX V5-B5-R3). Supernatants were harvested and IFN-γ secretion was quantified using an IFN-γ ELISA kit (cat# 31673539, ImmunoTools, Friesoythe, Germany).

In a 96 well plate (cat# 167008, Thermo Fisher Scientific), 100 μL of media containing primary human T cells with or without CD27xEGFR (10 µg/mL) were added to 100 μL of media containing 2x10³ ES-2^scFvCD3, DLD-1^scFvCD3, FaDu^scFvCD3, OVCAR-3^scFvCD3, or OVCAR-3^scFvCD3EGFR^KO cells incubated overnight in the E:T ratios indicated (0:1, 1:1, 2:1, 5:1, 10:1, 20:1). Experiments were imaged for mCherry fluorescence for up to 7 days using the Incucyte S3 live-imaging system (Essen BioScience, Royston, UK) and analyzed using Incucyte S3 software v2021A. Four pictures of each well for each of three technical replicates were acquired and analyzed based on the Top-Hat segmentation method (Radius 50 µm, Threshold 0.0950, Edge Split On, Edge Sensitivity 5, Hole Fill 0 µm², Adjusted Size 7 pixels, Filters: min Area 210 µm², min Integrated Intensity 50). As a measure of cytotoxicity, cell survival was calculated as the mCherry area (µm²/image) from the sample at the indicated time point/mCherry area (µm²/image) from the cancer cells only control at the indicated time point. To evaluate direct EGFR-blocking anti-carcinoma activity of CD27xEGFR, 2x10³ mCherry expressing FaDu^scFvCD3 cells were seeded in a 96 well plate (cat# 167008, Thermo Fisher Scientific) and treated with CD27xEGFR (10 µg/mL) or mAb425 (10 µg/mL) for 3 days.

Data are presented as mean + SD as stated in the figure legends. For the before-after plots, each pair of observations represents independent experiments with different T cell donors. Statistical significance was determined as indicated in the figure legends, with a p-value of < 0.05 considered statistically significant.

For the proportions of CD27⁺ cells, a two-sample test for equality of proportions with Bonferroni correction was applied for comparing the proportions of CD27 among different T cell types. The corresponding p-values are documented in Supplementary Table 2 .

For CD27xEGFR binding, binding blockade, doublet formation, T cell proliferation, cytotoxicity, and IFN-γ secretion, experiments were performed with T cells from different donors on different days, and each experiment was treated as independent. The normality of flow cytometry data was assessed through visual inspection of flow cytometry histograms. For the proliferation and IFN-γ secretion data, normal distribution was assumed. For the cytotoxicity experiments, the Shapiro-Wilk test was employed to determine the normality of the data on three independent experiments consisting of three technical replicates.

For the relationship between EGFR or scFvCD3 expression and difference of cancer cell survival between CD27xEGFR and medium control, a simple linear regression was performed. Each data point in the analysis represents the mean of all independent experiments at the E:T 5:1, conducted for each cell line with different T cell donors.

To delineate the potential applicability of EGFR-targeted activation of CD27 agonism, the TCGA PAN CANCER dataset was analyzed for concurrent CD27 and EGFR expression. Compared to non-epithelial cancers, lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), and uveal melanoma (UVM), all 20 epithelial cancers expressed high levels of EGFR mRNA ( Figure 1A , black for epithelial and gray for non-epithelial cancers). Furthermore, lymphocytic infiltrates across all 20 epithelial cancers expressed CD27 mRNA, demonstrating a clear correlation between lymphocyte infiltration score and CD27 (R² = 0.6895) within the whole epithelial cancer set, confirming that the strategy of EGFR-mediated crosslinking of CD27 could be employed within epithelial cancers ( Figure 1B , see Supplementary Figure 2 for individual tumor types). In an established single-cell tumor immune atlas from a range of cancer types, CD27 expression was prominent in a subpopulation of regulatory T cells (T_regs), terminally exhausted CD8⁺ T cells, as well as cytotoxic CD8⁺ T cells ( Figure 1C ). Notably, tumor-reactive CD8⁺ T cells - inclusive of cytotoxic, terminally exhausted, and pre-exhausted cells - present a higher proportion of CD27⁺ cells in comparison to all CD4⁺ T cells, Th17 cells, and naïve T cells ( Figure 1D ). Upon comparing CD27⁺ T cells proportions ( Figure 1D , Supplementary Table 2 ), T_regs and terminally exhausted CD8⁺ T cells have proportions of 63.25% and 51.14% that are significantly higher than those observed in other T cell types. Notably, the proportions of CD27⁺ cells within cytotoxic and pre-exhausted CD8⁺ T cells are similar, with no significant difference observed between these two groups ( Supplementary Table 2 ). Subsequent differential gene expression analysis in terminally exhausted CD8⁺ and cytotoxic CD8⁺ T cell subsets revealed that several exhaustion genes (GZMK, HAVCR2, TIGIT, and LAG3) and cytotoxicity genes, (TNFRSF9, CST7, and CD28) are significantly upregulated in the CD27⁺ fraction ( Figures 1E, F ). In addition, expression of genes associated with tissue residency, trafficking, adhesion, and migration (VCAM1, CXCR3, ITGA4, CXCL13, and CCR7) are also elevated ( Figures 1E, F ). Therefore, the concurrent expression of CD27 and EGFR in various epithelial cancers, along with the gene expression signatures related to exhaustion, cytotoxicity, and tissue residency of CD27-expressing T cells, underscores the potential of EGFR-targeted activation of CD27 agonism for enhancing the re-activation of tumor-infiltrated and tumor-reactive T cells in EGFR-expressing cancers.

To exploit the above-described concurrent expression of EGFR and CD27 for targeted activation of CD27 signaling, the ICS-bsAb CD27xEGFR was constructed. CD27xEGFR consists of an N-terminal CD27 targeting antibody fragment (scFv1F5) fused via a (GGGGS)₃ linker to the EGFR-targeting antibody fragment scFv425, with a silent human IgG1 containing LALAPG mutations ( Figure 2A ), which prevent Fc-FcR mediated antibody effector functions. The specific binding activity of CD27xEGFR to soluble CD27 and EGFR individually was confirmed using biolayer interferometry ( Figure 2B ). Moreover, the association rate (as defined by the gradient of the initial association curve) increased when both antigens were combined compared to a single antigen, supporting the proposed mechanism of action of dual binding ( Figure 2B ). Reversely, upon immobilization of CD27 or EGFR onto the biosensor, CD27xEGFR also specifically bound to both targets ( Supplementary Figure 1B ).

In a cell-based assay, CD27xEGFR dose-dependently bound to the EGFR⁺ epidermoid carcinoma cell line A431 ( Figure 2C ). Similarly, dose-dependent binding of CD27xEGFR was detected on CD27⁺ primary human T cells that expressed CD27 but not EGFR ( Figure 2D , Supplementary Figure 1C ). Further, CD27xEGFR also bound to the EGFR⁺ fibrosarcoma cell line HT1080, engineered to ectopically express CD27, with binding only partly inhibited by pre-incubation with mAb 425 (anti-EGFR mAb) or anti-CD27 mAb alone ( Figure 2E ). However, CD27xEGFR binding was abrogated after a combined pre-incubation with mAb 425 and anti-CD27 mAb ( Figure 2E ), demonstrating that CD27xEGFR binds to both antigens when they are present on the same cell surface. Moreover, CD27xEGFR induced the formation of doublets between EGFR⁺ and CD27⁺ target cells in a mixed culture of A431 and primary human T cells. The percentage of doublets increased significantly from ~15% in the medium control up to ~40% in the CD27xEGFR-treated condition ( Figure 2F ), demonstrating CD27xEGFR simultaneously interacted with EGFR and CD27 expressed on distinct cells. In conclusion, the designed bsAb CD27xEGFR exhibited selective and simultaneous binding to both targets (CD27 and EGFR).

To evaluate the EGFR-restricted co-stimulation of T cells by CD27xEGFR, the carcinoma cell lines MDA-MB-231, ES-2, DLD-1, and FaDu were engineered to express a UchtV1 anti-CD3 antibody fragment (scFvCD3) on their surface, which enabled activation of TCR signaling in allogeneic T cells independent of MHC presentation ( Figure 3A ). The system was validated by treating a culture of MDA-MB-231^scFvCD3 with peripheral blood mononuclear cells (PBMCs), which clearly activated T cells, as evidenced by cluster formation in the MDA-MB-231^scFvCD3 co-culture ( Figure 3A , bottom left). The addition of CD27xEGFR to this co-culture increased cluster formation further ( Figure 3A , bottom right). Using flow cytometry, an increase in CD25 expression was observed in T cells within the PBMC population upon treatment with CD27xEGFR compared to the medium control ( Supplementary Figure 1D ). In contrast, no cluster formation was detected in the co-culture of MDA-MB-231^WT with PBMCs with or without CD27xEGFR treatment ( Figure 3A , top left and right), validating that scFvCD3 can activate T cells MHC-independently. To specifically study the effects of CD27xEGFR on T cells, T cells were isolated from PBMCs in further studies. In line with the cluster formation, a prominent proliferation of T cells was detected in CD27xEGFR-treated mixed cultures with ES-2^scFvCD3 cells, with up to 5 proliferation peaks detected ( Figure 3B , bottom). In contrast, minimal proliferation was detected in the mixed cultures of ES-2^scFvCD3 and T cells in the absence of CD27xEGFR ( Figure 3B , top). Upon quantification, a significant increase in T cell proliferation was detected in CD27xEGFR treated ES-2^scFvCD3 co-cultures, as evidenced by a significantly reduced percentage of cells in the parental peak and an increased percentage of cells in the proliferation peaks compared to medium control co-cultures ( Figure 3C ). A similar co-stimulatory activity of T cells by CD27xEGFR was detected in mixed cultures with DLD-1^scFvCD3 and FaDu^scFvCD3, with a significant increase in T cell proliferation in either culture upon CD27xEGFR treatment ( Figures 3D, E , respectively). Consistent with this increase in proliferation, CD27xEGFR treatment of ES-2^scFvCD3, DLD-1^scFvCD3, and FaDu^scFvCD3 co-cultures increased the expression of CD25 on T cells compared to medium control co-cultures ( Figures 3F–H , respectively), both on CD4⁺ and CD8⁺ T cells, and at different E:T ratios. The largest increase was detected at an E:T ratio of 20:1, with a 20–25% increase. Finally, CD27xEGFR treatment increased pro-inflammatory cytokine IFN-γ secretion in FaDu^scFvCD3 co-cultures at 10:1 and 20:1 E:T ratios compared to medium control co-cultures ( Figure 3I ). Taken together, this data provides evidence that CD27xEGFR can effectively co-stimulate T cells in co-cultures with a wide range of EGFR⁺ cell lines.

In view of the clear co-stimulatory activity of CD27xEGFR, potential anti-tumor T cell activity induced by CD27xEGFR was evaluated next. As a measure of T cells cytotoxicity, cancer cell survival was determined by using the fluorescence of mCherry transduced cancer cells. After three days, treatment with CD27xEGFR strongly reduced the survival of mCherry-expressing ES-2^scFvCD3 cells compared to medium control in co-culture experiments with primary human T cells ( Figure 4A ). To quantify this data, cell survival was measured over time. After approximately 24 hours of co-culture of T cells with mCherry-expressing ES-2^scFvCD3 cells, T cell-mediated killing was observed, which continued to near complete eradication of cancer cells after 160 hours of treatment with CD27xEGFR ( Figure 4B ). In contrast, mCherry-expressing ES-2^scFvCD3 started to grow back after approximately 96 hours when co-cultured with medium control and T cells ( Figure 4B ). Cell survival at this time point was normalized to that of cancer cells only and measured at different E:T ratios (0:1, 5:1, 10:1, and 20:1) ( Figure 4C ). An E:T ratio dependent reduction of cancer cell survival was detected, in which treatment with CD27xEGFR significantly reduced cancer cell numbers, with a maximum decrease of ~40% at an E:T ratio of 5:1 ( Figure 4C ). A similar reduction in cancer cell numbers in mCherry-expressing DLD-1^scFvCD3 and mCherry-expressing FaDu^scFvCD3 cells upon CD27xEGFR treatment was detected, with a maximum effect of ~20% and ~30% at a 2:1 E:T ratio for mCherry-expressing DLD-1^scFvCD3 and mCherry-expressing FaDu^scFvCD3 cells, respectively ( Figures 4D, E ). Of note, treatment of a monoculture of mCherry-expressing cancer cells (ES-2^scFvCD3, DLD-1^scFvCD3, and FaDu^scFvCD3) with CD27xEGFR slightly reduced the survival of the cancer cells compared to the medium controls ( Figures 4C–E , 0:1 E:T ratios, ~2-7%). This effect is most likely caused by the EGFR-growth inhibitory activity of CD27xEGFR, which proved to be reminiscent of EGFR blocking with mAb 425 ( Figure 4F ). Notably, in a co-culture of T cells with mCherry-expressing EGFR⁺ OVCAR-3^scFvCD3, treatment with CD27xEGFR increased cancer cell killing by T cells ( Figure 4G , ~25% reduction at an E:T ratio of 5:1) whereas with the corresponding EGFR^KO cells, treatment with CD27xEGFR had no effect on cancer cell survival when compared to medium control ( Figure 4H ). The EGFR and scFvCD3 expression levels varied between the cell lines ( Supplementary Figure 3 ), with EGFR expression having a significant, positive correlation with the cytotoxicity of CD27xEGFR at an E:T ratio of 5:1 ( Figure 4I ). No correlation was identified for scFvCD3 levels ( Supplementary Figure 4 ). Taken together, CD27xEGFR has anti-cancer activity both by co-stimulation of T cells at the sites of EGFR expression as well as by directly blocking EGFR on cancer cells.