Previous studies from us and others have shown that chimeric molecules that engage either T cells (mostly via a scFv to CD3) or NK cells can enhance the therapeutic effect of an oncolytic virotherapy.14–17 Here, we report the design of a novel chimeric molecule that can engage both NK and T cells. As depicted in figure 1A, the key components of this chimeric molecule are at the N-terminus is a 152 amino acid OMCP and at the C-terminus is a mutant form of EGF (m123). OMCP is a small polypeptide encoded by monkeypox and cowpox virus that can selectively bind to NKG2D with an affinity equal to, or greater than, all other known NKG2D ligands.23 The mutated EGFα ligand (EGFm123) has an enhanced binding affinity and dynamic to both murine and human EGFR.24 It is engineered by directed evolution through yeast surface display for significantly enhanced affinity for the EGFR. Compared with the wild type EGF, m123 bound 8-fold and 33-fold more tightly to surface EGFR on NR6WT cells and BJ-5a cells, respectively. m123 also bounds 18-fold and 8-fold more tightly to human EGFR and murine EGFR, respectively.24 Interestingly, m123 showed stronger binding at low pH, which is beneficial given that the pH of the TME is universally acidic. Additionally, the binding of m123 to the EGFR may enhance its intracellular degradation, thus benefiting the overall antitumor activity.24

We constructed two versions of this chimeric molecule—the bispecific and trispecific binding engagers. In the bispecific construct, termed BiCEP, OMCP and the m123 are linked via a flexible 20 residue (Gly-Gly-Gly-Gly-Ser)₄ linker and a Myc-tag for ease of detection. In the tri-specific construct, termed TriCEP, a mutated form of IL-2 (mutIL-2) that has substitutions of alanine for arginine at the 38 position (R38A) and/or lysine for phenylalanine at the 42 position (F42K) is placed upstream of OMCP via a (Gly-Ser-Ser)₄ and a His8-tag for detection. These mutations decrease the affinity of IL-2 for IL-2Ra. This allows mutIL-2 to selectively activate IL-2-signaling only as a tethered form without broadly activating other IL-2R bearing cells and minimizing the unwanted toxicity.25 The hypothesized action mechanisms of BiCEP and TriCEP are illustrated in figure 1B. The simultaneous binding of OMCP to NKG2D and mEGFα to EGFR by the chimeric engagers will efficiently engage the NKG2D bearing NK and T cells with EGFR-bearing tumor cells, bringing the two cells in proximity and creating an immunological synapse. The mutIL-2 in TriCEP would flicker on the IL-2R on the engaged immune cells and enable the activation and proliferation of the engaged NK or T cell to further potentiate the immune response and improve the efficacy.

After the design and construction of BiCEP and TriCEP, we checked for the presence of steric-imposed conformational constraints on both chimeric molecules, and for that, we generated the predicted 3D structure using Robetta protein modeling software.25 The 3D structure in figure 1C predicts that all the individual components in both chimeric molecules are spatially separated by an intermittent linker, such that they can readily bind to their cognate targets without intradomain steric hindrance. Next, we examined the expression of both BiCEP and TriCEP by transfecting the plasmid constructs to mammalian cells (HEK293 and BHK cells), followed by a Western blot analysis. The results in figure 2A showed that both chimeric molecules are efficiently expressed and secreted to the supernatant after the molecules are synthesized.

Next, we conducted a series of in vitro experiments to test the binding specificity of the individual components in the chimeric molecules to their respective receptors. First, the binding affinity of OMCP to NKG2D was assessed by incubating the TALL-104 cells with the supernatants harvested from HEK293 transfected with the BiCEP and TriCEP constructs for 1 hour at RT. TALL-104 cells are a human leukemic cell line that expresses markers characteristic of both NK cells and cytotoxic T-lymphocytes with high expression of NKG2D.26 The binding of OMCP to NKG2D was determined by measuring the number of cells positive for both NKG2D and the Myc tag contained in both BiCEP and TriCEP (NKG2D⁺/Myc⁺) via flow cytometry analysis. Over 50%–60% of NKG2D⁺ cells are positive for Myc, indicating a good binding affinity of OMCP to NKG2D (figure 2B).

For determining the binding activity of BiCEP and TriCEP to EGFR, we initially incubated the supernatants with SKOV3 cells for 30 mins at RT. SKOV3 is a human ovarian cancer cell line with overexpression of EGFR.27 The binding of the chimeric molecules to EGFR on the surface of SKOV3 cells was detected by measuring the number of cells positive for both EGFR and Myc (EGFR+/Myc+), again via flow cytometry analysis. The result in figure 2C showed that over 80% of EGFR expressing SKOV3 cells are also positive for Myc, indicating that the mutant form of EGFm123 contained in both BiCEP and TriCEP can efficiently bind to EGFR. No binding was observed with the mock-transfected supernatants, confirming the specificity of this assay. Moreover, from figure 2C, it can be appreciated that approximately 50% of SKOV3 are EGFR negative. These EGFR^- SKOV3 cells would serve as an internal control. Importantly, there was no BiCEP or TriCEP binding to these EGFR^- cells, indicating the binding specificity of these chimeric molecules to the targeted EGFR.

To further confirm the binding specificity of OMCP and EGFm123 in the two chimeric molecules, we also determined the costaining positivity of EGFR in a murine colon cancer cell line CT26-EGFR that was established in our own lab and has been used in our previous studies in an EGFR-targeted immunotherapy.28 We repeated the binding assay on this cell line with the cell-free supernatants collected from the amplicon infected cells (figure 2D). From the binding assays, assessed using flow cytometry, there is increased binding of EGF to the truncated human EGFR on the CT26 cell line, evaluated by the cells positive for both Myc and the EGFR. Over 90% of the cells are double positives (EGFR⁺/Myc⁺), indicating that the EGFα binding affinity is retained in the supernatants collected from the amplicon infected cells (figure 2E). Since the mutIL-2 could only bind weakly to IL-2R, we did not perform any in vitro binding assays on the TriCEP. However, as presented in the following sections, the IL-2 dependent activation by TriCEP could be detected from in vivo studies by the scRNA-seq analysis.

To test whether BiCEP and TriCEP could engage NK cells (TALL-104) to kill tumor cells, a real-time in vitro tumor-killing assay was performed.29 Target tumor cells (SKOV3) were incubated with the effector TALL-104 cells at an increasing effector-to-target (E:T) ratio (1:1, 2:1, and 5:1) in the presence or absence of BiCEP or TriCEP. The tumor cell viability was monitored by IncuCyte, a real-time cell imaging device. Images were taken every 2 hours and the number of viable cells per well was quantified with the IncuCyte-FLR-Platform technology (figure 3A). The cytotoxicity is reported by the percentage of viable cells/percentage of confluence remaining at the end of 24 hours coculture (online supplemental file 1-Fig.S1A). The results show that at the lower E:T ratios (1:1 and 2:1), there is a significant increase in the percentage killing in the presence of BiCEP and TriCEP compared with the control well with the mock-transfected supernatants. However, at the high E:T ratio (5:1), this difference became insignificant. This is probably due to the high background killing activity of TALL-104 cells.30

Moreover, we conducted a highly sensitive FACS (Fluorescence-activated cell sorting) based cleaved caspase 3 cytotoxicity assay using the cleavage of caspase-3 as a readout of cytotoxicity.28 Briefly, the assay involved labeling of tumor cells (CT26-EGFR cells) with a cell tracker dye, which were then used to coculture with primary human NK cells at different E:T ratios (1:1, 3:1 and 5:1). The cells were permeabilized and stained with an antibody recognizing cleaved caspase 3 and analyzed by flow cytometry (figure 3B, with the detailed data in (online supplemental file 1-Fig.S1B). The results show that at the lowest E:T ratio (1:1), there is a significant increase on the tumor cell killing (represented as percentage caspase 3 positive cells) in the presence of BiCEP compared with the control well (NT). However, at the high E:T ratios (3:1 and 5:1), this difference became insignificant, probably due to the significant background killing in the control well. There was a significant increase in the tumor cell killing in the presence of TriCEP over the control at all the E:T ratios. Moreover, TriCEP resulted in a better killing than BiCEP at high ratios (3:1 and 5:1).

An obvious and common approach to codelivering the chimeric molecules during virotherapy is to insert their coding sequences into the backbone of the OV. However, as HSV has a large genome (over 150 kb) and recombination insertion of foreign genes is cumbersome and time-consuming, thus we chose to use an HSV amplicon vector to deliver these two transgenes. An HSV amplicon is a plasmid-like vector that contains a copy of HSV replication origin (ori-) and packaging signal (pac). In the presence of a helper HSV (eg, an oncolytic HSV), the plasmid gets amplified by a rolling-circle mechanism and the amplified DNA (a total of 150 kb) will be subsequently packaged into a viral particle. Depending on the size of the amplicon plasmid, many copies of the amplicon sequence (and hence multiple copies of the transgene) can be packaged into each viral particle. So it is an efficient and nimble gene delivery system that we have successfully used in several of our previous studies.31 We inserted the coding sequence of either BiCEP or TriCEP, together with a copy of the EGFP (Enhanced Green Fluoroscent Protein) Fgene into the amplicon construct. The inclusion of the EGFP gene allows for easy and convenient titration of the amplicon vector.

We examined the un-packaged amplicon (via transfection of the amplicon plasmid into HEK293 cells) and the packaged amplicon (via infection to BHK cells) for transgene expression (both GFP and the chimeric molecules). For packaging the amplicon plasmids into HSV particles, we initially transfected the amplicon plasmids into BHK cells, which were super-infected 24 hours later with Synco-2D, which is an HSV-1-based OV that has a clear fusogenic property.32 It was constructed by deletion of the ICP34.5. Additionally, it contains two membrane fusion mechanisms—the syn phenotype through mutagenesis and insertion of the truncated form of the gibbon ape leukemia virus envelope fusogenic membrane glycoprotein into the virus genome.32 The generated stock thus contains the mixture of the OV (Synco-2D) and the packaged amplicon. The titer of Synco-2D was determined by the conventional plaque assay and the titer of the packaged amplicon was determined by GFP positive cell counts. The results in figure 4A showed efficient GFP expression from the amplicon plasmids when they were transfected into BHK cells as they all contain the EGFP gene. The extensive appearance of GFP⁺ cells after infection in the bottom panel of figure 4B indicated that the amplicon plasmid had been efficiently packaged into viral particles when Synco-2D was used as a helper virus (as well as the OV for the in vivo studies) in this unique delivery system, and the estimated amplicon titer from the GFP⁺ cell counting is 1×10⁵ per milliliter. The Western blot analysis in figure 2D showed that the BiCEP and TriCEP molecules were sufficiently produced from the infection of the packaged amplicons.

We chose the CT26-EGFR tumor model for the in vivo studies to evaluate the therapeutic impact of these two chimeric engagers during Synco-2D virotherapy. The experiment process is summarized in figure 5A. Initially, CT26-EGFR tumors were subcutaneously established as reported.28 Once tumors reached the approximate size of 5–6 mm in diameter, they were treated with intratumoral injection of PBS (Phosphate buffered saline), Synco2D-BiCEP, Synco2D-TriCEP, or the control amplicon expressing GFP alone (Synco-2D -GFP) at 5×10⁶ pfu of Synco-2D per mice. Tumors were measured every other day using calipers and tumor volumes were calculated as described in the Materials and Methods. The results showed that, while Synco-2D-GFP only showed a marginally therapeutic effect against this murine tumor, both Synco2D-BiCEP and Synco2D-TriCEP produced a significantly better therapeutic effect compared with the PBS control (figure 5B). By the end of the experiment, all mice were euthanized, and the tumor explanted (figure 5C). The measurement of explanted tumors confirmed the enhanced therapeutic efficacy by the codelivery of both the BiCEP and TriCEP molecules. The transgene expression by the codelivered amplicons during virotherapy was confirmed by examining the GFP expression in tumor sections collected 2 days after virotherapy (figure 5D).

The previous studies on characterizing the infiltrating immune cells during virotherapy are fragmented, as they were designed to focus on certain populations of immune cells.12 13 33 Hence, we decided to use scRNA-seq to fully characterize the immune cell landscape as well as their activation status during Synco-2D virotherapy with or without the amplicon-mediated codelivery of TriCEP. We decided not to include BiCEP in this scRNA-seq analysis as it was similarly constructed as TriCEP. The scheme for this scRNA-seq is shown in figure 6A. BALB/c mice bearing subcutaneous CT26-EGFR tumors (approx. 8–10 mm in diameter) were injected intratumorally with either PBS, Synco2D-TriCEP, or Synco-2D-GFP at 5×10⁶ pfu per mice (figure 6A). Forty-eight hours later, the tumors were excised from the mice and dissociated into a single-cell suspension. Due to the rarity of immune cell filtration in the TME, we initially sorted the single cells into CD45⁻ and CD45⁺ populations, which were subsequently mixed at a 3:1 ratio for single cell capture, barcoding, and sequencing by the 10× Genomics Chromium pipeline.

For characterizing the types of infiltrating immune cells in the TME of the collected tumor samples after treatment, all cells were initially clustered into unbiased cell-type classification using the Seurat package,34–36 as shown in figure 6A. Tumor cells were clustered by CD45^- and hEGFR⁺ expression. CD45⁺ cells were clustered based on the assessment of known cell type markers into distinct lymphoid; monocyte/macrophage, T cells, NK cells, DCs, and neutrophils (figure 6B). For the purpose of this study, we restricted our analysis to NK and T cells, which are the main effector cells targeted by these chimeric engagers. Cell-type specific gene expression of Cd3d for T cells and Ncr1 for NK cells are shown in the violin plots (figure 6C). The T and NK cells are further subclustered into six distinct clusters (NK cells, Vd2 gd T cells, Th2 cells, T regulatory cells, Naïve CD8+ T cells and Th1/Th17 cells) (figure 6D).37 The subclustering was then stratified by samples to illustrate the relative composition of each subtype of these infiltrating NK and T cells in the different groups (figure 6E). The data showed that Synco-2D-TriCEP treatment increased the proportion of NK cells in TME. Considering that the proportion of CD45⁺ cells in tumors treated with Synco-2D-TriCEP was significantly higher than in the other two groups (data not shown), the increase in NK cells is quite significant. Moreover, treatment with Synco-2D-TriCEP generated more favorable T cell responses. First, it increased the proportion of Th1/Th17 cells by approximately twofold and fourfold compared with PBS and GFP, respectively. Second, it reduced the relative presence of both Th2 (approx. twofold and fourfold reduction compared with PBS and GFP, respectively) and Treg cells (approx. twofold reduction compared with both control groups). Both of these changes on T cells in TME are considered desirable for cancer immunotherapy.

To determine the activation status of the infiltrating NK and T cells in the TME, we analyzed the expression of the activation and major cytotoxic effector markers of NK and T cells, including Klrk1 (NKG2D), Cd69 (Cluster of Differentiation 69), Stat3 (Signal transducer and activator of transcription 3) Prf1 (perforin), and Gzmb (granzyme B) (figure 7A–E). Among them, Cd69 is an activation marker for both T and NK cells.38 Although NKG2D is constitutively expressed on both NK and CD8⁺ T cells, its expression is enhanced when these cells become activated.39 As such, it is also considered as an activation marker for both cell types. Stat3 is a transcription factor that is activated downstream of many key cytokine receptors expressed by lymphocytes. As such, the presence of Stat3 is indicative of the activated status of immune cells. Moreover, it plays an important role in regulating NK cell function and is thus considered as a NK cell activation marker (figure 7C). Perforin and granzyme B are classical markers for critical cytolytic enzymes for both NK and T cells and their expression levels indicate their cytolytic activity.40 The expression of all these genes was significantly elevated in the tumors treated by Synco-2D TriCEP as compared with the other two groups with the p<0.05 (figure 7D,E), indicating that TriCEP directly contributed to the activation and/or effector function of NK and T cells.

In addition to the above activation markers, we also analyzed the expression of key cytokines in the infiltrated NK and T cells shown in (online supplemental file 1-Fig.S2), including Icos (online supplemental file 1-Fig.S2A), Ifng (online supplemental file 1-Fig.S2B) and Tgfb1 (online supplemental file 1-Fig.S2C). The expression of these cytokines is significantly elevated in NK cells, whereas significant expression of Icos and Tgfb1 is only observed within the T cell clusters in the TriCEP treated group (compared to the PBS control. The p values for the violin plots showing activation genes and cytokine genes are shown in (online supplemental file 1-Fig. S2D).

HEK293, SKOV3, CT26, Vero, BHK, TALL-104 cells were obtained from ATCC. CT26-EGFR cells were established from CT26 cells by stably transducing the cells with a lentiviral vector that contains EGFR extracellular and transmembrane domains without the intracellular sequence.17 All cells were cultured as described in online supplemental file 2. Synco-2D is an HSV-1-based OV. Its construction has been described in our previous publications.32

For building the BiCEP, the coding sequence for OMCP (1–152) and the mutated form of EGFα (m123), together a glycine/serine linker and a Myc tag was synthesized by GenScript (U0596DB120; U0596DB130) is inserted in the frame for ease of detection. TriCEP coding sequence was similarly synthesized, except that the coding sequence for a mutated form of IL-2 was added to the 5’ end. Both synthesized sequences were cloned into pcDNA3.1 plasmid to generate pcDNA3.1-BiCEP and pcDNA3.1-TriCEP, respectively.

For constructing amplicon plasmids containing these two chimeric engagers, the key components of an HSV amplicon, the Ori and the Pac signals, together with the EGFP coding sequence, were cut from pW7-EGFP, which is an amplicon that our lab had constructed and used in many of our previous studies.31 The cut-out fragment containing the amplicon components was then cloned into pcDNA3.1-BiCEP and pcDNA3.1-TriCEP, to generate Amplicon-BiCEP and Amplicon-TriCEP, respectively. The procedure for packaging amplicons is described in online supplemental file 2. The generated stocks of viruses were labeled as Synco-2D-GFP, Synco-2D-BiCEP, and Synco-2D-TriCEP, respectively, and stored at −80°C until use.

For determining the transgene expression from either the amplicon plasmids or from the packaged amplicons, HEK293 cells were transfected with pW7-GFP, Amplicon-BiCEP and Amplicon-TriCEP, and BHK cells were infected with the corresponding packaged amplicons. The supernatants were collected 48 and 72 hours later. The collected supernatants were either used directly or concentrated using 10 000 MWCO Millipore spin-columns and stored at −80°C before they were used for Western blot detection or other quantitative assays. Ponceau staining was performed on the gels loaded with supernatant samples to ensure a near equal sample loading.

The binding of OMCP to NKG2D on immune cells (TALL-104) and EGFα to EGFR on tumor cells (SKOV3) were determined by incubating the cells with the supernatants collected from HEK293 cells transfected with the amplicon plasmids or BHK cells infected with the packaged amplicon as described above. The experimental procedure for conducting the assay is detailed in online supplemental file 2. The binding of OMCP to NKG2D and EGFα to EGFR on the cells is determined by the detection of NKG2D⁺/Myc⁺ and EGFR⁺/Myc⁺ double-positive cells by flow cytometry, respectively.

Ovarian cancer cells (SKOV3) were cocultured with TALL-104 cells, at a different E:T ratios (1:1, 2:1, and 5:1) for 2–3 days. Tumor cell lysis was monitored in real-time using real-time fluorescent microscopy (IncuCyte; Essen Biosciences). The cytotoxicity is reported by the percentage of viable cells/percentage of confluence remaining at the end of 48-hour coculture.

DDAO-SE (CellTrace Far Red dye- C34564) labeled target cells (CT26-EGFR) were seeded at 100, 000 cells per well in a 96-round bottom tissue culture plate and cocultured with human primary NK cells at E:T ratios of 1:1, 3:1 and 5:1. The cells were incubated O/N at 37°C, 5% CO2 in a humidified incubator for 3–4 hours. The cells were then washed, trypsinized, fixed and permeabilized with Fix/Perm solution (BD Biosciences) and stained for cleaved caspase 3. The stained cells were then prepared for flow cytometric analysis. The detailed procedure is described in online supplemental file 2.

Vero cells in 12-well plates were infected with serially diluted stocks in triplicates. The titer of Synco-2D is determined by plaque-forming units counted 24–48 hours later. The amplicon titer is determined by counting the number of GFP⁺ cells. In most stock preparations, the Synco-2D to amplicon ratio is approximately 8–10:1.

Whole-cell lysates and supernatants from either transfected or infected cells were analyzed by following standard procedures for western blot. The primary antibody used for the detection of chimeric proteins is Myc-tag (1:2000) (Cell Signaling Technology, Danvers, Massachusetts, USA) and an HRP-labeled secondary antibody (anti-rabbit IgG, HRP linked Antibody) at 1:1000 dilution.

Immune-competent female BALB/c mice (4–6 weeks old) were purchased from Charles River Laboratories. All animal experiments were approved by the University’s Institutional Animal Care and Use Committee. 3×10⁵ CT26-EGFR cells were injected subcutaneously to the shaved right flank of the mice. Once the tumor volumes reached the approximate size of 6 mm in diameter, eight mice were randomized into different groups to receive either PBS control or Synco-2D treatment with or without the chimeric molecule-containing amplicons, at the dose of 5×10⁶ pfu Synco-2D per mouse. Three mice from groups receiving the treatment of PBS, Synco-2D GFP and Synco-2D TriCEP were euthanized on day 3 after virotherapy to collect tumor tissues for scRNA-seq or histology exam and spleens for other immune assays. The rest of the mice (n=5) were kept for 2–3 weeks to monitor tumor growth by measuring two perpendicular tumor diameters with a caliper. Tumor volume was calculated by the formula: tumor volume (mm³) = [length (mm)] × [width (mm)]² × 0.52.

Tumor tissues were fixed and embedded in paraffin and sections were prepared. H&E staining was performed as per standard procedures and as detailed in online supplemental file 2. The primary antibody used to stain the sections is GFP (Santa Cruz Biotech, Dallas, Texas, USA).

For scRNA-seq studies, the freshly collected tumors were immediately immersed in a tissue storage medium (Miltenyi, San Diego, California, USA) and kept at 4°C until ready for dissociation. Within 24 hours, tissues were processed to single‐cell suspensions using the human tumor dissociation kit from Miltenyi and the gentleMACS apparatus and this was done by following the manufacturer’s protocol. Single-cell suspensions were then stained with a fluorescently conjugated antibody specific to CD45 (BioLegend) for 30 min at 4°C. The cells were washed with cell staining buffer (BioLegend) and CD45⁺ live cells were sorted on a FACS Melody cell sorter (BD) into 2%FBS in PBS, which were kept on ice until the cells were further processed for scRNA-seq (scRNA-seq library preparation and sequencing, transcriptome analysis and single-cell data analysis)48–50 are included as online supplemental file 2.

All quantitative results are displayed as the mean±SD.D. The statistical difference between the two groups was compared using a Mann-Whitney U test or a Student’s t-test. If more than two groups were compared, analysis of variance was used. Statistical analysis was determined using Prism V.5 software (GraphPad Software, La Jolla, California, USA). A p <0.05 was considered statistically significant.