Targeted delivery of TLR3 agonist to tumor cells with single chain antibody fragment-conjugated nanoparticles induces type I-interferon response and apoptosis

Application of Toll-like receptor (TLR) agonists is a promising approach to treat cancer. In particular, nucleic acid-based TLR agonists such as short ssRNA and dsRNA molecules, which activate endosomal TLRs, can be delivered to tumors by use of nanoparticle delivery systems. However, such delivery systems bear unspecific side effects and poor pharmacokinetics. To overcome these limitations we developed a system for targeted delivery of a 50 bp dsRNA TLR3 agonist (Riboxxol) to treat PSCA-positive tumor cells, which consists of neutravidin conjugated to mono-biotinylated dsRNA and to humanized mono-biotinylated anti-PSCA single chain antibody derivative scFv(h-AM1)-BAP. The assembly of the components resulted in the formation of nanoparticle-like immunoconjugates designated Rapid Inducer of Cellular Inflammation and Apoptosis (RICIA). Anti-PSCA-RICIA exclusively delivered Riboxxol to PSCA-positive tumor cells as well as subcutaneous tumors. Uptake of anti-PSCA-RICIA induced a type I-interferon response and apoptosis in HEK-BluehTLR3/PSCA reporter cells and PSCA-positive HT1376 bladder cancer cells in vitro. No such effects were observed when using RICIA coupled to an unspecific control antibody or when using Riboxxol alone. Treatment of HT1376 xenografts in immune-deficient hosts with targeted delivery of TLR3 agonist did not induce adverse effects and only modestly inhibited tumor growth when compared to controls. These results suggest promising activation of innate immune response and apoptosis upon selective delivery of TLR3 agonists in tumor cells. Yet, further studies using syngeneic and orthotopic tumor models are needed to fully exploit the potential of RICIA immunoconjugates.


Results
Synthesis and functional characterization of humanized anti-PSCA scFv(h-AM1) and mono-biotinylated scFv(h-AM1)-BAP. Recently, we described a biotin-neutravidin-based modular system for antibody-mediated delivery of polycationic (poly)propyleneimine siRNA carrier molecules to tumor cells 26 . For targeted delivery of TLR3 agonist to tumor cells we sought to employ this platform technology and to couple biotinylated nucleic acid-based TLR agonists, in particular dsRNA, to tumor-specific antibodies via consecutive bio-conjugation to neutravidin. As a model tumor-antigen we choose PSCA, a surface protein which is found on several tumor entities [18][19][20][21][22][23][24] . For a potential clinical application we furthermore engineered an humanized scFv based on our recently described mono-biotinylated anti-PSCA scFv(AM1)-P-BAP 26 . For this we engrafted the complimentary determining regions (CDRs) of the murine scFv(AM1) in human germline Ig genes as outlined in the method section. We furthermore, exchanged the Propionibacterium shermanii transcarboxylase-derived biotin acceptor peptide (P-BAP) with a short 22 amino acid synthetic BAP to minimize potential immune responses of the host. As control for our studies we furthermore generated scFv(MR1.1)-BAP, which specifically binds a neo-epitope of the oncogenic epidermal growth factor receptor variant III (EGFRvIII). EGFRvIII is not present in healthy normal cells and the target cell lines used for this study, respectively. The structures of the resulting scFv-BAPs as well of the parental scFv are shown in Fig. 1a. All antibody constructs contained an N-terminal Igκleader for the extracellular secretion as well as a C-terminal c-myc epitope and 6x histidine (His) tag for detection and purification, respectively.
As demonstrated in Coomassie-stained polyacrylamide gels, the recombinant scFv(h-AM1)-BAP and scFv(h-AM1), the latter devoid of the BAP, were produced in high purity as full-length proteins, with bands at 50 kDa and 38 kDa, which is slightly higher than the respective calculated molecular masses (Fig. 1b). The increased molecular sizes of the antibody derivatives might be due to posttranslational glycosylation. Furthermore, besides the detection of the c-myc epitope, a biotin-specific antibody demonstrated the efficient site specific biotinylation of the humanized scFv(h-AM1)-BAP. As expected, the parental scFv(h-AM1) was devoid of biotin residues (Fig. 1c). Similar results were obtained with scFv(MR1.1)-BAP control antibodies (data not shown). The engrafting of the CDRs of the murine antibody into the framework region of human Ig germline genes did not affect the specificity of scFv(h-AM1) towards PSCA, as investigated in flow cytometry using HEK293T PSCA cells (Fig. 1d). The EGFRvIII-specific control antibody bound to EGFRvIII-positive HEK293T EGFRvIII cells but as expected not to HEK293T PSCA cells. By using a PE-labeled biotin-specific antibody we furthermore demonstrated that the C-terminal biotin residue of scFv(h-AM1)-BAP and scFv(MR1.1)-BAP was accessible under native conditions, respectively (Fig. 1e). Unexpectedly, the murine scFv(AM1) acquired an improved affinity after humanization (Supplementary Tab. 1). A detailed comparison of the amino acid sequences of the murine and the humanized scFvs revealed subtle changes in potential O-glycosylation sites which in part might account for the improved affinity of the humanized anti-PSCA single chain antibody fragment ( Supplementary Fig. 2).

Crosslinking of scFv(h-AM1) on HEK293T PSCA cells causes internalization of PSCA.
Internalization of PSCA after crosslinking with nanoparticles is a prerequisite for the antibody-mediated delivery of TLR3 agonist. Therefore, it was of special interest whether crosslinking of at least two PSCA-molecules on the cell surface could induce the internalization of PSCA. Indeed, the crosslinking of scFv(h-AM1) with a biotinylated bivalent anti-c-myc-biotin antibody directed against the C-terminal c-myc epitope of antibody construct induced a significant time-dependent internalization of PSCA in HEK293T PSCA cells, as assessed by staining with a tertiary anti-biotin-PE antibody (Fig. 2a). In contrast, the treatment of HEK293T PSCA cells with monovalent scFv(h-AM1) antibodies alone did not induce a receptor internalization. The internalization of PSCA after crosslinking was calculated by the mean fluorescence intensity (MFI) for PSCA immunosignals (Fig. 2b). During the whole treatment period surface PSCA-expression levels did not completely disappeared after crosslinking as exemplified in representative flow cytometry histograms (Fig. 2a) which might be due to recycling of the receptor to the cell surface.
Assembly and analysis of "Rapid Inducer of Cellular Inflammation and Apoptosis" (RICIA) immunoconjugates. The proper conjugation of biotinylated scFv(h-AM1)-BAPs to (neutr)avidin was assessed by gel shift assays. Here, a constant amount of scFv(h-AM1)-BAP was incubated with decreasing numbers of avidin molecules, resulting in increased molar ratios of scFv(h-AM1)-BAP to avidin (Fig. 3a). The scFv(h-AM1)-BAP/avidin conjugates were detected at molecular masses of approximately 120 kDa and 180 kDa via the C-terminal c-myc epitope of the scFv(h-AM1)-BAPs. Noteworthy, the conjugation with avidin essentially depleted the pool of free biotinylated scFv(h-AM1)-BAP molecules at scFv(h-AM1)-BAP to avidin ratio of 1:1 to 4:1 indicating an efficient conjugation. The experiments which were performed with scFv(MR1.1)-BAP revealed similar results (data not shown).
To produce immunoconjugates delivering TLR3 agonist to target cells with cognate receptor expression neutravidin was consecutively bio-conjugated to mono-biotinylated scFv(h-AM1) and mono-biotinylated Riboxxol as depicted in Fig. 3b. Subsequent in situ atomic force microscopy (AFM) was done after 8 h of fabrication (Fig. 3c). The diameters of immunoconjugates were found in the range of 22-55 nm.
Anti-PSCA-RICIA specifically delivers 50 bp dsRNA TLR3 agonist into PSCA-expressing cells and induces a type I-interferon response and apoptosis. To evaluate whether the selective delivery of TLR3 agonist by RICIA induces activation of downstream signaling and type I-interferon response we generated HEK-Blue hTLR3/PSCA reporter cells with ectopic expression of PSCA. Using this reporter cell line it was possible to develop immunoconjugate nanoparticles by a step-by-step assembly approach including testing of different ratios of components. Since permanent prostate cancer cell lines with endogenous PSCA expression have not been described so far, HT1376 bladder carcinoma cells, which have been identified to express robust amounts of surface PSCA, were chosen for succeeding experiments 27 . HEK-Blue hTLR3/PSCA reporter cells show an improved expression of surface PSCA when compared to the HT1376 bladder cancer cells (Fig. 4a).
To assess the selective delivery of RICIA immunoconjugates we labeled the coupled dsRNA of the RICIA with FITC-mal19-PPI as described in the method section. In a previous study we have shown that the cationic glycodendrimer mal19-PPI was biocompatible, lost transfection capacity but still was competent in binding to nucleic acids 26 . Subsequent confocal laser scanning microscopy revealed that HEK-Blue hTLR3/PSCA cells internalized FITC-labeled scFv(h-AM1)-BAP-containing RICIA whereas no FITC signals were detected when cells were treated with RICIA immunoconjugates coupled to scFv(MR1.1)-BAP (Fig. 4b).
Notably, 24 h treatment with scFv(h-AM1)-BAP-containing RICIA induced a dose dependent expression and secretion of the reporter SEAP by the HEK-Blue hTLR3/PSCA reporter cells indicated by markedly increased OD of the cell culture medium at 655 nm indicating SEAP-mediated p-nitrophenyl phosphate substrate conversion (Fig. 4c). For the anti-PSCA-RICIA an half maximal inhibitory concentration (IC50) of 12.5 nM was calculated based on the SEAP reporter gene assay. Strikingly, treatment with scFv(MR1.1)-BAP-containing control RICIA www.nature.com/scientificreports www.nature.com/scientificreports/ as well as treatment with Riboxxol-biotin alone barely induced SEAP reporter gene expression. This even more indicates, the need for a cognate ligand for targeted delivery of RICIA immunoconjugates, while Riboxxol-biotin cannot easily pass cell membranes due to its net charge.
In order to address the route of the anti-PSCA-RICIA internalization in more detail, the same experiments were performed in presence of inhibitors of endocytosis. As shown in Fig. 4d the incubation of cells with chlorpromazine, which blocks clathrin-dependent endocytosis 22 , markedly inhibited SEAP activity in RICIA-treated cells. Weaker but still significant effects were observed when cells were treated with filipin III, an inhibitor of caveolae-dependent and lipid raft-mediated endocytosis 24 . Whether the targeted delivery of TLR3 into endosomes of target cells can induce a type I-interferon response was investigated by enzyme-linked immunosorbent assay. For this, HEK-Blue hTLR3/PSCA cells were treated for 36 h with increasing amounts of RICIA containing concentrations of Riboxxol-biotin ranging from 0 to 285.7 nM. Also poly(I:C), a TLR3 agonist commonly used for activation of innate immune responses, was included in the experiments. As shown in Fig. 4e, the treatment of HEK-Blue hTLR3/PSCA cells with increasing doses of anti-PSCA-RICIA induced a dose-dependent release of IFN-β. Interestingly, in the chosen experimental setting treatment with 50 µg/ml poly(I:C) did not result in IFN-β release by the HEK-Blue hTLR3/PSCA cells. This lack of an induction of immune response in the cells might be related to the net charge of poly(I:C) which precludes diffusion through cell membranes.
Since it was of further interest whether the targeted delivery of TLR3 agonist Riboxxol affects survival of cells we performed Annexin V-FITC/PI staining of treated HEK-Blue hTLR3/PSCA cells. As already seen in phase contrast microscopy depicted in Fig. 5a, cells treated with scFv(h-AM1)-BAP-containing RICIA showed dying cells and increased cellular debris which was not seen in cells treated with PBS or RICIA containing the EGFRvIII-specific scFv(MR1.1)-BAP or Riboxxol-biotin alone. When analyzing the fraction of Annexin V-positive apoptotic cells by flow cytometry it became obvious that treatment of HEK-Blue hTLR3/PSCA with the scFv(h-AM1)-BAP-conjugated RICIA induced apoptosis compared to the control-RICIA and to treatment with Riboxxol-biotin alone (Fig. 5b,c). Similar results showing induction of innate immune response and apoptosis were seen when using HT1376 bladder cancer cells with endogenous expression of PSCA (see Fig. 6b,c). This cell line also endogenously expresses physiological levels of intracellular TLR3 as determined by flow cytometry analysis (Fig. 6a). HT1376 cells treated with increasing concentrations of scFv(h-AM1)-BAP-conjugated RICIA shows weaker induction of apoptosis ( Fig. 6b) but an analogous robust release of IFN-β ( Fig. 6c) when compared to the results obtained in HEK-Blue hTLR3/PSCA reporter cells.

Targeted delivery of TLR3 agonist in vivo.
In further experiments we sought to address whether the anti-PSCA-RICIA immunoconjugates can be delivered to PSCA-expressing tumors in vivo. First stability of Riboxxol and RICIA-bound Riboxxol-biotin in body fluids was investigated. More specifically, a mixture of Riboxxol-biotin and Riboxxol at molar ratio of 10:1 was complexed with neutravidin and scFv-BAPs and subjected to agarose gel electrophoresis. This approach allowed the simultaneous semi-quantitative assessment www.nature.com/scientificreports www.nature.com/scientificreports/ of Riboxxol-biotin embedded in RICIA immunoconjugates as well of non-conjugated free Riboxxol using the dsRNA-intercalating dye RedSafe. As depicted in Fig. 7a Riboxxol signals were detectable and remained unchanged after treatment with serum and urine, respectively, when compared to untreated controls. Similar results concerning stability were revealed with Riboxxol-biotin embedded in RICIAs. The RICIA immunoconjugates were not visible as distinct band but were detected as smear in the agarose gels. This result most likely reflects various nanoparticle sizes ranging from 22-55 nm and in part different net charges of immunoconjugates. Interestingly, when incubated with serum, RedSafe-stained bands >4,000 bp and between 1,000 bp and  www.nature.com/scientificreports www.nature.com/scientificreports/ 1,500 bp were seen, which most likely indicates the development of a serum-protein corona associated with RICIA immunoconjugates. To demonstrate the capability of our scFv(h-AM1)-BAP-containing RICIA immunoconjugates to deliver Riboxxol-biotin dsRNA in vivo, FITC-mal19-PPI-labeled RICIA were injected intraperitoneally in NMRI Foxn1nu/Foxn1nu mice xenotransplanted with HT1376 bladder cancer cells. This immune-deficient mouse strain is suitable for xenografting human tumors and was chosen for initial experiments. As depicted in Fig. 7b, the scFv(h-AM1)-BAP-conjugated RICIA were able to mark and enter tumor cells in the outer rim of the tumor specifically. In contrast, no FITC signals in tumors were observed when using EGFRvIII-specific scFv(MR1.1)-BAP-conjugated RICIA (Fig. 7c), thus indicating the specificity mediated by the targeting moiety. When using PSCA-specific RICIA as well as EGFRvIII-specific RICIA a small amount of FITC signals was detected in a punctuated pattern in liver and spleen of mice. These organs contain cells of the reticuloendothelial system such as Kupffer cells and spleen monocytes/macrophages which have been described to phagocytose several types of nanoparticles 25 . Although in vitro the apoptotic effect of anti-PSCA-RICIA nanoparticles on HT1376 cells was low (Fig. 6b), it was of special interest whether induction of apoptosis in vivo was sufficient to induce tumor growth control in HT1376 xenografted NMRI Foxn1nu/Foxn1nu mice. Transplanted HT1376 tumors grew slowly and treatment of mice with three intraperitoneally injections every 48 h resulted in a diminished www.nature.com/scientificreports www.nature.com/scientificreports/ although statistically not significant reduction of tumor growth rates when compared to mice treated with PBS or Riboxxol-biotin alone (Supplementary Fig. 3).

Discussion
Engagement of the innate immune system via activation of TLRs in both tumor and immune cells combined with the triggering of adaptive immune responses contains interesting prospects regarding the treatment of cancer.
In our study, we focused on TLR3 agonist Riboxxol, a 50 bp dsRNA, for immunotherapy of cancer. So far, there is accumulating evidence that activation of TLR3 in tumor cells and tumor stroma preferentially results in anti-tumor effects. For instance, treatment of human renal cancer cell lines with the synthetic TLR3 agonist poly(I:C) induced IFN-1β release and caused cell death 28 . Also treatment of human breast cancer cells with poly(I:C) has been demonstrated to induce a type I-interferon release and was accompanied by increased pro-apoptotic Bax levels and caspase-dependent apoptosis 29 . Similar results were obtained in a panel of human prostate cancer cells which after treatment with poly(I:C) succumbed to apoptosis and showed an type I-interferon response, respectively 30,31 . Interestingly, TLR3 activation in prostate cancer cell lines resulted in secretion of chemokines capable of attracting natural killer cells, granulocytes, B cells and T cells 32 . Recently, also www.nature.com/scientificreports www.nature.com/scientificreports/ anti-tumor effects of poly(I:C) treated ex vivo cultivated microglia derived from glioblastoma has been described. Notably, such (poly(I:C)) pre-activated tumor-infiltrating microglia cells were programmed into an inflammatory M1-like immune phenotype which enables them to kill glioblastoma cells and to phagocytose dying cancer cells, respectively 33 . www.nature.com/scientificreports www.nature.com/scientificreports/ Yet, when considering a systemic application of natural or synthetic TLR3 agonists to treat tumor cells, the targeted delivery of TLR3 agonists appears advantageous since off-target effects are limited. Such targeted delivery of TLR3 agonists can be accomplished by implementing an antibody or specific ligand for a surface tumor antigen, which upon crosslinking is endocytosed by the tumor cell and delivers the agonist to endosomal TLR3. To reach this objective, a simple modular system for delivery of biotinylated nucleic acids, designated "Rapid Inducer of Cellular Inflammation and Apoptosis" (RICIA) was developed. For a proof-of concept we focused on the tumor-associated surface antigen PSCA and employed the novel humanized mono-biotinylated anti-PSCA scFv(h-AM1)-BAP. Noteworthy, due to RICIAs modular character the scFv-BAP antibody derivatives and nucleic acid-based TLR agonists can be easily exchanged to address tumor heterogeneity or to treat different cancers.
Notably, in our experiments we demonstrate that simultaneous binding of at least two PSCA molecules on the cells surface by a crosslinked scFv(h-AM1) antibody significantly increased endocytosis of PSCA when compared to monovalent binding of scFv(h-AM1). This also provided the basis for efficient targeted delivery by mechanisms of receptor-mediated, preferential clathrin-dependent endocytosis of scFv(h-AM1)-BAP-conjugated RICIA. The assembly of RICIA immunoconjugates was accomplished by consecutive bio-conjugation of neutravidin with mono-biotinylated scFv-BAPs and mono-biotinylated Riboxxol-biotin and resulted in nanoparticles with diameters of 22 nm to 55 nm. The unexpected formation of such nanoparticles might be due to aggregation events of the scFv-BAPs or to the formation of alternative tertiary dsRNA structures. Although not investigated so far, it appears conceivable that the nanoparticle-like format of the immunoconjugates increases avidity to PSCA-positive target cells which might further facilitate their endocytosis.
The data, obtained in the present investigation using the TLR3 agonist Riboxxol, are in line with the above mentioned promising results using poly(I:C). More specifically, anti-tumor effects including a potent induction of a type I-interferon response as well as apoptosis were observed after targeted delivery of Riboxxol to PSCA-positive tumor cells using RICIA immunoconjugates. No such effects were revealed when using RICIA conjugated to a control antibody, which demonstrates the feasibility of the targeted delivery approach. Noteworthy, treatment of HEK-Blue hTLR3/PSCA reporter cells with naked poly(I:C) and Riboxxol-biotin, respectively, did not result in induction of a type I-interferon response nor in increased apoptosis ( Fig. 4e and 5). Yet, in contrast to professional phagocytizing cells (i.e. macrophages, DCs), HEK-Blue hTLR3/PSCA cells have only a weak phagocytic capacity which limits uptake. In addition, under physiological conditions and the chosen experimental settings poly(I:C) and Riboxxol-biotin cannot easily cross membranes due to their size and negative net charge. Therefore, previous studies utilized polycationic carrier molecules to transfect poly(I:C) into cancer cells. Such transfection of poly(I:C) and other TLR3 agonists causes disruption of endosomal membranes and result in activation of both endosomal (TLR3) and cytosolic (i.e. RIG-I, MDA5) dsRNA receptors 31 . We suggest a predominant TLR3 activation utilizing receptor-mediated endocytosis for delivery of Riboxxol. However, additional activation of cytosolic dsRNA receptors due to leakage or disruption of endosomal membranes upon RICIA-treatment cannot be fully ruled out. However, this might be advantageous since it could further improve induction of a type I-interferon response and apoptosis in targeted cancer cells.
In line with a previous report addressing the stability of a similar type of dsRNA 14 , Riboxxol, when assembled in RICIA, remains stable when treated with serum or urine. When administered intraperitoneally in nude mice, anti-PSCA-RICIA specifically accumulated in xenografted HT1376 tumors, whereas the treatment with EGFRvIII-specific RICIA, which were used as negative control, did not deliver Riboxxol to PSCA-positive tumors (Fig. 7b,c). Noteworthy, the in vivo experiments showed no anti-PSCA-as well as anti-EGFRvIII-RICIA signals in lung and kidneys and only some weak punctuated signals in liver and spleen. In the latter organs this likely is due to phagocytizing cells of the reticuloendothelial system (i.e. spleen monocytes/macrophages, Kupffer cells of the liver). Collectively, the data show that TLR3 agonist Riboxxol can be selectively delivered to xenografted tumor cells expressing the cognate surface receptor in vivo by assembled RICIA immunoconjugates. Of note, there is only 60% homology of human and mouse PSCA protein and the humanized anti-PSCA scFv(h-AM1) does not bind to murine PSCA (data not shown). When considering treatment of patients suffering from PSCA-positive cancers with anti-PSCA-RICIA, it can be predicted that tumorigenic as well as normal cells expressing PSCA at sufficient surface levels and containing TLR3 are affected by anti-PSCA-RICIA. It remains to be investigated whether on-target/off-tissue effects are limited in normal epithelial cells having predominant PSCA expression at the apical cell membrane. Otherwise, in de-differentiated epithelial cancer cells displaying a more de-polarized cell type, PSCA might be more accessible for anti-PSCA-RICIA. However, from a general point of view on-target/ off-tissue effects affecting the prostate or other PSCA-positive normal tissues or cell types cannot be excluded and must be addressed in proper risk assessment. In our study, the observed anti-tumoral effects, in particular the induction of a type I-interferon response, after targeted delivery of Riboxxol substantiates TLR3 agonists as promising candidates for cancer immunotherapy. Yet, efficacy of anti-PSCA-RICIA in immune-deficient mice bearing PSCA-positive human xenografts was limited (Supplementary Fig. 3). Therefore, future experiments using syngeneic mouse tumor models are needed to address a possible induction of adaptive immune responses and to assess on-target/off-tissue effects when applying anti-PSCA-RICIA.

Methods
Cell lines. The human embryonic kidney cell lines HEK293T, HEK293T PSCA and HEK293T EGFRvIII have been described previously 26 and were cultured in DMEM complete with 4.5 g/l glucose, 10% v/v heat-inactivated FCS, 100 U/ml penicillin, 100 µg/ml streptomycin and 10 mM HEPES (all from Life Technologies). HEK293T huBirA cells stable expressing codon-optimized biotin ligase huBirA 26  www.nature.com/scientificreports www.nature.com/scientificreports/ Invivogen). The bladder cancer cell line HT1376 with endogenous expression of PSCA was maintained in DMEM complete supplemented with 1X non-essential amino acids (Life Technologies). FreeStyle ™ 293-F cells were incubated at 37 °C and 8% CO 2 and all other cell lines with 5% CO 2 in a humidified incubator .

Development of humanized scFv(h-AM1). The humanized scFv(h-AM1) was designed in silicio by
engrafting the CDRs of the murine scFv(AM1) into framework regions of human Ig germline genes. The CDRs of the murine scFv(AM1) were identified using an algorithm described by North et al. 34 and were used to identify suitable human Ig germline genes by alignment with the software "IgBLAST" (NIH) 35,36 . CDRs of scFv(AM1)-V L were engrafted into the IGKV1-39*01 germline gene. Since no suitable framework region was identified for the C-terminus of scFv(AM1)-V H , the V H was only partially humanized by engrafting CDR1 and CDR2 into the IGHV3-23*03 germline gene. In an additional step the partially humanized scFv(AM1)-V H was engrafted into the IGHV1-NL1*01 germline gene resulting in a fully humanized scFv(AM1)-V H containing framework regions from IGHV3-23*03 and IGHV3-23*03.
Vector constructs. DNA encoding anti-PSCA scFv(h-AM1) fused to a N-terminal Igκ leader sequence, to C-terminal c-myc epitope and 6xHis tag 37  Western blot analysis. Purity of scFvs was determined by SDS-PAGE with 2.5-5 µg protein and subsequent staining of proteins with SimplyBlue TM SafeStain (Invitrogen). For immunoblot analysis 1.5 µg of scFv were separated by SDS-PAGE electrophoresis under reducing conditions. After protein transfer to a PVDF membrane (Whatman), scFvs were detected using a primary mouse anti-c-myc antibody (Invitrogen) and a secondary polyclonal rabbit anti-mouse IgG HRP conjugate (Dako). Biotinylated scFv-BAPs were detected with a HRP-conjugated anti-biotin antibody (Sigma-Aldrich). Visualization of scFvs was performed by Luminata Forte Western HRP substrate (Merck Millipore) and LAS 3000 chemoluminescence imager (FujiFilm). For analysis of scFv-BAP complexation with avidin, the two components were mixed at various molar ratios ranging from 1:2 to www.nature.com/scientificreports www.nature.com/scientificreports/ 8:1 in PBS. After 30 min incubation at room temperature scFv-BAP-avidin conjugates as well as non-conjugated biotinylated scFv-BAPs were analyzed by non-reducing SDS-PAGE and detected by Western blot analysis as described above.
Assembly and functional analysis of RICIA. RICIA were generated in the first step by conjugation of biotinylated scFv-BAPs with neutravidin (Pierce) for 30 min in PBS and in a second step with TLR3 agonist Riboxxol-biotin (riboxx life science) for 30 min in complexation buffer (10 mM HEPES, 150 mM NaCl, pH 7.4, Fig. 3b) in a final molar ratio of 2:1:2. For functional analysis of RICIA HEK-Blue detection system according to the manufacturer's instructions (Invivogen) was used. Briefly, RICIA based on 1 µg/ml (28.6 nM), 2.5 µg/ ml (71.4 nM) and 5 µg/ml (142.9 nM) Riboxxol-biotin or Riboxxol-biotin only were incubated with 5 × 10 4 HEK-Blue hTLR3/PSCA cells for 24 h. In another experiment 0.3 µg/ml filipin III or 3 µg/ml chlorpromazine (both Sigma-Aldrich) was added to block caveolae-and chlatrin-mediated endocytotic pathways, as described previously 26 . Hydrolysis of substrate by SEAP was quantified at 655 nm by Synergy 2 Multi-Mode Microplate Reader (BioTek Instruments). Values of endocytotic pathway analysis were normalized to untreated cells. To investigate the stability of RICIA immunoconjugates as well as of Riboxxol itself in body fluids, RICIA were generated based on 2 µg Riboxxol-biotin/Riboxxol (10:1) and incubated with either 50% serum or urine or in complexation buffer at 37 °C. RICIA and Riboxxol-biotin controls were stored at 4 °C. The stability after 0 h, 4 h, 8 h, and 24 h was assessed via agarose gel shift assays. After adding 1X DNA gel loading dye (Thermo Fisher Scientific) and subsequent electrophoresis, fluorescence of RedSafe (iNtRON Biotechnology)-stained RNA was detected using the UV transilluminator AlphaImager (AlphaInnotech).

Atomic force microscopy (AFM).
For in-situ AFM, silicon wafers were treated with O 2 -plasma to obtain a hydrophilic surface for the adsorption of RICIA based on 1 µg/ml Riboxxol-biotin. The AFM measurements in fluids were done in the peak force tapping mode by a Dimension ICON (Bruker-Nano). Silicon nitride sensors SCANASYST-FLUID+ (Bruker-Nano) with a nominal spring constant of 0.7 N/m −1 and a nominal tip radius of 2 nm were used. The particle size distribution was calculated by the software NanoScope Analysis (Bruker-Nano).
Confocal laser scanning microscopy. Cellular uptake of RICIA was analyzed using 2.5 × 10 5 HEK-Blue hTLR3/PSCA cells and FITC-labeled RICIA. FITC-labelling of RICIA was accomplished by complexing with maltose-modified 4 th generation FITC-poly(propyleneimine)-dendrimer (FITC-mal19-PPI) at a molar ratio of 1:2.5. Cells were treated with FITC-stained-RICIA for 24 h in DMEM complete medium containing FCS. Afterwards cells were fixed, stained (cell membrane with Texas Red-X conjugate of WGA and DNA with Hoechst 33342) and analyzed by confocal laser scanning microscope Leica SP5.
Delivery of RICIA to bladder cancer xenografts. The animal experiments were approved by the Landesdirektion Sachsen (animal experiment permission number: DD24-5131/354/14) and the methods were carried out in accordance with the approved guidelines and regulations. 2 × 10 6 HT1376 cells were subcutaneously implanted in the right flank of athymic female nude mice (NMRI Foxn1nu/Foxn1nu ). After reaching an average tumor diameter of 5-8 mm, mice were injected intraperitoneally with FITC-labeled mal19-PPI-RICIA, described above, containing 5 µg (142.9 nM) Riboxxol-biotin. Tumor, liver, lung, spleen, and kidney were resected 18 h after injection and 10 µm thick cryosections were prepared. After staining of DNA with Hoechst 33342, the slides were analyzed by Axio Examiner Z1 fluorescence microscope (Zeiss). For treatment with anti-PSCA-RICIA, HT1376 xenografts were established as described above and groups of animals with similar mean tumor volumes were treated with three intraperitoneal injections of PBS, 142.9 nM Riboxxol-biotin in PBS or anti-PSCA-RICIA containing 142.9 nM Riboxxol in total volumes of 200 µl at days 1, 3 and 5. Tumor sizes were measured in both directions by a digital caliper every two days. After calculating the mean of the tumor diameter (d), the volume of each tumor was calculated by the formula V Tumor = 1/6 × Π × d 3 .

Statistical analysis.
All experiments were performed at least three times. Differences between groups were examined for statistical significance using Student's t-test. Values of p < 0.05 were considered as statistically significant: *p < 0.05, **p < 0.01, ***p < 0.001.

Data Availability
Data will be provided upon reasonable request.