4-1BB (CD137) is a key costimulatory immunoreceptor and promising therapeutic target in cancer. To overcome limitations of current 4-1BB–targeting antibodies, we have developed PRS-343, a 4-1BB/HER2 bispecific molecule. PRS-343 is designed to facilitate T-cell costimulation by tumor-localized, HER2-dependent 4-1BB clustering and activation.
PRS-343 was generated by the genetic fusion of 4-1BB–specific Anticalin proteins to a variant of trastuzumab with an engineered IgG4 isotype. Its activity was characterized using a panel of in vitro assays and humanized mouse models. The safety was assessed using ex vivo human cell assays and a toxicity study in cynomolgus monkeys.
PRS-343 targets 4-1BB and HER2 with high affinity and binds both targets simultaneously. 4-1BB–expressing T cells are efficiently costimulated when incubated with PRS-343 in the presence of cancer cells expressing HER2, as evidenced by increased production of proinflammatory cytokines (IL2, GM-CSF, TNFα, and IFNγ). In a humanized mouse model engrafted with HER2-positive SK-OV-3 tumor cells and human peripheral blood mononuclear cells, PRS-343 leads to tumor growth inhibition and a dose-dependent increase of tumor-infiltrating lymphocytes. In IND-enabling studies, PRS-343 was found to be well tolerated, with no overt toxicity and no relevant drug-related toxicologic findings.
PRS-343 facilitates tumor-localized targeting of T cells by bispecific engagement of HER2 and 4-1BB. This approach has the potential to provide a more localized activation of the immune system with higher efficacy and reduced peripheral toxicity compared with current monospecific approaches. The reported data led to initiation of a phase I clinical trial with this first-in-class molecule.
See related commentary by Su et al., p. 5732
This article is featured in Highlights of This Issue, p. 5727
Immunotherapy with checkpoint inhibitors, such as anti–PD-1 mAbs, has had a major impact on cancer therapy. Although such therapies afford durable responses or even cures, additional therapeutic strategies are still required for the majority of patients who do not respond or relapse. The activation of costimulatory pathways such as 4-1BB has long been acknowledged to hold great promise, but clinical efforts thus far have failed to demonstrate broad efficacy and have been associated with severe toxicity (1, 2). Herein, we describe the HER2/4-1BB bispecific molecule PRS-343, which aims to overcome these limitations by activating 4-1BB in a tumor-localized manner, sparing the periphery of unwanted toxicity. The data presented herein confirm the desired mode of action of PRS-343, providing a novel approach to target 4-1BB that may prove both safer and more efficacious compared with monospecific targeting. PRS-343 has the potential to offer an alternative therapeutic strategy to patients in multiple HER2-positive indications including bladder, breast, and gastric cancer.
4-1BB, also known as CD137, is a costimulatory immune receptor, a member of the TNF receptor (TNFR) superfamily, and is predominantly expressed on activated CD4+ and CD8+ T cells, activated B cells, and natural killer (NK) cells (3). 4-1BB plays an important role in the regulation of immune responses and together with the fact that it is expressed on tumor-infiltrating lymphocytes (TIL) makes it a promising target for cancer immunotherapy. The 4-1BB ligand (4-1BBL) is constitutively expressed on several types of antigen-presenting cells (APC; ref. 4). Upon pathway activation, 4-1BB facilitates enhanced proliferation, cytokine production, and cytolytic activity of T and NK cells (5). Recent studies have pointed to the pathway's impact on mitochondrial capacity and biogenesis of T cells to explain why 4-1BB agonism could help overcome the immunosuppressive landscape of the tumor microenvironment (6). While most research has focused on 4-1BB in the context of T effector cells, it should be noted that 4-1BB is also expressed on T-regulatory cells where its role remains contentious and requires further elucidation (7). 4-1BB has also been implicated in promoting a central memory T-cell response, which may support therapeutic persistence of tumor-specific T cells and resistance to exhaustion in patients treated with a 4-1BB agonist (8, 9).
The potential therapeutic benefit of 4-1BB costimulation has been demonstrated in multiple preclinical models. The forced expression of 4-1BBL on a tumor, for example, leads to tumor rejection (10). Likewise, the forced expression of an anti–4-1BB single-chain antibody fragment (scFv) on a tumor leads to a CD4+ T-cell and NK-cell–dependent elimination of the tumor (11–13). A systemically administered anti–4-1BB antibody has also been demonstrated to lead to retardation of tumor growth in mouse models (14).
Human ex vivo data support the potential of 4-1BB as a costimulatory receptor in cancer therapy. It has been reported that for T cells isolated from human tumors, 4-1BB is a marker for those that are tumor-specific (15). In line with this observation, anti–4-1BB antibodies can be utilized to improve adoptive T-cell therapy (ACT) by augmenting the expansion and activity of CD8+ melanoma TILs (16). Further clinical evidence for the importance of 4-1BB signaling for a sustained and successful anticancer T-cell response is provided in ACT with chimeric antigen receptors (CAR), where inclusion of 4-1BB–signaling elements in the cytoplasmic domain of CARs has been found to improve durability of clinical responses (17, 18).
Receptors from the TNFR family in general require higher-order clustering for efficient downstream signaling activation in the TNFR-expressing cell (19), which usually is not afforded by the soluble ligand alone, but requires engagement of another cell expressing the respective ligand on its surface. For example, despite being a homotrimer, soluble 4-1BBL is not capable of efficiently activating 4-1BB downstream signaling (20). Antibodies targeting 4-1BB are either inherently agonistic or often require a secondary means of clustering beyond bivalent binding to allow for agonistic activity. It is hypothesized that in vivo TNFR agonism by such antibodies is dependent on simultaneous binding to Fcγ receptor–positive cells (21, 22).
The intricate mode of activation of 4-1BB may underlie the modest clinical success obtained with monospecific anti–4-1BB antibodies to date. BMS-663513 (urelumab; ref. 23) has been hampered by dose-limiting hepatotoxicity, likely due to its tendency to systemically activate 4-1BB (1). Initial clinical trials with PF-05082566 (utomilumab; refs. 24, 25) have not shown the same safety issues as urelumab, yet preclinical characterization suggests it is a less potent 4-1BB agonist (2).
To overcome the challenges of targeting the 4-1BB pathway systemically, we have generated PRS-343, a bispecific molecule designed to activate the pathway in a tumor-localized and -dependent manner. PRS-343 binds 4-1BB and the tumor-associated antigen HER2, and was generated by the recombinant fusion of a 4-1BB–specific Anticalin protein to a HER2-specific antibody. Anticalin proteins are engineered variants of lipocalins, a family of natural extracellular binding proteins. Lipocalins are characterized by a highly conserved tertiary structure despite low amino acid identity. This property coupled to the fact that they possess free C- and N-termini, which are not required for target recognition, allow facile fusion to other proteins, defining a versatile basis for a multispecific biologics platform (26). The molecule, PRS-343, is designed to promote 4-1BB clustering by bridging T cells with HER2-positive tumor cells, providing a potent costimulatory signal to tumor antigen–specific T cells, further enhancing T-cell receptor–mediated activity and leading to tumor destruction. PRS-343–mediated 4-1BB activation is, therefore, biased toward locations in the body where T cells and tumor cells are colocalized, such as in primary tumors with TILs or in lymph nodes containing tumor metastases. Here, we describe the generation and preclinical characterization of PRS-343 with regard to target engagement, biological activity, and safety.
Materials and Methods
Engineering and production of Anticalin proteins and bispecific fusion proteins
Recombinant 4-1BB (R&D Systems) was used as the target for phage display selection and ELISA screening of cognate Anticalin protein candidates. A random library based on human neutrophil-gelatinase–associated lipocalin (NGAL) with high combinatorial complexity was prepared by concerted mutagenesis of multiple amino acid positions (27). 4-1BB target specificity was confirmed by ELISA screening and selected Anticalin proteins were optimized via partial mutagenesis of coding regions followed by biophysical and functional characterization, which resulted in selection of a lead candidate J10.
Bispecific antibody-Anticalin fusion proteins were generated by the recombinant fusion of the 4-1BB–specific Anticalin protein (J10) to either the heavy or light chain N- or C-terminus of a modified variant of trastuzumab. Specifically, the isotype of trastuzumab was changed to an IgG4 and additional mutations were introduced to reduce interactions with activating and inhibitory Fcγ receptors. Mutation S228P, which has been described to suppress half-antibody exchange (28), was introduced together with mutations F234A and L235A, which have been reported to reduce binding to FcγRI (29, 30). Bispecific proteins were obtained by recombinant expression in mammalian cells (HEK293 or CHO) using standard techniques.
Reagents and cell lines
Recombinant 4-1BB and HER2 protein (human and cynomolgus monkey) were obtained from R&D Systems and Sino Biological. The cell lines HT-29, SKBR-3, SKOV-3, BxPC-3, MCF-7, MDA-MB-231, MKN-7, NCI-N87, ZR-75-1, A375, A431, A549, JIMT-1, MDA-MB-453, MKN45, SUM-225, and LoVo were obtained from ATCC. Primary cells (human cardiac myocytes, human epidermal keratinocytes, human pulmonary fibroblasts, human cardiac fibroblasts, human dermal microvascular endothelial cells, human umbilical vein endothelial cells, human aortic smooth muscle cells, and human bronchial smooth muscle cells) were obtained from Promocell. The cells were expanded following cell bank instructions. The HER2 levels were confirmed by QifiKit (Dako) according to the provided protocol and were standardized to HER2 levels expressed by SKBR3. The list of the cells, corresponding cell bank catalogue number, and their respective HER2 expression levels standardized to SKBR3 HER2 levels are summarized in Supplementary Table S1. CHO cells expressing human 4-1BB (h 4-1BB) were established by stable transfection of h 4-1BB using the Flp-In System (Invitrogen). An NF-κB-Luc2/4-1BB Jurkat cell line was obtained from Promega. Human peripheral blood mononuclear cells (PBMC) from healthy volunteer donors were isolated from buffy coats by centrifugation through a polysucrose density gradient, and T cells were isolated from the resulting PBMCs using a pan T-cell isolation Kit (Miltenyi Biotec) according to the manufacturer's protocols. The 4-1BB agonist antibody 20H4.9 (corresponding to urelumab) was obtained by recombinant expression in CHO cells using standard techniques.
Binding of PRS-343 to 4-1BB or HER2 proteins (human or cynomolgus monkey) was analyzed by ELISA or surface plasmon resonance (SPR).
In the target-binding ELISA, the extracellular domain of the target proteins, 4-1BB (human or cynomolgus), HER2 (human) or control protein, was coated onto microtiter plates. Either PRS-343, the 4-1BB–specific reference Anticalin protein, or the reference anti-HER2 antibody was added and binding was detected with peroxidase-conjugated anti-human IgG or anti-NGAL antibody.
For the dual-binding ELISA, human HER2 (h HER2) was coated and PRS-343-binding was detected using biotinylated h 4-1BB detected by peroxidase-labeled ExtrAvidin.
In the SPR affinity assay, biotinylated h 4-1BB or h HER2 was captured on a sensor chip CAP using the Biotin CAPture Kit (GE). Determination of PRS-343–binding kinetics and affinity was performed by applying four dilutions of PRS-343 to the chip surface with a flow rate of 30 μL/minute; the sample contact time was 180 seconds and dissociation time was 1,800 seconds. Data were fit with a 1:1 binding model.
Binding capability to primary cells, tumor cells, and target-positive engineered cell lines was assessed by flow cytometry. Cells were incubated with PRS-343, reference antibody, or h IgG4 as negative control, and binding of the test molecules was detected using Alexa Fluor 488–labeled goat anti-human IgG antibody.
An IHC assay was performed to assess cross-reactivity of PRS-343 on human tissue. Briefly, cryo-sections histologically prepared at a nominal 5 μm from a panel of 40 frozen human tissues together with positive and negative control material were stained with biotinylated PRS-343 or control. A detailed microscopic examination was performed to assess each tissue for any sign of positive staining.
The impact of HER2 receptor density on PRS-343–mediated T-cell activation was assessed in coculture experiments using a panel of cell lines expressing different levels of HER2. Cancer cell lines representing a range of clinically relevant levels of HER2 receptor were tested for their ability to mediate clustering of PRS-343 and subsequent activation of T cells. To evaluate a potential therapeutic window, cell lines derived from healthy tissues known to express background levels of HER2 were also included. The level of T-cell activation was measured by quantification of human IL2, using an electrochemiluminescence (ECL) immunoassay (Mesoscale Discovery). Briefly, cancer cells or cells derived from healthy tissue pretreated with 10 μg/mL of mitomycin C (Sigma Aldrich) were seeded in culture plates precoated with anti-CD3 and incubated overnight at 37°C in a humidified 5% CO2 atmosphere. T-cell suspension (5 × 104 cells) together with test article was added. The IL2 concentration in the supernatant was assessed by an ECL assay (using IL2 DuoSet kit; R&D Systems) following 3-day incubation.
Specific activation of the 4-1BB pathway by PRS-343 was assessed using a luciferase reporter cell assay (Promega), where a 4-1BB–overexpressing reporter cell line (NF-κB-Luc2/4-1BB Jurkat cells) was cocultured with HER2-positive tumor cell lines and where 4-1BB pathway activation was measured by luminescence.
Cytokine release assay
The potential of PRS-343 to induce cytokine release syndrome was evaluated in vitro using a cytokine release assay (31), where soluble or coated (wet or dry-coated) PRS-343 was incubated with human PBMC for 72 hours followed by the quantification of a panel of proinflammatory cytokines (MSD). The anti-CD3 mAb OKT3 (Muromonab-CD3) was used as positive control and a human monoclonal IgG4 isotype antibody was used as negative control.
Cynomolgus toxicity study
A 4-week–repeated dose toxicity study (vehicle, 10 mg/kg or 120 mg/kg PRS-343 on days 1, 8, 15, 22, and 29) followed by a 4-week recovery period was performed in cynomolgus monkeys. Serum levels of PRS-343 were assessed by an ECL assay. Assessment of safety and toxicity was based on standard parameters.
Mouse and cynomolgus monkey pharmacokinetics
Single-dose pharmacokinetics studies were performed in mice and cynomolgus monkeys. Male CD-1 mice were administered with an intravenous injection of PRS-343 (10 mg/kg). Cynomolgus monkeys (Macaca fascicularis) received a 60-minute infusion of PRS-343 (3 mg/kg) For each study, plasma was collected predose and at multiple time points post administration.
PRS-343 plasma concentrations were determined via an ECL assay (MSD) using a dual-binding ELISA. The pharmacokinetic parameters were derived by noncompartmental analysis using Phoenix WinNonlin.
In vivo tumor efficacy model
PRS-343 in vivo activity was evaluated in a SK-OV-3 ovarian cancer model in human PBMC–reconstituted NOG female mice (NOD.cdPrkdcscidIL2rgtm1.Sug/JicTac, supplied by Taconic), ages 5–7 weeks. Tumor growth was monitored twice weekly and tumor volume was calculated as follows: (length × width2)/2. At study end, plasma was collected for all animals and lymphocyte phenotyping was carried out by flow cytometry assessing human cell markers such as CD45 (Invitrogen), CD4, and CD8 (BD Biosciences). To assess TILs, tumors were also collected, fixed, and paraffin-embedded followed by human CD45, CD3, CD4, and CD8 staining. Stained tumor sections were digitally scanned, and the resulting digitalized data were evaluated to determine the percentage of target-positive (tumor-infiltrating) cells.
Graft versus host disease model
The potential for PRS-343 to induce systemic graft versus host disease (GVHD) was assessed. Five- to 7-week-old female NOG mice (NOD.cdPrkdcscidIL2rgtm1.Sug/JicTac, supplied by Taconic) were injected with fresh human PBMC and treated with PRS-343 or controls. As a read-out for signs of GVHD, the animals were routinely checked for changes in body weight or general health. Mice who were euthanized (based on predefined criteria) or spontaneously expired mice were recorded for survival analysis by each treatment. Data were plotted using a Kaplan–Meier curve.
Flow cytometry analyses were carried out on an iQue Screener (Intellicyt Corporation) equipped with ForeCyt software or with an Attune Focusing Cytometer [blue (488 nm)/violet (405 nm) laser configuration].
All animal experiments and protocols were approved by the institutional animal welfare body and the relevant local authorities and were conducted according to all applicable international, national, and local laws and guidelines. All animal experiments were approved by the regional Ethics Committees (Germany or the United Kingdom).
All statistics were calculated using GraphPad Prism Version 5 for Windows. Statistical significance was determined using one-way or two-way ANOVA to compare differences among multiple groups. P values less than 0.05 were considered statistically significant.
Generation of PRS-343 by protein engineering and design
A 4-1BB–binding Anticalin protein was obtained by phage display selection and optimized by protein engineering as described in Materials and Methods. The 4-1BB–specific Anticalin protein (J10) has an affinity of 2 nmol/L as determined by SPR, binds to human 4-1BB transfected CHO cells, and does not compete with the binding of 4-1BB ligand to its receptor (data not shown). An ELISA screen of related TNFR superfamily proteins (4-1BB Ox40, GITR, TNFRI, TNFRII, CD30, RANK) confirmed J10 Anticalin protein specifically binds to 4-1BB (Supplementary Fig. S1)
As the HER2-binding, tumor-targeting building block of a 4-1BB/HER2 bispecific, we utilized a modified variant of trastuzumab. The formatting flexibility offered by the Anticalin technology facilitated the generation of various Antibody-anticalin fusion formats as depicted in Supplementary Fig. S 1. We hypothesized that the distance between T-cell and tumor cell–binding sites may impact the ability of the 4-1BB/HER2 bispecific to appropriately cluster the 4-1BB receptor and thus activate the pathway. Therefore, the 4-1BB–engaging Anticalin protein was genetically fused either to the N- or C-terminus of the anti-HER2 antibody heavy or light chain resulting in the generation of four different geometries of the fusion protein covering a range of binding distances. The antibody Fc region was engineered (see Materials and Methods for details) to exclude the risk of NK-driven antibody-directed cellular cytotoxicity (ADCC) against 4-1BB–positive cells or undesired, non–tumor-target activation of 4-1BB–positive lymphocytes via FcγR-derived cross-linking. The interaction with the neonatal Fc receptor (FcRn) was retained to support a prolonged terminal plasma half-life.
While all formats behaved similarly from the perspective of biophysical and target-binding properties, we observed a marked difference in their ability to elicit T-cell activation (Supplementary Fig. S2). On the basis of these data, a single format was selected and is described herein. The lead bispecific fusion molecule, PRS-343, was constructed via the genetic fusion of the 4-1BB–specific Anticalin protein to the C-terminus of the heavy chain of the trastuzumab IgG4 variant, connected by a flexible, nonimmunogenic (G4S)3 linker sequence, as depicted in Fig. 1.
Human target binding
The interaction of PRS-343 with its human targets was investigated by ELISA, SPR, and FACS. As assessed by SPR, PRS-343 binds to recombinant human HER2 with high affinity (Kd = 0.3 nmol/L), similar to the parental antibody trastuzumab. The binding affinity of PRS-343 to human monomeric 4-1BB was determined to be 5 nmol/L (Fig. 1).
In a FACS assay, PRS-343 was found to bind HER2-expressing MCF-7 cells with an EC50 of 7.4 nmol/L. Binding to CHO cells transfected with human 4-1BB was determined with an EC50 of 6.2 nmol/L (Fig. 1).
Using a dual-binding ELISA, PRS-343 was found to be capable of binding both targets simultaneously, which is a prerequisite for its envisioned mode of action.
PRS-343 displays reduced cross-reactivity to cyno 4-1BB in comparison with human 4-1BB as shown by ELISA (Fig. 1K). An SPR experiment also demonstrated that binding of PRS-343 to cyno 4-1BB strongly depends on target densities (data not shown).
Further SPR experiments demonstrated that PRS-343 does not bind to Fcγ receptors but binding to the neonatal Fc receptor (FcRn) is retained. In accordance with its lack of interaction with Fcγ receptors, PRS-343 does not elicit any ADCC activity when coincubated with human PBMCs and HER2-positive breast carcinoma cells (data not shown).
A tissue cross-reactivity study using a panel of 40 human tissues (n = 3 donors) each demonstrated that PRS-343 binds as expected for a molecule targeting both HER2 and 4-1BB in a bispecific manner. Epithelial cells of various tissues showed strong positivity for binding PRS-343, which is in accordance with published data on HER2 expression in human tissues (32). The observed lymphocyte staining is in agreement with the expected binding of PRS-343 to 4-1BB on this cell type because 4-1BB has been reported to be (inducibly) expressed on lymphocytes (33).
In vitro activity of PRS-343
The ability of PRS-343 to costimulate T cells in a HER2-dependent manner was investigated in vitro using a reporter assay.
The reporter cell assay was performed to investigate whether PRS-343 is capable of inducing HER2-dependent 4-1BB clustering on T cells. In this assay, NF-κB-Luc2/4-1BB Jurkat cells were cocultured with a HER2-positive tumor cell line. Successful clustering of 4-1BB on the Jurkat cell surface is expected to lead to TRAF-2 (and TRAF-1)-mediated activation of NF-κB (reviewed in ref. 34), which in the presence of a luciferase substrate can be detected by increased luminescence.
Indeed, coincubation of up to 10 nmol/L PRS-343 with the Jurkat reporter cell line and the HER2-positive NCI-N87 cell line led to a 20-fold increase of NF-κB luciferase reporter activity with an EC50 of 50 pmol/L (Fig. 2A). Furthermore, in a wash-out experiment where PRS-343 was added to various cancer cell lines overnight prior to a media exchange, HER2-specific 4-1BB activation was maintained (Supplementary Fig. S4). When using the low HER2-expressing MKN-45 or Hep-G2 cell lines, the NF-B activity was similar to background and control values. A bell-shaped response was observed when concentrations of up to 1 μmol/L PRS-343 were applied in this type of assay (data not shown).
These results demonstrate that PRS-343 facilitates 4-1BB pathway activation only in the presence of HER2-overexpressing cells.
HER2-dependent costimulatory T-cell activation
To investigate whether this mechanism can be applied to effectively costimulate T cells, a coculture assay was developed using human T cells and HER2-expressing cell lines. Here, the T cells receive a primary T-cell receptor stimulus via anti-CD3 antibody activation, and PRS-343 induced costimulation can be quantified by measuring supernatant levels of proinflammatory cytokines. Using a multiplex assay, we first investigated which cytokines were produced by T cells costimulated by PRS-343 in the presence of HER2-expressing cells (Supplementary Fig. S5). The strongest fold increase over background was observed for IL2. Statistically significant increase in GM-CSF, IFNγ, and TNFα was observed. IL4 and IL6 supernatant levels were increased albeit absolute cytokine levels were low. There was no PRS-343–dependent response observed for IL1β, IL8, IL10, and IL12. Given the robust nature of the IL2 response, it was selected as a marker of activation for future assays. In the presence of HER2-positive cell lines, a dose-dependent induction of IL2 was observed with PRS-343 (Fig. 2B and C). When the experiment was performed with cell lines expressing basal levels of HER2, no PRS-343–dependent IL2 induction was observed. To further confirm that PRS-343 activity is dependent on HER2 binding, additional assays were performed in the presence of an excess of trastuzumab (200 nmol/L). Under these conditions, the excess trastuzumab competed for binding to HER2 and was found to abrogate PRS-343–induced T-cell costimulation, further supporting its envisaged mode of action (Fig. 2C), whereas the control 4-1BB agonistic antibody 20H4.9 consistently induced IL2 secretion irrespective of the coincubated cell lines (Fig. 2C).
Impact of HER2 expression level
To examine the HER2 expression level threshold required for successful PRS-343–induced T-cell costimulation, the assay was performed with a panel of cell lines of variable HER2-positivity (Table S1). The HER2 cell surface expression level was determined by quantitative FACS and is reported as the specific anti–HER2-antibody–binding capacity. For the purpose of comparison, HER2 expression was normalized relative to a reference cell line, SK-BR-3. The cancer cell lines selected represent a wide range of HER2 expression levels ranging from approximately 1% to 200% of the reference value. In contrast, cell lines from healthy tissues express HER2 at a much lower and more confined level, with relative HER2 expression around 1 % of HER2 expression on SK-BR-3 (Table S1). These cancer- and healthy tissue–derived cell lines were then applied in the previously described costimulatory T-cell activation assay to identify a potential HER2 threshold required for T-cell costimulation.
A selection of experimental results based on IL2 as the readout and utilizing tumor cell lines and primary cell types is shown in Fig. 3. These data confirm that IL2 secretion strongly correlates with HER2 expression. Assessing the maximum level of IL2 secretion compared with control, we show that cancer cells with a high level of HER2 induced significantly increased IL2 secretion. For cell lines described as having a more intermediate level of HER2 (e.g., MCF-7 or MKN-45), we could also observe significantly increased IL2 secretion in a proportion of the donors tested. Both the cancer cell lines expressing basal levels of HER2 (such as Bx-PC-3 or MDA-MB-231) and the tissue-derived primary cells did not induce a significant amount of IL2 secretion, supporting the hypothesis that PRS-343 requires HER2 expression at supraphysiologic levels on target cells to costimulate T cells.
Cytokine release assay
To investigate the risk of systemic cytokine release in patients induced by PRS-343, an in vitro cytokine release assay (31) was carried out with PBMCs from 12 healthy control donors. In contrast to the anti-CD3–positive control antibody OKT3, PRS-343 did not induce a significant increase of cytokines over background when wet coated onto plates or when incubated with PBMCs in solution. When PRS-343 was air-dried, a modest but statistically significant increase of the cytokines IFNγ, GM-CSF, IL4, and IL8 was observed only at the highest concentration (i.e.,100 μg/mL). Relative to the maximum increase induced by OKT-3 (10 μg/mL, air dried), the levels induced by PRS-343 were 0.2% for GM-CSF, 0.1% for IFNγ, 6% for IL4, and 13% for IL8 (Supplementary Fig. S3).
Pharmacokinetics in mice and cynomolgus monkeys
Pharmacokinetics was evaluated following a single intravenous administration of PRS-343 by bolus injection of male CD-1 mice at 10 mg/kg, or upon a 60-minute intravenous infusion to male cynomolgus monkeys at 3 mg/kg. The plasma concentrations in both species were found to decline in a biphasic manner (Fig. 4). The terminal elimination half-life was >14 days in mouse. In cynomolgus monkeys, a terminal elimination half-life of approximately 4 days was observed. In conclusion, the determined PK parameters for PRS-343 demonstrate that it behaves similar to a conventional mAb in preclinical models.
Humanized mouse tumor model
The in vivo activity of PRS-343 was evaluated in a humanized mouse model using immunocompromised mice subcutaneously engrafted with the SK-OV-3 cell line as a HER2-positive tumor xenograft. This cell line was chosen based on HER2 positivity, trastuzumab sensitivity, and the ability to grow homogenously as a subcutaneous xenograft in the presence of human PBMC without being directly eliminated. Mice with established tumors (>100 mm3) were grouped and treated with test articles on a weekly dosing schedule. Median tumor volumes over time for each of the treatment groups are shown in Fig. 5A. PRS-343 displayed dose-dependent antitumor efficacy with doses ranging from 4 μg to 100 μg (approximately 0.2 mg/kg to 5 mg/kg), while the 200-μg dose (approximately 10 mg/kg) did not further enhance tumor regression. Equimolar dosing of the trastuzumab-IgG4 control antibody displayed comparable antitumor efficacy to that of PRS-343. The model has been previously described as being responsive to anti-HER2 therapy and helps confirm that HER2-mediated antitumor activity is preserved in PRS-343. Interestingly, the 4-1BB agonist (20H4.9) antibody was unable to block tumor growth in this model.
Flow cytometry analysis of the human immune cell population in the peripheral blood of the mice revealed very little differences between the PRS-343–treated groups and control groups. However, in mice treated with the 4-1BB agonist antibody (20H4.9), a peripheral expansion of human CD8+ cells was observed (Fig. 5B), demonstrating a systemic impact of the agent.
To assess the 4-1BB–mediated effect of PRS-343, excised tumors from treated mice were analyzed for immune infiltration. PRS-343 administration led to a significant increase in human CD45+ lymphocytes in tumor tissue (Fig. 5C) compared with the control groups. While assessing the immune cell subtypes infiltrated within the tumors, we could observe that PRS-343–induced infiltrate consisted predominantly of CD8+ T cells (Fig. 5C). Importantly, there was no significant increase in immune cell infiltrates in the excised tumors from mice treated with the trastuzumab-IgG4 when compared with controls.
In immune-compromised mice engrafted with human PBMCs, GVHD is expected to occur due to the human PBMCs destroying mouse tissue and organs resulting in progressive weight loss and mortality (35). In this humanized xenograft study, we observed that the 4-1BB agonist antibody accelerated GVHD-induced mortality, while the administration of PRS-343 or the isotype control antibody did not influence weight loss or the median survival of the mice. To confirm this observation, we carried out a dedicated experiment using larger group sizes (n = 15) of non–tumor-bearing mice. This study confirmed that weight loss and survival following PRS-343 treatment were comparable with that observed in the IgG4 isotype control–treated animals, while treatment with the systemically active 4-1BB agonist led to both significantly greater weight loss and shorter survival compared with control (Fig. 5D).
Taken together, these data show that PRS-343 provides dual activity by increasing the number of TILs coupled with direct tumor growth inhibition by bispecific targeting of 4-1BB and HER2. Notably, the tumor growth inhibition provided by targeting HER2 did not require any ADCC, as both PRS-343 and the trastuzumab-IgG4 control lack the ability to interact with Fc-gamma receptors on NK cells that ADCC would require.
Safety of PRS-343 in cynomolgus monkeys
The safety of PRS-343 was investigated in a 4-week GLP-compliant toxicity study in cynomolgus monkeys. It is important to take into account the reduced cross-reactivity of PRS-343 to cynomolgus 4-1BB, which impacts the ability of this study to predict 4-1BB–related toxicity. PRS-343 was administered weekly as an intravenous infusion of 120-minute duration at doses of 10 and 120 mg/kg over 4 weeks, followed by a recovery phase in the control and high-dose group to evaluate any potential delayed onset or reversibility of toxicity.
Overall, PRS-343 was well tolerated at both doses tested, with no significant findings. No unscheduled deaths occurred. There were no changes in standard parameters such as clinical observations, body weight, ophthalmology, rectal temperatures, clinical chemistry, hematology, coagulation tests, urinalysis parameters, or serum cytokines, as well as ECG. Furthermore, no toxicologically relevant organ weight or organ weight ratio changes, and no macroscopic or microscopic findings were observed, indicating that treatment with PRS-343 over four weeks did not lead to any systemic toxicity. In addition, no evidence of delayed onset toxicity was noted at the end of the 4-week recovery phase. Toxicokinetic analysis demonstrated dose-proportional systemic exposure at both dose levels upon first and last dose. Two of sixteen animals developed anti-drug antibodies (ADAs) that persisted throughout the study while three others showed a transient ADA response. No gender-related toxicity differences were noted in the study.
4-1BB (CD137) is a key costimulatory immunoreceptor and a member of the TNFR superfamily. The preclinical and clinical demonstration of the potential therapeutic benefit of 4-1BB costimulation has spurred the development of therapeutic antibodies targeting 4-1BB, utomilumab (24, 25), and urelumab (23). Utomilumab is a fully humanized IgG2 mAb that binds 4-1BB in a manner that blocks the binding of endogenous 4-1BBL to 4-1BB and that appears well tolerated as a monotherapy (36) and in combination with rituximab; however, modest activity has been observed (37). Urelumab is an IgG4 mAb that, in contrast to utomilumab, binds 4-1BB in a manner that does not interfere with the 4-1BB/4-1BBL interaction. While an initial trial reported modest clinical activity (38), a follow-up study was stopped because of hepatotoxicity (1). These data indicate that systemic activation of 4-1BB with a potent agonist may lead to prohibitive toxicity, supporting the rationale for tumor-localized targeting of the pathway.
Although multiple lines of evidence suggest that 4-1BB is a highly promising therapeutic target in cancer, current systemic antibody-based approaches are not designed to achieve a tumor-target–driven activation and are likely to display toxicity due to peripheral T-cell and NK-cell activation. We hypothesized that biotherapeutics addressing this pathway should efficiently activate the immune costimulatory target, but its activation should be restricted to the tumor microenvironment (TME) to avoid systemic effects and unwanted toxicity.
PRS-343 is an Anticalin-antibody fusion protein with dual specificity for both 4-1BB and the tumor antigen HER2. PRS-343 was designed to promote 4-1BB clustering by bridging 4-1BB–positive T cells with HER2-positive tumor cells, thereby providing a potent costimulatory signal to tumor antigen–specific T cells. On the basis of its differentiated mechanism of action, PRS-343 has the potential to expand therapeutic options to HER2-positive tumors including those who may not be responsive to conventional antibody or small-molecule–based HER2 inhibitors.
Anticalin proteins are 18 kDa protein therapeutics derived from human lipocalins. We utilized phage display to generate an Anticalin protein binding to 4-1BB with high affinity and specificity. PRS-343 was generated by genetic fusion of the 4-1BB–specific Anticalin protein to a variant of the HER2-targeting mAb trastuzumab with an engineered IgG4 backbone. We demonstrated the benefits of our bispecific platform's flexible formatting, allowing for functional testing of multiple bispecific geometries and the demonstration that the 4-1BB Anticalin protein fused to the heavy-chain C-terminus of the antibody was functionally the most active in cell-based assays.
Binding studies using SPR, ELISA, and FACS showed that PRS-343 displays similar potency against each target compared with parental building blocks. Simultaneous binding of both targets was also demonstrated. PRS-343 displays cross-reactivity to cynomolgus HER2 similar to trastuzumab but with reduced cross-reactivity to cynomolgus 4-1BB. PRS-343 induces 4-1BB clustering and downstream signaling in a Jurkat NF-κB reporter cell line in the presence of HER2-positive cells with a potency of approximately 50 pmol/L (EC50) as well as IL2 production in a costimulatory T-cell activation assay in the presence of HER2-positive NCI-N87 cells, with a potency of 35 pmol/L. We observed a bell-shaped response both in the Jurkat NF-κB reporter assay and the primary T-cell activation assay, which is in accord with expectations as a response requires the formation of a ternary complex of the tumor cell target HER2, the drug PRS-343 and the T-cell receptor 4-1BB, that can be disrupted when components are individually saturated with drug. The effect can be rationalized by a mathematical model recently described for ternary complex formation in biological systems and depends on the concentrations of all binding partners and their affinities to each other (39).
The pharmacology of PRS-343 was investigated by further ex vivo T-cell costimulation assays based on mixed culture of human T cells and cell lines. The cell panel was selected to cover a broad range of HER2-expressing cells derived from cancer tissue as well as from different healthy tissue origins. Of note, three of the cell lines capable of costimulating T cells in a HER2-dependent manner have been described as being resistant [(40–42)] in preclinical models to trastuzumab (SUM-225) or even both trastuzumab and lapatinib (JIMT-1, MDA-MB-453), demonstrating the potential of PRS-343 to provide a therapeutic option for patients with innate or acquired resistance to HER2-targeted therapy. The risk of PRS-343–mediated systemic 4-1BB activation and concomitant toxicity was investigated in a cytokine release assay, indicating that 4-1BB clustering leads to negligible cytokine release by T cells in the absence of a primary T-cell receptor stimulus. Together with the costimulation experiments, these results indicate that PRS-343 is able to activate T cells only when these are engaging a target cell that expresses HER2 at a level that is usually only encountered in malignant, tumorous tissue, and when the T cells are activated at the same time via the T-cell receptor, for example, by recognizing a tumor antigen.
In vivo proof-of-concept data utilizing a humanized SK-OV-3 mouse model support the desired mode of action of PRS-343 in vivo. PRS-343 exhibits both direct cytotoxicity via monospecific targeting of tumor-expressed HER2 and activates 4-1BB via bispecific targeting of tumoral HER2 and 4-1BB on human lymphocytes. These data also highlight the differentiating features over a monospecific 4-1BB–targeting constitutive agonist. PRS-343 leads to an increase in the frequency of human lymphocytes in the tumor, but does not affect the human lymphocyte frequency in the peripheral blood, which correlates with an unchanged time course of GVHD-induced morbidity and mortality compared with controls. In contrast, the monospecific 4-1BB–targeting agonist antibody 20H4.9 leads to an expansion of human lymphocytes in the peripheral blood and a concomitant acceleration of GVHD-induced morbidity and mortality, despite lacking any activity in the TME, as evidenced by no significant increase in T-cell infiltration in the TME when compared with control. The data on PRS-343 reported here support its further evaluation either as a single-agent or combination therapy. Indeed, a preclinical rationale for combining 4-1BB agonism with checkpoint blockade has been demonstrated (43, 44). PRS-343 is the first bispecific 4-1BB agonist to enter the clinic, and a phase I dose escalation study in patients with HER2-positive advanced or metastatic solid tumors is ongoing (NCT03330561). A clinical trial evaluating PRS-343 in combination with atezolizumab (anti–PD-L1) has also commenced (NCT03650348).
Disclosure of Potential Conflicts of Interest
M. Hinner, R.S. Bel Aiba, T.J. Jaquin, S. Berger, M.C. Dürr, C. Schlosser, A. Allersdorfer, G. Matschiner, U. Moebius, C. Rothe, L. Matis, and S.A. Olwill hold ownership interest (including patents) in Pieris Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.
Conception and design: M.J. Hinner, R.S. Bel Aiba, S. Berger, U. Moebius, C. Rothe, L. Matis, S.A. Olwill
Development of methodology: M.J. Hinner, R.S. Bel Aiba, T.J. Jaquin, C. Schlosser, A. Allersdorfer, A. Wiedenmann, J. Schüler, U. Moebius, C. Rothe, S.A. Olwill
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.J. Hinner, R.S. Bel Aiba, T.J. Jaquin, S. Berger, C. Schlosser, A. Allersdorfer, A. Wiedenmann, J. Schüler
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.J. Hinner, R.S. Bel Aiba, T.J. Jaquin, S. Berger, M.C. Dürr, C. Schlosser, A. Allersdorfer, A. Wiedenmann, J. Schüler, C. Rothe, S.A. Olwill
Writing, review, and/or revision of the manuscript: M.J. Hinner, R.S. Bel Aiba, T.J. Jaquin, S. Berger, M.C. Dürr, C. Schlosser, A. Allersdorfer, G. Matschiner, J. Schüler, L. Matis, S.A. Olwill
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Schlosser, A. Allersdorfer, A. Wiedenmann, J. Schüler, L. Matis
Study supervision: M.J. Hinner, U. Moebius, C. Rothe, L. Matis, S.A. Olwill
We would like to thank the extended Pieris team for their support in generating and reviewing data related to the PRS-343 project. We would like to specifically thank Tanya Aneichyk for her assistance in data interpretation and presentation. The research funding was provided by Pieris Pharmaceuticals.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.