Fusion of Bacterial Flagellin to a Dendritic Cell-Targeting αCD40 Antibody Construct Coupled With Viral or Leukemia-Specific Antigens Enhances Dendritic Cell Maturation and Activates Peptide-Responsive T Cells

Conventional dendritic cell (DC) vaccine strategies, in which DCs are loaded with antigens ex vivo, suffer biological issues such as impaired DC migration capacity and laborious GMP production procedures. In a promising alternative, antigens are targeted to DC-associated endocytic receptors in vivo with antibody–antigen conjugates co-administered with toll-like receptor (TLR) agonists as adjuvants. To combine the potential advantages of in vivo targeting of DCs with those of conjugated TLR agonists, we generated a multifunctional antibody construct integrating the DC-specific delivery of viral- or tumor-associated antigens and DC activation by TLR ligation in one molecule. We validated its functionality in vitro and determined if TLR ligation might improve the efficacy of such a molecule. In proof-of-principle studies, an αCD40 antibody containing a CMV pp65-derived peptide as an antigen domain (αCD40CMV) was genetically fused to the TLR5-binding D0/D1 domain of bacterial flagellin (αCD40.FlgCMV). The analysis of surface maturation markers on immature DCs revealed that fusion of flagellin to αCD40CMV highly increased DC maturation (3.4-fold elevation of CD80 expression compared to αCD40CMV alone) by specifically interacting with TLR5. Immature DCs loaded with αCD40.FlgCMV induced significantly higher CMVNLV-specific T cell activation and proliferation compared to αCD40CMV in co-culture experiments with allogeneic and autologous T cells (1.8-fold increase in % IFN-γ/TNF-α+ CD8+ T cells and 3.9-fold increase in % CMVNLV-specific dextramer+ CD8+ T cells). More importantly, we confirmed the beneficial effects of flagellin-dependent DC stimulation using a tumor-specific neoantigen as the antigen domain. Specifically, the acute myeloid leukemia (AML)-specific mutated NPM1 (mNPM1)-derived neoantigen CLAVEEVSL was delivered to DCs in the form of αCD40mNPM1 and αCD40.FlgmNPM1 antibody constructs, making this study the first to investigate mNPM1 in a DC vaccination context. Again, αCD40.FlgmNPM1-loaded DCs more potently activated allogeneic mNPM1CLA-specific T cells compared to αCD40mNPM1. These in vitro results confirmed the functionality of our multifunctional antibody construct and demonstrated that TLR5 ligation improved the efficacy of the molecule. Future mouse studies are required to examine the T cell-activating potential of αCD40.FlgmNPM1 after targeting of dendritic cells in vivo using AML xenograft models.

Conventional dendritic cell (DC) vaccine strategies, in which DCs are loaded with antigens ex vivo, suffer biological issues such as impaired DC migration capacity and laborious GMP production procedures. In a promising alternative, antigens are targeted to DCassociated endocytic receptors in vivo with antibody-antigen conjugates co-administered with toll-like receptor (TLR) agonists as adjuvants. To combine the potential advantages of in vivo targeting of DCs with those of conjugated TLR agonists, we generated a multifunctional antibody construct integrating the DC-specific delivery of viral-or tumorassociated antigens and DC activation by TLR ligation in one molecule. We validated its functionality in vitro and determined if TLR ligation might improve the efficacy of such a molecule. In proof-of-principle studies, an aCD40 antibody containing a CMV pp65derived peptide as an antigen domain (aCD40 CMV ) was genetically fused to the TLR5binding D0/D1 domain of bacterial flagellin (aCD40.Flg CMV ). The analysis of surface maturation markers on immature DCs revealed that fusion of flagellin to aCD40 CMV highly increased DC maturation (3.4-fold elevation of CD80 expression compared to aCD40 CMV alone) by specifically interacting with TLR5. Immature DCs loaded with aCD40.Flg CMV induced significantly higher CMV NLV -specific T cell activation and proliferation compared to aCD40 CMV in co-culture experiments with allogeneic and autologous T cells (1.8-fold increase in % IFN-g/TNF-a + CD8 + T cells and 3.9-fold INTRODUCTION Dendritic cells (DCs) play a key role at the interface between innate and adaptive immunity, and therefore hold potential for use in the immunotherapy of diseases such as cancer. In particular, the high capacity of DCs for processing and presenting antigens makes them suitable candidates for manipulating with antigens of choice in a vaccine approach. The most common strategy is to load DCs ex vivo with major histocompatibility complex (MHC)-binding peptides. This has been investigated in numerous clinical trials in different cancer entities, which have so far shown safety and feasibility, but often lack efficacy (1,2). Improvements have been achieved with the development of personalized neoantigen-based DC vaccines, which elicited potent neoantigen-specific T cell responses with remarkable efficacy in melanoma patients (3)(4)(5). However, this type of DC vaccination features drawbacks. The ex vivo engineering and GMP production of DCs is costly and laborintensive, and standardization is difficult as vaccines are generated individually for each patient (6). In addition, the efficacy of these vaccines can be limited by inefficient migration of administered DCs to the lymph nodes, wherein DCs activate antigen-specific T cells (7).
An alternative approach consists of delivering an antigen to target DCs in vivo using an antibody-antigen fusion construct. Such vaccines can be applied to a larger patient cohort and thus be manufactured on a larger scale. More importantly, this technique has biological advantages as it exploits the intricate migratory capacity of DCs in situ and directly activates natural DC subsets at multiple sites in vivo thereby producing a more physiological DC maturation (8,9). Although clinical data are still scarce, in vivo DC vaccination is considered a promising strategy for eliciting strong and sustained T cell responses (2,10,11).
Different DC surface receptors have been proposed as targets for in vivo DC vaccines. These differ widely in their expression levels, intracellular trafficking pathways and antigen presentation capacity. Among those, CD40 is of high therapeutic interest.
Indeed, previous pre-clinical studies showed that the delivery of antigens to DCs by CD40-targeting antibodies was more efficient in eliciting MHC-I cross-presentation and inducing CD8 + T cell responses compared to other receptors such as Dec205 (12)(13)(14). The 48 kDa type I transmembrane protein CD40 is a critical mediator of immune cell communication, for example, by initiating T cell priming, and is a costimulatory surface receptor of the tumor necrosis factor receptor (TNFR) family (15). Importantly, agonistic aCD40 antibodies not only facilitate DC-targeting, but also exhibit adjuvant function by inducing CD40 signaling to transduce an intrinsic stimulatory signal to DCs.
The use of adjuvants is particularly important for DC vaccination. At steady state, immature DCs tend to induce tolerogenic T cell responses (16). However, DCs mature and upregulate co-stimulatory molecules in the presence of adjuvants, thereby enhancing cross-talk with T cells. In addition to CD40-activating agents, ligands for toll-like receptors (TLRs) are commonly used that potently activate innate immunity and are critical for optimizing T cell responses (17). If a single adjuvant is insufficient for DC activation, using a combination of adjuvants has been proposed, especially for targeting multiple intracellular signaling pathways (18,19). Ahonen et al. have shown that coadministration of CD40 activators together with various TLR agonists induced higher antigen-specific CD8 + T cell responses than either agonist alone, demonstrating the synergy between TLR-derived stimuli and the CD40 pathway (20). TLR3 or TLR7/ 8 agonists are being used as adjuvant drugs in various clinical trials, and TLR5 agonists have been recently investigated (21,22). As TLR5 detects bacterial flagellin-the protein that polymerizes to form flagella-flagellin or constitutive domains can be used as activators (23). A clinically advanced example is entolimod, a pharmacologically optimized truncated version of flagellin from Salmonella enterica serovar Dublin, which retains the two domains (D0/D1) essential for TLR5 binding (24). Entolimod was initially established as a potential treatment for lethal radiation exposure in patients with advanced solid tumors due to its tissue protective activity (24)(25)(26). It is also reported to elicit direct anti-tumoral effects through activation of the NK-DC-CD8 + T cell axis and demonstrated anti-metastatic activity through immune stimulatory mechanisms involving NK cells (26)(27)(28)(29). More recently, entolimod is investigated as a vaccine adjuvant by its developing company Cleveland BioLabs.
Importantly, not only does the nature and optimal combination of adjuvants determine the outcome of DC vaccination strategies but also the administration regimen. In practice, adjuvants are typically co-administered with DC-targeting antibodies, which has been shown to improve the therapeutic efficacy of DC-targeted vaccines. However, the use of soluble adjuvants can lead to antigenindependent activation of bystander immune cells, which carries the risk of adverse events occuring, such as cytokine release or autoimmunity (30,31). Indeed, targeted delivery of TLR agonists to DCs was not only associated with a reduced serum cytokine release and related toxicity, but also reduced their dose requirement by 100-fold (31). Furthermore, co-delivery of antigen and adjuvant into the same antigen-presenting cell (APC), for example, in the form of antigen-adjuvant conjugates, resulted in superior crosspresentation and peptide-specific T cell activation compared to administration of separate molecule in vivo (31)(32)(33)(34).
To combine the benefits of in vivo co-delivery of a TLR agonists and an antigen, we generated novel multifunctional antibody constructs (aCD40.Flg CMV and aCD40.Flg mNPM1 ) to target viral-and tumor-specific antigens to DCs via an aCD40 antibody. These simultaneously activate TLR5 with a genetically fused truncated entolimod-like flagellin domain (Flg). As antigen domain, we either fused a cytomegalovirus (CMV)-specific epitope or a neoantigen derived from mutated NPM1 (mNPM1), which occurs in 30-35% of acute myeloid leukemia (AML) (35,36). We addressed the question of whether TLR5 ligation ameliorates the in vitro efficacy of aCD40 CMV and aCD40 mNPM1 by evaluating these molecules based on their binding specificity, DC maturation and induction of viral-and tumor-specific T cell responses. Since the addition of a flagellin domain was indeed able to enhance DC maturation and to increase both CMV-and mNPM1-specific T cell responses compared to aCD40 CMV or aCD40 mNPM1 alone, we consider this novel antibody-TLR agonist fusion format a promising therapeutic approach worthy of investigation in in vivo studies.

Expression and Purification
The variable domains of aCD40 were derived from the IgG2 antibody CP-870,893 (selicrelumab, clone 21.4.1., Hoffman-La Roche) (37,38). The aHer2 non-specific binding control contained variable domains from the 4D5-8 clone (trastuzumab, Hoffman-La Roche) (39). To produce IgG1 molecules, we cloned the coding sequences for both antibodies into the mammalian expression vectors pFUSE2-CLIg-hk and pFUSE-CHIg-hG1 (InvivoGen). P329G, L234A, and L235A (PGLALA) mutations were inserted to silence the Fc function (40). The flagellin D0/D1 domain (Flg), with a GSGGG linker introduced in place of D2/D3 of full-length flagellin, was cloned from genomic DNA of Salmonella enterica Typhimurium strain SL1344. The antigen domain consisted of either a sequence of the cytomegalovirus (CMV)specific pp65 protein (CMV 487-508 ) including the HLA-A*02:01restricted CMV 495-503 epitope (NLVPMVATV), or the mutated NPM1 derived mNPM1 277-298 protein that comprises the HLA-A*02:01-restricted mNPM1 288-296 neoepitope (CLAVEEVSL). The CMV 487-508 domain is hereafter abbreviated as CMV, and the mNPM1 277-298 domain as mNPM1; the epitope peptides are denoted CMV NLV and mNPM1 CLA , respectively. The antigen domains were cloned to the C terminus of the light chain and flagellin to the heavy chain, both separated by (G 4 S) 4 linkers. As a control, an Fc-flagellin fusion was generated, which consisted of an IgG1 Fc region followed by a (G 4 S) 4 linker and flagellin on the C terminus.
All molecules were produced over 5 days in Expi293F cells (Thermo Fisher Scientific) according to the manufacturer's i n s t r u c t i o n s . F o r p u r i fi c a t i o n , p r o t e i n A a ffi n i t y chromatography (nProtein A Sepharose 4 Fast Flow, Thermo Fisher Scientific) was performed followed by size-exclusion chromatography (SEC) on Superdex 200 Increase 10/200 GL columns (GE Healthcare) in Dulbecco's phosphate-buffered saline (DPBS; Thermo Fisher Scientific). Proteins were analyzed by analytical SEC (Superdex 200 Increase 5/150 GL column, GE Healthcare). Thermal stability was measured by nano differential scanning fluorimetry (nanoDSF) on a Tycho NT.6 instrument (NanoTemper Technologies).

Cell Lines
The L-428 cell line was purchased from the "German Collection of Microorganisms and Cell Cultures" (DSMZ, Leibniz Institute, Braunschweig, Germany). Cells were cultured in RPMI 1640 medium supplemented with GlutaMAX (Gibco, Thermo Fisher Scientific) and 10% fetal bovine serum (FBS; Gibco, Thermo Fisher Scientific). The Expi293F cell line was cultured in Expi293 Expression Medium (Thermo Fisher Scientific).

Preparation of Peripheral Blood Mononuclear Cells From Whole Blood
Peripheral blood samples were collected from healthy donors (HDs) giving written informed consent according to a clinical protocol ("In vitro studies to establish new immunotherapies for AML and other hematological neoplasias") approved by the Ethics Committee at the LMU Munich. Peripheral blood mononuclear cells (PBMCs) from HDs were separated from peripheral blood using the Histopaque-1077 Hybri-Max separating solution (Sigma-Aldrich) and Leucosep tubes (Greiner Bio-One) according to the manufacturer's instructions. In brief, 15 ml of Histopaque solution was preloaded into a 50 ml Leucosep tube by centrifugation for 30 s at 1,000 g. The heparinized whole-blood samples were diluted with equal volumes of DPBS, and 30 ml of the diluted blood was added to the Leucosep tube. Tubes were centrifuged for 10 min at 1,000 g with the brakes off. The buffy coat containing PBMCs was collected, and cells were washed and re-suspended in RPMI 1640 containing 10% FBS for immediate use.

Generation, Maturation, and Peptide-Loading of Monocyte-Derived DCs
Monocyte-derived DCs (moDCs) were generated within 3 days as previously described (41). Monocytes were enriched from PBMCs by plastic adherence in flat-bottom 6-or 12-well plates with surface treatment for maximum adhesion (Nunc, Thermo Fisher Scientific) at a concentration of 0.5-1 × 10 7 cells/ml in VLE RPMI (Biochrom) supplemented with 1.5% human serum (HS, serum pool of AB positive adult males; Institute for Transfusion Medicine, Suhl, Germany)-hereafter termed DC medium. If required, non-adherent cells (NACs) were kept at 37°C and 5% CO 2 until further use three days later. For microscopy experiments requiring pure monocytes, monocytes were isolated from PBMCs using the Classical Monocyte Isolation kit (Miltenyi Biotec) according to the manufacturer's instructions. Next, cells were seeded in 24-well Nunc-plates and cultured for 48 h at 37°C and 5% CO 2 in DC medium supplemented with 800 IU/ml of GM-CSF (Peprotech) and 580 IU/ml of IL-4 (Peprotech). These cells were then loaded with 200 nM antibody with or without flagellin fusion for 24 h, optionally pre-incubated for 30 min with a fourfold excess of aTLR5-IgA2 (Q2G4) or corresponding isotype control (T9C6, both InvivoGen). If moDCs were to be maintained at the immature stage (iDCs), cells were incubated with 800 IU/ml of GM-CSF, 580 IU/ml of IL-4, and 250 ng/ml of PGE 2 (Sigma-Aldrich) for a further 24 h. To generate fully mature moDCs (mDCs), maturation was achieved within 20-24 h after addition of 800 IU/ml of GM-CSF, 580 IU/ml of IL-4, 250 ng/ml of PGE2, 2000 IU/ml of IL-1b (R&D Systems), 1,100 IU/ml of TNF-a (Peprotech), 5,000 IU/ml of IFN-g (Peprotech), and 1 µg/ ml of R848 (InvivoGen) (41,42). For DC-T cell co-culture experiments, iDCs and mDCs were pulsed with the processed HLA-A*02:01-restricted peptides (CMV NLV , mNPM1 CLA ) for 1.5 h at 37°C.

Surface Phenotyping of DCs and IL-6 Secretion
Immunofluorescent staining of DC surface antigens was performed using a panel of fluorescence-conjugated monoclonal antibodies and analyzed by flow cytometry: CD80 (BV510, 2D10), CD83 (PerCP-Cy5.5, HB15e), CD86 (APC, BU63), and HLA-DR (Pacific Blue, L243, all BioLegend). DCs could be separated from non-differentiated monocytes according to their higher FSC/SSC intensities. Mean fluorescence intensity (MFI) ratios were determined by dividing the MFI value measured for the antibody by that of the corresponding isotype control. In parallel, IL-6 secretion into the supernatant was quantified via cytometric bead array (BD Biosciences) according to the manufacturer's instructions.

Quantitative Analysis of Binding by Flow Cytometry
To determine equilibrium binding constants (K D ; as an avidity measurement), L-428 cells were incubated with antibodies in concentrations ranging from 0.005 to 200 nM and stained with an ahuman IgG secondary antibody (FITC, HP6017, BioLegend). The assay was performed in triplicates, and data points were normalized to the maximum MFI and analyzed by non-linear regression using a one-site specific binding model.
To determine maximum binding, L-428 cells, iDCs or mDCs were incubated with the antibodies at a concentration of 200 nM followed by secondary antibody staining and MFI ratios were calculated.

Quantitative Analysis of Binding by Surface Plasmon Resonance
For determining K D values by surface plasmon resonance (SPR; Biacore X100, GE Healthcare), a CM5 chip (GE Healthcare) and the ahuman Fc capture (GE Healthcare) were used. The extracellular domain of CD40 was passed over the antibodycoated chip at concentrations ranging from 15.62 to 1,000 nM. K D values were determined as the ratio of the association rate constant (k on ) and the dissociation rate constant (k off ).
Interaction and Signaling Studies Using the hTLR5 Reporter Cell Line HEK293-T cells were seeded in 24-well plates and transfected at 70% confluency the next day. 1 h before transfection, the medium was exchanged for 0.5 ml of OptiMEM medium (Gibco, Thermo Fisher Scientific) containing 5% FBS. 200 ng of human TLR5 plasmid was transfected using Lipofectamine 2000 (Invitrogen, Thermo Fisher Scientific) according to the manufacturer's instructions (hTLR5-HEK293). At 22 h post-transfection, the medium was replaced with 1 ml of fresh DMEM medium supplemented with 5% FBS and further incubated for TLR5 expression (43). At 44 h post transfection, cells were incubated for 4 h with 10 nM/well of CD40 C MV , aCD40.Flg C M V , aCD40.mFlg CMV , or Fc.Flg. Before the coincubation, all proteins were prepared briefly by sonication for 1 min at 4°C and power 5, to release monomeric flagellin, which is the active molecule binding to TLR5 (Branson Ultrasonics). Purified endotoxin-free native Salmonella Typhimurium full-length flagellin (100 ng/well) was used as a positive control protein to evaluate the TLR5-dependent activation in each experiment. IL-8 secretion in the supernatants of the activated cells was determined using an IL-8 OptEIA ™ ELISA system (BD Biosciences), according to the manufacturer's recommendations. Empty vector-transfected cells were used as control conditions for each experiment. Those conditions did not produce any significant elevation of IL-8 secretion over mockcoincubated with any of the tested proteins, indicating that the response was strictly TLR5-dependent.

Expansion of CMV NLV and mNPM1 CLA Peptide-Specific T Cells
To generate CMV NLV -specific T cells that recognize the pp65derived CMV 495-503 epitope (NLVPMVATV) in the context of HLA-A*02:01, PBMCs from an HLA-A*02:01 + and CMV + donor were isolated. Mature DCs were produced as described above, pulsed with 1 µM CMV NLV (JPT Peptide Technologies) for 90 min, then irradiated with 30 Gy (XStrahl RS225, XStrahl). Autologous CD8 + T cells were isolated from NACs using the CD8 + T Cell Isolation Kit (Miltenyi Biotec). T cells and pulsed DCs were co-cultivated at a T cell:DC ratio of 4:1 in RPMI 1640 containing 5% HS and 30 ng/ml of IL-21 (Peprotech) for 72 h. On days 3 and 6, co-cultures were expanded 1:1 by the addition of medium supplemented with 10 ng/ml of IL-15 and 10 ng/ml of IL-7 (both Peprotech). On day 9, CMV NLV -specific T cells were sorted on a FACSAria III cell sorter (BD Biosciences) by staining for CD8 (FITC, SK1, BioLegend) and with HLA-A*02:01-pCMV NLV dextramer (PE, Immudex). mNPM1 CLA -specific T cells that recognize the mNPM1 288-296 epitope (CLAVEEVSL) in the context of HLA-A*02:01 were generated as previously described by the transduction of CD8 + T cells with a mNPM1 CLA -specific T cell receptor (TCR) (kindly provided by Dr. Marieke Griffioen, Leiden University) (44).
For expansion of specific T cells, PBMC feeders of two HLA-A*02:01 + and two HLA-A*02:01 − donors were mixed in equal amounts and pulsed with 1 µM CMV NLV or mNPM1 CLA peptide (JPT Peptide Technologies) in X-VIVO 15 medium (Lonza) for 2 h at 37°C. After irradiation with 30 Gy, feeders were cultivated at a concentration of 2 × 10 6 cells/ml with specific T cells at 0.4 × 10 6 cells/ml in X-VIVO 15 containing 5% HS supplemented with 10 ng/ml of IL-7, 10 ng/ml of IL-15, and 0.5 µg/ml of PHA-L (Sigma-Aldrich) in a 6-well plate. After 3 days, cultures were fed by replacing half the volume of medium with fresh X-VIVO 15 containing 5% HS supplemented with 50 U/ml of IL-2 (Peprotech), 20 ng/ml of IL-7, and 20 ng/ml of IL-15. Using this medium, T cells were expanded every 3 days. Experiments were performed 9-21 days after expansion. Feeding was necessary every 14-21 days to re-challenge the cells with peptide.

Internalization Assay by Structured Illumination Microscopy
To assess internalization of antibodies by DCs, moDCs differentiated from magnetically isolated monocytes were used. aCD40 CMV and the aHer2 CMV non-specific binding control were labeled with Alexa Fluor (AF)594 Labeling Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. iDCs and mDCs were incubated for 90 min at 4°C or 37°C with 200 nM labeled antibodies. For samples incubated at 4°C and the aHer2 controls at either temperature, membrane staining with an AF488-labeled aHLA-DR antibody was performed. Next, moDCs were transferred onto coverslips using a Shandon Cytospin 3 cytocentrifuge (Thermo Fisher Scientific). Fixation was performed with 4% paraformaldehyde (Sigma-Aldrich) followed by permeabilization with 0.1% Triton X-100 (Sigma-Aldrich). Lysosomes were stained using aLAMP1 (polyclonal, Novus Biologicals) or aEEA1 antibody (F.43.1, Thermo Fisher Scientific), which were detected using a secondary donkey arabbit AF488 antibody (polyclonal, Thermo Fisher Scientific). Post-fixation took place with 4% paraformaldehyde followed by nuclear staining with 1 µg/ml of DAPI (Thermo Fisher Scientific). Finally, the coverslips were mounted onto glass slides using Vectashield (Vectorlabs).
3D structured illumination microscopy (SIM) acquisition was done on a Deltavision OMX V3 microscope (General Electric) equipped with a 100 × 1.4 oil immersion objective UPlanSApo (Olympus), 405, 488, and 593 nm diode lasers and Cascade II EMCCD cameras (Photometrics). Raw data were first reconstructed and corrected for color shifts with the provided software softWoRx 6.0 Beta 19 (unreleased). A custom-made macro in Fiji finalized the channel alignment and established composite TIFF stacks (45).

Autologous DC-T Cell Co-Culture Using Non-Adherent Cells as the T Cell Source
DCs were washed prior to co-culturing with T cells to remove antibodies, maturation reagents and peptides. Peptide-or antibody-loaded iDCs and mDCs were incubated with NACs of an HLA-A*02:01 + and CMV + donor in a 1:10 ratio in DC medium for 6 days. Dead cells were excluded as 7-AAD-positive cells (BioLegend) and the percentage of CMV NLV -specific CD8 + T cells was determined by staining CD8 + cells (FITC, SK1, BioLegend) with HLA-A*02:01-pCMV NLV dextramer (PE, Immudex) and analyzed by flow cytometry.
Allogeneic DC-T Cell Co-Culture Using Expanded CMV NLV -and mNPM1 CLA -Specific T Cells Peptide-or antibody-loaded iDCs and mDCs were cultivated with allogeneic expanded CMV NLV -and mNPM1 CLA -specific T cells in a 1:5 ratio for 4-6 h in DC medium containing 25 µM monesin and 10 µg/ml of brefeldin A (both Sigma-Aldrich) at 37°C and 5% CO 2 . Viable Zombie Green-negative (BioLegend) cells were stained for CD8 (PerCP-eFluor710, SK1, eBioscience), subsequently fixed and permeabilized using the BD Cytofix/Cytoperm Kit (BD Biosciences) according to the manufacturer's instructions followed by intracellular IFN-g (PE, B27) and TNF-a (APC, Mab11, both BioLegend) staining. The percentage of cytokine-secreting CD8 + T cells was determined by flow cytometry.

Flow Cytometry
Flow cytometry measurements were performed either on a Guava easyCyte 6HT flow cytometer (Merck Millipore) with data analyzed using GuavaSoft version 3.1.1 (Merck Millipore) or on a CytoFLEX LX flow cytometer (Beckmann Coulter) with FlowJo 10.6 (Tree Star Inc.).

RESULTS
aCD40 CMV and aCD40.Flg CMV are Pure and Stable Proteins That Bind to DCs To deliver immunogenic peptides specifically to CD40 on DCs, we first generated proof-of-principle constructs with a CMV-specific peptide as a model antigen. It consisted of a CMV peptide-coupled aCD40 antibody (aCD40 CMV ) with an Fcsilenced IgG1 backbone ( Figure 1A). The variable regions are derived from the therapeutic IgG2-based aCD40 antibody CP-870,893 (clone 21.4.1) (37,38). To combine DC targeting with TLR5 agonist-mediated stimulation, we fused the D0/D1 domain of Salmonella Typhimurium flagellin (Flg), structurally similar to the therapeutic molecule entolimod, to the C-termini of the heavy chains of the antibody portion using a flexible linker of four repeating polyglycine-serine (G 4 S) units (aCD40.Flg CMV ). The peptide domain is connected to the C-termini of the light chains by a (G 4 S) 4 linker and consists of a partial sequence of the CMV-specific pp65 protein including the HLA-A*02:01restricted CMV 495-503 epitope (NLVPMVATV, CMV NLV ). Purified proteins were analyzed by analytical SEC (Supplementary Figure S1A) and protein stability was assessed by nanoDSF (Supplementary Figure S1B). Both  homogeneity without noticeable aggregation and exhibit two melting transitions at around 72°C and 84°C, independent of the flagellin domain. We evaluated the binding properties of the antibody constructs to APCs by flow cytometry. An antibody directed against Her2 served as a non-specific binding control since DCs are Her2 negative. No differences in MFI ratios for aCD40 CMV and aCD40.Flg CMV were measured with either L-428 cells or moDCs. However, MFI ratios were higher on mDCs compared to iDCs due to upregulation of CD40 during the maturation process ( Figure 1B). Further maximum binding analysis on iDCs from different donors confirmed that the flagellin domain neither impaired nor contributed to DC binding, suggesting that flagellin does not interfere with the CD40 binding domain ( Figure 1C). Furthermore, apparent dissociation constants (K D ) were determined by flow cytometry of L-428 cells ( Figure  1D) and SPR analysis ( Figure 1E). By flow cytometry, we measured affinities in the low nanomolar range (K D aCD40 CMV = 4.9 ± 0.3 nM, aCD40.Flg CMV = 3.0 ± 0.2 nM; for n = 3), whereas K D values measured by SPR were higher due to monovalent binding (K D aCD40 CMV = 27.4 ± 3.7 nM, aCD40.Flg CMV = 21.8 ± 1.1 nM; for n = 3). Both experiments showed that the flagellin fusion did not substantially alter K D values, again indicating no noticeable perturbation of CD40binding functionality.

aCD40 CMV Internalizes Into DCs Upon CD40 Binding
To ensure endosomal processing of the attached peptide and efficient cross-presentation, DC-targeting antibodies need to internalize after binding their cognate receptors. Therefore, we analyzed the ability of DCs to internalize aCD40 CMV by fluorescence microscopy (Figure 2). For this, we incubated iDCs and mDCs with an aCD40 CMV antibody directly labeled with AF594. Specific internalization by targeting CD40 was compared to non-specific internalization using the aHer2 CMV control. At 4°C, no internalization occurred, as indicated by a colocalization with HLA-DR on the cell surface. At 37°C, specific internalization mediated by CD40 was detected within 1.5 h, suggesting a specific mechanism mediated by CD40. aCD40 CMV translocated specifically to EEA1 + early endosomal compartments, but not to LAMP1 + late endosomes ( Figure  2A). This agrees with data reported for another aCD40 clone (S2C6) (12,13). Only minimal non-specific internalization was visible using the aHer2 CMV -AF594 control for all conditions ( Figure 2B). We also observed no obvious difference between iDCs and mDCs.

The Low aCD40 CMV -Mediated Upregulation of Maturation Markers on iDCs Is Highly Amplified by Genetic Fusion With a Flagellin Domain
The antibody clone used in this work was previously characterized to bind agonistically to CD40 in its initial IgG2 format (37,38). Therefore, we expected that maturation of DCs is induced by the aCD40 antibody on its own and, if our approach works as anticipated, can be further enhanced by ligation of TLR5 by aCD40.Flg CMV . To investigate TLR5specific effects, the mutations R90A and E114A (mFlg) were introduced to the flagellin domain, which have been individually or in combination with further mutations shown to drastically reduce flagellin binding to TLR5 (46)(47)(48). To test DC-maturating activities, we incubated iDCs for 24 h with aCD40 CMV , aCD40.Flg CMV , aCD40.mFlg CMV or the respective aHer2 controls. TLR7/8-maturated mDCs were included as a positive control (41,42). Subsequently, we analyzed the DC surface markers CD80, CD83, CD86, and HLA-DR, as well as the secretion of IL-6 by flow cytometry as responses to the maturating stimuli ( Figure 3A). aCD40 CMV itself led to a slight, but significant upregulation of most maturation markers compared to the aHer2 control, indicating a perceptible intrinsic agonistic activity. Treatment of iDCs with aCD40.Flg CMV , in contrast, caused a strong and significant upregulation of all maturation markers and IL-6 secretion compared to aCD40 CMV . These effects were abolished by introducing the two crucial mutations into the flagellin portion.
To investigate whether the genetic fusion of aCD40 CMV and flagellin maintains functionalities of both fusion partners, we compared the activities of aCD40.Flg CMV to co-administered Fc.Flg and aCD40 CMV (Supplementary Figure S2A). Even if a difference between the two variants was observed for the aHer2 control molecule, treatment with aCD40.Flg CMV and also the combination of Fc.Flg and aCD40 CMV led to similar DC maturation states, as reflected by the increased expression of most of the surface markers. This indicates that fusion of flagellin to aCD40 CMV does not greatly affect its interplay with TLR5 and that neither antibody nor flagellin integrity is impaired.
To verify that the fused flagellin domain activates DCs specifically via TLR5, we studied the activation potential of aCD40.Flg CMV with and without TLR5-receptor blockage and the induction of TLR5-specific downstream signaling. First, DC maturation triggered by flagellin was investigated in the presence of an aTLR5-blocking and neutralizing IgA antibody. To this end, iDCs were pre-incubated with either the blocking antibody or the isotype control followed by the addition of aCD40 CMV , aCD40.Flg CMV , aCD40.mFlg CMV , or Fc.Flg for 24 h. The flagellin-induced upregulation of DC maturation markers was greatly diminished by the presence of the aTLR5 antibody, but not by the isotype control ( Figure 3B). Therefore, the TLR5 interaction is necessary for stimulation of DC maturation by aCD40.Flg CMV and Fc.Flg. Next, the activation of TLR5-specific downstream signaling processes by the proprietary flagellin fusion molecules was studied. For this purpose, a hTLR5-HEK293 reporter cell line, transiently transfected with full-length human TLR5, was incubated with abovementioned fusion proteins and TLR5-specific signal transduction was quantitated based on IL-8 secretion. As expected, aCD40.Flg CMV and Fc.Flg induced TLR5 signaling as shown by similar levels of IL-8 secretion, whereas aCD40.mFlg CMV did not ( Figure 3C).
Taken together, the results show that aCD40.Flg CMV significantly enhanced DC maturation compared to aCD40 CMV by the specific interaction of flagellin with TLR5.

CMV-Derived Peptides Targeted to DCs via aCD40 are Efficiently Cross-Presented and Induce Activation and Proliferation of Antigen-Specific T Cells in DC-T Cell Co-Cultures
We addressed the question of whether the CMV-derived peptide fused to aCD40 is correctly processed into the epitope sequence NLVPMVATV (CMV NLV ) and subsequently cross-presented on the DC surface via MHC-I. The interaction with peptideresponsive T cells was validated based on T cell activation and proliferation in autologous and allogeneic DC-T cell co-culture experiments ( Figure 4A). Prior to co-culturing with T cells, iDCs or mDCs from HLA-A*02:01 + donors were pre-loaded with either aCD40 CMV or aHer2 CMV . To rule out the possibility of T cell activation triggered by aCD40-mediated DC maturation and to confirm antigen specificity, an aCD40 antibody coupled with a non-stimulating peptide (aCD40 mNPM1 , vide infra) was used as a control in the allogeneic setting. T cell functionality was validated by pulsing the DCs with the already processed CMV NLV peptide. First, cross-presentation and T cell interaction were investigated by incubating pre-loaded DCs with allogeneic and expanded A B FIGURE 2 | aCD40 CMV internalized into early endosomal compartments upon binding to CD40. Internalization of aCD40 CMV -AF594 (A) or of the non-specific binding control aHer2 CMV -AF594 (B) into iDCs and mDCs after incubation for 1.5 h at 4°C and 37°C. Co-staining with aHLA-DR-AF488 (membrane) was performed at 4°C and with aHer2 CMV at 37°C. aCD40 CMV -treated DCs were stained with aLAMP1-AF488 (late endosomes) and aEEA1-AF488 (early endosomes) at 37°C. Internalization was visualized by structured illumination microscopy. Fluorescent signal within enlarged images is enhanced until saturation of the endosomal signals to better visualize co-localization. The scale bar represents 5 µm and in the enlarged images 0.5 µm. CMV NLV -specific T cells ( Figure 4B). After co-culture for 4 h, T cell activation was measured by intracellular IFN-g and TNF-a staining of CD8 + T cells. Independently of the maturation state, aCD40 CMVloaded DCs elicited significantly higher T cell activation compared to aCD40 mNPM1 and aHer2 CMV controls. As expected, a high level of secretion of pro-inflammatory cytokines was achieved with CMV NLV peptide-pulsed DCs that do not require internalization, processing and cross-presentation. Next, we examined, whether T cells were stimulated sufficiently to proliferate after encountering CMV NLV -presenting DCs ( Figure 4C). To this end, autologous DCs from an HLA-A*02:01 + donor with a preceding CMV infection were cultivated together with the non-adherent PBMC fraction (NACs) as a source of T cells. After 6 days of culture, CMV NLVspecific dextramer staining was performed to determine the percentage of expanded CMV NLV -responsive CD8 + T cells. In line with the results obtained for allogeneic co-cultures, loading of iDCs and mDCs with aCD40 CMV was accompanied by significantly greater CMV NLV -specific T cell proliferation compared to control  molecules. T cell proliferation was at the highest level after interaction with CMV NLV -loaded DCs. In summary, targeting CMV peptides to DCs via aCD40 leads to antigen processing into the correct epitope and crosspresentation and induces activation and proliferation of antigenspecific T cells. This validates the targeting approach and that the design affords functional constructs.

Incubation of DCs With aCD40.Flg CMV Elicits a Significantly Higher CMV NLV -Specific T Cell Activation and Proliferation Compared to aCD40 CMV in DC-T Cell Co-Cultures
Having validated the cross-presentation approach, we further analyzed whether the maturating activity of flagellin on DCs enhances cross-presentation and the T cell response. Accordingly, co-culture experiments were performed with allogeneic CMV NLVresponsive T cells or autologous NACs and iDCs that were preloaded with aCD40 CMV , aCD40.Flg CMV , aCD40.mFlg CMV , or the respective aHer2 controls. Fusion of flagellin to aCD40 elicited a significantly higher level of IFN-g and TNF-a secretion of antigenresponsive allogeneic T cells compared to the control that does not ligate TLR5 ( Figure 5A). Moreover, loading of DCs with aCD40.Flg CMV significantly increased the percentage of CMV NLV -specific CD8 + T cells in co-cultures with autologous NACs in comparison to aCD40 CMV ( Figure 5B). No difference in T cell proliferation was measured between aCD40.Flg CMV and the co-administered Fc.Flg and aCD40 CMV , highlighting that coupling of the activating flagellin domain to the DC-targeting aCD40 antibody does not alter its functionality (Supplementary Figure S2B). In all experiments, mutations in the TLR5-binding region (aCD40.mFlg CMV ) abolished benefits attributed to the flagellin domain. Conclusively, the delivery of antigens to DCs by a CD40-targeting antibody and the simultaneous activation by flagellin shows a clear advantage over aCD40 CMV , not only in terms of DC maturation but also with respect to T cell activation and proliferation.

Flagellin Fusion Also Enhances the Activation of mNPM1-Specific T cells in DC-T Cell Co-Cultures
To replicate these results of successfully enhanced T cell activation by flagellin fusion proteins in the AML setting, we A B C FIGURE 4 | aCD40 CMV loaded DCs efficiently cross-presented the peptides on the surface and induced activation and proliferation of antigen-responsive T cells.
(A) Co-cultures of iDCs/mDCs and allogeneic CMV NLV -specific T cells or with autologous NACs from HLA-A*02:01 + donors with a preceding CMV infection. DCs were pre-incubated with different antibody constructs or with peptide controls. In the allogeneic setting, T cell activation was analysed by counting IFN-g-and TNF-aproducing CD8 + cells. In the autologous setting, T cell proliferation was determined by CMV NLV -specific dextramer staining of CD8 + T cells. (B) Co-culture of allogeneic CMV NLV -specific T cells and HLA-A*02:01 + iDCs or mDCs. The DCs were pre-incubated with aHer2 CMV , aCD40 CMV , aCD40 conjugated to a control peptide (aCD40 mNPM1 ) or with the processed CMV NLV peptide. As a readout, T cell activation was measured by flow cytometry (n = 10 donors). (C) Co-culture of autologous NACs with iDCs and mDCs pre-loaded with aHer2 CMV , aCD40 CMV and the processed CMV NLV peptide. As readout, T cell proliferation was analysed by flow cytometry (n = 10 donors). T cell activation and proliferation are normalized to w/o Ab for each donor, respectively. Bars represent mean ± SEM. For statistical analysis, a Wilcoxon-signed rank test was applied. *p < 0.05, **p < 0.01.

Schmitt et al.
Antibody-Antigen-Flagellin Conjugate for AML Frontiers in Immunology | www.frontiersin.org November 2020 | Volume 11 | Article 602802 exchanged the model antigen CMV NLV for an AML-specific neoantigen. Accordingly, we fused the C-terminal sequence of the mutated NPM1 (mNPM1)-derived extended protein as a n t i g e n d o m a i n t o t h e C D 4 0 -t a r g e t i n g a n t i b o d y (aCD40 mNPM1 ). The protein results from a frameshift mutation and includes the mNPM1 288-296 neoepitope (CLAVEEVSL, mNPM1 CLA ), characteristic for AML (44). In a second step, the aCD40.Flg mNPM1 construct was generated to explore the effect of TLR5 ligation on the efficiency of induced specific T cell responses. First, we investigated whether iDCs and mDCs incubated with aCD40 mNPM1 cross-present the processed mNPM1 CLA on their surface and activate peptide-responsive T cells. For target cells, we used expanded allogeneic T cells transduced with an mNPM1 CLA -specific TCR generated by Lee et al. that were shown to recognize HLA-A*02:01 + and mNPM1 + AML cell lines and primary patient cells (44). We co-cultured those T cells with DCs that had been pre-incubated with aHer2 mNPM1 , aCD40 mNPM1 or aCD40 conjugated with a non-stimulating control peptide (aCD40 CMV ). T cell functionality was validated using mNPM1 CLA peptide-pulsed DCs. aCD40 mNPM1 -loaded iDCs and mDCs elicited significantly greater T cell activation compared to those treated with aCD40 and aHer2 controls ( Figure 6A). A high level of cytokine secretion was observed for the mNPM1 CLA peptide control, further confirming the specificity of the T cells. Finally, we investigated whether simultaneous TLR5 activation by aCD40.Flg mNPM1 favors a T cell response as previously observed in the CMV setting. In fact, in co-culture experiments with mNPM1 CLA -responsive T cells, iDCs loaded with aCD40.Flg mNPM1 significantly increased the percentage of IFN-g-and TNF-a-secreting CD8 + T cells compared to aCD40 mNPM1 ( Figure 6B).
These results demonstrate that mNPM1 peptides targeted to DCs by an aCD40 mNPM1 antibody were processed and crosspresented efficiently on MHC-I molecules, thereby activating antigen-specific CD8 + T cells. Importantly, DC stimulation by TLR5 ligation was able to boost the T cell response. Thus, the proof-of-principle based on CMV was replicated in an AML setting with aCD40 mNPM1 and aCD40.Flg mNPM1 showing that this approach has the potential to be developed into a therapy for AML.

DISCUSSION
The aim of this study was to generate a multifunctional antibody construct to co-deliver an antigen and adjuvant to CD40 + DCs for the induction of a peptide-specific T cell response. This construct was designed to combine the potential advantages of in vivo DC vaccination with those of targeted delivery of TLR agonists, in this case a protein agonist of TLR5, which is easily amenable to protein fusion constructs.
Our in vitro experiments demonstrated that aCD40 CMV presenting CMV as a model antigen for proof-of-principle studies binds specifically to CD40. However, the simple fusion construct induces only a low upregulation of DC maturation markers upon binding. In contrast, the enhanced fusion construct containing a flagellin domain as TLR5 agonist A B FIGURE 5 | Flagellin fusion significantly enhanced CMV-specific T cell activation and proliferation. (A) Allogeneic DC-T cell co-culture of expanded and pre-activated CMV NLV -specific T cells and HLA-A*02:01 + iDCs. DCs were pre-incubated with aHer2.Flg CMV , aCD40.Flg CMV and controls with mutated or no flagellin. As readout for T cell activation, IFN-g-and TNF-a-producing CD8 + T cells was counted by flow cytometry (n = 10 donors). (B) iDCs of an HLA-A*02:01 + and CMV + donor were pre-incubated with aCD40 CMV , aCD40.Flg CMV , and aCD40.mFlg CMV or the respective aHer2 controls and subsequently co-cultivated with autologous NACs. As a readout for T cell proliferation, CMV NLV -specific dextramer staining was performed and analyzed by flow cytometry (n = 10 donors). T cell activation and proliferation are normalized to the aHer2 control for each donor, respectively. Bars represent mean ± SEM. For statistical analysis, a Wilcoxon-signed rank test was applied. **p < 0.01. provided a much higher activation potential of DCs and a better activation of antigen-specific T cells. The variable regions of the aCD40 portion used in this study are derived from the IgG2-type aCD40 clone CP-870,893 (selicrelumab) with high agonistic activity. Severe treatment-related adverse events such as cytokine release syndrome and hepatotoxicity have been reported for the parental antibody, which has been investigated previously as a therapy against advanced solid tumors, limiting the dosage of treatment (37,49). In addition, selicrelumab also induced chronic B cell activation associated with diminished circulating T cell numbers. This effect potentially resulted from activation-induced cell death as observed for other agonistic aCD40 antibodies along with T cell tolerance (50)(51)(52). This side effect might be particularly undesirable if those antibodies are to be used in the context of DC vaccines and limit their efficacy. In contrast, in the Fc-silenced IgG1 format as used in our assays, the aCD40 antibody only exhibits a low activating potential. This is consistent with the data from White et al. and Dahan et al. that indicate that the potency of agonistic CD40 antibodies is not only dependent on the epitope recognized, but can be influenced by different hinge regions and the level of Fc-mediated crosslinking (53,54). Furthermore, we observed by structured illumination microscopy that aCD40 CMV internalized into early endosomal compartments after binding, which is in line with published data for a different aCD40 clone (12,13). Chatterjee et al. have shown that the beneficial effect of antigen-delivery to CD40 on MHC-I cross-presentation and CD8 + T cell responses is predominantly promoted by the intracellular trafficking pathway and not by its activating potential (12). Thus, concerning the possibility of adverse events and impaired T cell function, our aCD40 Fcsilenced IgG1 antibody should combine the advantage of low intrinsic agonistic activity with the benefit of targeting of antigens to CD40 and their intracellular destination.
Since the low aCD40-mediated DC activation is likely to be insufficient for lasting T cell responses, a TLR agonist was included in the molecule as an adjuvant. We chose to target TLR5 since flagellin and its derivatives have been successfully used as vaccine adjuvants and have been shown to induce potent anti-viral and anti-tumor immune responses in animal studies and clinical trials (55)(56)(57)(58)(59)(60)(61). Vicente-Suarez et al. reported the ability of flagellin to convert tolerogenic DCs into activating APCs that preferentially induce Th1 responses in pre-clinical models (62). We selected a truncated version of flagellin similar to the clinically advanced molecule entolimod that retains the essential terminal TLR5-binding domains (63). In contrast to other TLR agonists, entolimod exhibited a unique and beneficial safety profile due to the restricted pattern of TLR5 expression in distinct organs and the nature of cytokines induced following TLR5 stimulation. In our experiments, the fusion of the truncated flagellin domain to aCD40 CMV was functional, upregulated maturation markers on DCs via TLR5 activation, and was much more effective than the simple aCD40 construct.
Autologous and allogeneic DC-T cell co-culture experiments demonstrated that aCD40 CMV -loaded DCs effectively crosspresented CMV-derived peptides on their surfaces and interacted with CMV NLV -responsive T cells. This effect was significantly enhanced if T cells were cultivated with TLR5stimulated DCs that had been pre-incubated with A B FIGURE 6 | mNPM1-specific peptides fused to aCD40 were efficiently cross-presented on the DC surface and induced activation of mNPM1-responsive T cells that was significantly enhanced by TLR5 ligation. (A) Allogeneic DC-T cell co-culture of mNPM1 CLA -specific T cells and HLA-A*02:01 + iDCs. DCs were pre-incubated with aHer2 mNPM1 , aCD40 mNPM1 , aCD40 conjugated to a control peptide (aCD40 CMV ) or with the processed mNPM1 CLA peptide. As a readout for T cell activation, IFN-g-and TNF-a-producing CD8 + T cells were counted by flow cytometry (n = 10 donors). (B) Allogeneic DC-T cell co-culture of mNPM1 CLA -specific T cells and HLA-A*02:01 + iDCs. DCs were pre-incubated with aCD40 mNPM1 , aCD40.Flg mNPM1 , aCD40.mFlg mNPM1 or the respective aHer2 controls. T cell activation was measured by flow cytometry (n = 10 donors) and normalized to w/o Ab or to the aHer2 control for each donor, respectively. Bars represent mean ± SEM. For statistical analysis, a Wilcoxon-signed rank test was applied. **p < 0.01. aCD40.Flg CMV . When we compared the immuno-stimulatory capacity of aCD40.Flg CMV with separate administration of aCD40 CMV and an Fc-fused variant of flagellin, no clear difference was detected. Thus, there was no benefit in specific targeting in vitro under optimal conditions, which was an outcome to be expected. However, targeted delivery of TLR agonists in the context of protein fusion is likely of higher relevance in vivo minimizing dilution and buffering effects and potentially permitting lower dosages. As potential side effects at therapeutically relevant doses need to be investigated in vivo, mouse studies are indicated to support our hypothesis of reduced side effects and enhanced therapeutic efficacy. Still, in vitro data showed that the structural integrity of both aCD40 and flagellin is unaffected by their genetic fusing. This was further confirmed by binding studies and thermal unfolding measurements, which revealed that aCD40 CMV and aCD40.Flg CMV have the same maximum binding and melting behavior. This is an important and novel finding, since other sites of genetic fusion or chemical modification of the TLR agonist have been shown to affect its physicochemical protein properties. In a similar approach, Kreutz et al. developed an antibody-antigen-adjuvant conjugate consisting of an aDec205-specific antibody conjugated via sulfo-SMCC linkers to the model antigen ovalbumin and CpG oligodeoxynucleotides. They observed that the CpG-conjugate was more potent at inducing cytotoxic T cell responses compared to the separate components when co-administered in vivo. Nevertheless, they encountered the problem that antibody binding and uptake was altered by the fusion with CpG allowing the delivery of both antigen and adjuvant to cells to be partially independent of the DC-targeting antibody (32). This was not observed for our antibody, probably due to its sitespecific C-terminal fusion of the TLR agonist to the heavy chains with linkers of adequate length. The multifunctional antibody construct was further validated using a neoantigen-derived antigen domain, making this the first study to investigate neoantigen delivery to a DC-targeting antibody construct. Applying neoantigens in a vaccination approach is exceptionally interesting as they arise from tumorspecific mutations and reactive T cells have not undergone clonal deletion. In the case of AML, neoantigens derived from NPM1 mutations are promising targets, as they occur in 30-35% of all AML patients and thus cover a broad spectrum of patients. Recently, Van der Lee et al. identified the mNPM1-derived HLA-A*02:01-specific neoepitope CLAVEEVSL among others by mass spectrometry analysis (44). Specific T cells were found in healthy individuals and a TCR with high reactivity was isolated. As a therapeutic approach, TCR-transduced T cells elicited antitumor efficacy in a mouse model and were able to recognize and kill mNPM1 + AML cell lines and primary patient samples in vitro. We generated DC-targeting antibody constructs with mNPM1 as an antigen domain, which is the first investigation of mNPM1 in a DC vaccination concept. In our experiments, DCs loaded with aCD40 mNPM1 , aCD40.Flg mNPM1 or with the processed mNPM1 CLA peptide potently activated mNPM1 CLAspecific T cell responses. Further experiments are required to evaluate whether our molecules are able to expand the low abundance of mNPM1-specific T cells present in AML patients.
We demonstrated that the multifunctional antibody constructs aCD40.Flg CMV and aCD40.Flg mNPM1 are highly potent activators of peptide-specific T cells. In particular, targeting mNPM1 to DCs is of utmost therapeutic significance. The in vitro experiments performed in this study show promising initial results for the development of aCD40.Flg mNPM1 as an in vivo DC vaccine strategy for AML. Future studies should examine the anti-tumor effects of this molecule as monotherapy and in combination with potential synergistic immunotherapies in AML xenograft models (64). The conjunction of neoepitope-based vaccines with immune checkpoint inhibition has resulted in sustained progression-free survival of cancer patients in clinical trials (4,5). In addition, several clinical trials have explored the potential of post-transfer vaccination to enhance the clinical efficacy of adoptively transferred T cells expressing a TCR specific for an intracellular tumor antigen (65,66). In line with this, combining aCD40.Flg mNPM1 with TCR gene therapy, in particular with mNPM1-specific T cells that recognize primary AML cells, would potentially show great promise to expand and enhance the efficiency of transferred T cells in vivo and to generate longlasting responses (44).
Given the encouraging results of the molecule format in both the CMV and the mNPM1 setting, it is noteworthy, that these constructs could also serve as a vaccination platform to be potentially applied to any other (tumor) peptide as antigen domain.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Ethics Committee at the LMU Munich. The patients/participants provided their written informed consent to participate in this study.