Introduction

Despite advances in diagnosis and treatment, breast cancer (BC) is still a leading cause of death among women with the 5-year survival rate of patients diagnosed with recurrent disease at only 20%.1 2 This situation is particularly dire for women with triple-negative BC (TNBC), an aggressive BC subtype characterized by high recurrence, metastasis, and high mortality rates.2 3 While immune checkpoint inhibitors (ICIs) have demonstrated enhanced clinical benefit in advanced solid tumors like melanoma, non-small cell lung cancer, and urothelial carcinoma, a similar treatment strategy has not been successful for the majority of patients with BC.4–6 However, recent studies suggest that TNBC may be among the BC subtypes that may benefit from immunotherapy approaches, given its expression of programmed death-ligand 1 (PD-L1) and a higher proportion of tumor-infiltrating lymphocytes compared with other BC subtypes.7 8 Although Food and Drug Administration-approved for patients with metastatic TNBC, pembrolizumab did not significantly improve overall survival in patients with positive PD-L1-positive disease versus chemotherapy alone.9 In the early stage of TNBC, the results were better. The pathological complete response rate within the neoadjuvant setting approached 63% (95% CI: 59.5% to 66.4%) for patients who received pembrolizumab in combination with chemotherapy compared with 56% (95% CI: 50.6% to 60.6%) for patients who received chemotherapy alone. Nonetheless, while initially responsive, the 5-year disease-free survival event rate remains low (11.6%) for high-risk, early disease patients,10 and acquired resistance to ICI-based therapies remains a clinically emerging, major challenge.11

Epithelial membrane protein-2 (EMP2) is a tetraspan protein that mediates the formation and integrity of multireceptor lipid raft microdomains on the cell membrane, and dysregulation of EMP2 has been shown in a number of gynecologic cancers.12–15 While expression of EMP2 is limited in non-pathological breasts, EMP2 expression is increased in greater than 75% of all human BCs sampled relative to normal breast glandular and ductal epithelial cells. High expression was observed in>75% of women with TNBC and in 100% of metastatic lesions.12 16 Thus, EMP2 has emerged as a potential target for precision cancer therapeutic approaches.

While antibody-drug conjugates (ADCs) have revolutionized the era of targeted therapy, their potential is yet to be fully realized due to off-target toxicity from drug release and the development of an immunogenic response against the therapeutic agent used.17–19 Fully human fusion proteins and immunotoxins are an emerging class of therapeutics that represent an innovative approach to the delivery of a deadly payload to a tumor cell with more specificity and potentially fewer avenues for the development of resistance than ADCs. Constructs developed in our laboratory incorporate the serine protease granzyme B (GrB), which has been validated as a cytotoxic payload when delivered as a targeted therapeutically in various formats including ligand-based and single chain antibody.20–24

Several features of GrB make it a natural conjugate for immunotoxin therapy. A natural component of the body’s defense against viral infection and tumor development, GrB has a well-understood mechanism of action that includes induction of caspase-dependent and independent apoptosis.25–27 Unconjugated GrB internalizes into cells in a perforin-dependent manner, significantly limiting off-target toxicity of GrB-based therapies that would be expected with chemotherapeutic agents delivered by ADCs.28 Further, as GrB is an endogenous protein present in plasma in both normal and pathological states, it is unlikely to engender an immunogenic response.20

Since GrB is a well-established molecular mediator delivered by cytotoxic T lymphocytes (CTL) and natural killer (NK) cells to target cells, we hypothesized that direct targeting of EMP2+TNBC may simultaneously induce apoptosis of the tumor cell and activate immune cells within the tumor microenvironment (TME). In this study, we developed a novel construct, GrB-Fc-KS49, comprised of the active GrB fused to the immunoglobulin G (IgG) heavy chain (Fc) domain containing a single-chain antibody against EMP2. This new construct showed a high, specific affinity for EMP2 using both flow cytometry and ELISA assays. Moreover, treatment with this agent induced cell death in EMP2-positive cells. Preclinical assessment of GrB-Fc-KS49 showed no acute toxicity and prolonged plasma retention after administration. In vivo efficacy against syngeneic model systems produced a significant reduction in tumor burden and a striking increase in CD45+immunopermissive cells with a notable rebalancing of tumor-associated macrophage (TAM) populations. Our results suggest that continued development of this agent may be successful at targeting advanced BC.

Materials and methodsCell lines

SKBR3, HEC-1-A, MDA-MB-231, MDA-MB-468, MDA-MB-361, 4T1/FLuc, and EMT6 cells were obtained from the American Type Culture Collection (Manassas, Virginia, USA) and cultured for up to 3 months before replacing. BTTR cells are trastuzumab-resistant BT474 cells (kind gift of Neil O’Brien, University of California Los Angeles). All human cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) medium (Mediatech, Manassa, Virginia, USA) except for HEC-1-A, which was cultured in McCoy’s media (MilliporeSigma, St. Louis, Missouri, USA). Both media were supplemented with 10% fetal calf serum (FCS; Hyclone Laboratories, Logan, Utah, USA), 1 mM sodium pyruvate, 2 mM L-glutamine, 100 U/mL penicillin, and 100 U/mL streptomycin (Life Technologies, Carlsbad, California, USA). MEF3.5−/− (MEF3.5) cells are mouse embryo fibroblasts (kind gift of Matthew Hayden, Dartmouth University) and were cultured in DMEM media supplemented with 15% FCS. The endometrial cell line HEC-1-A expressing EMP2 and HEC-1-A KD cells stably expressing a short hairpin RNA (shRNA) lentiviral knockdown of EMP2 were used as controls and have been previously described.29 C8161-GFP cells were a kind gift from Claire Lugassy (Institut Curie, Paris, France). The murine BALB/c syngeneic cells EMT6 and 4T1/FLuc cells were cultured in Roswell Park Memorial Institute medium and supplemented as described above. 4T1/FLuc were stably infected with firefly luciferase (FLuc) as previously described.12 Cells were cultured in 5% humidified CO2 chamber at 37°C. Cells were passaged weekly and tested periodically for mycoplasma contamination using the MycoAlert Kit (Lonza, Hayward, California, USA).

Animal research

The work performed was completed under approved protocol # 491-RN03 by the Animal Research Committee at MD Anderson or # 2004–182 by the Animal Research Committee at the University of California, Los Angeles. Experimental details are provided in the online supplemental methods and materials.

Construction of the GrB-Fc-KS49 expression vector

The GrB-Fc-KS49 DNA cassette was generated by overlapping PCR, as previously described.20 The construct is illustrated in figure 1A. Initially, three PCR fragments were generated: Human GrB fused to the N-terminus of the Fc (hinge CH2 and CH3) region of a human IgG1 heavy chain (GrB-Fc), and the previously described variable heavy (VH) and light (VL) sequences for the human anti-EMP2 diabody KS4913 flanked by a universal 218 linker, resulting in VH-218 and 218-VL fragments, respectively.21 Human GrB can cross-react with other species.30 The fusion construct was completed by two splice overlap extension PCR reactions, first with VH-218 and 218-VL to generate the VH-218-VL scFv (KS49scFv), followed by GrB-Fc with KS49scFv to generate GrB-Fc-KS49. The cassette was cloned into the mammalian expression vector pSecTag (Life Technologies, Waltham, Massachusetts, USA).

Figure 1

GrB-Fc-KS49 construction and purification. (A) The GrB-Fc-SK49 sequence is preceded by a hexa-histidine tag for purification by metal affinity chromatography, and an enterokinase cleavage site to remove the purification tag and activate granzyme B enzymatic activity. The DNA cassette encoding the GrB-Fc-KS49 sequence was cloned into the mammalian expression vector pSECTag. (B) Purified non-cleaved (dashed arrow) and cleaved (solid arrow) GrB-Fc-KS49 as assessed by SDS-polyacrylamide gel electrophoresis under reduced and non-reduced conditions. C, active GrB-Fc-KS49; GrB, granzyme B; NC, inactive (pro-) form of GrB-Fc-KS49; SDS, sodium dodecyl sulfate.

Expression and purification of GrB-Fc-KS49

GrB-Fc-KS49 was expressed in HEK-293E cells, as previously described.20 22 Briefly, cells (9×106 cells/mL) were transiently transfected with a mixture of the expression vector (4.5 µg/mL) with polyethylenimine (1:4.5) for 4 hours. Cells were diluted 1:1 with growth media and grown overnight at 37°C and 120 rpm. Following the addition of 3.8 mM valproic acid and 1:1000 anti-clumping agent, cells were incubated for a further 72 hours prior to harvesting the conditioned media, which contained the inactive (pro-) form of GrB-Fc-KS49. After purification of the dimer by cobalt-metal affinity chromatography, the protein was activated by cleavage of the purification tag with recombinant enterokinase (20 U/mg).24 Further purification by SP Sepharose chromatography to remove the cleaved tag resulted in purified, active GrB-Fc-KS49.

Additional methods are described in the online supplemental methods and materials.

ResultsGrB-Fc-KS49 expression and purification

Immunotoxins are antibody-toxin bifunctional molecules that combine the specificity of an antibody with a potent intracellular payload to kill target cells. We have previously characterized diabodies to EMP2 which show specificity and sensitivity to its expression. To create a second generation of molecules targeting EMP2, we fused the KS49 scFv with GrB as a toxin payload to ultimately determine its therapeutic efficacy (figure 1A). One liter of HEK-293E conditioned medium yielded approximately 20 mg of purified GrB-Fc-KS49 protein. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis showed one band at∼250 kDa and 100 kDa under non-reducing and reducing conditions, respectively, clearly demonstrating that the final construct is present as a disulfide-linked dimer as expected (figure 1B). Comparison of GrB-Fc-KS49 with its expected dimeric molecular weight of∼158 kDa suggests significant glycosylation when produced in HEK-293E cells.

Assessment of cellular EMP2 levels

We have previously shown that EMP2 is expressed in many human and murine breast and endometrial cancer cell lines.29 31 Within different subtypes, EMP2 is expressed in both hormone receptors, HER2+positive, as well as in TNBC. Western blot analysis indicated a range of cellular EMP2 expression (figure 2A), with all BC cell lines tested showing moderate to high levels of EMP2 expression. The highest EMP2 levels were observed in the human HER2+BC cell lines MDA-MB-361 and SKBR3 and the mouse TNBC cell line 4T1/FLuc. The human TNBC cell line MDA-MB-468 also showed high expression of EMP2. Two additional cell lines, the human melanoma cell C8161-GFP as well as murine embryonic fibroblasts MEF3.5 expressed low to undetectable levels of EMP2. For comparison, expression was also measured in the endometrial cancer cell line HEC-1-A, with the specificity of the response confirmed using the HEC-1-A shRNA knockdown cell line HEC-1-A KD.

Figure 2

Validation of binding characteristics of the GrB-Fc-KS49 immunotoxin to EMP2 cells. (A) Left, EMP2 expression in a panel of breast cancer cells was compared with the endometrial cancer cell line HEC-1-A and the melanoma cell line C8161 in a panel of human cells. Right, the level of EMP2 expression was measured in the murine 4T1/FLuc, EMT6 and MEF cells. (B) Left, binding of the immunotoxin to 4T1/FLuc and EMT6 cells show similar surface expression of the EMP2 antigen. The experiment was performed three times, with one representative experiment shown. The Kd was calculated using non-linear regression with specific binding. Right, surface binding of the GrB-Fc-KS49 immunotoxin to EMP2 was measured in HEC-1-A cells or cells engineered to stably express a shRNA lentiviral knockdown vector (HEC-1-A KD). Results are the average of two independent experiments, with error bars indicating SEM. EMP2, epithelial membrane protein-2; shRNA, short hairpin RNA. Kd, equilibrium dissociation constant. MFI, mean fluorescence intensity.

Binding affinity of GrB-Fc-KS49

To confirm that the scFv portion of GrB-Fc-KS49 immunotoxin recognized EMP2, we first determined the binding affinity of GrB-Fc-KS49 by Octet binding. Using a 1:1 binding model, the GrB-Fc-KS49 immunotoxin showed a strong equilibrium affinity constant (Kd) of 1×10E−12 M to an EMP2 peptide with no significant binding to a non-specific control (online supplemental figure S1). Next, we sought to confirm the specificity of the GrB-Fc-KS49 immunotoxin to cell surface EMP2 on cell lines with validated EMP2 levels using flow cytometry. The correlation of cellular EMP2 levels to cell surface expression levels was validated by demonstrating high cell surface binding of GrB-Fc-KS49 to two murine TNBC lines EMT6 and 4T1/FLuc. While 4T1/FLuc cells showed increased total EMP2 expression, both cell lines showed similar binding and saturation levels, suggesting that they may have a similar surface expression of the antigen. The calculated Kd for the GrB-Fc-KS49 were 2.063±0.35 and 1.469±0.89 nM for EMT6 and 4T1/FLuc, respectively, which is similar to binding from the KS49 IgG1.32 Binding was also tested on the HEC-1-A cell line which was used as a positive control as well as on cells engineered to knockdown EMP2 expression in HEC-1-A KD (figure 2B). While the Kd was similar for HEC-1-A and HEC-1-A KD (1.3 vs 1.0 nM, respectively), the calculated maximum number of binding sites orBmax showed the expected reduction, going from 5462 to 3716 for the two cell lines, respectively. Thus, GrB-Fc-KS49 appears to be an excellent candidate to target EMP2 on TNBC cells and other EMP2+cancers.

GrB activity assay

Our results thus far demonstrated that the binding of the GrB-Fc-KS49 immunotoxin was dependent on EMP2 levels. Next, an enzymatic activity assay was performed to evaluate the activity of the GrB component. The specific activity of commercial GrB was 122±10 U/nmol, and that of GrB-Fc-KS49 ranged from 98±5 U/nmol GrB (p=0.08, Student’s t-test) indicating that while some GrB enzymatic activity is lost, the molecule retained a high degree of its enzymatic function within the fusion construct (online supplemental figure S2).

Cytotoxicity of GrB-Fc-KS49 in vitro

The in vitro cytotoxicity of GrB-Fc-KS49 was assessed against a panel of BC and control cell lines (table 1 and online supplemental figure S3). Using an endometrial cancer cell line as a positive control, HEC-1-A was sensitive to GrB-Fc-KS49 treatment in vitro, with an average half-maximal cytotoxicity(IC50) of 44±5 nM. HEC-1-A KD cells with downregulated EMP2 were∼4 fold less sensitive, indicating that the cytotoxicity of the GrB-Fc-KS49 was receptor-mediated. The IC50 of the GrB-Fc-KS49 construct ranged from 40 to 83 nM against all tested human and mouse TNBC cell lines as well as the HER-2+ cell line SKBR3. In contrast, cell lines with very low to undetectable EMP2 levels were less sensitive to GrB-Fc-KS49. For all cells, free GrB had low cytotoxicity, indicating that the cytotoxicity is driven by the construct. Because internalization of antibodies and fusion proteins can result in their entrapment in the endo/lysosomal compartments and result in diminished cytotoxicity, we examined the effect of chloroquine, a known lysosomotropic agent, on cytotoxicity against MDA-MB-231 cells.21 A cytotoxicity profile of various concentrations of chloroquine against MDA-MB-231 cells established an IC50 of approximately 50 µM over 72 hours (online supplemental figure S3B) and 10 µM as the minimal level required to demonstrate toxicity. The presence of 10 µM chloroquine in combination with GrB-Fc-KS49 did not improve cytotoxicity towards these cells (table 1), suggesting that GrB-Fc-KS49 is efficiently released from intracellular vesicles and can access cytosolic GrB substrates.

Table 1

Cytotoxicity of GrB-Fc-KS49 on various cell lines

Internalization of GrB-Fc-KS49 into TNBC cells

We assessed the internalization of GrB-Fc-KS49 by immunofluorescence microscopy. The fusion protein internalized within 4 hours in the EMP2+BC cell lines, while internalization of the untargeted GrB was not observed, suggesting receptor-mediated internalization (online supplemental figure S4). To confirm its specificity to EMP2-positive cells, internalization of GrB-Fc-KS49 or GrB was also evaluated in MEF3.5 cells. No internalization was observed in this cell line.

Effects of GrB-Fc-KS49 on clinical chemistry parameters

Clinical chemistry parameters tested at 40 mg/kg GrB-Fc-KS49 delivered over 2 weeks were evaluated in CD-1 mice. Interestingly, whereas an increase in alkaline phosphatase (ALP) is indicative of toxicity,33 the levels of ALP showed a slight but significant decrease in GrB-Fc-KS49 treated mice. Nonetheless, ALP levels in both control and GrB-Fc-KS49 treated mice fall within the serum chemistry reference range for CD-1 mice.34 For all other parameters tested, a comparison of the control versus GrB-Fc-KS49 treated mice showed no significant alterations, suggesting a lack of toxicity associated with the GrB construct (table 2).

Table 2

Toxicity profile of GrB-Fc-KS49 compared to the vehicle control. N=4/group

Ex vivo stability and in vivo pharmacokinetics

The relative stability of GrB-Fc-KS49 under ex vivo conditions was assessed using sterile mouse serum at 37°C for 96 hours. As shown in figure 3A, Western blot analysis indicated a gradual reduction in GrB-Fc-KS49 over time. Quantitation of the SDS-PAGE gel by densitometric analysis of the Western blots indicated that GrB-Fc-KS49 was relatively stable at 37°C in mouse serum with an overall half-life exceeding 96 hours (figure 3B). Next, we determined the pharmacokinetics (PK) of the GrB-Fc-KS49 immunotoxin in vivo by injecting mice with the GrB-Fc-KS49 and assessing the Fc portion of the immunotoxin using Western blot analysis. In vivo PK indicated a biexponential plasma clearance, with a moderate initial clearance (t1/2α=18.4 hours) and a much slower terminal clearance rate (t1/2β=73.1 hours), confirming that GrB-Fc-KS49 is stable in vivo (figure 3C). The drug concentration at time 0 was determined to be 156 µg/mL. The volume of distribution was a relatively low 2.58 mL, suggesting the limited distribution of GrB-Fc-KS49 into peripheral sites outside the vasculature immediately after administration.

Figure 3

Ex vivo serum stability and in vivo pharmacokinetics. (A) Stability of GrB-Fc-KS49 in mouse serum at 37°C was assessed by Western blot (representative blot shown). (B) The Western blots were quantitated by densitometric analysis. The overall half-life of GrB-Fc-KS49 exceeded 96 hours. Data represents the average from four mice, with error bars indicating SEM. (C) In vivo pharmacokinetics of the GrB-Fc-KS49. Injection of 400 µg of GrB-Fc-KS49 indicated a two-phase clearance with a t1/2α of 18.4 hours and a t1/2β of 73.1 hours. GrB, granzyme B.

In vivo efficacy

To next evaluate the efficacy of GrB-Fc-KS49, syngeneic orthotopic models were created using 4T1/FLuc cells. When tumors approached 100 mm3, they were treated with the immunotoxin or control IgG two times per week. Treatment with GrB-Fc-KS49 significantly reduced tumor load (p=0.002, figure 4A; Online supplemental figure S5A), and throughout the treatment period, no signs of animal toxicity were observed (online supplemental figure S5B). This included no signs of chronic pain (eg, reluctance to move or failure to groom) as well as no changes in appetite as evaluated by fluctuations in weight (online supplemental figure S5B). To evaluate the histology of the tumors, they were initially fixed and analyzed by H&E, and increased cell death and immune infiltrates were observed (figure 4B and online supplemental figure S5C). To confirm these features, we stained the tumors for Ki-67 expression and performed the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay to assess tumor cell proliferation and cell death, respectively. Compared with the control treatment, GrB-Fc-KS49 reduced cellular proliferation as measured by Ki-67 expression by more than twofold (p=0.001; figure 4C, top). To next distinguish necrosis from apoptosis within these specimens, we performed a TUNEL assay.35 Samples were stained using the TUNEL kit to detect DNA strand breaks (fluorescein isothiocyanate; FITC) and Isolectin GS-IB4 Alexa Fluor 594 to assess vascular changes (figure 4C, bottom; higher magnification images are provided in online supplemental figure S5D). Compared with the control, treatment with the GrB-Fc-KS49 immunotoxin significantly enhanced cell death while reducing the presence of tumor-associated vasculature (p=0.02).

Figure 4

GrB-Fc-KS49 treatment reduces tumor load in a 4T1/FLuc breast cancer model. (A) 4T1/FLuc cells were implanted in the mammary fat pad of BALB/c animals. Tumors (50–100 mm3) were treated 3 times over 10 days with control IgG or the GrB-Fc-KS49 immunotoxin at 10 mg/kg. Results are expressed as the mean±SEM from n=4 for control or n=5 GrB-Fc-KS49 groups, and statistical differences are determined using a two-way ANOVA with Bonferroni’s multiple comparison test used for post hoc analysis. *p<0.05. (B) Tumor histology was evaluated on FFPE tumor sections using H/E staining. (C) For each group, all tumors were also stained for Ki-67 as a measure of cellular proliferation, with quantitation of staining presented to the right. All tumors were also stained by the TUNEL assay to measure apoptosis (FITC), Isolectin GS-IB4 Alexa Fluor 594 to assess vascular changes, and counterstained with DAPI. n=4 with results expressed as the mean±SEM. Statistical differences were determined using Student’s t-test. (D) Immunohistochemical staining of tumors for CD45, iNOS (M1 macrophage marker), and CD206 (M2 macrophage marker). Representative 400× magnification images are shown with quantitation of staining to the right. n=4 with results expressed as the mean±SEM. Statistical differences were determined using Student’s t-test. ANOVA, analysis of variance; DAPI, 4′,6-diamidino-2-phenylindole; FFPE, formalin-fixed paraffin-embedded; FITC, fluorescein isothiocyanate; GrB, granzyme B; IgG, immunoglobulin G; mAb, monoclonal antibody; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling.

GrB-Fc-KS49 treatment increases tumor infiltration of CD45+ cells

Traditionally, breast tumors are considered immunologically quiescent,36 but post-GrB-Fc-KS49 treatment, an altered histological distribution could be observed. To confirm that these cells were of immune origin, tissue sections were stained using anti-CD45 antisera. GrB-Fc-KS49 treatment significantly increased the percentage of CD45+cells within the tumor (p=0.003, figure 4D, top). Within the TME, TAMs have been reported to be the most abundant immune cells, and these cells have been shown to regulate BC regression or progression through their ability to polarize into proinflammatory M1 macrophages or immunosuppressive M2 macrophages, respectively.37 38

It has been reported that macrophages exist in high numbers at the margins of breast tumors but decrease within the tumor stroma.38 39 Within the tumor, macrophages may be found in association with blood vessels where they are believed to help orchestrate the migration of tumor cells.40 To evaluate M1 and M2 macrophages within these 4T1/FLuc tumors, we stained tumors for iNOS to detect M1 macrophages or CD206 to evaluate M2 macrophages. While M1 macrophages were typically present at low numbers throughout the tumor, GrB-Fc-KS49 significantly increased the number of cells observed per field (p=0.0006, figure 4D, middle panel). In contrast, M2 macrophages were readily observed along the periphery of control tumors, but these numbers were reduced with immunotoxin treatment (p=0.01, figure 4D, bottom panel).

GrB-Fc-KS49 treatment of EMT6 xenografts significantly increases immunopermissive M1 TAMs around the tumor vasculature

To validate these effects, we created a second TNBC syngeneic model using EMT6 cells. Treatment with GrB-Fc-KS49 began 1 week after tumor placement (50–100 mm3) with GrB-Fc-KS49 or with control IgG administered intraperitoneally two times per week. As in the previous model, the study was terminated 2 weeks after onset of treatment, due to the control group tumors approaching 1500 mm3. Treatment with the immunotoxin reduced tumor load by twofold over the short duration of the study (p=0.03; figure 5A). The growth rate from individual mice is shown in online supplemental figure S6A). Similar to the 4T1/FLuc study, these tumors exhibited marked cellular death and immune infiltration by standard H&E staining (figure 5B and online supplemental figure S6B). We next analyzed the tumors by immunohistochemistry to probe for EMP2 expression, tumor proliferation, and immune composition. EMT6 strongly expresses EMP2 in a cytoplasmic/membrane distribution, and treatment with the GrB-Fc-KS49 showed a slight, non-significant reduction in expression (figure 5C, top panel). We next examined Ki-67 expression as an estimate of tumor growth and used the TUNEL assay to evaluate cell death. Treatment with GrB-Fc-KS49 significantly reduced Ki-67 expression (p=0.005, figure 5C, middle panel) while samples stained by TUNEL (FITC) and Isolectin GS-IB4 Alexa Fluor 594 showed DNA strand breaks and vascular changes, respectively. Treatment with the GrB-Fc-KS49 immunotoxin relative to the control showed significantly more positive staining by TUNEL while reducing the presence of tumor-associated vasculature (p=0.02; figure 5C, bottom). Higher magnification images are shown in online supplemental figure S6C).

Figure 5

GrB-Fc-KS49 treatment improves survival in the EMT6 breast cancer model. (A) EMT6 cells were implanted in the mammary fat of BALB/c animals. Tumors (50–100 mm3) were treated two times per week with control IgG or the GrB-Fc-KS49 immunotoxin at 10 mg/kg. Average tumor load±SEM is shown from n=5 control or n=5 GrB-Fc-KS49 groups. Statistical differences were determined using a two-way ANOVA with Bonferroni’s multiple comparison test used for post hoc analysis. *p<0.05. (B) Tumor histology was evaluated on FFPE tumor sections using H&E staining. Images from a representative tumor from each group are shown. (C) For each group, all tumors were also stained for EMP2 and Ki-67 with representative images shown. Samples were also stained by the TUNEL assay (FITC), Isolectin GS-IB4 Alexa Fluor 594, and nuclei counterstained with DAPI, with a representative image shown. Quantitation of staining is shown to the right. n=3 with results expressed as the mean±SEM. Statistical differences were determined using an unpaired Student’s t-test. (D) Immunohistochemical staining of tumors for CD45, iNOS (M1 macrophage marker) and CD206 (M2 macrophage marker). Representative images at the indicated magnifications are shown with quantitation of staining to the right. n=4 with results expressed as the mean±SEM. Statistical differences were determined using an unpaired Student’s t-test. ANOVA, analysis of variance; DAPI, 4′,6-diamidino-2-phenylindole; EMP2, epithelial membrane protein-2; FFPE, formalin-fixed paraffin-embedded; FITC, fluorescein isothiocyanate; GrB, granzyme B; IgG, immunoglobulin G; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling.

As in the study with 4T1/FLuc tumors, EMT6 tumors showed notable cytologic heterogeneity. To characterize these cells, tumors initially were stained for CD45 to evaluate the immune composition within the microenvironment. GrB-Fc-KS49 treated cells showed a significant increase in CD45+cells compared with control-treated tumors (p=0.002, figure 5D, top panel). Within this population of immune cells, we characterized the TAMs and found a significant increase in M1 macrophages within the immunotoxin-treated tumor, particularly concentrated around tumor vasculature (figure 5D, middle panel). In contrast, along the periphery of the tumor, a reduction in M2 macrophages as measured by CD206 expression was observed (figure 5D, bottom panel). These results collectively suggest that GrB-Fc-KS49 alters the TME and may allow it to become more immune-permissive.

Immunotoxin-treated animals show an upregulation in immune signatures compared with control animals

To validate and further understand how the immunotoxin altered the TME, we performed RNA sequencing of whole tumor tissues on EMT6 tumors treated with either a control antibody or the GrB-Fc-KS49 immunotoxin as above. In these experiments, to limit ulceration, treatment ceased when tumors approached 1000 mm3. Similar to the results presented above, EMT6 tumors showed a reduction in tumor load when treated with the immunotoxin (online supplemental figure S7A). Initial exploration of whole-transcriptome data revealed expected variations in cell-type composition across samples as estimated through deconvolution analysis (see online supplemental methods). These variations are common confounding factors in the analysis of the tumor transcriptomics.41 To minimize the effect of cellular heterogeneity in our analyses, we regressed out the estimated TME scores from the gene expression matrix (see online supplemental methods). As a result, the main sources of gene expression variability in the corrected data fully segregated the control and GrB-Fc-KS49 treated samples (online supplemental figure S7B), and this transformed data was used for all downstream analysis (online supplemental figures S6–S8).

We first unitized gene set enrichment analysis to evaluate the macrophage signature in immunotoxin versus control-treated tumors. Analysis revealed an increased score in canonical macrophage expression after treatment with the immunotoxin, and this increase was mostly associated with a preferential M1 signature (online supplemental figure S7C) in GrB-Fc-KS49-treated samples as compared with control samples. More specifically, we next evaluated the ratio of M1:M2 macrophages using the NOS2:Mrc1 (iNOS:CD206) ratio and cellular proliferation using murine Ki-67 expression following treatment with the GrB-Fc-KS49 immunotoxin or control (figure 6A). In agreement with the immunohistochemistry results presented above, treatment with the GrB-Fc-KS49 immunotoxin induced a significant reduction in mKi-67 expression. M1 and M2 macrophages are often defined by multiple markers such as CD86, CD80, and iNOS for M1 macrophages or Arg1 and CD206 for M2 macrophages. To evaluate if the immunotoxin produced a consistent effect, we examined its ability to affect the M1 and M2 ratio across these markers. The GrB-Fc-KS49 immunotoxin increased the ratio of M1 to M2 macrophages when evaluating CD86:Arg1 or CD80:Arg1 levels (online supplemental figure S7D).

Figure 6

GrB-Fc-KS49 increases immune infiltration within the tumor microenvironment. (A) Estimated gene expression of the proliferation marker mKi67 (left) and expression ratio of M1 (Nos2 or iNOS) and M2 (Mrc1 or CD206) macrophage markers (right) for the control and immunotoxin-treated mice (from RNA-seq data). (B) Volcano plot summarizing RNA-seq differential expression results between immunotoxin and control IgG-treated tumors. Selected genes identified as significantly regulated are highlighted. Upregulated genes correspond to those showing higher expression after treatment with the GrB-Fc-KS49 immunotoxin. Key: down (downregulated), ns (not significant), up (upregulated). (C) Functional enrichment of upregulated and downregulated genes. Selected, non-redundant pathways are highlighted. Key: GO:MF (Gene Ontology, Molecular Function); GO:BP (Gene Ontology Biological Processes), GO:CC (Gene Ontology Cellular Component); REAC (Reactome Pathways). (D) Expression barplots for the top 20 most significant genes between immunotoxin and control IgG-treated tumors (empirical Bayes t-test p value, from RNA-seq data, n=4 per group). The y-axis represents the gene expression residuals after correcting for the estimated tumor heterogeneity (see online supplemental methods). GrB, granzyme B; IgG, immunoglobulin G.

Further differential expression analysis between the GrB-Fc-KS49 treatment and control revealed a small yet distinct set of significantly regulated genes (empirical Bayes t-test p value<0.05, figure 6B). In fact, functional enrichment of genes upregulated by the immunotoxin revealed a signature of inflammation and adaptive immune response pathways (figure 6C) while genes downregulated after GrB-Fc-KS49 treatment control-treated tumors showed enrichment in genes associated with chemical homeostasis and positive regulation of vascular associated smooth muscle cell migration, among others (figure 6C, online supplemental figure S8). Specifically, 9 of the top 20 significantly regulated genes were involved in the regulation of inflammatory cells including dendritic cells (Siglech), B cells (Spib, CD22) and NK cells (Kird1) (figure 6D). In addition, genes involved in caspase-dependent cell death were upregulated (Rabac1), while genes in pathways associated with tumor progression were downregulated (Sema3c, Osmr, Plekhh2, Dclk1).

Discussion

In BC, our previous results have shown that EMP2 is upregulated in over 75% of patients, including those with the triple-negative subtype. Loss of EMP2 reduces tumor growth, in part likely due to its regulation of HIF1α and tumor-mediated neoangiogenesis.31 We have previously created diabodies to EMP2 and shown that they have the potential to internalize.16 In this study, we aimed to improve and optimize recombinant scFv fragments for targeted therapy by creating an immunotoxin.42 Immunotoxins make up a category of protein-based therapeutics. They typically consist of at least two domains, where one part allows for receptor binding and the other part induces apoptosis after internalization.

GrB is a well-known serine protease that is part of the body’s natural defense against viral infection and tumor development, delivered by CTL and NK cells to the target cells.25 Once endocytosed, GrB remains trapped in endocytic vesicles until released by perforin into the cytosol, where it induces caspase-dependent and independent apoptosis, as well as directly cleaving nuclear matrix material.27 Capitalizing on this ability, we developed a fully human GrB-based immunotoxin to the tetraspan protein EMP2 termed GrB-Fc-KS49. In this study, we characterized its ability to bind to human and murine EMP2 as we have previously shown that this scFv cross-reacts to both species.43 44 Like those studies, the immunotoxin to EMP2 displayed sensitivity to EMP2 levels in both mouse and human cell lines, and this specificity translated to its internalization and subsequent apoptosis of EMP2-positive cells. This phenotype was further observed in vivo, where in both EMT6 and 4T1/FLuc syngeneic models, treatment with the GrB-Fc-KS49 reduced tumor load and increased immune cell recruitment to the tumor.

To begin its preclinical characterization, the immunotoxin was evaluated for both its PK and for acute toxicity as we and others have documented that EMP2 is expressed in type 1 pneumocytes of the lung as well as in keratinocytes of the epidermis.16 45 46 Preliminary characterization revealed no visible toxicity or death in animals, suggesting that the 10 mg/kg dosage is well tolerated. To evaluate the PK of the immunotoxin, we evaluated its presence through the detection of the Fc region and binding to an EMP2 peptide. Our studies show an initial clearance of t1/2α of 18.4 hours and a much slower terminal t1/2β of 73.1 hours, which is in line with other GrB-immunotoxins published.20 22

While we did observe significant efficacy of GrB-Fc-KS49 in vivo, the overall length of each study was limited primarily due to the growth rate of control tumors. Given that the doses of other GrB-based constructs in our studies exceed doses administered here by at least threefold, it is likely that efficacy can be markedly improved by this construct at higher doses.20 While the maximum tolerated dose of GrB-Fc-KS49 has not yet been established, an increased dose at this schedule is unlikely to result in toxicity as doses of GrB-based therapeutics approaching 500 mg/kg (over 10 times the total dose used in these studies) have been shown to be non-toxic (manuscripts in preparation). In addition, our recent studies with other GrB-based constructs have shown that a weekly dose schedule (QW×5) improves efficacy compared with schedules with shorter intervals.20 Studies are underway to test this.

BC is traditionally considered an immunologically suppressive tumor, with the majority showing a lack of lymphocytic infiltration and consequently, modest response rates to immunotherapeutic agents such as anti-programmed cell death protein-1/PD-L1.36 The recruitment of myeloid cells is believed to play a key role in the establishment of this suppressive microenviro