**Preclinical Evaluation of Recombinant Human IL15 Protein Fused with Albumin Binding Domain on Anti-PD-L1 Immunotherapy Efficiency and Anti-Tumor Immunity in Colon Cancer and Melanoma**

**Fei-Ting Hsu 1,†, Yu-Chang Liu 2,3,4,† , Chang-Liang Tsai 5,†, Po-Fu Yueh 1,6, Chih-Hsien Chang 5,7 and Keng-Li Lan 6,8,\***

	- <sup>3</sup> Department of Radiation Oncology, Show Chwan Memorial Hospital, Changhua 500, Taiwan

**Simple Summary:** In this manuscript, we reported that a newly developed recombinant human IL15 fused with albumin binding domain (hIL15-ABD) showed superior biological half-life, pharmacokinetic and anti-tumor immunity than wild-type (WT) hIL15. Our hIL-15-ABD can effectively enhance anti-tumor efficacy of anti-PD-L1 on colon cancer and melanoma animal models. The anti-tumor potential of hIL-15-ABD was associated with tumor microenvironment (TME) regulation, including the activation of NK cells and CD8<sup>+</sup> T cells, the reduction of immunosuppressive cells (MDSCs and Tregs) and the suppression of immunosuppressive factors (IDO, FOXP3 and VEGF). In conclusion, our new hIL15-ABD combined with anti-PD-L1 antibody increased the activity of anti-tumor effector cells involved in both innate and adaptive immunities, decreased the TME's immunosuppressive cells, and showed greater anti-tumor effect than that of either monotherapy. We suggested hIL15-ABD as the potential complementary agent may effectively augment the therapeutic efficacy of anti-PD-L1 antibody in colon cancer and melanoma model.

**Abstract:** Anti-PD-L1 antibody monotherapy shows limited efficacy in a significant proportion of the patients. A common explanation for the inefficacy is a lack of anti-tumor effector cells in the tumor microenvironment (TME). Recombinant human interleukin-15 (hIL15), a potent immune stimulant, has been investigated in clinical trial with encouraging results. However, hIL15 is constrained by the short half-life of hIL15 and a relatively unfavorable pharmacokinetics profile. We developed a recombinant fusion IL15 protein composed of human IL15 (hIL15) and albumin binding domain (hIL15-ABD) and explored the therapeutic efficacy and immune regulation of hIL-15, hIL15-ABD and/or combination with anti-PD-L1 on CT26 murine colon cancer (CC) and B16-F10 murine melanoma models. We demonstrated that hIL15-ABD has significant inhibitory effect on the CT26 and B16-F10 tumor growths as compared to hIL-15. hIL-15-ABD not only showed superior half-life and pharmacokinetics data than hIL-15, but also enhance anti-tumor efficacy of antibody against PD-L1 via suppressive effect on accumulation of Tregs and MDSCs and activation of NK and CD8+T cells. Immune suppressive factors including VEGF and IDO were also decreased by combination treatment. hIL15-ABD combined with anti-PD-L1 antibody increased the activity of

**Citation:** Hsu, F.-T.; Liu, Y.-C.; Tsai, C.-L.; Yueh, P.-F.; Chang, C.-H.; Lan, K.-L. Preclinical Evaluation of Recombinant Human IL15 Protein Fused with Albumin Binding Domain on Anti-PD-L1 Immunotherapy Efficiency and Anti-Tumor Immunity in Colon Cancer and Melanoma. *Cancers* **2021**, *13*, 1789. https:// doi.org/10.3390/cancers13081789

Academic Editors: Subree Subramanian and Xianda Zhao

Received: 30 January 2021 Accepted: 7 April 2021 Published: 9 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

anti-tumor effector cells involved in both innate and adaptive immunities, decreased the TME's immunosuppressive cells, and showed greater anti-tumor effect than that of either monotherapy.

**Keywords:** PD-L1; IL15; colon cancer; melanoma; tumor microenvironment

#### **1. Introduction**

Active immune system possesses fighting ability against tumor development and progression. An example of this is cytotoxic T lymphocytes (CD8<sup>+</sup> T cells) and natural killer (NK) cells attacking tumor cells that can be elicited by antitumor immune signaling, resulting in tumor destruction [1,2]. However, tumors can escape immune surveillance through immunosuppressive tumor microenvironment (TME), restricting antitumor immunity. Immune checkpoints, which are composed of immunosuppressive molecule receptors or their ligands such as programed death receptor 1 (PD-1)/programed death ligand 1 (PD-L1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) modulate inactivation of CD8<sup>+</sup> T and nature killer (NK) cells, resulting in tumor immune evasion in TME. Evasion of immune surveillance is conductive to tumor survival and progression [3–5].

Immunotherapy, an innovative therapeutic method that treats cancer by evoking antitumor immunity, is a promising strategy for treatment of solid tumors and hematologic malignancies [6–8]. Increased expression of PD-1/PD-L1 pathway is linked to T cell exhaustion and poor survival in multiple types of cancers. Blockade of PD-1/PD-L1 interaction with monoclonal antibodies reverses T cell exhaustion and prolongs survival benefit in patients with cancers such as melanoma, hepatocellular carcinoma (HCC), nonsmall-cell lung cancer (NSCLC), gastric, and urothelial cancers [7,9–11]. Furthermore, many preclinical and clinical studies have demonstrated that the therapeutic efficacy of PD-1/PD-L1 blocking antibodies can be enhanced with immunologic or non-immunologic agents [4].

Interleukin-15 (IL15), the immuno-oncology agent, potentiates antitumor immune via enhancement of CD8<sup>+</sup> T and NK cells proliferation and cytotoxic activity. IL15 therapy has been demonstrated to attenuate tumor growth and improve survival rates in murine tumor models. IL15 is also recognized as a potential complementary agent to immunotherapy, effectively increasing anticancer immune response. The first clinical trial of hIL15 was conducted in patients with metastatic renal cell carcinoma and melanoma by daily intravenous administration for 12 consecutive days of recombinant hIL15 expressed by *Escherichia coli* [12–14]. Although some encouraging clinical results were observed, the bioactivity of IL15 is limited due to short in vivo half-life. N-803, formerly ALT-803, composed of N72D IL15 mutant, sushi domain of IL15Rα, and Fc domain of human IgG1, has been demonstrated to have longer serum half-life and more potent stimulatory effect on NK cells and T-lymphocytes than that of WT hIL15 [15,16]. N-803 has been shown to boost antitumor response of anti-PD-L1 antibody in triple negative breast and colon cancers in vivo and its combination with anti-PD-1 monoclonal antibody, Nivolumab, has been verified in safety to treat refractory metastatic non-small cell lung cancer patients with observed tumor responses [17].

We have generated a recombinant fusion protein (hIL15-ABD), which is composed of human IL15, albumin binding domain (ABD), and hexahistidine tag (his6). hIL15-ABD could be expressed by *E. coli* and refolded into active fusion protein, which simultaneously binds to albumin and stimulate CTLL-2 proliferation as well as downstream signaling pathway evidenced by enhanced STAT5 phosphorylation. Fusion of hIL15 with ABD greatly enhanced pharmacokinetic parameters, including half-life, Cmax and area under curve (AUC) as compared with those of hIL15 in experimental mice. hIL15-ABD also displayed significant inhibitory effect on the tumor growths of CT26 murine colon cancer (CC) and B16-F10 murine melanoma models. Moreover, combination of hIL15-ABD with a rat antibody against murine PD-L1 antibody, 10F.9G2, demonstrated greater anti-tumor effect than that of either monotherapy, by enhancing the activity of anti-tumor effector cells associated with both innate and adaptive immunities as well as decreasing the TME's immunosuppressive cells.

#### **2. Materials and Methods**

#### *2.1. Reagents and Antibodies*

FITC Rat Anti-Mouse CD3 (#561798), PerCP-Cy™5.5 Rat Anti-Mouse CD4 (#561115), FITC Rat Anti-Mouse CD8a (#561966), PE Rat Anti-Mouse CD25 (#561065), FITC Rat Anti-CD11b (#561688), PE Rat Anti-Mouse CD49b (DX5, #561066), PerCP-Cy™5.5 Rat Anti-Mouse CD335 (#560800), Alexa Fluor® 488 Rat anti-Mouse Foxp3 (#560407), PE Rat Anti-Mouse Ly-6G and Ly-6C (Gr-1, #561084), PerCP-Cy™5.5 Mouse Anti-Mouse NK-1.1 (#561111) and Foxp3 Fixation/Permeabilization Buffer Set (#560409) were all purchased from BD Pharmingen™ (BD Biosciences, San Diego, CA, USA). Cleaved-caspase-3 (E-AB-30004, Elabscience Biotechnology Inc, Houston, TX, USA), BAX (#50599-2-lg, Proteintech Inc., Rosemont, IL, USA), Ki-67 (#E-AB-2202, Elabscience Biotechnology Inc.), granzyme B, Indoleamine 2,3-dioxygenase (IDO), Forkhead box protein P3 (FOXP3), CD49b and Interferon-gama (IFN-γ Rat Anti-Mouse PD-L1 (10F.9G2, Bioxcell, Lebann, NH, USA) antibodies were all purchased from different companies as listed.

#### *2.2. In Vitro Characterization for CTLL2 Stimulation and Albumin Binding of hIL15-ABD*

CTLL-2 cells in logarithmic phase were harvested, washed and resuspended in RPMI-1640 medium supplemented with 10% FBS at a concentration of 8 × 10<sup>3</sup> cells per well of 96-well plate. hIL15-ABD in 1 µL of various refolding buffers was incubated with 100 µL of CTLL-2 cell culture of 96-well at 37 ◦C for 2 days followed by addition of 20 µL MTS reagent (CellTiter 96® AQueous Non-Radioactive Cell Proliferation Assay, Promega, Madison, WI, USA) according to manufacturer's instructions. The viable CTLL-2 cells were measured at 490 nm on a TECAN Sunrise™ multichannel microtiter plate reader. For STAT5 phosphorylation assay, CTLL2 cells were treated with increasing concentrations of either hIL15 or hIL15-ABD for 20 min followed by fixation with formaldehyde (2% *v*/*v* final concentration) for 15 min at room temperature. Fixed cells were then spun down (×500 g), and cell pellets were permeabilized with ice cold 100% methanol and incubated on ice for 20 min. Cells were rehydrated by washing twice with 250 µL of PBS in the presence of 0.8% BSA. Cells were incubated with antibody against phosphorylated STAT5 for 16 h at 4◦ , washed twice with PBS with 0.5% BSA, and lastly treated with FITC-conjugated anti-rabbit IgG for 60 min at room temperature in the dark. The positive events were detected with a BD FACSCalibur flow cytometer and analyzed with CellQuest Pro and Cytexpert software (Beckman Coulter). To examine the ability of hIL15-ABD for binding to albumin, human or murine albumin was diluted in PBS and immobilized on an ELISA plate by incubation at 4 ◦C overnight. Albumin coated well were blocked with 300 µL 3% milk for 2 h at room temperature, followed by washing the plate three times with wash buffer (0.05% Tween-20 in PBS). Various concentrations of hIL15 or hIL15-ABD were incubated with immobilized human or murine albumin at room temperature for one hour, followed by washing the plate three times with wash buffer. The *in vitro* binding of his6 tagged hIL15 and hIL15-ABD with human albumin was detected using an HRP-tagged, anti-his6 antibody and developed by the addition of the HRP substrate (100 µL/well), 3,3′ ,5,5′ -tetramethylbenzidine (TMB). The peroxidase reaction was stopped 20 min after the addition of 0.5 M H2SO<sup>4</sup> (50 µL/well), and the absorbance was measured at 450 nm with a multichannel microtiter plate reader.

#### *2.3. Expression, Refolding and Purification of hIL15 and hIL15-ABD*

BL21 (DE3) *E. coli* strain transformed with plasmids encoding hIL15 or hIL15-ABD was cultured with terrific broth (TB)/ampicillin (100 µg/mL) and grew at 37 ◦C in a shaker at 250 rpm. When reaching logarithmic phase with OD<sup>600</sup> at 0.7, the *E. coli* culture was treated with increasing concentration of IPTG ranging from 0, 0.01, 0.05, 0.1, 0.5, to 1 mM. Additionally, the cultures were further incubated at either 30 or 37 ◦C overnight. The induced cultures were centrifuged at 6000 rpm for 15 min, and supernatant was removed followed by re-suspending the pellet in PBS. The pellet was subjected to continuous highpressure cell disrupter twice at pressure of 28 kpsi followed by centrifugation at 4500 rpm for 15 min for removal of supernatant. The pellets were washed six-time with either 200 mL of H2O or Tris-HCl buffer in the presence or absence of 1%SDS, and 0.5 M NaCl. Each washing steps were followed by centrifugation at 6000 rpm for 15 min. The inclusion body pellets were solubilized with by 8 M urea buffer containing 25 mM imidazole and loaded onto Ni-NTA resin column. Elution of immobilized protein was conducted by stair-wise increase of imidazole from 75, 300, to 500 mM of imidazole in PBS buffer (pH 7.5). Fractions containing hIL15 or hIL15-ABD were pooled and dilution refolded to a concentration of 0.1 mg/mL by slow dripping into refolding buffer matrix with various combinations of Triton (0.05%), EDTA (2 mM), NaCl (250 mM), GSH (1 mM)/GSSG (0.1 mM), and L-Arginine (0.4 M). The refolding process lasted for 24 h at 4◦ . After examining activities of hIL15ABD and hIL15 for albumin binding and STAT5 phosphorylation, buffer 26 containing NaCl (250 mM), GSH (1 mM) and GSSG (0.1 mM) in Tris-HCl buffer (50 mM, pH 8.5) was selected for later large-scale protein refolding.

#### *2.4. In Vitro Characterization for CTLL2 Stimulation and Albumin Binding of hIL15-ABD*

The ability of hIL15 and hIL15-ABD for T-cell activation was examined for either proliferation of or phosphorylation of STAT5 in CTLL2 cells using MTS assay and flow cytometry, respectively. The detail procedure was described in material and methods section.

#### *2.5. Pharmacokinetics*

Female Balb/c mice (*n* = 3, 6 weeks, 20 g) received an intraperitoneal injection of either hIL15 (1 µg) or equimolar hIL15-ABD (1.5 µg) in 300 µL of PBS. For mice injected with hIL15, in time intervals of 15 min, 30 min, 45 min, 1 h, 2 h, 4 h, 6 h, whereas for mice treated with hIL15-ABD, in time intervals of 15 min, 30 min, 45 min, 1 h, 2 h, 4 h, 8 h, 24 h, 36 h, and 52 h, blood samples were withdrawn from the tail and placed on ice. Serum samples were obtained by centrifuging clotted blood at 800 g for 10 min at 4 ◦C. Serum concentrations of hIL15 and hIL15-ABD were determined by ELISA specific for human hIL15 (DY247- 05, R&DSystem, Minneapolis, MN, USA). Pharmacokinetic parameters were determined using the Phoenix® WinNonlin software version 7.0 (Certara USA Inc., Princeton, NJ, USA). Noncompartmental analysis (extravascular input) was used with the log/linear trapezoidal rule. Parameters, including terminal half-life (T1/2λ<sup>z</sup> ), Tmax, Cmax and area under the curve (AUC) were determined. Pharmacokinetic parameters associated with the terminal phase were calculated using the last four measured time points to estimate the terminal half-life.

#### *2.6. Cell Culture*

CT26 mouse colon cancer (BCRC #60447) cell line and B16-F10 (BCRC #60031) mouse melanoma were purchased from the Bioresource Collection and Research Center (Hsinchu, Taiwan). Cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium and Dulbecco's Modified Eagle medium (DMEM) supplemented with 10% heat-inactivated FBS, 2mMl-glutamine, 100units/mL penicillin and 100µg/mL streptomycin in a humidity atmosphere containing 5% CO<sup>2</sup> and at 37 ◦C. Cells culture related reagents and medium were all purchased from Gibco BRL, Grand Island, NY, USA.

#### *2.7. Transfection and Stable Clone Selection*

The vector containing CMV-luciferase2 vector (pGL4.50[luc2/CMV]) (Promega, Madison, WI, USA) and transfection reagent (Polyplus transfection, France) were prepared in advance. B16-F10 cells were seeded in 6 cm plate one day before transfection. Cells density is around 70% during transfection procedure. The jetPEITM reagent (10 µL) dissolved in

250 µL of NaCl buffer was then added into DNA buffer (5 µg plasmids with 250 µL of NaCl buffer), the mixture was then incubated at 25 ◦C for 25 min. The mixture was finally added to the B16-F10 cells for an incubation period of 1 day. Luc2 expression cells were selected by hygromycin B 200 µg/mL for another two weeks and named as B16-F10/*luc2* cells [18].

#### *2.8. Immune Cells (CD8+T Cells and NK Cells) Validation*

The CD8+T and NK cell percentages and functions were used to evaluate immune activation status. CD8<sup>+</sup> T cell percentage and function were identified by CD8, IL-2, and IFN-γ markers in tumor-draining lymph node (TDLN) and spleen. Intracellular staining was performed with Fixation/Permeabilization kit following the manufacturer's protocol. In addition, NK cells on two different strains of animal models were also identified by various markers, CD3-/CD49b+/CD335<sup>+</sup> on BALB/c and CD3-/CD49b+/NK1.1<sup>+</sup> on C57BL/6, respectively [19]. The percentages of these cell types were acquired by NovoExpress® flow cytometry (Agilent, Santa Clara, CA, USA) and data was analyzed by FlowJo software (BD Pharmingen™).

#### *2.9. Immune Suppressive Cells (Treg Cells and MDSCs) Validation*

The percentages of regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs) cells were used to evaluate immunosuppressive function. Immune suppressive cells isolated from tumor-draining lymph node (TDLN), spleen [20], and bone marrow (BM), were stained with anti-FOXP3-Alexa Fluor 488/CD4-PerCP-Cy™5.5/CD25-PE antibodies using a Mouse Treg Flow Kit according to manufacturer's protocol. CD11b-FITC/Gr-1-PE antibodies were used to detecting Tregs and MDSCs [21], respectively. The percentages of these cell types were acquired by NovoExpress® flow cytometry and data were analyzed by FlowJo software (BD Pharmingen™).

#### *2.10. Animal Experiments*

The animal experiments were performed in accordance with the protocols approved by the Animal Care and Use Committee at China Medical University (approval number: CMU IACUC-2019-208). Six-week-old male BALB/c and C57BL/6 mice were purchased from the National Laboratory Animal Center and housed in a pathogen-free animal facility. The establishment of animal model was described in material and methods section. All experiment was repeated at least twice (*n* = 6).

#### *2.11. Animal Treatment Procedure*

The animals were anaesthetized with 1–2% isofluorane during surgery and imaging. The animals were fed sterilized mouse chow and water. Five million of CT26 or B16F10/luc2 cells were administered to mice (20–25 g) by subcutaneous injection on right thigh. The body weight and tumor volume were measured 3 times per week. Tumor volume was calculated by following formula: volume = length × width<sup>2</sup> × 0.523. The animals were separated into various groups and administered with 100 µL of indicated treatment by i.p.: control (DMSO 0.1%), hIL15 (5 µg/injection), hIL15-ABD (5 µg/injection), or 10F.9G2 alone (anti-PD-L1, 100 µg/injection), co-treatments of hIL15-ABD and 10F.9G2. The drugs for animal treatment were dissolved in 100 µL H2O with 0.1% DMSO.

#### *2.12. Enzyme-Linked Immunosorbent Assay (ELISA)*

Secreted IL15, and VEGF were collected from serum and assayed by ELISA. All the procedures followed commercially provided protocol. IL15, and VEGF ELISA kits were all purchased from Elabscience (Houston, TX, USA). ELISA readings were determined by OD scanning at 450 nm using SpectraMax iD3 microplate reader from (Molecular Devices, Downingtown, PA, USA).

#### *2.13. Bioluminescence Imaging (BLI)*

Mice bearing B16-F10/*luc2* tumors of each group (*n* = 6) were intraperitoneally injected with 200 µL of 150 mg/kg D-luciferin in PBS before anesthetization with 1–2% isoflurane 10 min prior to imaging. Mice were then set onto the imaging platform and continuously exposed to 1–2% isoflurane throughout the time. The luc2 signal from tumor region was collected by BLI using IVIS50 Imaging System (Xenogen) once per week. The photons emitted from the tumor were assayed using IVIS50 Imaging System with an acquisition time of 1 min. Regions of interest (ROIs) were drawn around the tumor and quantified with the Living Image software as photons/s/cm2/sr.

#### *2.14. Immunohistochemistry (IHC)*

Formalin-fixed and paraffin-embedded tissues from mice were subjected to IHC staining. In brief, sections of paraffin-embedded tumor tissue on slides obtained from each group was deparaffinized in xylene, rehydrated with decreasing concentrations of ethanol (100%, 70%, 30%, 0%), and then incubated in 3% H2O<sup>2</sup> for 10 min. After washing, the slides were blocked with 5% normal goat serum for 5 min in a tight container, followed by incubation with different primary antibodies in a dilution of 1:100–500 at 4 ◦C overnight. Finally, slides were counterstained with hematoxylin. At least three slides from each group were studied. Slides were photographed at 200 × magnifications by Nikon ECLIPSE Ti-U microscope and quantified by ImageJ software (National Institutes of Health, Bethesda, MD, USA).

#### *2.15. Statistical Analysis*

Statistical analysis was performed utilizing excel 2017 software (Microsoft, Redmond, WA, USA) and GraphPad *Prism* 8.0 (GraphPad Software, Inc., San Diego, CA, USA). Values were expressed as means ±SD. Comparison of means between several groups were performed by one-way analysis of variance (ANOVA) and independent-test was used to compare between two groups. Tukey's test was used to compare all groups as post-hoc test. Values were considered statistically significant at *p* ≤ 0.05.

#### **3. Results**

#### *3.1. Expression and Purification of Active hIL15-ABD*

hIL15-ABD expression by transformed *E. coli*, BL21(DE3), were initiated when OD of culture reached 0.7, in the presence of increasing concentration of IPTG, ranging from 0, 0.05, 0.1, 0.3, 0.5, to 1 mM) at either 30 or 37 ◦C and 200 rpm. SDS-PAGE (upper panel, Figure 1A) and Western blot (lower panel, Figure 1A) analysis displayed comparable expressions of hIL15-ABD (21.0 kDa) induced by all the IPTG concentrations and two temperatures tested. Large scale protein expression was initiated by addition of 0.1 mM IPTG to 1 L of transformed *E. coli* culture when OD600 value reached 0.7 and kept in rotating shaker at 37 ◦C and 200 rpm (Figure 1). The majority of expressed hexahistidine-tagged hIL15-ABD was in the inclusion bodies, which were washed and dissolved in an 8 M urea denaturing buffer (Figure 1B) before being loaded onto a Ni-NTA resin column. The immobilized hexahistidine-tagged hIL15-ABD was eluted sequentially using 75, 300, to 500 mM of imidazole in PBS buffer (Figure 1C). The most significant portion of hexahistidine-tagged hIL15-ABD was eluted with 300 mM as displayed by SDS-PAGE (Figure 1D) and Western blotting using anti-hexahistidine antibody (Figure 1E). The purified denatured hIL15-ABD was investigated for optimal refolding condition using buffer matrix as listed in Table 1, and the bioactivities of the refolded hIL15-ABD were examined for binding to human albumin (Figure 1F) as well as stimulation CTLL-2 proliferation (Figure 1G). It turns out that refolding of ABD moiety of hIL15-ABD was quite robust in most of the buffers examined, whereas there is more significant difference in stimulatory effects of hIL15-ABD refolded in individual buffers. We selected buffer 26 composed of NaCl (250 mM), GSH (1 mM) and GSSG (0.1 mM) in Tris-HCl buffer (50 mM, pH 8.5) for later large-scale protein refolding (Figure 2A). We were able to obtain approximately 60 mg of recombinant hIL15ABD with purity higher than 90% per liter of TB culture in shake flasks IPTG-induced *E. coli.* The bioactivities of the refolded hIL15-ABD were examined for binding to either human or murine albumin as well as stimulation of STAT5 phosphorylation in CTLL-2 cells. hIL15-ABD demonstrates similar binding affinity for both human and murine albumin with Kd values of 3.0 and 2.8 nM, respectively (Figure 2B), whereas it displayed comparable stimulatory effect on STAT5 phosphorylation with that of hIL15 positive control with EC50 of 0.17 and 0.10 nM, respectively (Figure 2C).


**Table 1.** Combinations of buffer used for protein refolding.

<sup>1</sup> 1 mM GSH/0.1 mM GSSG; <sup>2</sup> 0.4 M L-Arginine.

**Figure 1.** hIL15-ABD expression, refolding and purification. (**A**) SDS-PAGE (upper panel) and Western blot (lower panel) analysis of hIL15-ABD expression induced with increasing concentrations of IPTG, ranging from 0, 0.05, 0.1, 0.3, 0.5, to 1 mM in transformed *E. coli*, BL21(DE3) at either 30 or 37 ◦C. (**B**) SDS-PAGE analysis of fractions from transformed *E. coli* lysates (lane 1); supernatant of the lysates after centrifugation at 10,000 rpm for 20 min (lane 2), supernatant after washing with H2O (lane 3); supernatant after washing with 20 mM Tris-HCl (lane 4); supernatant after washing with 50 mM Tris-HCl buffer containing 2 mM EDTA and 0.1% SDS (lane 5); supernatant after washing with 50 mM Tris-HCl, 150 mM NaCl and 2 mM EDTA (lane 6); supernatant after washing with H2O (lane 7); supernatant after washing with H2O (lane 8); denatured inclusion body in 8 M urea (lane 9). Lane 1 to 8 each are loaded protein equal to 75 µL and lane 9 equal to 37.5 µL of culture medium. (**C**) Purification of solubilized inclusion body from *E. coli* expressing hIL15-ABD through Ni-column. The blue curve indicates the absorption at 280 nm in mAU, whereas the green line represents the concentration of imidazole. (**D**) SDS-PAGE and (**E**) western blot analysis of elution fractions number 3–6 (lane 1–4), 10, 11, and 12 (lane 5–7), and 21 (lane 9). (**F**, **G**) The purified denatured hIL15-ABD was refolded with buffer matrix listed in Table 1 and resulted L15-ABD is examined for (**F**) human albumin binding and (**G**) stimulation of CTLL-2 proliferation. The dot line indicates the average OD490 values representing the extents of viable CTLL-2 in 96-well plates cultured with refolded hIL5-ABD.

μL

μL

– – –

**0 m**

at 1.5 and 1.0 μg/mouse. Data are presented as mean ± **Figure 2.** *In vitro* characterizations and pharmacokinetics of hIL15-ABD purified in large scale. (**A**) Lane 1 of the SDS-PAGE indicates the purified hIL1-ABD in refolding buffer number 26 (NaCl (250 mM), GSH (1 mM) and GSSG (0.1 mM) in Tris-HCl buffer (50 mM, pH 8.5)) and lane 2 represents hIL15-ABD after being condensed following the process of refolding. (**B**) Refolded purified hIL-15-ABD displays similar affinity for binding to both human and murine albumin, whereas there is no measurable specific binding to albumin by hIL15. (**C**) hIL15 and IL15-ABD demonstrate comparable EC50 values of stimulation of STAT-5 phosphorylation in CTLL-2 cells, which are 0.10 and 0.17 nM, respectively. (**D**) Serum concentration-time curves of hIL15 in Balb/c mice following single intraperitoneal injection of hIL15-ABD and hIL15 at 1.5 and 1.0 µg/mouse. Data are presented as mean ± SD. N = 3 at each time point.

#### *3.2. Pharmacokinetics Studies*

**inine inine Arg** The serum concentration-time curves from derived from Balb/c mice intraperitoneally injected with 1 µg hIL15 and equimolar of 1.5 µg hIL15-ABD are shown in Figure 2D. The pharmacokinetic parameters are summarized in Table 2. The maximum serum concentrations (Cmax) and times to reach Cmax (Tmax) were determined as 13.59 ng/mL at 0.75 h for hIL15 and 51.28 ng/mL at 4 h for hIL15ABD, respectively. The terminal half-lives (T1/2λ<sup>z</sup> ) of hIL15 and hIL15-ABD were 0.88 h and 23.37 h after injection, respectively. The results showed that T1/2λ<sup>z</sup> of hIL15-ABD was 26-fold longer than that of hIL15 in serum, which confirmed that the long circulation of the hIL15-ABD has been achieved. The AUC (0→∞) of hIL15 and hIL15-ABD were 18.8 ng/mL×h and 1602.4 ng/mL×h, respectively. The AUC (0→∞) of hIL15-ABD in serum was 180-fold larger than that of hIL15.

**Table 2.** Pharmarcokinetic parameters of hIL-15 and hIL-15-ABD after intraperitoneal injection in BALB/c mice.


Calculated with WinNonlin 7.0 for a noncompartmental model.

#### *3.3. hIL15-ABD Showed Superior Tumor Growth Inhibition and Positive Regulation of Immune Response on CC Model*

After confirming that hIL15-ABD has more than 20- and 80-fold increase of biological half-life and AUC, respectively, compared to those of hIL15 (Table 2), we further validated the treatment efficacy of both on colon cancer-bearing animal model (Figure 3A). In light of the superior pharmacokinetic profiles of hIL15-ABD and to demonstrate the potent in vivo anticancer effect of hIL15-ABD, we used 5 µg for each injection, which represent 0.36 and 0.24 nanomole of hIL15 and hIL15-ABD, respectively, instead of using equimolar proteins. As shown in Figure 3B and C, hIL15-ABD displayed better tumor growth inhibition ability as compared to hIL15 past day 9 post-treatment. Additionally, hIL15-ABD showed potential to suppress the accumulation of MDSCs, immunosuppressive cells, in spleen and bone marrow (Figure 3D). Percentage of CD11b+/Gr-1<sup>+</sup> cells that are recognized as MDSCs were effectively decreased in hIL15-ABD treated group (Figure 3E,F). Percentage of another group of immunosuppressive cells, regulatory T cells (Tregs), from TDLN and spleen was also identified by flow cytometry after treatment (Figure 3G). Number of CD4+/CD25+/FOXP3<sup>+</sup> cells was more effectively reduced around one of two by hIL15- ABD as compared to hIL15 (Figure 3H,I). The Treg population was reduced by more than a half in hIL15-ABD-treated group compared to non-treatment group, and also had significantly larder reduction compared to hIL15-treated group. Furthermore, we observed that hIL15-ABD may also increase the population of CD8<sup>+</sup> T cells in TDLN and spleen (Figure 3J). Two times more percentage of CD8<sup>+</sup> T cells was detected in hIL15- ABD group compared to non-treatment group (Figure 3K,L). Other than induction of adaptive immunity, NK cells, which plays role in innate immunity, was also effectively triggered in hIL15-ABD-treated group (Figure 3M). CD3-/CD49b<sup>+</sup> , CD3-/CD335<sup>+</sup> and CD3-/CD49b+/CD335<sup>+</sup> cells population were all significantly elevated in hIL15-ABDtreated group compared to hIL15-treated group (Figure 3N). Decrease in mice body weight was only observed in hIL15-treated group (Figure 3O), indicating the possibility of toxicity caused by prolonged treatment with unmodified form of hIL15. In sum, hIL15-ABD not only demonstrates better tumor inhibition, but also provides a positive microenvironment for cells involved in innate and adaptive immunities to function.

**Figure 3.** *Cont*.

**Figure 3.** hIL15-ABD induced the accumulation of CD8 <sup>+</sup> T cells and NK cells, but diminished Tregs and MDSCs, resulting in colon cancer growth inhibition. (**A**) Animal flow chart of different treatment materials is displayed. (**B**) Tumor volume is recorded every 3 days and (**C**) tumor weight is weighed after isolation from mice on day 21. (**D**) Flow cytometry pattern of CD11b+/Gr-1+ MDSCs isolated from BM and SP. Percentage of CD11b <sup>+</sup>/Gr-1 <sup>+</sup> MDSCs from (**E**) BM and (**F**) SP are gated and quantified by FlowJo software. (**G**) Flow cytometry pattern of CD4+/CD25+/FOXP+ Tregs isolated from TDLN and SP. Percentage of CD4 <sup>+</sup>/CD25 <sup>+</sup>/FOXP3 <sup>+</sup> Tregs from (**H**) TDLN and (**I**) SP. (**J**) Flow cytometry pattern of CD8 <sup>+</sup> T cells isolated from TDLN and SP. Percentage of CD8 <sup>+</sup> T cells from (**K**) TDLN and (**L**) SP. (**M**) Flow cytometry pattern of CD3-/CD49b <sup>+</sup>/CD335 <sup>+</sup> NK cells isolated from TDLN and SP. Percentage of CD3 -/CD49b <sup>+</sup>/CD335 <sup>+</sup> NK cells from (**N**) SP. (**O**) Mice body weight are measured 3 time per week. [BM = bone marrow, TDLN = tumor-draining lymph node and SP = spleen] (a <sup>1</sup> *p* < 0.05, a <sup>2</sup> *p* < 0.01 vs. CTRL; b <sup>1</sup> *p* < 0.05, b <sup>2</sup> *p* < 0.01 vs. hIL15).

#### *3.4. hIL15-ABD Enhanced Tumor Inhibition Capacity and Triggered Apoptosis Effect of Anti-PD-L1 Therapy on Both CC and Melanoma Models*

Though hIL15-ABD monotherapy demonstrated tumor inhibition potential, the inhibition ability remained limited. Therefore, we further validated whether hIL15-ABD may positively augment the function of checkpoint inhibitor-related therapy. In Figure 4A, we show the effects of hIL15-ABD and anti-PD-L1 anti-body (10F.9G2) monotherapies as well as their combined therapy effect. Combined therapy not only showed superior tumor growth inhibition in colon cancer (CC) (Figure 4B), but also melanoma bearing animal model (Figure 4D). Tumors isolated from combined therapy groups in CC and melanoma models on day 21 displayed significant tumor shrinking effect as compared to those isolated from monotherapy groups (Figure 4C,E). In addition, tumor weight of combined therapy groups also showed more significant decreases compared to either hIL15-ABD or anti-PD-L1 anti-body monotherapy groups (Figure 4F,G). Luc2 signal emitted from melanoma (B16-F10/*luc2*) was recognized as amount of living cells within tumor region that also presented the minimal signal intensity in combined therapy group (Figure 4H). Quantification result from BLI (Figure 4I) was corresponded to tumor volume, and the combined therapy group was found to exhibit superior tumor growth inhibition compared to the monotherapy groups. No obvious body weight loss of each treatment procedure was found in both CC and melanoma models (Figure 4J,K). Ki-67, a cell proliferation marker, was showed to be effectually suppressed by combination therapy (Figure 4L,M). Finally, we measured BAX and cleaved caspase-3 protein expression levels in CC and melanoma (Figure 4N). BAX and cleaved caspase-3 stain signals were markedly increased in combination therapy group (Figure 4O,P). Taken together, these results demonstrate that hIL15-ABD can successfully enhance tumor inhibition ability of anti-PD-L1 by disrupting proliferation effect and induction of apoptosis signaling.

**Figure 4.** *Cont*.

– – μ **Figure 4.** hIL15-ABD facilitated anti-tumor efficacy of anti-PD-L1 antibody via enhancing apoptosis mechanism. (**A**) Animal flow chart of hIL15-ABD, anti-PD-L1 and combination treatment is presented. (**B**,**C**) Colon cancer (CC) tumor growth from day 0–18 and tumor photographed on day 21 are displayed. (**D**,**E**) Melanoma tumor growth from day 0–18 and tumor photographed on day 21 are displayed. Tumor weight from (**F**) CC and (**G**) melanoma on day 18 are summarized. (**H**) BLI and (**I**) quantification results from B16-F10/*luc2* bearing mice are presented. Mice body weight from (**J**) CC (**K**) melanoma model is recorded every 3 days during therapy. (**M**,**N**) IHC staining images and (**L**,**O**,**P**) relative proteins quantification level on CC and melanoma are presented. (a <sup>1</sup> *p* < 0.05, a <sup>2</sup> *p* < 0.01 vs. CTRL; b <sup>1</sup> *p* < 0.05, b <sup>2</sup> *p* < 0.01 vs. hIL15-ABD and anti-PD-L1; scale bar = 100 µm).

#### *3.5. hIL15-ABD Strengthened Anti-PD-L1-Induced Function of CD8 <sup>+</sup> T Cells on Both CC and Melanoma Models*

γ and IL γ were increased to 50% in combination treatment in TDLN. Both CC and melanoma γ in CD8 γ in CD8 γ and IL To further investigate the effect of hIL15-ABD and anti-PD-L1 combination on tumor microenvironment, we determined the function of CD8 <sup>+</sup> T cells by observing activation of intracellular IFN-γ and IL-2. As shown in Figure 5A, CD8 + cells with the expression of IFN-γ were increased to 50% in combination treatment in TDLN. Both CC and melanoma displayed an increasing percentage of IFN-γ in CD8 + cells from TDLN after combination therapy (Figure 5B). The expression level of IFN-γ in CD8 <sup>+</sup> Tcells from SP has showed similar elevations, especially in combination therapy group (Figure 5C,D). Furthermore, we found that IL-2 activation in CD8 <sup>+</sup> T cells from TDLN (Figure 5E,F) and SP (Figure 5G,H) were both effectually increased in the hIL15-ABD + anti-PD-L1 combined therapy group as compared to monotherapy groups. Lastly, we performed IHC staining on CC and melanoma tumor to validate granzyme B and CD8 protein expression after therapy (Figure 5I). The results show that not only CD8, but also granzyme B, key indicators of cytotoxic T cells (CD8 <sup>+</sup> T), were raised by combination therapy (Figure 5J,K). Higher activation levels of IFN-γ and IL-2 in CD8 <sup>+</sup> T cells from TDLN and SP in CC and melanoma models were also found in the combination treatment group relative to those of the monotherapy groups.

**Figure 5.** *Cont*.

γ γ μ **Figure 5.** hIP-15-ABD offer a reinforce role of increasing anti-PD-L1 antibody induced CD8+ T cells activation. (**A**) Flow cytometry pattern and (**B**) quantification results of CD8+/IFN-γ + cells from TDLN. (**C**) Flow cytometry pattern and (**D**) quantification results of CD8 <sup>+</sup>/IFN-γ + cells from SP. (**E**) Flow cytometry pattern and (**F**) quantification results of CD8 <sup>+</sup>/IL-2 + cells from TDLN. (**G**) Flow cytometry pattern and (**H**) quantification results of CD8 <sup>+</sup>/IL-2 + cells from SP. (**I**) Granzyme B and CD8 immunohistochemistry (IHC) staining images and relative proteins quantification level of (**J**) CC and (**K**) melanoma are displayed. (a <sup>1</sup> *p* < 0.05, a <sup>2</sup> *p* < 0.01 vs. CTRL; b <sup>1</sup> *p* < 0.05, b <sup>2</sup> *p* < 0.01 vs. hIL15-ABD and anti-PD-L1; scale bar = 100 µm).

#### *3.6. hIL15-ABD Increased Anti-PD-L1 Antibody Induced Accumulation and Activation of NK Cells in Both CC and Melanoma Models*

To identify whether combining hIL15-ABD with anti-PD-L1 promote the function of NK cells, we measured NK cell population and activity in the spleen after treatment. Results from Figure 6A indicate that CD3 -/CD49b + , CD3 -/CD335 + , CD3 -/NK1.1 + , CD3 -/NK1.1 <sup>+</sup>/CD335 <sup>+</sup> and CD3 -/CD49b <sup>+</sup>/CD335 + cells were all dramatically increased after combination therapy. Based on different species of animal, we separated NK cells according to their specific markers as indicated in Figure 6B,C. Highest amount of CD3 -/NK1.1 +/ CD335 <sup>+</sup> and CD3 -/CD49b <sup>+</sup>/CD335 + triple positive cells were found in the combined therapy group. Next, we identified whether these NK cells possessed function by detecting

intracellular IFN-γ. The activation of IFN-γ was observably increased in CD3 -/NK1.1 <sup>+</sup> and CD3 -/CD49b + cells from the combined therapy group (Figure 6D,E). Furthermore, we also investigated the expression levels of CD49b and IFN-γ proteins expression on tumor tissue from CC and melanoma models by IHC staining (Figure 6F). As shown in Figure 6G,H, the protein expression levels of both CD49b and IFN-γ were increased in treated groups. Finally, we checked VEGF (Figure 6I) secretion level in mouse serum to demonstrate the decreasing of immunosuppressive factor after combination therapy. Most importantly, the level of IL15 secretion was also effectively triggered by hIL15-ABD combined with anti-PD-L1 (Figure 6J). These results support the hypothesis that hIL15-ABD combined with anti-PD-L1 may develop a positive regulation of immune response for defending against tumor. γ. The activation of IFN γ was observably increased in CD3 γ proteins expression γ were

**Figure 6.** *Cont*.

γ γ γ μ **Figure 6.** hIP-15-ABD combined anti-PD-L1 antibody effectively trigger the accumulation and function of NK cells. (**A**) Flow cytometry pattern and (**B**) quantification results of CD3 -/CD49b + , CD3 -/CD335 <sup>+</sup> and CD3 -/CD49b <sup>+</sup>/CD335 <sup>+</sup> NK cells from SP on CC bearing BALB/c animal model. (**C**) Quantification results of CD3 -/CD49b + , CD3 -/NK1.1 <sup>+</sup> and CD3-/CD49b <sup>+</sup>/NK1.1 <sup>+</sup> NK cells from SP on melanoma bearing C57BL/6 animal model. (**D**,**E**) CD3 -/CD335 <sup>+</sup>/IFN-γ <sup>+</sup> and CD3-/NK1.1 <sup>+</sup>/IFN-γ <sup>+</sup> NK cells from SP on CC and melanoma model is displayed. (**F**) CD49b and IFN-γ IHC staining images and relative proteins quantification level on (**G**) CC and (**H**) melanoma are shown. Expression level of secreted (**I)** VEGF and (**J**) IL15 are shown as quantification results. (a <sup>1</sup> *p* < 0.05, a <sup>2</sup> *p* < 0.01 vs. CTRL; b <sup>1</sup> *p* < 0.05, b <sup>2</sup> *p* < 0.01 vs. hIL15-ABD and anti-PD-L1; scale bar = 100 µm).

#### *3.7. hIL15-ABD Combined Anti-PD-L1 Antibody Diminished the Accumulation of Immunosuppressive Cells in Both CC and Melanoma Models*

– – Next, we determined whether hIL15-ABD promotes anti-tumor capacity of anti-PD-L1 by reducing accumulation of Tregs and MDSCs. Flow cytometry from mice TDLN showed that CD4 <sup>+</sup>/CD25 <sup>+</sup>/FOXP3 + triple positive cells amount was significantly reduced in combination therapy group (Figure 7A). The amount of Tregs was decreased by around 5–10 fold as compared to non-treated control (Figure 7B). At the same time, the percentage of Tregs decreased the most in the combination treatment group (Figure 7C,D). Moreover, we also detected the population of MDSCs within BM and SP of CC and melanoma mice by flow cytometry. The obtained results of flow cytometry from BM indicated the effective diminishment of CD11b <sup>+</sup>/Gr-1 <sup>+</sup> MDSCs in combination treatment group (Figure 7E,F). The percentage of CD11b <sup>+</sup>/Gr-1 <sup>+</sup> MDSCs within CC and melanoma mice SP was also decreased after combination therapy (Figure 7G,H). Subsequently, we validated the protein expression level of FOXP3 and IDO in mice tumor by IHC staining (Figure 7I). FOXP3 and IDO are known to be important immunosuppressive factors that allow the tumor to escape immunosurveillance. As indicated in Figure 7J,K, proteins expression levels of FOXP3 and IDO in combination therapy group were decreased to 10–30% of that in non-treated

an environment that inhibits the tumor's ability to escape from immune surveillance.

control. Our results illustrate that combination of hIL15-ABD and anti-PD-L1 may develop an environment that inhibits the tumor's ability to escape from immune surveillance.

**Figure 7.** *Cont*.

; scale bar = 100 μm **Figure 7.** hIP-15-ABD combined anti-PD-L1 antibody successfully suppress the accumulation of immunosuppressive cells. (**A**,**C**) Flow cytometry pattern and (**B**,**D**) quantification results of CD3 -/CD49b + , CD4 <sup>+</sup>/CD25 <sup>+</sup>/FOXP3 <sup>+</sup> Tregs from TDLN and SP, respectively. (**E**,**G**) Flow cytometry pattern and (**F**,**H**) quantification results of CD11b <sup>+</sup>/Gr-1 <sup>+</sup> MDSCs from BM and SP, respectively. (**I**) FOXP3 and IDO IHC staining images and (**J**,**K**) relative quantification of CC and melanoma are presented. (a <sup>1</sup> *p* < 0.05, a <sup>2</sup> *p* < 0.01 vs. CTRL; b <sup>1</sup> *p* < 0.05, b <sup>2</sup> *p* < 0.01 vs. hIL15-ABD and anti-PD-L1; scale bar = 100 µm).

#### **4. Discussion**

μ In the first human clinical trial, hIL15, as a wild-type (WT) recombinant protein was administrated for 12 consecutive days to patients with metastatic melanoma and renal cell carcinoma [12]. Dose-limiting toxicities of WT hIL15, included grade 3 hypotension, thrombocytopenia, and elevated values of ALT and AST and 0.3 µg/kg per day was determined as the maximum tolerable dose. Although greatly altered homeostasis of lymphocyte subsets, such as NK cells and memory CD8 T cells, as well as anticancer efficacy observed in the first in-human trial of recombinant WT hIL15, it becomes evident that alternative dosing strategies is needed to enhance efficacy while reducing toxicity. Nonhuman primate pharmacokinetic study verified that constant administration regimens of recombinant IL15 through either continuous intravenous infusion or subcutaneous injection achieve remarkable immune stimulation in the absence of obvious toxicity, indicating potentially better clinical result than the previous bolus intravenous regimen [22]. Clinical

trial of recombinant hIL15 administrated subcutaneously daily (Monday through Friday) for two weeks was conducted in patients with refractory solid tumor cancers. This dosing regimen resulted in markedly enhanced circulating CD56bright NK and CD8<sup>+</sup> T cells as well as an encouraging safety profile [23].

Although hIL15 displayed encouraging results in early clinical trials, its short half-life suggests potential improvement in anticancer efficacy through engineering hIL15 with prolonged half-life. N-803, the novel hIL15 superagonist complex, comprises N72D mutant IL15 and IL15Rα-IgG Fc fusion protein and displays enhanced affinity for IL-2Rβ and prolonged half-life. It is under multiple clinical trials, including advanced melanoma, renal cell, non-small cell lung, head and neck, hematologic malignancies who relapse after allogeneic hematopoietic cell transplantation and showing encouraging results [16,17,24]. N-803 has been shown to exhibit greater anti-CC activity compared to hIL15 in CT26 bearing model [25]. In this study, CT26 bearing model was also used to evaluate differences in therapeutic efficacy and anticancer immune response between hIL15 and hIL15-ABD treatments. Our results demonstrate that hIL15-ABD group has higher tumor growth inhibition capability and anticancer immunity than hIL15 group (Figure 3). Immunosuppressive cells such as Tregs and MDSCs restrain antitumor immunity through the downregulation of effector T cells and NK cells [26–28]. The increased abundance of Treg or MDSCs in peripheral blood and tumor are associated with poor prognosis in different types of cancer [29–31]. Although the relationship between Treg population and prognosis in patients with colorectal cancer remains uncertain [32,33], depletion of Tregs and MDSCs has been indicated to promote anticancer immunity in colorectal cancer [34,35]. In our results, we present that hIL15-ABD not only significantly increased percentage of CD8<sup>+</sup> T and NK cells (Figure 3J–O), but also effectively reduced population of Tregs and MDSCs compared to hIL-15 treatment (Figure 3D–I).

Anti-PD-L1 therapies have been shown promising results as a member of an increasing number of immunotherapies against cancer [36]. However, despite its potential, anti-PD-L1 antibody has failed to elicit objective response in a majority of patients treated [37]. A common explanation for the lack of response is the lack of anti-tumor effector cells in the TME. Both tumor cells and MDSCs express PD-L1, which binds to PD-1 on T cells and causes T cell exhaustion as well as conversion of T helper type 1 (Th1) cells to Tregs. The combination of N-803 and anti-PD-L1 therapy reduced numbers of Tregs and MDSCs in lung [38,39]. Having verified that hIL15-ABD is superior to hIL15 in inhibiting tumor growth and regulating anti-cancer immunity, we investigated the anticancer efficacy and immune response induction of hIL15-ABD combined with anti-PD-L1 in both CC and melanoma models. Our results indicate an obvious enhancement of tumor growth inhibition in CT26 or B16-F10 bearing mice after hIL15-ABD and anti-PD-L1 combined therapy (Figure 4B–G). Furthermore, the combination group had significantly smaller population of Tregs (within TDLN and SP, Figure 7A–D) and MDSCs (within BM and SP, Figure 7E–H) compared to hIL15-ABD or anti-PD-L1 therapy monotherapies.

Both CD8<sup>+</sup> T and NK cells are critical executors that mediate tumor cell apoptosis through secretion of granzyme-B and IFN-γ in immunotherapy modulating tumor regression. In addition to hIL15, anti-PD-1/-L1 therapy has also been indicated to enhance anti-tumor efficacy of CD8<sup>+</sup> T and NK cells [4,9,40–42]. The combination of N-803 and anti-PD-L1 therapy significantly induced the activated CD8<sup>+</sup> T cell phenotype compared to N-803 or anti-PD-L1 monotherapies in murine breast cancer models [43]. The increased number and function of CD8<sup>+</sup> T or NK cells were linked to favorable prognosis in patients with colorectal cancer or melanoma [44–46]. Therefore, it is worthwhile to investigate whether hIL15-ABD promotes anti-PD-L1 therapy-elicited activity and percentage of CD8<sup>+</sup> T and NK cells in CT26 or B16-F10 bearing mice. In our results, the combination of hIL15-ABD and anti-PD-L1 monotherapy effectively increased percentage of CD8<sup>+</sup> IFN-γ or CD8<sup>+</sup> IL-2 cells in spleen and TDLN compared to hIL15-ABD or anti-PD-L1 therapy (Figure 5A–H). Human IL15-ABD also significantly promoted anti-PD-L1 therapy-induced accumulation and function of NK cells in spleen and TDLN (Figure 6A–H).

Granzyme-B, the granule protease secreted by NK and CD8<sup>+</sup> T cells, induces apoptosis through BAX/BAK-mediated mitochondrial apoptotic pathway [47]. The increased level of granzyme-B in serum or tumor was correlated with favorable outcomes in patients with colorectal cancer or NSCLC [48,49]. In our results indicated the combination group presented significantly higher expression of granzyme-B and apoptotic proteins (BAX and cleaved-caspase-3) in CT-26 or B16-F10 tumor tissues compared to hIL15-ABD or anti-PD-L1 therapy (Figure 5I–K and Figure 4N–P). VEGF, the major angiogenic mediator, contributes to tumor growth and metastasis through promoting new vessel formation. VEGF participates in regulation of Tregs and MDSCs leading to restriction of anti-tumor immunity. The high level of serum VEGF was correlated with poor overall survival of melanoma patients treated with the immune checkpoint inhibitor [50]. Indoleamine 2,3-dioxygenase (IDO), immunosuppressive protein, attenuates anti-tumor function of T cells by regulating the conversion of tryptophan to kynurenine [51]. The decreased expression of IDO was associated with better prognosis in patients with colorectal cancer or melanoma [52]. Our results demonstrated expression of IDO and VEGF was obviously reduced by hIL15-ABD, anti-PD-L1, or combination therapy (Figures 6I and 7I–K). The combination group had lower expression of IDO or VEGF compared to hIL15-ABD or anti-PD-L1 monotherapy in CT26 or B16-F10 bearing mice.

In light of the great potential of IL15 as one of the critical weapons in the arsenal of anticancer immunotherapy, many related therapeutics, ranging from the wild-type IL15, IL15 superagonist (N-803) to PD-L1–targeting IL15 (KD033 [53] and N-809 [54]), are actively being developed in either preclinical or clinical settings. In our study, hIL15-ABD displays a much-extended half-life and superior inhibitory effect on CT26 and B16 growth in experimental mice than WT hIL15, which has shown encouraging results in early human trial [12]. hIL15-ABD could be easily purified and refolded into active form with yields of approximately 60 and 300 mg/L of transformed *E. coli* TB cultures in shake flask and fermenter, respectively, indicating a relatively lower production cost than those of modified hIL15, such as N-803, KD033 and N-809, expressed by mammalian cells. Intriguingly, while displaying anticancer effect as a monotherapy (Figure 3B), hIL15 treated CT26-bearing mice showed statistically significant body weight loss as compared with those treated with vehicle and hIL15-ABD (Figure 3O), suggesting a better therapeutic window of hIL15-ABD comparing with hIL15. Given that albumin-based carriers for anticancer therapeutics has shown promising results in both preclinical and clinical studies not only through half-life prolongation but also enhanced tumor localization [55], it is of great interest to investigate whether albumin associated hIL15-ABD will obtain a more favorable biodistribution profile, thereby increasing anticancer effects while reducing toxicity to normal organs.

#### **5. Conclusions**

In conclusion, for the first time, we presented the hIL15-ABD, the novel recombinant IL15 protein, was superior to in induction of tumor regression and antitumor immunity. hIL15-ABD may suppress the accumulation of MDSCs and Treg at the site of the tumor. In addition, hIL15-ABD can also promote the activity of IL-2 and IFN-γ in CD8<sup>+</sup> T cells or NK cells, supporting more effective anti-tumor activity by effector cells. Importantly, hIL15-ABD can trigger innate immunity by enhancement of NK cells toxicity effect. Furthermore, the combination of hIL15-ABD and anti-PD-L1 therapy significantly inhibited tumor growth and promoted anti-tumor immune response compared to either monotherapy in mouse models of CC or melanoma. We demonstrated enhancement of CD8<sup>+</sup> T and NK cells accumulation and cytotoxic function and reduction of Tregs and MDSCs population are associated with antitumor properties of hIL15-ABD combined with anti-PD-L1 therapy in CC or melanoma. We suggested the combination of hIL15-ABD and anti-PD-L1 therapy as potential immune therapy may offers therapeutic activity for treatment of CC or melanoma.

**Author Contributions:** Conceptualization, F.-T.H. and K.-L.L.; data curation, F.-T.H., Y.-C.L., C.-L.T. and P.-F.Y.; funding acquisition, F.-T.H.; investigation, P.-F.Y.; methodology, F.-T.H.; project administration, F.-T.H.; supervision, F.-T.H.; validation, F.-T.H. and C.-L.T.; visualization, F.-T.H. and K.-L.L.; writing—original draft, F.-T.H. and Y.-C.L.; writing—review and editing, F.-T.H., Y.-C.L., C.-H.C. and K.-L.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financially supported by a grant from the Ministry of Science and Technology (Taipei), (grant number: MOST 108-2314-B-039-007-MY3, 109-2623-E-010-002-NU and MOST 109-2314-B-758-001). This work was also financially supported by the "Drug Development Center, China Medical University" from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan.

**Institutional Review Board Statement:** The animal experiments were performed in accordance with the protocols approved by the Animal Care and Use Committee at China Medical University (approval number: CMU IACUC-2019-208).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data generated and analyzed will be made available from the corresponding author on reasonable request.

**Acknowledgments:** Experiments and data analysis were performed in part through the use of the Medical Research Core Facilities Center, Office of Research & Development at China Medical University, Taichung, Taiwan.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Vaccine Increases the Diversity and Activation of Intratumoral T Cells in the Context of Combination Immunotherapy**

**Lucas A. Horn <sup>1</sup> , Kristen Fousek <sup>1</sup> , Duane H. Hamilton <sup>1</sup> , James W. Hodge <sup>1</sup> , John A. Zebala <sup>2</sup> , Dean Y. Maeda <sup>2</sup> , Jeffrey Schlom <sup>1</sup> and Claudia Palena 1,\***


**Simple Summary:** Innovative strategies to reduce immune suppression and activate tumor-specific immunity are needed to help patients who do not respond or become resistant to immune checkpoint blockade therapies. In this study, we demonstrate that the addition of a cancer vaccine targeting a tumor-associated antigen to a checkpoint inhibitor-based immunotherapy induces greater numbers of proliferative, activated, and cytotoxic tumor-infiltrating T cells, leading to improved antitumor activity in tumors otherwise resistant to immunotherapy. Our results provide the rationale for the addition of cancer vaccines in combination immunotherapy approaches being evaluated in the clinic.

**Abstract:** Resistance to immune checkpoint blockade therapy has spurred the development of novel combinations of drugs tailored to specific cancer types, including non-inflamed tumors with low T-cell infiltration. Cancer vaccines can potentially be utilized as part of these combination immunotherapies to enhance antitumor efficacy through the expansion of tumor-reactive T cells. Utilizing murine models of colon and mammary carcinoma, here we investigated the effect of adding a recombinant adenovirus-based vaccine targeting tumor-associated antigens with an IL-15 super agonist adjuvant to a multimodal regimen consisting of a bifunctional anti-PD-L1/TGF-βRII agent along with a CXCR1/2 inhibitor. We demonstrate that the addition of vaccine induced a greater tumor infiltration with T cells highly positive for markers of proliferation and cytotoxicity. In addition to this enhancement of cytotoxic T cells, combination therapy showed a restructured tumor microenvironment with reduced Tregs and CD11b+Ly6G<sup>+</sup> myeloid cells. Tumor-infiltrating immune cells exhibited an upregulation of gene signatures characteristic of a Th1 response and presented with a more diverse T-cell receptor (TCR) repertoire. These results provide the rationale for the addition of vaccine-to-immune checkpoint blockade-based therapies being tested in the clinic.

**Keywords:** cancer vaccine; combination immunotherapy; TCR diversity

#### **1. Introduction**

Immune checkpoint blockade therapies have led to successful and durable responses in patients with various tumor types [1,2]. Despite this great success, only a small percentage of patients with solid malignancies experience complete responses with antibodies directed against programmed cell death protein 1 (PD-1), programmed death ligand 1 (PD-L1), or cytotoxic T-lymphocyte associated protein 4 (CTLA-4) as monotherapies [3]. Expanding knowledge of the mechanisms of immunoregulation and resistance to immune checkpoint blockade therapy has allowed researchers to better formulate combinations of drugs aimed at simultaneously targeting the numerous inhibitory factors and cell types responsible for tumor-induced immune suppression and treatment failure [4,5].

**Citation:** Horn, L.A.; Fousek, K.; Hamilton, D.H.; Hodge, J.W.; Zebala, J.A.; Maeda, D.Y.; Schlom, J.; Palena, C. Vaccine Increases the Diversity and Activation of Intratumoral T Cells in the Context of Combination Immunotherapy. *Cancers* **2021**, *13*, 968. https://doi.org/10.3390/ cancers13050968

Academic Editors: Subree Subramanian and Xianda Zhao

Received: 22 January 2021 Accepted: 20 February 2021 Published: 25 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Immunologically "cold" or non-inflamed tumors present with a series of unique problems that cannot be overcome by immune checkpoint blockade or modification of the tumor microenvironment (TME) [6,7], including deficiencies in T-cell recognition of tumor antigens, dendritic cell priming, and lymphocyte homing to the tumor tissue. One approach being investigated to potentially address these additional problems is the incorporation of a therapeutic cancer vaccine to other immunotherapeutic regimens. Studies in murine models have demonstrated that checkpoint blockade antibodies are more effective when combined with cancer vaccines than checkpoint blockade alone, even in tumors that are refractory to checkpoint blockade monotherapy [8,9]. Other studies have shown that addition of a cancer vaccine can promote epitope spreading and antigen cascade [10]; this increase in T-cell receptor (TCR) diversity has been shown to drive more potent antitumor immunity and tumor clearance [11]. Furthermore, cancer vaccines targeted to cancerassociated antigens or neoantigens have had success in the clinic and have been shown to be safe and well tolerated by patients [12–14].

Bintrafusp alfa is a first-in-class bifunctional fusion protein composed of the extracellular domain of the human transforming growth factor β receptor II (TGF-βRII) fused to the C-terminus of each heavy chain of an IgG1 antibody blocking PD-L1. This agent is currently being evaluated in multiple clinical studies, showing clinical activity with a confirmed objective response rate of 30.5% in patients with human papillomavirus-associated malignancies [15,16]. In a previous study, we showed that the combination of bintrafusp alfa with SX-682, a small molecule inhibitor of the chemokine receptors CXCR1 and CXCR2 that blocks signaling initiated by IL-8 and other chemokines of the CXCL family, synergizes to mediate antitumor activity in murine models of breast and lung cancer [17]. To test our hypothesis that a vaccine could help overcome some of the challenges presented by tumors that are refractory to checkpoint blockade, in the present study we investigated the effect of adding a vaccine consisting of a recombinant adenovirus serotype-5 (Ad5) vector encoding a tumor-associated antigen in combination with N-803 as an adjuvant [18] to the bintrafusp alfa/SX-682 combination. N-803 is an IL-15 super agonist that helps activate antigen-specific T cells and has shown clinical activity in combination with checkpoint blockade in non-small cell lung cancer [19,20].

Using murine models of colon and breast cancer, we demonstrate that the addition of vaccine to bintrafusp alfa/SX-682 significantly increases tumor infiltration with T cells, enhances T-cell activation and TCR diversity at the tumor site, and diversifies the number of tumor antigens being recognized by TCRs through the phenomenon of antigen cascade or epitope spreading. These results provide the rationale for the addition of cancer vaccines as integral components in combination immunotherapy approaches being evaluated in the clinic.

#### **2. Materials and Methods**

#### *2.1. Cell Lines*

BALB/c-derived 4T1 mammary carcinoma cells were obtained and cultured as recommended by the American Type Culture Collection (ATCC, Manassas, VA, USA). MC38-CEA cells were previously obtained by retroviral transduction of C57BL/6-derived MC38 colon cancer cells to overexpress human carcinoembryonic antigen (CEA) [21]. Cell lines were tested to be mycoplasma free using a MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland) and used at low passage number.

#### *2.2. Mice*

Female BALB/c mice were obtained from the NCI Frederick Cancer Research Facility. Mice expressing human CEA on a C57BL/6 background (CEA.Tg) were generously provided by Dr. John Shively (City of Hope, Duarte, CA, USA). Mice were approximately 4 to 6 weeks old at start of experiments and were maintained under pathogen-free conditions in accordance with the Association for Assessment and Accreditation of Laboratory Animal

Care guidelines. All animal studies were approved by the NIH Intramural Animal Care and Use Committee (LTIB-038) on 9 January 2018.

#### *2.3. Tumor Inoculation, Treatment Schedule, and Metastasis Assay*

BALB/c mice were injected in the abdominal mammary fat pad with 3 × 10<sup>4</sup> 4T1 cells. CEA transgenic mice (CEA.Tg) were injected subcutaneously (s.c.) in the flank with 3 × 10<sup>5</sup> MC38-CEA cells. Control diet feed or SX-682-containing feed (1428.5 mg/kg, equivalent to a dose of 200 mg/kg body weight/day; Research Diets, New Brunswick, NJ, USA) were administered to mice starting on day 7. SX-682 was provided by Syntrix Pharmaceuticals under a Cooperative Research and Development Agreement (CRADA) with the NCI. In tumor volume experiments, intraperitoneal injections (i.p.) of bintrafusp alfa (kindly provided by EMD Serono under a CRADA) were given at a dose of 200 µg per mouse starting on day 14 and every 7 days thereafter, as noted. The vaccine utilized in this study consisted of a recombinant Ad5 encoding either the tumor antigen murine Twist1, a transcription factor that is overexpressed in 4T1 tumors [22], or human CEA, which is over-expressed in MC38-CEA tumors. The Ad-vector was combined with the IL-15 super agonist N-803 as an adjuvant. The antitumor efficacy of this vaccine formulation was previously described [18], and its optimized performance was confirmed here in terms of induction of higher levels of the Th1 cytokine, TNFα, in the serum of animals in the combined Ad-vector + N-803 group versus each single agent (Figure S1). Adenovirus vaccine was administered s.c. (1 × 10<sup>10</sup> viral particles) on day 7 (prime) followed by s.c. adenovirus vaccine (1 × 10<sup>10</sup> viral particles) plus N-803 (1 µg, s.c.) every 7 days as noted (boosts).

Metastasis assays were performed as previously described with some modifications [17]. Lungs were harvested from 4T1 tumor-bearing mice under sterile conditions, rinsed in phosphate buffer saline (PBS), transferred to gentleMACS C tubes (Miltenyi Biotec, Waltham, MA, USA) in RPMI-1640 medium containing 5% fetal bovine serum (FBS), 5 mg/mL collagenases IV and I (Gibco, Gaithersburg, MD, USA), and 40 U/mL DNase, and dissociated using a gentleMACS tissue dissociator (Miltenyi Biotec), following the manufacturer's recommended procedure. Cells were passed through a 70 µm filter, pelleted and washed with PBS, and resuspended in 10 mL RPMI-1640 medium supplemented with 10% FBS, 1% Na pyruvate, 1% Hepes, 1× glutamine, 1× gentamicin, and 1× penicillin-streptomycin. A 250 µL aliquot of this suspension, representing 1/40 of the total lung, was cultured in the same medium containing 60 µM 6-thioguanine for 14 days. Colonies were fixed with methanol, stained with 0.05% (*w*/*v*) methylene blue, air-dried, and counted. The number of metastases per lung was calculated as the number of colonies counted per flask ×40.

In mouse experiments quantifying TCR diversity, control or SX-682-containing feed were administered to mice starting on day 7 with i.p. injections of bintrafusp alfa given at a dose of 492 µg per mouse on days 9 and 11. The vaccine was administered s.c. (1 × 10<sup>10</sup> viral particles) plus s.c. N-803 (1 µg) on day 9. Tumors were collected on day 17 post-tumor injection for subsequent TCR sequence analysis, as indicated below. Adenovirus vaccines and N-803 were kindly provided by ImmunityBio under a CRADA. In all experiments, tumors were measured every 2–3 days in two perpendicular diameters. Tumor volume = (short diameter<sup>2</sup> × long diameter)/2.

#### *2.4. Depletion Studies*

To deplete CD8<sup>+</sup> T cells from MC38-CEA tumor-bearing mice, 100 µg of anti-CD8 (clone 2.43, BioXcell, Lebanon, NH, USA) depletion antibodies were administered i.p. starting on days 5, 6, and 7 post-tumor implantation and then once per week for the duration of the experiment. Blood was obtained from all animals upon termination of the experiment to determine immune cell population depletion efficiency by flow cytometry.

#### *2.5. Flow Cytometry*

Prior to staining, tumors were weighed, mechanically dissociated, incubated in a shaker at 37 ◦C for 30 min at a speed of 300 rpm in RPMI-1640 medium containing 5% FBS, 5 mg/mL collagenases IV and I (Gibco), and 40 U/mL DNase, and then passed through a 70 µm filter as a single-cell suspension. Spleens were crushed through a 70 µm filter and red cell lysis was performed with ammonium-chloride-potassium (ACK) buffer (Gibco). All antibodies used for flow cytometry were purchased from Thermo Fisher Scientific (Waltham, MA, USA), BioLegend (San Diego, CA, USA), or BD Biosciences (San Jose, CA, USA). Cells were stained for cell surface expression in flat-bottom 96-well plates on ice in phosphate buffered saline with 2% FBS. Intracellular markers were stained using the eBioscience Foxp3/Transcription Factor Staining Buffer Set according to the manufacturer's instructions. Fluorescently conjugated antibodies for CD45 (30-F11), CD3 (500A2), CD4 (RM4-5), CD8 (53-6.7), CD44 (IM7), CD62L (MEL14), Foxp3 (150D), Ki67 (16A8), GzmB (QA18A28), Ly6G (1A8), Ly6C (HK1.4), CD11b (M1/70), F4/80 (BM8), and CD11c (N418) were used as per the manufacturers' instructions. LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific) was used to gate on live cells. Data were acquired on an Attune NxT Flow Cytometer (Thermo Fisher Scientific) and analyzed via FlowJo (FlowJo, Ashland, OR). Immune cell subsets were defined as: CD4 = CD3+CD4<sup>+</sup> ; CD8 = CD3+CD8<sup>+</sup> ; TCM = CD3+CD44+CD62L<sup>+</sup> ; TEff&EM = CD3+CD44+CD62L−; Tregs = CD4+Foxp3<sup>+</sup> .

#### *2.6. ELISPOT Assays*

CEA.Tg mice bearing MC38-CEA tumors were fed an SX-682-containing diet starting on day 7; on days 14 and 21, mice received i.p. injections of bintrafusp alfa, with a priming vaccine dose of s.c. Ad-CEA administered on day 7 and boosting doses of Ad-CEA/N-803 vaccine on days 14 and 21. Control mice were left untreated and fed a base diet without SX-682. Splenocytes were harvested from control versus treated mice and assayed ex vivo on day 24 for antigen-dependent cytokine secretion using an IFNγ ELISPOT assay (BD Biosciences), according to the manufacturer's instructions. Briefly, 0.5 × 10<sup>6</sup> splenocytes were incubated overnight with 10 µg/mL of CEA526–533, p15e604–611, the MC38 neoepitope PTGFR, or a negative control peptide [10]. Spot-forming cells were quantified using an ImmunoSpot analyzer (Cellular Technology, Ltd, Shaker Heights, OH, USA). The amount of CD8<sup>+</sup> T cells added per well was calculated by flow cytometry analysis. Data were adjusted to the number of spots/0.5 × 10<sup>5</sup> CD8<sup>+</sup> T cells present in the assay, subtracting the number of spots in paired wells containing the control peptide.

#### *2.7. Real-Time PCR, Nanostring and TCR Analysis*

Total RNA from flash-frozen tumor sections was prepared using the RNeasy Mini Kit (Qiagen, Hilden, Germany). For some experiments, RNA was then reverse-transcribed using SMARTer® PCR cDNA Synthesis Kit (Takara Bio Inc, Mountain View, CA, USA) or the High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific) as per the manufacturer's instructions. cDNA was amplified in triplicate using TaqMan Master Mix in an Applied Biosystems 7500 Real-Time PCR System (ThermoFisher Scientific). The following Taqman gene expression assays were used (ThermoFisher Scientific): Cd247 (Mm00446171\_m1), Gzmk (Mm00492530\_m1), CD8a (Mm01182107\_g1), Prf1 (Mm00812512\_m1), Gzmb (Mm00442837\_m1), Cd3e (Mm01179194\_m1), Pdcd1 (Mm004349 46\_m1), Tbx21 (Mm00450960\_m1). NanoString analysis was performed on purified RNA samples from indicated tumors by using the PanCancer Immune Profiling Gene Expression Panel. The nSolver analysis software was used for data normalization (NanoString Technologies, Seattle, WA, USA). Further clustering and pathway analyses were performed using Ingenuity Pathway Analysis (Qiagen). To assess TCR diversity, genomic DNA was purified from whole tumor using the QIAamp DNA Mini Kit (Qiagen). TCRβ chain sequencing was then performed by Adaptive Biotechnologies and analyzed using the Immunoseq analyzer. Simpson clonality (square root of sum over all observed rearrangements of the square fractional abundances of each rearrangement) was calculated as a measure-

ment of the observed TCRβ repertoire. The number of clones representing the top 25% of TCR sequences was used as a metric of the relative diversity of the immune response.

#### *2.8. OPAL Immunofluorescence*

Tumor tissue was fixed in Z-fix (Anatech, Battle Creek, MI, USA), embedded in paraffin, and sectioned onto glass slides (American HistoLabs, Gaithersburg, MD, USA). Slides were stained using the Opal 4-Color Manual IHC Kit (PerkinElmer, Waltham, MA, USA). Antigen retrieval was performed with Rodent Decloaker (BioCare Medial, Pacheco, CA, USA) antigen retrieval solution and blocked with BLOXALL Blocking Solution (Vector Laboratories, Burlingame, CA, USA). All other steps, including staining with primary and secondary antibodies and OPAL fluorophore working solution, were conducted following the manufacturer's instructions. Antibodies used included anti-CD4 (4SM95, Invitrogen, Carlsbad, CA) and anti-CD8a (4SM16, Invitrogen). Slide scanning was performed on an Axio Scan.Z1 and Zen software (Zeiss, Oberkochen, Germany).

#### *2.9. Statistical Methods*

All statistical analyses were performed using GraphPad Prism V.7 for Windows (GraphPad Software, La Jolla, CA, USA). Analysis of tumor growth curves was conducted using two-way analysis of variance (ANOVA). Statistical differences between two sets of data were determined through a two-tailed Student's *t*-test, while one-way ANOVA with Tukey's post hoc test was used to determine statistical differences among three or more sets of data. Statistical differences between survival plots were determined using Log-rank (Mantel-Cox) test. Error bars represent SEM where noted. Asterisks indicate that the experimental *p* value is statistically significantly different from the associated controls at \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.001, \*\*\*\* *p* ≤ 0.0001.

#### **3. Results**

#### *3.1. Addition of Vaccine to Checkpoint Blockade-Based Therapy Enhances Immune T-Cell Infiltration and Promotes a Th1 Tumor-Infiltrating Lymphocyte (TIL) Phenotype*

The effect of adding a cancer vaccine to the combination bintrafusp alfa/SX-682 was first evaluated with CEA.Tg mice, where CEA is a self-antigen [23,24], bearing subcutaneous MC38-CEA tumors. To model a scenario where tumors do not respond to checkpoint-based immunotherapy, control feed or SX-682-containing feed were administered to mice starting on day 7, while administration of bintrafusp alfa at a low dose was delayed until day 14 to ensure response failure. In the vaccine treatment groups, mice were administered a priming vaccine dose of Ad-CEA on day 7 and a boosting dose of Ad-CEA/N-803 given on day 14 (hereafter designated "Vaccine"). As expected, the modified schedule of bintrafusp alfa plus SX-682 (Bintrafusp/SX) was unable to exert tumor control (Figure 1A). The use of vaccine as a monotherapy also failed to control tumors; the average tumor growth in the Vaccine group was statistically not different from that of the Control group (Figure 1A). Although the addition of vaccine to the Bintrafusp/SX therapy was able to induce a significant albeit modest delay in primary tumor growth in this experiment, the triple combination Vaccine/Bintrafusp/SX resulted in significant changes in the composition of the tumor immune infiltrate when compared with the other groups (Figure 1B). Overall, Vaccine/Bintrafusp/SX showed a significant enhancement of CD4<sup>+</sup> and CD8<sup>+</sup> T cells characterized by an effector and effector-memory phenotype (CD4Eff&Em and CD8Eff&Em TIL) above the levels achieved in the Vaccine monotherapy, Bintrafusp/SX, and Control groups (Figure 1B). Also remarkable was the ability of vaccine to decrease the percentage of regulatory T cells (Tregs) in the CD4<sup>+</sup> TIL population, compared to the Control and Bintrafusp/SX groups (Figure 1B). Previously, we demonstrated that Bintrafusp/SX therapy can significantly reduce tumor infiltration with suppressive granulocytic myeloid-derived suppressor cells (G-MDSC), defined as CD11b+F4/80−Ly6CloLy6G<sup>+</sup> , an effect attributed to the ability of SX-682 to block the CXCR1/2-mediated migration of G-MDSC into the tumor. The effect was not observed with monocytic MDSC, defined as

CD11b <sup>+</sup>F4/80 <sup>−</sup>Ly6G−Ly6C + . Here, CD11b <sup>+</sup>F4/80 <sup>−</sup>Ly6C loLy6G<sup>+</sup> cells were significantly reduced in the tumors of mice treated with both Bintrafusp/SX and Vaccine/Bintrafusp/SX, an effect that was not observed with CD11b <sup>+</sup>F4/80 <sup>−</sup>Ly6G−Ly6C + fractions (Figure 1C). Neither fraction of myeloid cells was altered in the spleen of mice in any of the treatment groups (Figure 1D). As shown in Figure 1E, only Vaccine/Bintrafusp/SX treatment induced a significant increase in the ratio of CD8 <sup>+</sup> TIL to both Tregs and CD11b <sup>+</sup>F4/80 <sup>−</sup>Ly6C loLy6G<sup>+</sup> cells in the TME compared to Control mice.

**Figure 1.** Vaccine synergizes with Bintrafusp alfa/SX-682 and increases TIL in MC38-CEA tumors. (**A**) CEA.Tg mice were injected s.c. with 3 × 10 <sup>5</sup> MC38-CEA in the flank. On day 7, mice were started on a control or SX-682 diet (200 mg/kg body weight/day), and on days 14, 17, and 21 mice received i.p. injections of 200 µg bintrafusp alfa. Priming vaccine dose of s.c. Ad-CEA (1 × 10 <sup>10</sup> viral particles) was administered on day 7 with a boosting dose of Ad-CEA/N-803 (1 × 10 <sup>10</sup> viral particles, N-803, 1 µg, s.c.) on day 14. Graph shows average tumor growth and error bars indicate SEM of biological replicates; *n* = 8 mice/group. \* *p* ≤ 0.05; \*\*\* *p* ≤ 0.001 for two-way ANOVA in (**A**). Control indicates mice that were left untreated and fed a base diet without SX-682. Tumors (**B**,**C**) and spleens (**D**) were harvested and analyzed by flow cytometry on day 23 for lymphocytes (**B**) and myeloid cells (**C**,**D**). (**E**) Cell ratios comparing the number of cells per mg tumor weight were also calculated. Individual points represent data from one tumor. ns, not significant; \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.001; \*\*\*\* *p* ≤ 0.0001 for one-way ANOVA followed by Tukey's post hoc test in (**B**–**E**). i.p. = intraperitoneal. s.c. = subcutaneous. TIL = tumor-infiltrating lymphocyte. Tregs = regulatory T cells.

To understand whether both bintrafusp alfa and SX-682 were needed for the antitumor efficacy of the combination Vaccine/Bintrafusp/SX, in the next study we also

evaluated the addition of vaccine to SX-682 (Vaccine/SX) or bintrafusp alfa alone (Vaccine/Bintrafusp). In this experiment, an additional boosting dose of vaccine was administered on day 21. While the growth of MC38-CEA tumors was not delayed with Vaccine/SX or Vaccine/Bintrafusp combinations, there was a significant delay in tumor growth in the Vaccine/Bintrafusp/SX group (Figure 2A). Interestingly, some tumors began to completely regress in the Vaccine/Bintrafusp/SX group immediately after the final dose of vaccine plus bintrafusp alfa administered on day 21. Sections of tumor tissue stained by immunofluorescence revealed high levels of infiltrating CD4 <sup>+</sup> and CD8 <sup>+</sup> T cells in the Vaccine/Bintrafusp/SX group that were distributed uniformly throughout the tumors, compared to the other groups (Figure 2B).

**Figure 2.** Vaccine combination immunotherapy is dependent on CD8 <sup>+</sup> TIL. (**A**) CEA.Tg mice were injected s.c. with 3 × 10 5 MC38-CEA in the flank. On day 7, mice were started on a control or SX-682 diet (200 mg/kg body weight/day). On days 14 and 21, mice received i.p. injections of 200 µg bintrafusp alfa. A priming vaccine dose of s.c. Ad-CEA (1 × 10 <sup>10</sup> viral particles) was administered on day 7 with a boosting dose of Ad-CEA/N-803 vaccine on days 14 and 21 (1 × 10 <sup>10</sup> viral particles, N-803, 1 µg, s.c.). Shown are the individual tumor growths for mice in the Control, Vaccine/SX, Vaccine/Bintrafusp, and Vaccine/Bintrafusp/SX groups; *n* = 7 mice/group. Control indicates mice that were left untreated and fed a base diet without SX-682. (**B**) Representative images of indicated tumors stained for CD4 + (green) and CD8 + (red) T cells and DAPI (blue) by immunofluorescence. (**C**) MC38-CEA tumor-bearing CEA.Tg mice received Vaccine/Bintrafusp/SX as in (**A**). Additionally, mice receiving Vaccine/Bintrafusp/SX also received depleting antibodies for CD8 + cells starting on day 5; *n* = 7 (Control and Vaccine/Bintrafusp/SX – CD8 Depleted) or 5 (Vaccine/Bintrafusp/SX) mice/group. (**D**) Flow profiles confirming efficacy of CD8 depletion antibodies from (**C**). Error bars indicate SEM of biological replicates. \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.001; \*\*\*\* *p* ≤ 0.0001 for two-way ANOVA in (**A**,**C**). i.p. = intraperitoneal. s.c. = subcutaneous. TIL = tumorinfiltrating lymphocyte.

The importance of the CD8<sup>+</sup> T-cell fraction for the effectiveness of the multimodal therapy was evaluated with CEA.Tg mice bearing MC38-CEA tumors that were either left untreated and fed a base diet without SX-682 (Control group), treated with Vaccine/Bintrafusp/SX multimodal therapy, or treated with multimodal therapy with simultaneous depletion of CD8<sup>+</sup> T cells (Vaccine/Bintrafusp/SX – CD8 Depleted group, Figure 2C,D). As shown in Figure 2C, depletion of CD8<sup>+</sup> T cells completely abrogated the antitumor efficacy of Vaccine/Bintrafusp/SX treatment. The triple combination also had a modest yet significant effect on the survival of MC38-CEA tumor-bearing mice over that of Bintrafusp/SX-treated or Control mice (Figure S2).

It has been previously reported that combination therapy consisting of vaccine and various immune modulatory agents, including immune checkpoint blockade, can enhance antitumor immunity by diversifying the number of tumor antigens being recognized by TCRs through the phenomenon of antigen cascade or epitope spreading [10]. In this study, splenocytes from Control and Vaccine/Bintrafusp/SX-treated mice were evaluated for potential epitope spreading by quantifying on an ELISPOT assay the number of CD8<sup>+</sup> T cells specific for CEA, the MC38-neoantigen PTGFR [10], or P15e, compared to a negative control peptide (Figure 3A). While there was a modest enhancement of the number of T cells specific for CEA in the spleens of vaccinated mice (~2-fold increase), high numbers of both PTGFR-specific and P15e-specific T cells were observed in the Vaccine/Bintrafusp/SX-treated mice, compared to the Control group (2.9-fold and 3.6-fold, respectively) (Figure 3A).

To understand how the combination of these agents restructures the immune profile of the TME in Vaccine/Bintrafusp/SX-treated tumors, NanoString gene expression analysis was performed on whole tumor tissue-derived RNA. Table 1 lists genes that were found to be up- or down-regulated more than 2.0-fold in Vaccine/Bintrafusp/SX-treated mice compared to Control tumors. Ingenuity Pathway Analysis demonstrated an upregulation of many immune-specific canonical pathways, with Th1 and Th2 being the two most significantly upregulated pathways (Figure 3B) in Vaccine/Bintrafusp/SX versus Control tumors. In addition, strong upregulation of inducible T-cell costimulator (ICOS) signaling, nuclear factor of activated T cells (NFAT) regulation, CTL-mediated apoptosis of target cells, and CD28 signaling were observed in tumors treated with the multimodal therapy Vaccine/Bintrafusp/SX versus Control. Figure 3C shows genes that were up- or downregulated >2.5-fold in the triple combination group, with some of them being confirmed by PCR analysis in tumors of mice treated with Vaccine/Bintrafusp/SX versus Control (Figure 3D). There was a significant upregulation of Cd3e, Cd8a, Tbx21, Pdcd1, Cd247, and genes encoding for the effector molecules, Prf1, Gzmb, and Gzmk, suggesting a highly cytotoxic phenotype in TIL isolated from Vaccine/Bintrafusp/SX-treated tumors. Additional PCR analysis of expression of CD8a, Tbx21, Gzmk, and Prf1 mRNA was conducted in individual tumors from the Control, Vaccine, Bintrafusp/SX and Vaccine/Bintrafusp/SX groups. While vaccine used as monotherapy induced only a modest upregulation of these genes in some of the tumors compared with Control tumors, a stronger upregulation was observed in the Bintrafusp/SX group, though the level of upregulation was variable among genes and across tumor samples (Figure 3E). Supporting the benefit of adding all agents together, tumors in the Vaccine/Bintrafusp/SX group exhibited a more robust upregulation of all four genes in the majority of samples evaluated (Figure 3E). These data indicated that addition of vaccine can further enhance immune infiltration and activation above the induction mediated by blockade of PD-L1, TGF- β and CXCR1/2.

**Figure 3.** Immune activation signature observed in MC38-CEA tumors treated with Vaccine/Bintrafusp/SX combination. CEA.Tg mice were injected s.c. with 3 × 10 <sup>5</sup> MC38-CEA in the flank. On day 7, mice were started on a control or SX-682 diet (200 mg/kg body weight/day). On days 14 and 21, mice received i.p. injections of 200 µg bintrafusp alfa. A priming vaccine dose of s.c. Ad-CEA was administered on day 7 (1 × 10 <sup>10</sup> viral particles) with a boosting dose of Ad-CEA/N-803 vaccine on days 14 and 21 (1 × 10 <sup>10</sup> viral particles, N-803, 1 µg, s.c.). (**A**) IFNγ ELISPOT analysis of spleens collected on day 24 from Control and Vaccine/Bintrafusp/SX-treated mice against MC38-CEA tumor antigens. Control indicates mice that were left untreated and fed a base diet without SX-682; n = 7 (Control) or 5 (Vaccine/Bintrafusp/SX) mice/group. Tumors collected on day 24 were used for RNA preparation and NanoString analysis as described in the Materials and Methods. Shown in (**B**) is an Ingenuity Pathway Analysis performed on genes that were found to be up- or down-regulated more than 2-fold in Vaccine/Bintrafusp/SX-treated tumors compared to Control tumors; *n* = 3 mice/group. (**C**) Heat map of genes differentially expressed >2.5-fold in Vaccine/Bintrafusp/SX-treated tumors compared to Control tumors; *n* = 3 mice/group. (**D**) Real-time PCR analysis confirming selected genes upregulated in Vaccine/Bintrafusp/SX-treated tumors compared to Control tumors; *n* = 3 (Control) or 4 (Vaccine/Bintrafusp/SX) mice/group. Individual points represent data from one tumor. ns, not significant; \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.001 for two-tailed Student's *t*-test in (**A**,**D**). (**E**) Heat map expression of indicated genes in MC38-CEA tumors treated as per the schedule of administration in Figure 1. Tumor RNA was prepared at day 23; RNA expression of indicated genes was evaluated by real-time PCR as described in the Materials and Methods.


**Table 1.** Genes that were found to be up- or down-regulated more than 2.0-fold in Vaccine/ Bintrafusp/SX-treated mice compared to Control tumors.

#### *3.2. Addition of Vaccine to Checkpoint Blockade-Based Therapy Enhances Immune T-Cell Activation and TCR Diversity*

To corroborate the results in a different tumor model, a single dose of bintrafusp alfa in combination with SX-682 was given to 4T1 tumor-bearing mice which, as expected, failed to control tumor growth (Bintrafusp/SX, Figure 4A). In this mammary carcinoma model, vaccine was administered as a priming dose of Ad-Twist on day 7 with a boosting vaccine on day 14 consisting of Ad-Twist plus N-803. Twist1, a transcription factor that drives metastasis, was identified and characterized as a targetable "self" tumor-associated antigen in 4T1 tumor cells [22]. Addition of vaccine to Bintrafusp/SX therapy induced only a modest delay in primary tumor growth (Vaccine/Bintrafusp/SX, Figure 4A), and a trend towards reduced number of lung metastases (Figure 4B), with a 76% reduction of metastases in the Vaccine/Bintrafusp/SX group compared with the Control (Figure 4C). Two caveats with these results, however, are the low number of mice evaluated in each

group, and the reduction of primary tumor volume in the Vaccine/Bintrafusp/SX group that could directly impact the number of disseminated cells.

**Figure 4.** Vaccine synergizes with Bintrafusp alfa and SX-682 and increases TIL in 4T1 tumors. (**A**) BALB/c mice bearing 4T1 tumors in the mammary fat pad received control or SX-682 diet on day 7 (200 mg/kg body weight/day), with a priming vaccine dose of s.c. Ad-Twist (1 × 10 <sup>10</sup> viral particles). On day 14, mice received an i.p. injection of 200 µg bintrafusp alfa with a boosting vaccine dose of Ad-Twist/N-803 (1 × 10 <sup>10</sup> viral particles, N-803, 1 µg, s.c.). Graph shows average tumor growth and error bars indicate SEM of biological replicates; *n* = 6 (Control) or 7 (Bintrafusp/SX, Vaccine/Bintrafusp/SX) mice/group. Control indicates mice that were left untreated and fed a base diet without SX-682. \* *p* ≤ 0.05; \*\*\* *p* ≤ 0.001 for two-way ANOVA. (**B**) Number of metastases quantified in the lungs of 4T1 tumor-bearing mice on day 21; individual points represent data from one mouse. (**C**) Table depicting the number and percentage of mice with the indicated range of lung metastases in each group, the mean number of metastases in each group, and the % reduction of the mean in each group vs. the Control group. Data are pooled from 2 independent experiments. (**D**) Tumors were harvested and analyzed by flow cytometry on day 21. Individual points represent data from one tumor. \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.001; \*\*\*\* *p* ≤ 0.0001 for one-way ANOVA followed by Tukey's post hoc test. i.p. = intraperitoneal. s.c. = subcutaneous. TIL = tumor-infiltrating lymphocyte.

Similar to the results observed with MC38-CEA tumors, addition of vaccine had a marked impact on the composition of 4T1 primary tumor T-cell infiltrates. As shown in Figure 4D, flow cytometry analysis of tumors collected at 1 week post-bintrafusp alfa ± vaccine administration (day 21 post-tumor injection) revealed significantly higher frequencies of CD8 <sup>+</sup> T cells characterized by an effector and effector-memory phenotype (CD8Eff&EM) in the Vaccine/Bintrafusp/SX group compared with the Bintrafusp/SX group or Control tumors. In contrast, the frequency of CD4 <sup>+</sup> T cells and central memory CD8 <sup>+</sup> T cells (CD8CM) were similar among the two treatment groups, irrelevant of vaccine. In agree-

ment with the flow cytometry data, immunofluorescence-based analysis of TIL in sections of Formalin-Fixed Paraffin-Embedded (FFPE) tumor tissues (Figure S3A) showed large clusters of CD4<sup>+</sup> and CD8<sup>+</sup> T cells homogenously distributed throughout the tumor in Vaccine/Bintrafusp/SX-treated tumors and not solely contained to the tumor boundaries. Consistent with previous findings, immune subset profiling of Vaccine/Bintrafusp/SXtreated tumors also revealed a significant decrease in the frequency of tumor-infiltrating CD11b+F4/80-Ly6G+Ly6Clo myeloid cells and CD11b+F4/80hi macrophages, together with a marked increase of CD4<sup>+</sup> and CD8<sup>+</sup> T cells (Figure S3B). Additionally, no adverse events or toxicity were observed with the total combination of therapeutics. These results suggested that addition of a prime-boost vaccine to a checkpoint blockade-based immunotherapy can further enhance frequency of effector T lymphocytes in the TME.

The quality of the T-cell infiltrates in 4T1 tumors of Bintrafusp/SX ± vaccine-treated mice was further evaluated. Intracellular flow cytometry-based analysis of tumor-infiltrating T cells from Vaccine/Bintrafusp/SX-treated mice revealed significantly higher frequencies of proliferative (CD8<sup>+</sup> Ki67<sup>+</sup> ) and cytotoxic (CD8<sup>+</sup> Granzyme B<sup>+</sup> ) TIL compared to tumors in the Bintrafusp/SX and Control groups (Figure 5A). TCRβ sequencing analysis was also performed on whole tumor lysates from 3 individual tumors per group; addition of vaccine to Bintrafusp/SX resulted in reduced clonality (Figure 5B) and expanded the T-cell repertoire compared with Control and Bintrafusp/SX-treated tumors, with an average of 481 ± 240, 907 ± 372, and 1897 ± 1469 productive TCRβ rearrangements in the Control, Bintrafusp/SX and Vaccine/Bintrafusp/SX groups, respectively (Figure 5C).

In addition, analysis of sequence similarities revealed a higher number of TCRβ sequences shared among tumors in the Vaccine/Bintrafusp/SX > Bintrafusp/SX > Control group, as shown by the numbers in the regions of intersection. Analysis of the top 25% of TCRβ sequences present in tumors from 3 mice in each group revealed a more diversified TCR repertoire in the Vaccine/Bintrafusp/SX-treated mice (Figure 5D) comprising 21, 17, and 13 clones per individual, while tumors from Control and Bintrafusp/SX-treated mice contained 5, 7, 6 and 6, 18, and 3 different TCRβ clones, respectively. These data indicated that the addition of a vaccine consisting of Ad-vector plus N-803 adjuvant to bintrafusp alfa plus SX-682 therapy has the potential to increase the proliferation and cytotoxic functionality of tumor-infiltrating CD8<sup>+</sup> T cells, while promoting a more diversified TCR repertoire in the tumor (Figure 6).

**Figure 5.** Vaccine enhances activation and TCR diversity of TIL when incorporated into combination immunotherapy in the 4T1 carcinoma model. (**A**) BALB/c mice bearing 4T1 tumors in the mammary fat pad received control or SX-682 diet on day 7 (200 mg/kg body weight/day), with a priming vaccine dose of s.c. Ad-Twist (1 × 10 <sup>10</sup> viral particles). On day 14, mice received an i.p. injection of 200 µg bintrafusp alfa with a boosting vaccine dose of Ad-Twist/N-803 (1 × 10 <sup>10</sup> viral particles, N-803, 1 µg, s.c.). Graphs show immune subsets determined by flow cytometry analysis of tumors at day 21. Individual points represent data from one tumor. \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.001 for one-way ANOVA followed by Tukey's post hoc test. (**B**) Simpson clonality score for individual tumor samples in each indicated group determined as indicated in the Materials and Methods. (**C**) Number of productive TCRβ rearrangements per individual tumor in the indicated groups, showing the number of overlapping TCRβ sequences among individuals. (**D**) The number of TCRβ clones comprising the top 25% of detected sequences. *n* = 3 mice/group. i.p. = intraperitoneal. s.c. = subcutaneous. TCR = T-cell receptor. TIL = tumor-infiltrating lymphocyte.

**Figure 6.** Schematic representation of the mechanism of action of the combination Ad5-vaccine, N-803, Bintrafusp alfa and SX-682. G-MDSC = granulocytic myeloid-derived suppressor cells. TCR = T-cell receptor. TIL = tumor-infiltrating lymphocyte. Tregs = regulatory T cells.

#### **4. Discussion**

In this study, we demonstrate the effect of adding a cancer vaccine to immune checkpoint blockade therapy. Our data show that a vaccine consisting of a recombinant adenovirus with a target antigen transgene coupled with an IL-15 super agonist adjuvant is able to contribute to checkpoint-based immunotherapy by increasing T-cell migration to the tumor, enhancing T-cell activation and cytotoxicity, and promoting TCR diversity and antigen cascade.

The mechanism of action and immunological benefits of both bintrafusp alfa and SX-682 have been extensively studied as monotherapies and in combination by our group and others. Bintrafusp alfa, designed as a checkpoint inhibitor and to "trap TGF-β" in the TME, has been shown to promote T- and NK-cell killing of tumor cells, promote antibodydependent cell cytotoxicity, revert TGF-β-induced epithelial-mesenchymal phenotypic changes in cancer cells (tumor cell plasticity), and delay tumor growth in numerous mouse models of cancer [15,25–27]. There are numerous ongoing clinical studies of bintrafusp alfa in patients with a variety of cancer types, with several of these studies investigating its use in combination with other immunotherapies, chemotherapy or radiation [15]. SX-682 is a small molecule inhibitor that allosterically binds to the CXCR1 and CXCR2 receptors to irreversibly inhibit downstream signaling from CXC family ligands CXCL1-3 and CXCL5-8. One of the most notable CXCR1/2 ligands, IL-8 (CXCL8), is a known inducer of tumor cell plasticity, attractant of suppressive myeloid-derived suppressor cells to the tumor, and correlates with failure of treatment in numerous cancer types, including failure to checkpoint inhibitor therapy [28–31]. SX-682 has been shown to inhibit tumor growth, block migration of G-MDSC to tumors in vivo, and decrease markers of tumor cell plasticity in human xenografts and murine tumors [17,32,33], and is currently undergoing clinical evaluation in several clinical trials [29]. In a previous study, we demonstrated that the combination of bintrafusp alfa and SX-682 reduces mesenchymal tumor features and increases epithelial protein expression in murine models of breast and lung cancer, reduces tumor infiltration with G-MDSC, and enhances T-cell infiltration and activation in tumors [17].

Tumor immunologists have been attempting to develop highly specific yet off-theshelf immune activating vaccines for the treatment of cancer patients prior to the immune checkpoint blockade revolution. These vaccines often targeted tumor-associated antigens and were combined with immune-activating adjuvants or costimulatory molecules to promote T-cell infiltration into tumors and kick-start antitumor immunity [8,34,35]. More

recent studies have also found efficacy with the use of neoantigen-based vaccines and irradiated cancer cell vaccines. However, the subsequently activated T-cell population can still be rapidly inhibited by immune checkpoint pathways or immune suppressive cells once arriving to the tumor. Additionally, many tumor types with low degree of T-cell infiltration which respond poorly to immunotherapy such as pancreatic, colon, and prostate cancers upregulate additional immune suppressive mechanisms including TGFβ, MDSC, and mesenchymal features [36–39]. In this study, we lowered the dose and delayed the administration of bintrafusp alfa in combination with SX-682 with the idea of preventing antitumor activity to mimic the situation of non-responsive tumors. We were able to demonstrate that the addition of vaccine in this context promoted further T-cell infiltration and activation, and enhanced TCR diversity in the tumor above what was induced by bintrafusp alfa/SX-682 treatment (Figure 6). We also showed here that addition of vaccine further enhanced the expression of genes indicative of immune activation and T-cell infiltration in the TME (CD8a, Tbx21, Gmzk, Prf1). These data are in agreement with the flow cytometric analysis of MC38-CEA tumors, which demonstrated an increased number of infiltrating CD4<sup>+</sup> effector/effector memory T cells as well as CD8<sup>+</sup> effector/effector-memory T cells in Vaccine/Bintrafusp/SX-treated tumors versus tumor in the Bintrafusp/SX group. Similarly, infiltration with CD8<sup>+</sup> effector/effector-memory T cells was significantly enhanced in 4T1 tumors treated with Vaccine/Bintrafusp/SX versus Bintrafusp/SX treatment. Additionally, increased proliferation and cytolytic effect of T cells was observed in the TME of Vaccine/Bintrafusp/SX-treated 4T1 tumors, denoted by a higher percentage of CD8<sup>+</sup> T cells positive for Ki67 or Granzyme B, compared with tumors in the Bintrafusp/SX group.

Analysis of splenocytes via ELISPOT assay also revealed epitope spreading in the Vaccine/Bintrafusp/SX-treated mice, with an increase in the number of T cells specific for antigens found in the tumor but not in the vaccine (PTGFR and P15e), compared with the Control group. One could hypothesize that these activated, tumor-specific T cells from spleens of Vaccine/Bintrafusp/SX-treated mice could mediate some degree of tumor control if adoptively transferred into MC38-CEA tumor-bearing mice; however, such experiments would not be able to reveal the full potential of this combination immunotherapy, which relies on tumor-localized effects mediated by SX-682 and bintrafusp alfa. As we have previously shown, inhibition of CXCR1/2 via SX-682 significantly reduces the migration of suppressive CXCR2<sup>+</sup> G-MDSC into tumors. At the same time, SX-682 directly affects the phenotype of the tumor cells resulting in reduced mesenchymal features which, in turn, improves tumor susceptibility to immune-mediated lysis [17]. Similarly, bintrafusp alfa is able to mediate neutralization of PD-L1 and TGF-β in the TME, leading to alleviation of local tumor immunosuppression mediated by both pathways, including the reversion of tumor mesenchymal features for improved susceptibility to immune attack [15,17].

Despite increased infiltration of tumors with activated T cells and increased numbers of tumor-specific T cells in the Vaccine/Bintrafusp/SX group, the treatment schedules investigated here did not result in a significant number of tumor cures. We hypothesize that this could have been due to various factors, including the limited therapeutic window in which the human drugs employed here could be administered to immune competent mice without production of anti-drug antibodies. Another possibility is the very rapid tumor growth characteristic of the two murine models utilized in this study, combined with a delayed initiation of therapy, which limited time for treatment. Notably, in the clinical setting, multiple agents can be administered continuously with optimal dosing over an extended period of time for maximum benefit, as in the case of the combination of Adenoviral-based vaccines, N-803, and bintrafusp alfa currently being tested in the clinic [40]. Alternatively, other mechanisms of immune suppression may have limited tumor control in the combination group, even in the presence of activated, infiltrating T cells. Interestingly, one of the genes most upregulated in MC38-CEA tumors treated with Vaccine/Bintrafusp/SX was Ido1, suggestive of the possibility of adding an IDO inhibitor to this therapeutic regimen. Overall, the combination Vaccine/Bintrafusp/SX therapy was more effective at controlling MC38 compared with 4T1 tumor growth, an effect that could be related to the higher mutational burden and neoepitope expression in MC38 versus 4T1 tumors.

In conclusion, this study highlights the mechanistic synergy between vaccine and combination checkpoint immunotherapy and provides rationale for an ongoing clinical trial combining a cancer vaccine with bintrafusp alfa plus SX-682 therapy in patients with advanced solid tumors (NCT04574583).

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-6 694/13/5/968/s1, Figure S1: Optimization of the combination Ad-CEA plus N-803, Figure S2: Survival of CEA.Tg mice bearing MC38-CEA tumors in response to indicated treatments, Figure S3: Multimodal therapy effect on 4T1 tumor immune cell infiltration.

**Author Contributions:** L.A.H., D.H.H., J.W.H., C.P. and J.S. conceived various aspects of the project, designed experiments, and interpreted the results. L.A.H., K.F. and D.H.H. were responsible for performing experiments, data compilation and analysis. J.A.Z. and D.Y.M. provided reagents and in vivo dosing and formulation recommendations. L.A.H., C.P. and J.S. were responsible for manuscript writing and coordination. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Intramural Research Program of the Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH), as well as through Cooperative Research and Development Agreements (CRADA) between the NCI/NIH and Syntrix Pharmaceuticals, the NCI/NIH and EMD Serono, and the NCI/NIH and ImmunityBio.

**Institutional Review Board Statement:** Mice were maintained under pathogen-free conditions in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care guidelines. All animal studies were approved by the NIH Intramural Animal Care and Use Committee (ACUC); protocol LTIB-038.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study will be provided upon reasonable request.

**Acknowledgments:** The authors thank Haiyan Qin for her technical assistance with animal studies, Masafumi Iida for help with tumor collections, and Debra Weingarten for editorial assistance in the preparation of this manuscript.

**Conflicts of Interest:** The NCI/NIH authors do not have any competing interests to disclose. J.A.Z. and D.Y.M. are paid employees of Syntrix Pharmaceuticals. The NCI/NIH has ongoing Collaborative Research and Development Agreements (CRADA) with Syntrix Pharmaceuticals, EMD Serono, and ImmunityBio.

#### **References**


## *Article* **Facile Generation of Potent Bispecific Fab via Sortase A and Click Chemistry for Cancer Immunotherapy**

**Xuefei Bai <sup>1</sup> , Wenhui Liu 1,2, Shijie Jin <sup>1</sup> , Wenbin Zhao <sup>1</sup> , Yingchun Xu <sup>1</sup> , Zhan Zhou 1,3, Shuqing Chen 1,3,4,\* and Liqiang Pan 1,5,\***

	- <sup>2</sup> Hangzhou Biosun Pharmaceutical Co., Ltd., Liangzhu International Life Science Town, 268 Tongyun Street, Yuhang District, Hangzhou 310015, China

**Simple Summary:** The formats of bispecific antibody have been investigated for many years to enhance the stability of the structure and anti-tumor efficacy. One of the formats combining two Fabs at their C termini provides unmodified variable region and comparable activity to other fragmentbased bispecific antibodies that are usually combined in a head-to-tail manner. However, the current strategy to produce the BiFab molecule is limited to a semisynthetic method that introduces unnatural amino acid to antibodies' sequences during production. To improve the application of BiFab format in investigational biodrugs, we have applied sortase A-mediated "bio-click" chemistry to generate BiFab, for facile assembly of Fab molecules that have been expressed and stored as BiFab module candidates. The BiFabs made by our method stimulate T cell proliferation and activation with favorable in vitro and in vivo anti-tumor activit. Our results indicate that BiFab made by sortase A-mediated click chemistry could be used to efficiently generate various BiFabs with high potency, which further supports personalized tumor immunotherapy in the future.

**Abstract:** Bispecific antibodies (BsAbs) for T cell engagement have shown great promise in cancer immunotherapy, and their clinical applications have been proven in treating hematological malignance. Bispecific antibody binding fragment (BiFab) represents a promising platform for generating non-Fc bispecific antibodies. However, the generation of BiFab is still challenging, especially by means of chemical conjugation. More conjugation strategies, e.g., enzymatic conjugation and modular BiFab preparation, are needed to improve the robustness and flexibility of BiFab preparation. We successfully used chemo-enzymatic conjugation approach to generate bispecific antibody (i.e., BiFab) with Fabs from full-length antibodies. Paired click handles (e.g., N<sup>3</sup> and DBCO) was introduced to the C-terminal LPETG tag of Fabs via sortase A mediated transpeptidation, followed by site-specific conjugation between two click handle-modified Fabs for BiFab generation. Both BiFabCD20/CD3 (EC<sup>50</sup> = 0.26 ng/mL) and BiFabHer2/CD3 exhibited superior efficacy in mediating T cells, from either PBMC or ATC, to kill target tumor cell lines while spared antigen-negative tumor cells in vitro. The BiFabCD20/CD3 also efficiently inhibited CD20-positive tumor growth in mouse xenograft model. We have established a facile sortase A-mediated click handle installation to generate homogeneous and functional BiFabs. The exemplary BiFabs against different targets showed superior efficacy in redirecting and activating T cells to specifically kill target tumor cells, demonstrating the robustness of sortase A-mediated "bio-click" chemistry in generating various potent BiFabs. This approach also holds promise for further efficient construction of a Fab derivative library for personalized tumor immunotherapy in the future.

**Citation:** Bai, X.; Liu, W.; Jin, S.; Zhao, W.; Xu, Y.; Zhou, Z.; Chen, S.; Pan, L. Facile Generation of Potent Bispecific Fab via Sortase A and Click Chemistry for Cancer Immunotherapy. *Cancers* **2021**, *13*, 4540. https://doi.org/10.3390/ cancers13184540

Academic Editors: Subree Subramanian and Xianda Zhao

Received: 14 August 2021 Accepted: 6 September 2021 Published: 10 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Keywords:** bispecific antibody; sortase A; chemo-enzymatic approach; anti-CD20 antibody; Fab; BiFab

#### **1. Introduction**

Immunotherapies, such as chimeric antigen receptor T cells (CAR-Ts) and T-cellengaging bispecific antibodies (T-BsAbs), have revolutionized cancer treatments by leveraging the immune system [1,2]. T-BsAbs usually refer to bifunctional antibodies with one arm targeting T cell receptors (e.g., CD3) to engage T cells and another arm targeting antigen on tumor cells, for the purpose of bridging and redirecting T cells to tumor cells. Compared with CAR-T cells, which are autologous T lymphocytes that are genetically engineered to express chimeric antigen receptor for specific tumor cell targeting [3], bispecific antibody can be produced relatively easier and provide off-the-shelf treatment [4,5]. This strategy has generated great interest with more than 50 T-BsAb candidates in clinical trials for a range of indicators nowadays [6].

One representative T-BsAb is Blinatumomab, a bi-specific T cell engager (BiTE) targeting CD19 and CD3 that was approved by FDA in 2014 for the treatment of Acute Lymphoblastic Leukemia (ALL). The flexible tandem arrangement of this single chain bispecific antibody accounts for its superior efficacy in inducing lytic synapse and thereby high T cell activity in comparison with its IgG-based and Fab-based format [7,8]. Despite the high efficacy, BiTE molecule has a very short half-life of ~2 h in blood circulation in the absence of Fc domain [9]. In order to increase stability and activity of fragment-based T-BsAbs, Dual-Affinity Re-Targeting (DART®) protein and tandem diabody (TandAb) were designed to further improve the half-life and stability in vivo [10–13]. However, the variable region spanning engineered constant scaffold might result in the loss of affinity and stability for Fc-free T-BsAbs, such as single-chain variable fragment (scfv) molecules [14,15]. For example, variable regions assembled to a format that deviate significantly from its cognate high stable IgG might compromise its affinity, especially when the N-terminus of Fvs have additional polypeptide chains that function as linkers [14,16,17].

Bispecific antibody binding fragment (BiFab) represents another promising platform for generating bispecific antibodies. Two Fab fragments providing different binding specificities are usually chemically linked in a tail-to-tail manner to generate BiFab. The intact structure of Fab fragments is parallelly grafted into the BiFab format, which maintains a natural association of four domains (VL, CL, VH and CH1) and thus ensures stability [14,18,19]. The BiFabs could also avoid Fc-related side effects since they lack a Fc region. However, the site-specific conjugation of two Fab molecules remains challenging during BiFab preparation [20]. One of the well-known chemical approaches for BiFab generation is the application of click chemistry, in which the click handle is installed through the introduced noncanonical amino acid (ncAA) on Fabs, to realize site-specific conjugation of two Fabs [21–23].

To achieve site-specific click handle installation, an alternative approach is sortase A-mediated transpeptidation. Sortase A is a bacterial enzyme that recognizes C-terminal LPXTG motif (X represents any amino acid) of proteins or peptides, which is used to anchor building blocks of cell walls of Gram-positive bacteria. The enzyme cleavages between Thr and Gly residues and then yields an acyl-enzyme intermediate. Subsequently, the nucleophilic primary amine of oligo-glycine modified substrates resolved the intermediate and then form a covalent bond between oligo-glycine modified substrates and LPETG-tagged protein [24–27]. Therefore, the paired click handles could be modified with oligo-glycines, such as GGG, before installation to the C-terminus of the target protein (e.g., Fab). Herein, we applied sortase A-mediated two-step chemo-enzymatic conjugation to generate BiFabs. The paired click handles that comprising azide and dibenzocyclooctyne function groups was firstly attached to the Fabs by sortase A mediated transpeptidation between LPETG-tagged Fab and click chemistry-functionalized GGG, and subsequently the Fab-linkers are conjugated via click chemistry to form BiFabs. Using this strategy, we

successfully constructed homologous BiFabCD20/CD3 and BiFabHer2/CD3. We have demonstrated the potent in vitro and in vivo efficacy of BiFabCD20/CD3, and its ability to stimulate resting PBMC to proliferate and degranulate. In addition, functional BiFabHer2/CD3 was generated by simply replacing FabCD20 arm with FabHer2, further suggesting the potential of this chemo-enzymatic approach on preparing various BiFabs based on prestored Fab derivative library.

#### **2. Materials and Methods**

#### *2.1. Reagents and Cell Lines*

The human CD20-positive cell lines Ramos, Raji, Daudi and the human CD20-negative cell line K562 were purchased from the American Type Culture Collection (ATCC, San Francisco, CA, USA), and were cultured in 1640 medium (Gibco) with 10% fetal bovine serum (FBS, Gibco). The human HER2-positive cell line SK-OV-3 and HER2-negtive cell line MDA-MB-468 were purchased from ATCC and were cultured in McCoy's 5A or DMEM (Gibco) with 10% FBS, respectively. The expression plasmids of the full-length anti-CD20 antibody Ofatumumab and sortase A enzyme were constructed in our laboratory [28]. The HEK-293F cell line was from Qilin Zhang's laboratory in Tsinghua University. The HEK293F cells were grown in 250 mL SMM-293-TI medium (Sinobiological, Beijing, China) supplemented with 100 U/mL ampicillin, 100 µg/mL streptomycin (Sorlabio), and 1% FBS and the cells were shaking cultured at 37 ◦C and 210 rpm (Eppendorf). Anti-CD3 Fab sequence was derived from the humanized OKT3 antibody [29]. Anti-Her2 Fab sequence was derived from the Trastuzumab [30].

Triple glycine-modified linker Gly3-(PEG)3-N<sup>3</sup> (GPN) were synthesized by Concortis (San Diego, CA, USA). Triple glycine-modified linker Gly3-(PEG)4-dibenzocyclooctyne (DBCO) (GPD) was purchased from Lumiprobe (Hunt Valley, MD, USA).

#### *2.2. Sortase A-Mediated Click Handle Installation*

We previously showed that sortase A was used to specifically conjugate LPETG tagged IgG with GGG modified toxins [24], and the enzyme was kept by our lab. Briefly, we used a sortase A mutant (△N59) derived from *Staphylococcus aureus*, which is subcloned into pET28a(+) before a six Histidine polypeptide (His6). The expression vector of sortase A was then transfected into BL21 (DE3) Competent Cells (Sangon, Shanghai, China) and the expression is induced by 0.5 M IPTG for 16 h. After incubation, cells were harvested and disrupted by French Press (ThermoFisher Scientific Inc., Shanghai, China). The soluble fraction was collected and purified by Ni-NTA (HiTrap Ni-NTA column, GE) with instruction of the manufacturer's protocol. The purified sortase A protein was buffer exchanged to 50 × 10−<sup>3</sup> M Tris–HCl (pH 7.5), 150 × 10−<sup>3</sup> M NaCl by ultrafiltration (Amicon Ultra-10k, Millipore, MA, USA), sterile filtered and stored at −80 ◦C. Sequences of light chain and heavy chain of antibody fragments (Fabs) were, respectively, inserted into pMH3 expression vector behind human signal peptide sequence, and Fabs of heavy chain were C-terminally tagged with nucleotide sequence that express polypeptide GGGGSGGGGSGGGGS-LPETG-6 × His ((G4S)3-LPETG-His6). G4S linker was used to facilitate sortase A mediated transpeptidation. The expression vector of Fabs was transiently expressed in HEK293F cells for 3–4 days.

To optimize the reaction conditions of the sortase A-mediated conjugation, the reaction molar ratio of antibody fragments to glycine modified linkers (e.g., GPD and GPN) was explored. The reaction molar ratios (1:25 and 1:50) and different reaction time (6 h, 12 h or 24 h) at 37 ◦C were investigated in reaction buffer (50 mM Tris-HCl, 150 mM NaCl, 5 mM CaCl2, pH 7.4) solution in the presence of 50 µM sortase A enzyme (the molar ratio of sortase A/Fab was 1:8.3). To evaluate the conjugation efficiency, the reverse-phase high pressure liquid chromatography (RP-HPLC) with a Varian PLRP-S 100 ´Å column was used as previously described [28,31]. The conjugation reaction was scaled up under optimal reaction condition. Since the His tag was cut off by sortase A during transpeptidation, the flow-through fluid containing modified Fabs (e.g., FabCD3-DBCO and FabCD20-N3) was collected during HiTrap Ni-NTA affinity chromatography. All modified Fabs were buffer exchanged to PBS (pH 7.4) by ultracentrifugation (Millipore Amicon Ultra Filters, 10 kDa cut-off).

#### *2.3. Click Chemistry Mediated Generation of Bispecific Fab (BiFab)*

The copper-free click reaction between Fab-GPN and Fab-GPD was reacted in a buffered solution contained 50 mM Tris-HCl, 150 mM NaCl (pH 7.4). FabCD3-DBCO was reacted with FabCD20-N<sup>3</sup> or FabHer2-N<sup>3</sup> at a molar ratio of 1:1 at 4 ◦C for 12 h. After reaction, BiFabs were purified from free Fab by size exclusion chromatography (SEC) (Superdex 200 increase 10/300 GL, GE) on AKTA purifier (Amersham Biosciences, MA, USA). Sample from each peak was analyzed by SDS-PAGE under reducing condition and non-reducing condition. The purified protein from SEC was also analyzed by RP-HPLC with the following condition, a linear gradient elution starting from 75% buffer A (1.5 M (NH4)2SO4, 25 mM Na3PO4, pH 7.0), 25% buffer B (25 mM Na3PO4, pH 7.0) and 0% isopropanol, to 0% buffer A (1.5 M (NH4)2SO4, 75 mM Na3PO4, pH 7.0), 75% buffer B (25 mM Na3PO4, pH 7.0) and 25% isopropanol.

#### *2.4. Flow Cytometry*

All flow cytometry studies were conducted on ACEA NovoCyteTM (ACEA Biosciences Inc., San Diego, CA, USA). Data were processed with FlowJo 10.1 (FlowJo, LLC, Ashland, OR, USA) and Prism 8.0.1 (GraphPad Software Inc., San Diego, CA, USA).

To evaluate the binding ability of BiFabCD20/CD3, 1 × 10<sup>6</sup> CD20-positive cells or 1 × 10<sup>6</sup> CD3-positive Jurkat cells were incubated with serial concentrations of FabCD20 , FabCD3 and BiFabCD20/CD3 in ice-cold PBS (pH 7.4) for 30 min, followed by incubation with the primary anti-human IgG-Fab fragment (Abcam, Cambridge, UK) for 30 min. After washing three times with cold PBS (pH 7.4), cells were incubated with secondary goat anti-mouse IgG-FITC (Beyotime, Shanghai, China) for 30 min. After washing step, immune-stained cells were analyzed by flow cytometry.

#### *2.5. Preparation of Active T Cells (ATC) from Peripheral Blood Mononuclear Cells (PBMC)*

Human blood samples were obtained from healthy volunteers. PBMC were extracted from fresh blood samples by density centrifugation (Ficoll-Paque) following manufacturer's instruction.

PBMC were stimulated with Dynabeads™ Human T-Activator CD3/CD28 (Thermo Fisher) for T cell expansion and activation to generate active T cells (ATC). Briefly, PBMC were mixed with dynabeads at a cell-to-bead ratio of 1:1, and co-incubated for 4 days in the presence of 30 U/mL recombinant IL-2.

#### *2.6. Cell Apoptosis*

PBMCs were used as effector cells in all experiments. For LDH releasing assay, 96-well plates were seeded with 3 × 10<sup>4</sup> tumor cells (e.g., Ramos or Daudi cells) and 6 × 10<sup>4</sup> ATC per well, and then added with serial concentrations of BiFabs for a 24 h incubation at 37 ◦C. After incubation, the release of the intracellular enzyme lactate dehydrogenase (LDH) was determined by LDH cytotoxicity assay kit (Beyotime, Shanghai, China) to measure cell death. The percentage of necrotic cells was calculated according to the absorbance of each well at 450 nm.

For flow cytometry studies on cell apoptosis, ATC were prestained with Carboxyfluorescein succinimidyl ester (CFSE), and then co-cultured with 2 × 10<sup>5</sup> tumor cells at an effector: target (E:T) ratio of 2:1 for 24 h. When PBMC were used as effector cells, the E:T ratio was 5:1. Cells were then stained with Annexin-Cy5/Propidium Iodide (PI), and the percentage of apoptotic and necrotic cells were determined by flow cytometry.

#### *2.7. T Cell Activation*

CD69 and CD25 are early and late activation markers for T cells, respectively. We therefore used flow cytometry to evaluate T cell activation via measuring cell surface CD69 and CD25 expression. Fresh PBMC were mixed with target tumor cells (e.g., Ramous, Raji, Daudi and K562 cells) at E: T ratio of 5:1 before adding serial concentrations of BiFabs (BiFabHer2/CD3 or BiFabCD20/CD3) to initiate specific killing, and the co-incubation lasted 48 h. Naïve T cells were labeled with FITC-αCD4 and FITC-αCD8, and active T cells were further labelled with APC-αCD25 and PE-αCD69 (BD Biosciences). When CD20 positive tumor cells were used as target cells, fresh PBMC were pre-treated with anti-CD20 antibody-coated magnetic beads to deplete CD20-positive B cells.

Enzyme-linked immunosorbent assay (ELISA) was used to detect interferon gamma (IFN-γ) that was secreted from the activated T cells. Fresh PBMC were co-incubated with target tumor cells (e.g., K562 cells) at an E: T ratio of 5:1, and then treated with BiFabs or IgG format bispecific antibodies for 48 h. Supernatants were collected for IFNγ detection through Human IFN-γ CytoSetTM KIT (Invitrogen, Shanghai, China). The absorbance at 450 nm was measured by 680 Microplate reader (Bio-Rad, Hercules, CA, USA).

To evaluate T cell proliferation after activation, fresh PBMC were pre-labelled with CFSE and then mixed with target tumor cells at an E: T ratio of 5:1. The cell mixtures were treated with different concentrations of BiFabs for 48 h. T cell proliferation was further determined by flow cytometry. Ramos cells were used in experimental groups as CD20-positive cells, while K562 cells served as CD20-negative cell control.

#### *2.8. In Vivo Antitumor Activity of BiFab in Mouse Xenograft Model*

Eight-week-old female SCID Beige mice were inoculated subcutaneously with 2.5 × 10<sup>6</sup> Ramos cells and 1 × 10<sup>7</sup> PBMC (E:T = 4:1) into the right flank of the nude mice. Inoculated mice were randomly divided into 4 groups: vehicle group, FabCD3 group, FabCD20/CD3 (3 mg/kg) group and BiFabCD20/CD3 (1 mg/kg) group. Mice in the experimental groups received BiFabCD20/CD3 (1 mg/kg or 3 mg/kg) by intravenous (i.v.) injection into tail vein. Mice in the control groups received FabCD3 (1 mg/kg) or saline. Each treatment was given four times at 2-day intervals (q2d × 4). The mean tumor volume and mouse body weight were measured using calipers and an electronic balance, respectively. The mean tumor volume was calculated using the formula: tumor volume (mm<sup>3</sup> ) = tumor length × tumor width × tumor width/2.

#### *2.9. Statistical Analysis*

Statistical analysis was performed by using GraphPad Prism 6.01 software. Student's *t*-test was used when two independent groups are compared, while Dunnett's multiple comparison test was used for comparison of multiple groups. Statistical significance was determined by the *p* value (\* *p* < 0.05, \*\* *p* < 0.01, and \*\*\* *p* < 0.001).

#### **3. Results**

#### *3.1. Generation of Bispecific Fab via Sortase-Mediated Transpeptidation and Click Chemistry*

The whole procedure to generate BiFabs was summarized in Figure 1a. Fabs targeting CD20, CD3 or HER2 were first expressed with LPETG-His<sup>6</sup> tail at C terminus of heavy chains (Figure 1b) and stored for future assembly after purification. GGG-PEG3-N<sup>3</sup> or GGG-PEG4-DBCO was linked onto Fabs via sortase A transpeptidation, and His-tag was released from Fabs, which spared linker-Fab components from the reaction mixture when purified by Ni-NTA affinity chromatography. Before click reaction, the optimal molar ratio and reaction time for sortase A-catalyzed reaction was investigated. According to peak shifting of H-DBCO, the optimal reaction condition is 1:25 of FabCD3 and GPD and reacted for 12 h (Figure 1c), in which there is much less unconjugated heavy chain (peak "H") compared to other reaction conditions. Click reaction between FabCD3-DBCO and FabCD20-N<sup>3</sup> at a molar ratio of 1:1 efficiently generated BiFabCD20/CD3. After click reaction, homogenous BiFabCD20/CD3 was obtained by size exclusion chromatography purification and further

confirmed by SDS-PAGE (Supplementary Figure S1). The assembly of FabHer2 and FabCD3 was conducted in the same way to generate homologous BiFabHer2/CD3 (Figure 1d). The purity of BiFabCD20/CD3 was further confirmed by RP-HPLC analysis (Figure 1e). According to the peak area, the content of BiFabCD20/CD3 in the final buffered solution is about 95% after SEC purification and ultraconcentration.

**Figure 1.** Generation and characterization of BiFabs. (**a**) Schematic diagram of sortase A-mediated click chemistry installation for BiFab preparation. (**b**) Characterization of the purified Fabs by SDS-PAGE. Lane 1, high molecular weight protein marker; Lane 2, the reduced FabCD20; Lane 3, the intact FabCD20; Lane 3, the reduced FabCD3; Lane 4, the intact FabCD3 . (**c**) Reverse-phase HPLC analysis of Fab-click handle conjugation through sortase A-mediated transpeptidation, under different reaction conditions. (**d**) Characterization of BiFabs by SDS-PAGE. Lane 1, high molecular weight protein marker; Lane 2, the reduced BiFabHer2/CD3; Lane 3, the intact BiFabHer2/CD3; Lane 4, the intact FabHer2; Lane 5, the intact FabCD3 . (**e**) Reverse phase high-performance liquid chromatography (RP-HPLC) analysis of the purity of BiFabCD20/CD3 .

#### *3.2. The Binding Ability of BiFabs with Target and Effector Cells*

To confirm whether BiFabCD20/CD3 maintained the binding ability of two Fabs, we used Jurkat cells (CD3 positive) and Ramos cells (CD20 positive) for flow cytometric analysis of BiFabCD20/CD3. The BiFabCD20/CD3 showed concentration-dependent binding with CD20-positive Ramos cells and CD3-positive Jurkat cells (Figure 2a). Interestingly, BiFabCD20/CD3 had a higher binding affinity compared to that of FabCD20 or FabCD3 monomers to target cells (Figure 2a). Upon binding with CD3 on T cells and CD20 on tumor cells, the BiFabCD20/CD3 could efficiently activate T cells according to the measurement of cell-surfaceCD69 and CD25, which represent early and late activation markers on T cells, respectively (Figure 2b). At the same concentrations of BiFabCD20/CD3, the expression level of CD69 was much higher than that of CD25, which exhibited a quicker response curve of CD69 comparing to CD25. Similarly, the BiFabHer2/CD3 generated by replacing FabCD20 could also bind to HER2-positive SK-OV-3 cells and CD3-positive Jurkat cells (Figure 2c).

#### *3.3. BiFab Efficiently and Specifically Induced Cytokine Release and Proliferation of T Cells*

The release of interferon-γ (INF-γ) was evaluated as this cytokine is essential for mediating the antitumor activity. We measured INF-γ release by ELISA kit with CD20+ Daudi and Raji cells as target cells and unstimulated PBMC as effector cells. In both types of target cells, high level of IFN-γ release was detected in the culture supernatants in the presence of BiFabCD20/CD3 (400 and 2000 ng/mL) (Figure 2d,e). We also noticed that BiFabCD20/CD3 induced stronger T cell activation at a concentration of 80 ng/mL when the target cells were Daudi cells in comparison to Raji cells. Almost no T cell activation was observed in the absence of BiFabCD20/CD3, suggesting the specific mode of action underlying BiFab mediated T cell engaging.

For the analysis of T cell proliferation after BiFab stimulation, fresh PBMC were prestained with CFSE, a cell permeant green fluorescent molecule whose succinimidyl ester group reacts indiscriminately and covalently with primary amines of intracellular proteins, to facilitate fluorescent labeling of T cell population. Upon incubating with BiFabCD20/CD3 , PBMC significantly proliferated after 48 h in the presence of target cells (i.e., CD20-positive Ramos Cells) (Figure 2e). No obvious T cell proliferation was observed in negative control group (CD20-negative K562 cells). We further studied the proliferation rate with various concentrations of BiFabCD20/CD3 measured by flow cytometry. BiFabCD20/CD3 triggered T cells proliferation in a concentration-dependent manner (Figure 2f).

#### *3.4. BiFabs Redirected T Cells to Kill Target Tumor Cells*

In vitro cytotoxicity of BiFabCD20/CD3 was measured by LDH releasing assay on Daudi and Ramos cell lines. BiFabCD20/CD3 efficiently induced tumor cell apoptosis at an E:T ratio of 2:1, achieving half maximal-apoptosis rate at a concentration of 0.262 ng/mL (2.62 pM) on Daudi cells and 0.275 ng/mL (2.75 pM) on Ramos cells (Figure 3a). The apoptosisinducing efficacy of BiFabCD20/CD3 was further assessed by FITC-Annexin V/PI staining assay on Daudi and Ramos cell lines. BiFabCD20/CD3 could induce maximal apoptosis, including early (Annexin V+/PI−) and late (Annexin V+/PI+) apoptotic cells, on Daudi cells at various concentrations (10 ng/mL–10 ug/mL) (Figure 3b). For Ramos cells, the apoptosis rate, ranging from 70–100%, was concentration-dependent and lower than that of Daudi cells (Figure 3b). Fc-mediated nonspecific activation through binding to Fc receptors on immune cells could probably cause toxicity [32]. Comparing to IgGCD20/CD3 which could elicit Fc-mediated non-specific killing, the BiFabCD20/CD3 was demonstrated to have minimal killing towards CD20 negative K562 cells, suggesting the advantage of Fc truncation in eliminating Fc-mediated side effects (Figure 3c). Similar to BiFabCD20/CD3 , the BiFabHer2/CD3 exhibited remarkable killing efficacy on HER2-positive SK-OV-3 cells while spared HER2 negative MDA-MB-468 cells with marginal cell killing (Figure 3d). The results showed here suggested that sortase A-mediated chemo-enzymatic approach was successfully applied to the generation of other BiFabs.

γ **Figure 2.** In vitro efficacy of BiFabs. (**a**) The binding abilities of Fabs and BiFab with CD20-positive Ramos and Jurkat cells. (**b**) The in vitro efficacy of the BiFabCD20/CD3 on T cell activation. After CD20-positive B cell depletion, fresh PBMCs were treated with serial concentrations of BiFabCD20/CD3 in the presence of target tumor cells at an E:T ratio of 5:1 for 48 h. The expression levels of CD69 and CD25 on T cells, two biomarkers for T cell activation, were evaluated after immuno-staining via flow cytometry. (**c**) Evaluation of the binding abilities of BiFabHer2/CD3 with CD3-positive Jurkat cells and HER2-positive SK-OV-3 cells by flow cytometry. (**d**) The quantification of interferon-γ release from T cells activated by BiFabCD20/CD3 . Fresh PBMCs were incubated with Daudi or Raji cells at an E:T ratio for 5:1 for 48 h. The secreted interferon-γ from T cells was quantified by ELISA Kit. (**e**) BiFabCD20/CD3 mediated T cell proliferation in the presence of CD20-negative K562 cells or CD20-positive Ramos cells at an E:T ratio of 5:1 for 48 h. (**f**) After treatment with various concentrations of BiFabCD20/CD3 with an E:T ratio of 5:1 for 48 h, T cell proliferation was analyzed by flow cytometry.

#### *3.5. BiFabCD20/CD3 Eliminated B-Cell Lymphoma in Xenograft Mouse Model*

We next evaluated the in vivo efficacy of BiFabCD20/CD3 with mouse xenograft model of B-cell lymphoma. Mouse xenograft tumor model was successfully established by coinjection of Ramos and PBMC cells (E:T ratio = 4:1). The intravenous administration of

BiFabs was initiated 24 h after inoculation to facilitate T cell activation. The administration was repeated every two days for a total of four injections. Strikingly, the BiFabCD20/CD3 completely suppressed the tumor growth at a dosage of 3 mg/kg, and there was only one mouse that underwent a recurrence in the 1 mg/kg group (Figure 3e). In contrast, the anti-CD3 Fab group did not show any significant efficacy in vehicle group, in which tumor grew rapidly. These results demonstrated that the BiFabCD20/CD3 could efficiently mediate T cell killing in vivo.

**Figure 3.** The in vitro and in vivo antitumor activities of BiFabs. (**a**) The in vitro efficacy of BiFabCD20/CD3. Target cells (Ramos and Daudi) and active T cells (E:T = 2:1) were incubated with serial diluted BiFabCD20/CD3 for 24 h (data shown as mean ± SD, *n* = 3). LDH release was determined by ELSIA kit and used to calculate cell viability. (**b**) The in vitro cytotoxicity of BiFabCD20/CD3 was analyzed by Annexin V/PI apoptosis detection kit, by using the same condition as described in (**a**). (**c**) Study on potential Fc-related cytotoxicity of BiFabCD20/CD3. The K562 cells and PBMCs were co-cultured with serial concentrations of non-binding IgG-based bispecific antibody or BiFab. The apoptosis rate was determined by Annexin V-Cy5 Apoptosis Detection Kit. (**d**) The in vitro cytotoxicity of BiFabHer2/CD3. Target tumor cells (SK-OV-3 or MDA-MB-468) and PBMC (E:T = 4:1) were incubated with serial concentrations of BiFabHer2/CD3 for 72 h, and the LDH release in the supernatant was determined by LDH detection kit. All data were shown as mean ± SD, *n* = 3. (**e**) The in vivo antitumor activities of BiFabCD20/CD3 in mouse xenograft model. Mice were inoculated subcutaneously with 2.5 × 10<sup>6</sup> Ramos cells in the presence of 1 × 10<sup>7</sup> fresh human PBMC from healthy donors at an E:T ratio of 4:1. All samples were administered intravenously via the tail vein at following dosages, 1 mg/kg of FabCD3 and 1 mg/kg or 3 mg/kg of BiFabCD20/CD3 at every two days for four times.

#### **4. Discussion**

We have presented here a facile approach utilizing sortase A-mediated bio-click chemistry to generate BiFabs with potent antitumor activity. Paired click handles (e.g., N<sup>3</sup> and DBCO) was conjugated to the C-terminal LPETG tag of Fabs via sortase A mediated transpeptidation, followed by site-specific conjugation between two click handles-modified Fabs for BiFab generation. We have presented exemplary BiFabs against two different targets. First, the BiFabCD20/CD3 exhibited superior efficacy in mediating T cells, from either PBMC or ATC, to kill multiple CD20-positive lymphoma cell lines while spared CD20 negative tumor cells in vitro (Supplemental Figure S2). The BiFabCD20/CD3 also efficiently inhibited CD20-positive tumor growth in the mouse xenograft model (Figure 3e). Second, the BiFabHer2/CD3 also showed potent in vitro antitumor activity against HER2-positive tumor cell lines (Figure 3d), demonstrating the robustness of sortase A-mediated bio-click chemistry in generating various potent BiFabs.

The first BiFab construct, termed as BsF(ab')2, was first generated by chemical conjugation of Fab'-SH with the thionitrobenzoate derivative of another Fab (Fab'-TNB), which was described by Paul Carter et al. from Genentech Inc. [19]. The BiFabCD20/CD3 was also generated without Fc region. Fc region of IgG-based bispecific antibodies could potentially induce nonspecific T cell activation [33], causing off-target T cell engaging-related side effects. However, cytokine-related adverse effects, such as cytokine release syndrome (CRS), are probably inevitable for T cell engaging and activation related immunotherapy, e.g., CAR-T, BiTE [34,35]. At present, T-cell engagers targeting CD20 are mostly based on classical IgG-based antibody. Sun et al. [36] reported a T-cell recruiting bispecific antibody CD20-TDB with EC<sup>50</sup> of 0.22–11 ng/mL at the E: T ratio of 10:1. Smith et al. [37] reported another anti-CD20/CD3 T cell engagers REGE2280 and REGN 1979, which showed favorable EC<sup>50</sup> of 2.25–12.6 ng/L at the E:T ratio of 10:1 (ATC as effector cells). FBTA05 is a trifunctional chimeric rat/mouse CD3 × CD20 targeting bispecific antibody, and therefore it has higher immunogenicity [38]. In comparison with above anti-CD20/CD3 bispecific antibodies, our BiFabCD20/CD3 showed a potent apoptosis-inducing ability at an E:T ratio of 2:1 using ATC as effector cells (EC<sup>50</sup> = 0.26 ng/mL for Daudi cell lines and 0.275 ng/mL for Ramos) (Figure 3a). The in vivo antitumor efficacy of BiFabCD20/CD3 was consistent with its in vitro efficacy, since four intravenous injections (3 mg/kg) of BiFabCD20/CD3 completely suppressed tumor growth in the mouse tumor xenograft models (Figure 3e).

We previously reported a nucleic acid (i.e., left-handed DNA, L-DNA) mediated protein-protein assembly (NAPPA) approach to offer a general approach for preparing antibodies with higher-order specificity [39]. Similar to the NAPPA approach, our two-step conjugation strategy allows the preparation of modular Fab derivatives and the generation of customized Fab library thereof, which is the major difference comparing with the conventional BiFab construction methods (Figure 4). In addition, both BiFabCD20/CD3 and BiFabHer2/CD3 showed potent antitumor efficacies, regardless of different tumor target, suggesting the effectiveness and robustness of sortase A mediated chemo-enzymatic approach. Lawrence G Lum et al. [40] explored the application of anti-CD20/CD3 bispecific anti-body-armed activated T cells (aATC). The anti-CD20/CD3 aATC was a cell-based therapy that activated T cells from patients were armed with chemically conjugated anti-CD3 × anti-CD20 bispecific antibody, and then expanded and re-infused into patients. The aATC therapy was demonstrated to be safe and effective in a phase I clinical trial [40]. Inspired by this study, the sortase A-mediated bio-click chemistry could be further applied to personalized immunotherapy through ATC armed with combination-optimized BiFab. Since the efficacy of BiFab varies when using Fabs with different affinities or paratopes, the sortase A mediated transpeptidation reaction during BiFab generation facilitates the construction of Fab library for rapid efficacy evaluation of different BiFabs.

**Figure 4.** Schematic diagram of modular BiFab generation. Fabs could be adapted from full-length IgGs targeting tumor antigens or T cell/NK cell activating receptors. Fabs are genetically modified to have a C-terminal sortase A recognition motif (e.g., LPETG). Then, the paired click chemistry could be installed to the Fabs via sortase A mediated transpeptidation, followed by click reaction between two Fabs to generate BiFab.

#### **5. Conclusions**

We constructed BiFabCD20/CD3 and BiFabHer2/CD33 via sortase A-mediated bio-click chemistry and demonstrated their anti-tumor activity through engaging human immune cells. Our results shown here indicates that Sortase A-mediated click handle installation holds promise for facile generation of potent bispecific Fabs and further efficient construction of Fab derivative library for personalized tumor immunotherapy in the future.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/cancers13184540/s1, Figure S1: (**a**) Size exclusion chromatography (SEC) purification of BiFabCD20/CD3. (**b**) SDS-PAGE analysis of peaks from (**a**); Line 1, high molecular weight protein marker; Line 2, reduced protein product from peak 1; Line 3, reduced protein product from peak 2. Figure S2: BiFabCD20/CD3 activated T cells in the presence of target cell lines with different antigen expression level and mediated target cells killing in a T cell-dependent manner. (**a**) Target cell lines with different antigen expression level are measured by flow cytometry; (**b**) Target cell lines of different CD20 expression level and PBMC isolated from a healthy donor (1:5 cell ratio) were incubated with serial concentrations of BiFabCD20/CD3 for 48 h.

**Author Contributions:** Conceptualization, L.P., S.C., W.L. and W.Z.; methodology, X.B. and W.Z.; validation, X.B.; formal analysis, W.L., X.B. and S.J.; resources, Y.X.; writing—original draft preparation, X.B.; writing—review and editing, L.P. and X.B.; supervision, L.P. and S.C.; funding acquisition, L.P., S.C. and Z.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Joint Funds of the National Natural Science Foundation of China (Grant No. U20A20409), National Natural Science Foundation of China (Grant No. 82073750), key research and development project of Zhejiang province (No.2018C03022), the Fundamental

Research Funds for the Central Universities (No.2020QNA7005) and Zhejiang Province "Qianjiang Talent Plan".

**Institutional Review Board Statement:** For PBMC extraction assay, the experiment protocol was reviewed and approved by medical ethic committee in College of Pharmaceutical Sciences, Zhejiang University, China (2018-003). The animal experiments were carried out in compliance with the Public Health Service Policy on Human Care and Use of Laboratory Animals. The protocol was approved by the committee on the Ethics of Animal Experiments of Zhejiang University, China (19699).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study. Blood donors were recruited and informed of risks and discomforts of the donation process, and a signed informed consent document was obtained. The blood collection was done by standard phlebotomy.

**Acknowledgments:** We thank Ning Hu from Sun Yat-Sen University for helpful discussion.

**Conflicts of Interest:** The authors have filed a patent for the BiFab generation. Wenhui Liu is an employee of Hangzhou Biosun Pharmaceutical Co., Ltd.

#### **References**

