Design, expression, and purification of a GrBmut-TRA AFP

For modeling purposes, we incorporated the anti-HER2 antibody trastuzumab (TRA) as an AFP in order to demonstrate proof-of-concept and compare efficacy relative to the FDA-approved ADC trastuzumab emtansine (T-DM1). Our AFP design incorporated the pro-form of GrB on the N-terminus of the TRA light chain (LC) to conveniently provide a mature and catalytically active form of GrB following processing/polishing steps (as detailed in the Materials and Methods). In its native pro-form, GrB contains Gly-Glu dipeptide residues on the N-terminus (i.e., designated a pro-peptide) that renders the molecule inert (and non-toxic during expression in cells) and must first be removed by a selective enzyme such as dipeptidyl peptidase I/cathepsin C under physiological settings. In this sense, the resulting N-terminal isoleucine allows GrB to fold into a catalytically active molecule [25]. We, therefore, engineered an experimental pro-peptide that modeled these constraints and also contained an EK cut site and 6X HIS tag to provide controlled release of the pro-peptide following AFP expression/purification and negative selection of AFP variants retaining the pro-form of GrB, respectively (see Fig. 1).

Fig. 1

GrBmut-TRA construct design and physical characterization. A Schematic representation of DNA vectors required for expressing Pro-GrBmut-TRA as a human IgG1 antibody in Expi293 cells. Following affinity chromatography purification, EK cleavage was required to generate a catalytically active GrBmut-TRA molecule. B SDS-PAGE and Coomassie Blue staining of purified TRA and GrBmut-TRA under reducing and non-reducing conditions. Arrow insets indicate HC and LC bands for each molecule. C Western blot analysis of TRA, Pro-GrBmut-TRA, and GrBmut-TRA using reagents specific to human IgG and GrB. Abbreviations used: EK, enterokinase; GrB, granzyme B; rGrB, recombinant GrB; GrBmut-TRA, catalytically active AFP; HC, heavy chain; LC, light chain; NR, non-reduced; Pro-GrBmut-TRA, catalytically inert molecule; R, reduced; TRA, trastuzumab

For our ADC screening approach, though, it is critical to utilize a GrB molecule that has been further engineered for improved functional characteristics. To demonstrate this importance, when initially conjugating WT GrB to TRA (designated GrB-TRA) (as similarly described in Fig. 1A), we observed unacceptable AFP targeting properties as a result of non-specific binding. More explicitly, although GrB-TRA interacted with the extracellular domain of HER2 by ELISA and flow cytometry, the AFP also bound to irrelevant proteins, particularly at higher concentrations (Additional file 1: Figures S1A, S1B). These binding effects prevented any discrimination in cytotoxicity between B16 or B16.HER2 target cells (Additional file 1: Figure S1C). Due to several cationic residues, GrB’s overwhelming positive charge likely contributes to electrostatic interactions that manifest in non-specificity as first addressed by others [26,27,28,29]. We confirmed such suspicions by disrupting charged interactions with a chelator in our assays. In the presence of EDTA, GrB-TRA interacted specifically with HER2 by ELISA and flow cytometry while non-specific interactions were drastically diminished (Additional file 1: Figures S2A, S2B). However, considering that the enzymatic activity of GrB lessened with increasing concentrations of EDTA (Additional file 1: Figure S2C) and EDTA usage in cell-based in vitro assays would disrupt target cell behavior/viability over several days in culture, we elected to pursue a more robust scenario for AFP modeling by engineering potentially problematic GrB amino acids as described in the Materials and Methods. Overall, we successfully replaced suspected charged residues (R110A, R114A, R116A, K239A, K240A, K243A, K244A), disrupted a putative aggregation site (C210A) (as previously noted [30]), and included a serpin B9 resistance mutation (R201K) (endowing resistance to serine protease scavengers), changes collectively resulting in GrBmut. Ultimately, the nature of incorporating a conventional antibody (to more closely mirror FDA approved ADCs) as an AFP with GrBmut has not been described before.

TRA and GrBmut-TRA were expressed in Expi293 cells and purified by Protein G affinity chromatography. The GrBmut-TRA construct was designed to express an N-terminal pro-peptide (6 × HIS-GGS-FLAG tag) (designated Pro-GrBmut-TRA), which prevents GrBmut function (Fig. 1A) and toxicity to Expi293 cells. To activate the AFP, EK (conjugated to 6 × HIS) is admixed to cleave the pro-peptide and release the enzymatically active chain of GrBmut. The entire mixture was then subjected to IMAC chromatography to remove unwanted impurities such as uncut Pro-GrBmut-TRA, cut pro-peptide, and EK. Additional file 1: Figure S3 demonstrates the integrity of this refining process with our purified GrBmut-TRA protein losing HIS and FLAG tag reactivity (compared to Pro-GrBmut-TRA) when analyzed by flow cytometry and western blot.

Figure 1B details the SDS-PAGE analysis of TRA and GrBmut-TRA that confirmed the purity and predicted molecular weights of the constructs: TRA at 150 kDa (50 kDa heavy chain [31], 25 kDa LC) and GrBmut-TRA at 200 kDa (50 kDa HC/LC). The GrBmut component of the fusion protein contains various glycosylation sites, increasing the apparent molecular weight of the GrBmut-LC to approximately 60–65 kDa under reducing conditions and 250–260 kDa for the full construct. Western blotting further supported molecular weights as the reduced unmodified HC was evident at 50 kDa for TRA and GrBmut-TRA (Fig. 1C) while anti-GrB reactivity demonstrated the AFP’s LC band slightly above 50 kDa (Fig. 1D). These data, overall, demonstrate an ability to reliably produce and purify sufficient quantities of GrBmut-TRA with predicted structural characteristics.

Binding and enzymatic activity of GrBmut-TRA

We next assessed the binding profiles of the AFP relative to TRA and T-DM1. Using indirect ELISA, binding to immobilized HER2 versus an irrelevant protein was initially determined. The observed Kd values indicated similar binding profiles between the constructs (TRA 0.140 nM, 95% CI [0.131, 0.150] vs. T-DM1 0.266 nM, 95% CI [0.226, 0.313] vs. GrBmut-TRA 0.439 nM, 95% CI [0.399, 0.483]) while non-specific interactions were negligible (Fig. 2A). Flow analysis also demonstrated antibody specificity to membrane-anchored HER2. In scenarios with cell lines expressing HER2 (i.e., B16.HER2, SK-BR-3, and SK-OV-3), GrBmut-TRA, TRA, and T-DM1 binding was readily apparent and closely aligned (Fig. 2B). As a control, antibody binding to B16 cells lacking HER2 expression was not observed. A separate experimental AFP created against SIINFEKL/H-2 Kb also demonstrated selective target binding, helping support the generalizability of our approach (Additional file 1: Figure S4). Lastly, we incorporated an absorbance assay to assess the enzymatic activity of GrBmut against a defined caspase substrate. Across conditions that included rGrB, GrB-TRA, GrBmut-TRA, and TRA, specific activity was only detected in constructs bearing GrB (Fig. 2C). In separate studies, AFPs not previously subjected to EK digestion were enzymatically inactive (data not shown). Importantly, these results also demonstrated that specific activity of our engineered GrB variant (GrBmut) was indistinguishable from WT rGrB (165 U/nM, 95% CI [162, 176] vs 176 U/nM, 95% CI [170, 182], respectively).

Fig. 2

GrBmut-TRA target binding and GrB activity. A ELISA determination of TRA, T-DM1, and GrBmut-TRA binding to immobilized HER2 or irrelevant protein at various concentrations. B Assessment of TRA, T-DM1, and GrBmut-TRA binding to HER2-expressing (B16, SK-OV-3, SK-BR-3) and non-expressing (B16) cell lines by flow cytometry. C Enzymatic activity of GrB was determined through absorbance using a GrB-specific chromogenic substrate (Ac-IEPD-pNA). Abbreviations used: CI, 95% confidence intervals; ECD, extracellular domain; GrB, granzyme B; rGrB, recombinant GrB; GrB-TRA, catalytically active AFP with WT GrB; GrBmut-TRA, catalytically active AFP with mutated GrB; SA, specific activity; T-DM1, ado-trastuzumab emtansine; TRA, trastuzumab. Bars ± STDEV. Select results are based on technical replicates of 3 samples per treatment group. Individual data values are provided in supplementary information (Additional file 2)

GrBmut-TRA target cell internalization and apoptosis

GrBmut-TRA internalization was subsequently determined by IF using the B16 and B16.HER2 cell lines to document the role of AFP binding HER2. Cells were incubated under various conditions involving activated rGrB alone, TRA, or GrBmut-TRA and acid washed to remove surface-bound proteins (confirmed in part by an absence of TRA staining of B16.HER2 cells). As expected, GrB-reactivity was only observed in B16.HER2 cells provided GrBmut-TRA, indicating selective target binding as well as internalization of the AFP likely via HER2 receptor-mediated endocytosis (Fig. 3A). The lack of signal in wells treated with an equimolar concentration of rGrB further supports the notion that internalization of the AFP is not mediated by non-specific cellular uptake mechanisms such as pinocytosis. Such data is further corroborated by a report by Cienfuegos and colleagues who documented internalization and cytosolic accumulation of GrB conjugated to a specialized fusion protein [30].

Fig. 3

Internalization and apoptosis-inducing effects of GrBmut-TRA. A B16 and B16.HER2 cells were treated with molar equivalents of rGrB, TRA, or GrBmut-TRA before being acid washed, fixed/permeabilized, stained for GrB (green) and nuclei (blue), and assessed by IF. B HER2-expressing and non-expressing cells were treated overnight under various conditions that included TRA, T-DM1, or GrBmut-TRA. Camptothecin was utilized as a positive control for inducing cell death. All cells were collected and stained with reagents detecting caspase 3/7 and dead cells by flow cytometry. Bar graphs indicate the frequency of caspase 3/7 positive events. Representative histogram plots across treatments are also provided. Abbreviations used: GrB, granzyme B; rGrB, recombinant GrB; GrBmut-TRA, catalytically active AFP; Pro-GrBmut-TRA, catalytically inert molecule; T-DM1, ado-trastuzumab emtansine; TRA, trastuzumab. *P < 0.05, bars ± STDEV. Select results are based on technical replicates of 2 samples per treatment group. Individual data values are provided in supplementary information (Additional file 2)

After introduction into a cell, GrB is capable of inducing apoptosis through various intracellular processes, including the direct cleavage of pro-caspase 3 as well as the disruption of general mitochondrial integrity [32]. Therefore, to ensure that internalized GrB potentiated previously documented effects of caspase upregulation [30, 33, 34], we evaluated the caspase 3/7 activation of target cells subjected to the AFP through flow cytometry. B16, B16.HER2, and SK-BR-3 cells were incubated overnight with 500 ng/mL of TRA, T-DM1, or GrBmut-TRA. All cells were then collected, stained, and analyzed to determine relative populations bearing caspase-3/7 activity. Interestingly, GrBmut-TRA and T-DM1 demonstrated similar levels of caspase induction across HER2-expressing cell lines with minimal activity in the HER2-deficient B16 cell line (Fig. 3B). Identical treatment schemes with TRA, or Pro-Grbmut-TRA failed to induce significant caspase-3/7 incidence in cells. These data help confirm the apoptotic-inducing mechanism of internalized GrBmut-TRA that was comparable to the microtubule inhibitor payload of T-DM1 [35].

In vitro cytotoxicity following GrBmut-TRA treatment

Finally, target cell cytotoxicity was evaluated in vitro to more conclusively demonstrate AFP-induced killing. Using our panel of tumor cell lines (B16, B16.HER2, SK-OV-3, and SK-BR3), the cytotoxic effects of TRA, T-DM1, and GrBmut-TRA were determined after 48 h in culture by crystal violet staining as detailed in the Materials and Methods. Overall, trends in target cell killing were comparable between T-DM1 and GrBmut-TRA—although differences existed in general cell line susceptibility between the constructs that could be a result of HER2 copy number on target cells, discrepancies in drug-to-antibody ratios, and/or general sensitivity to payloads (Fig. 4). EC50 values for T-DM1 and GrBmut-TRA were as follows: B16.HER2: T-DM1—5.319 nM, 95% CI [4.252, 6.647] vs AFP—8.887 nM, 95% CI [6.491, 12.21]; SK-OV-3: T-DM1—3.305 nM, 95% CI [2.609, 4.175] vs AFP—11.920 nM, 95% CI [7.344, 19.420]; and SK-BR-3: T-DM1—0.041 nM, 95% CI [0.034, 0.050] vs AFP—0.417 nM, 95% CI [0.325, 0.536]. Similar to T-DM1 [36, 37], GrBmut-TRA’s internalized payload does not easily pass the plasma membrane to inflate cell killing in vitro. Indeed, an advantage to modeling ADC potential with our described AFP is the determination of direct target cell binding and killing without complications from a membrane permeable payload. The relative absence of killing of HER2-negative B16 cells (i.e., > 100 nM) further indicated the targeted nature of our GrBmut-TRA to incite cell cytotoxicity. These data do not appear to be a phenomenon relevant only to GrBmut-TRA since an AFP targeting the SIINFEKL/H-2 Kb molecule was also capable of selectively inducing cell cytotoxicity in vitro (Additional file 1: Figure S4).

Fig. 4

Target cell cytotoxicity following GrBmut-TRA treatment. Tumor cell lines deficient in or expressing HER2 were plated overnight and subsequently exposed to various concentrations of rGrB, TRA, T-DM1, or GrBmut-TRA. After 48 h, cell viability was determined by crystal violet staining. Target cell viability (%) was calculated based on cells untreated and exposed to 10 μM camptothecin (i.e., maximum cell death). The table inset summarizes EC50 values (ng/mL) across tumor cell lines and treatment conditions. Abbreviations used: CI, 95% confidence intervals; GrB, granzyme B; rGrB, recombinant GrB; GrBmut-TRA, catalytically active AFP; T-DM1, ado-trastuzumab emtansine; TRA, trastuzumab. Bars ± STDEV. Select results are based on technical replicates of 3 samples per treatment group. Individual data values are provided in supplementary information (Additional file 2)

Ultimately, for ADC modeling purposes, our results help establish the credibility of a GrB-conjugated AFP to selectively interact with an antigen of interest, internalize, and induce target cell cytotoxicity.