A universal reporter cell line for bioactivity evaluation of engineered cytokine products

Engineered cytokine products represent a growing class of therapeutic proteins which need to be tested for biological activity at various stages of pharmaceutical development. In most cases, dedicated biological assays are established for different products, in a process that can be time-consuming and cumbersome. Here we describe the development and implementation of a universal cell-based reporter system for various classes of immunomodulatory proteins. The novel system capitalizes on the fact that the signaling of various types of pro-inflammatory agents (e.g., cytokines, chemokines, Toll-like receptor agonists) may involve transcriptional activation by NF-κB. Using viral transduction, we generated stably-transformed cell lines of B or T lymphocyte origin and compared the new reporter cell lines with conventional bioassays. The experimental findings with various interleukins and with members of the TNF superfamily revealed that the newly-developed “universal” bioassay method yielded bioactivity data which were comparable to the ones obtained with dedicated conventional methods. The engineered cell lines with reporters for NF-κB were tested with several antibody-cytokine fusions and may be generally useful for the characterization of novel immunomodulatory products. The newly developed methodology also revealed a mechanism for cytokine potentiation, based on the antibody-mediated clustering of TNF superfamily members on tumor-associated extracellular matrix components.

Proteins were produced by transient transfection of CHO-S cells and purified by protein A affinity chromatography as described previously [35][36][37] . Quality control of the purified products included SDS-PAGE, size exclusion chromatography and, where applicable, mass spectrometry [Supplementary Figs. S4 and S5].
Virus production and stable transduction. For the virus production, 5 million HEK293T cells were seeded at a density of 300,000 cells/mL on the day prior to transfection. They were then transiently co-transfected with the reporter plasmid pJM046 as well as the packaging plasmid psPAX2 (Addgene, #12260; http://n2t.net/ addgene:12260; RRID:Addgene_12260) and the envelope plasmid pCAG-VSVG (Addgene, #35616; http://n2t. net/addgene:35616; RRID:Addgene_35616) using the jetPRIME reagent (Polyplus transfection). The packaging and the envelope plasmid were kindly provided by the group of Prof. Dr. Patrick Salmon. The medium was replaced on the day after the transfection and the virus was harvested on the following day. The virus was aliquoted and snap-frozen in an ethanol dry ice mixture. An aliquot was thawed and used for the transduction of 500,000 target cells. For the transduction 500,000 cells in 1 mL of medium were seeded in a 24 well plate and 1 mL of virus was added. Polybrene (Santa Cruz Biotechnology, #134220) was added to a final concentration of 8 μg/ mL. The cells were then centrifuged for 90 min at 1000 x g, 32 °C. Afterwards, the cells were incubated at 37 °C, 5% CO 2 . The medium was exchanged daily for the following three days. The cells were then expanded, activated and mCherry-positive cells were sorted by FACS (BD FACS AriaIII).
IL-12 assay. The method was adapted from previous publications 14,38 . Briefly, NK-92 cells were seeded at a density of 100,000 cells per well in a 96-well plate and 100 μL of medium containing varying concentrations of L19-IL12 was added. The total volume per well was 200 μL which corresponds to a cell density of 500,000 cells/ mL. After 24 h of incubation at 37 °C, 5% CO 2 , the concentration of IFN-γ in the cell culture supernatant was determined by ELISA (Invitrogen, #EHIFNG2). An IFN-γ standard was included in the ELISA and linear curve fitting using the GraphPad Prism 7.0 a software was used to derive the IFN-γ concentration in the samples from the absorbance at 450 nm (A450) [ Supplementary Fig. S7]. The data was fitted using the [Agonist] vs. response (three parameters) fit of the GraphPad Prism 7.0 a software to estimate the EC 50 .
tnf assay. L-M fibroblasts were seeded at a density of 20,000 cells/well in a 96-well plate and incubated for 24 h at 37 °C, 5% CO 2 . The medium was replaced by 100 μL fresh medium containing 2 μg/mL actinomycin D (BioChemica, #A1489,0005) and varying concentrations of F8-TNF. The cells were seeded at a concentration of 100,000 cells/mL. One day after seeding, the cell density should correspond to approximately 200,000 cells/mL, in a total of 200 µL, but this concentration was not measured immediately prior to the assay. The cells were incubated at 37 °C, 5% CO 2 for another 24 h before 20 μL of CellTiter 96 Aqueous One Solution (Promega, #G3582) was added and absorbance at 490 nm and 620 nm was measured. The % cell survival was calculated using the formula: % cell survival = (OD 490-620 treated − OD 490-620 medium )/(OD 490-620 untreated − OD 490-620 medium ) × 100%. The data was fitted using the [Agonist] vs. response (three parameters) fit of the GraphPad Prism 7.0 a software to estimate the EC 50 .
NF-κB response assay. CTLL-2 reporter cells were starved for 6-9 h as described above prior to use in order to reduce the background signal. If necessary, 100 μL 100 nM 11-A-12 fibronectin in phosphate buffered saline (PBS) was added to each well to be coated with EDA and the plate was incubated at 37 °C for 90 min. Cells were seeded in 96-well plates (50,000 cells/well) and growth medium containing varying concentrations of the cytokine to be tested was added. The cells were incubated at 37 °C, 5% CO 2 for several hours. To assess luciferase production, 20 μL of the supernatant was transferred to an opaque 96-well plate (PerkinElmer, Optiplate-96, white, #6005290) and 80 μL 1 μg/mL Coelenterazine (Carl Roth AG, #4094.3) in phosphate buffered saline (PBS) was added. Luminescence at 595 nm was measured immediately. mCherry expression was measured by flow cytometry (CytoFLEX, Beckman Coulter) and the data was analyzed using FlowJo (v.10, Tree Star). Therefore, the cells were resuspended in growth medium and transferred to a 96-well U bottom plate (Greiner BioOne, Cellstar, #650180) and harvested by centrifugation. The medium was discarded and the cells were washed with FACS buffer (0.5% BSA, 2 mM EDTA, PBS) and resuspended in FACS buffer. The relative luminescence and the relative fluorescence were calculated by dividing the obtained results by the results obtained when no inducer was added. Figure 1 describes the strategy followed for the generation of a universal reporter system for cytokine activity. The signaling of many different pro-inflammatory mediators (e.g., chemokines, Toll-like receptor agonists, members of the TNF superfamily and various other cytokines) involves the activation of NF-κB, in addition to other signaling pathways [ Fig. 1a] 23,30,39,40 . We constructed a vector for virus-mediated stable cell transduction, which incorporated an NF-κB response element upstream of a secreted luciferase (NanoLuc) and mCherry reporter genes [Supplementary Figs. S1 and S2]. The two reporter proteins are separated by a T2A peptide. The luciferase is secreted due to the presence of an IL-6 secretion signal while the mCherry is retained in the cytoplasm. A second-generation lentivirus was used for transduction experiments. The lentivirus was produced by the simultaneous transfection of HEK293T cells with the envelope plasmid pCAG-VSVG, the packaging plasmid psPAX2 and pJM046 (harboring the reporter construct) [Supplementary Figs. S1 and S2]. The transduction strategy was used to stably integrate the NF-κB reporter into CTLL-2 and A20 cell lines, of T-cell and B-cell origin (respectively) [ Fig. 1].

Results
We investigated the possibility of using the newly developed transduction method as a universal reporter for cytokine activity by performing a comparison with established dedicated test systems. For this experiment, antibody-cytokine fusions that have previously been developed in our lab were used [35][36][37] . We used fusions of the L19 antibody targeting the extradomain B of fibronectin 41 Table T1]. The activity of interleukin-12 (IL-12) is often measured in terms of interferon-γ production by NK-92 cells of Natural Killer cell origin 14 . In this case, the EC 50 values obtained using the NK-92 cell-based system were roughly ten times lower than those obtained by monitoring NF-κB reporter activity [ Fig. 2b and Supplementary Table T1]. The performance of the new methodology for the measurement of TNF activity was compared to the results of a conventional cell killing assay, using a transformed fibroblast cell line which is particularly sensitive to the action of TNF. EC 50 values in the pM range were observed for the killing assay, whereas the EC 50 values obtained with the universal assay were in the 100 pM range [ Fig. 2c and Supplementary Table T1].
We then used the of NF-κB reporter assay for the characterization of three novel fusion proteins, featuring the F8 antibody (specific to the alternatively-spliced EDA domain of fibronectin, a tumor-associated antigen) 42 fused to three different members of the TNF superfamily [ Fig. 3]. 4-1BB and GITR (glucocorticoid-induced TNFR-related gene) are expressed on activated CD8 + T-cells. Activation of each of these two receptors has been shown to prolong longevity of the activated T cells and transition towards a memory phenotype 43,44 . By contrast, CD40 is expressed on antigen-presenting cells 45 . The interaction between CD40 and its ligand, CD154, is important for the licensing of antigen-presenting cells 46 . A variety of agonists to these three members of the TNF superfamily are being developed and tested both at a preclinical stage and in clinical trials as agents for the immunotherapy of cancer 47 . Our lab has had a long-standing interest in the characterization of the tumor-homing properties 48 and anti-cancer activity of antibody fusion proteins with members of the TNF superfamily 36,49 . F8-4-1BBL, F8-GITRL and F8-CD154 featured the TNF superfamily member as single-chain polypeptide in order to stabilize the homotrimeric structure 20 while the antibodies were used as recombinant diabody moieties 50 . A concentration-dependent NF-κB reporter activity could be measured for all three antibody-cytokine fusions, monitoring both luciferase activity and mCherry expression [Fig. 3]. The EC 50 values obtained for 4-1BBL and GITRL constructs were in the nanomolar range whereas the values obtained for the CD154 construct were in the 100 pM range [Supplementary Table T2].
Antibody-cytokine fusions are increasingly being used for the therapy of cancer 5,6,21 and of chronic inflammatory conditions 51,52 with the aim to concentrate cytokine activity at the site of disease and help spare normal organs. We used derivatives of the F8 antibody to assess whether a localized concentration of cytokine activity in close proximity to target cells of interest may lead to a potentiation of biological activity. The alternatively-spliced extra-domain A (EDA) of fibronectin (i.e., the target antigen of the F8 antibody) is typically found as an abundant component of the modified extracellular matrix associated with newly-formed tumor blood vessels 53 , but is otherwise undetectable in most normal adult tissues, exception made for the female reproductive system 54 . We mimicked the localized deposition of EDA(+)-fibronectin by coating plastic wells with 11-A-12 recombinant fibronectin fragments, containing the EDA domain 42 [Fig. 4a]. The system was used to compare the biological activity of F8-4-1BBL, F8-GITRL and F8-CD154 fusion proteins, featuring the F8 antibody either in diabody format linked to a single-chain trimer of the cytokine or as single-chain Fragment variable (scFv) linked to a single unit of the cytokine [ Fig. 4b-d]. In the case of the F8 antibody in a diabody format linked to a single-chain trimeric 4-1BBL, only a small increase in stimulatory activity was observed when the agonist was clustered on EDA. www.nature.com/scientificreports www.nature.com/scientificreports/ However, in the case of the single-chain Fv fragment linked to a single 4-1BBL unit, no activity was observed in the absence of clustering. Activity could in this case be restored by clustering the agonist on EDA coated wells [ Fig. 4b]. By contrast, both F8-GITRL constructs were constitutively active [ Fig. 4c]. In the case of the F8 antibody in a diabody format linked to the single-chain trimer of GITRL, no activity enhancement was observed by clustering on EDA. The EC 50 of the construct consisting of the F8 antibody in an scFv format linked to a single unit of GITRL, a roughly 5-fold reduction in the EC 50 was observed when the experiment was performed in the presence of EDA. Some modest increase in activity in the presence of EDA was also observed in the case of the single-chain trimeric CD154 linked to F8 in a single-chain diabody format [ Fig. 4d].

Discussion
In this work, we presented the development of two reporter cell lines that can be used to measure the biological activity of a variety of cytokines. The new methodology capitalizes on the fact that many cytokine-triggered signaling events converge at the level of NF-κB activation.
An excellent agreement could be found between bioactivity measurements of IL-2 performed with the conventional cell-based cytokine activity assay and the newly developed method [Fig. 2a]. A discrepancy between the conventional cell-based cytokine assay and the new method could be seen for two fusion proteins with murine payloads (L19-IL12 and F8-TNF), for which the new methodology indicated a reduction in potency [Fig. 2b,c]. It is possible that NF-κB signaling may not fully capture the molecular events triggered by murine IL-12 and murine TNF, but the new methodology could still enable a comparative evaluation of multiple pro-inflammatory payloads and cytokine variants (e.g., wildtype and mutated versions).
The observation that the activity of murine 4-1BBL can be potentiated by clustering of antibody fusions on specific tumor-associated extracellular matrix components is surprising and potentially useful for pharmaceutical applications. Murine 4-1BBL is not able to form stable homotrimers, but rather forms low-activity homodimeric structures 55,56 . It is possible that the high-density binding of F8 fusions on EDA-containing fibronectin promotes the formation of supermolecular assemblies, which gain signaling activity in a concentration-dependent manner [ Fig. 4b]. This approach mimics on the extracellular matrix what happens on murine cells where surface display www.nature.com/scientificreports www.nature.com/scientificreports/ of multiple copies of dimeric ligands turns an inactive homodimer into an active multimeric assembly 55,56 . Some potentiation upon antigen binding had previously been reported for certain murine TNF fusions 57 and for other members of the TNF superfamily 58,59 . In principle, it would be attractive to engineer antibody-cytokine fusions which gain activity at the site of disease (e.g., upon antigen binding) while sparing normal tissues, as this approach could lead to biopharmaceuticals with improved therapeutic index. Our group has recently described a conceptually similar strategy, based on the assembly of heterodimeric split cytokine fusions (e.g., interleukin-12 superfamily members) 60 . Other strategies for the conditional potentiation of antibody-cytokine conjugates include the allosteric regulation of cytokine activity 61 or the attenuation of cytokine potency ("attenukine") 62 . Strategies aimed at potentiating the activity of antibody therapeutics based on conditional oligomerization include the development of hexameric IgG antibodies to augment complement-mediated cytotoxicity 63,64 .
Not all TNF superfamily members seem to need antigen-dependent clustering in order to gain activity [ Figs. 2c and 4c]. This is also reflected by the fact that some ligands of the TNF superfamily including TNF and CD154 are enzymatically shed in vivo and act as soluble ligands 65,66 . Although GITRL is so far not known to be present as soluble ligand 67 our data indicates that the recombinant soluble ligand shows similar behavior [ Fig. 4c].
In this article we have mainly focused on recombinant and engineered cytokine products. While many interleukins mainly signal through JAK/STAT activation 27 , NF-κB-driven transcriptional events are also induced, in line with previous reports on this matter [28][29][30][31] . NF-κB reporters should be broadly applicable also to other classes of pro-inflammatory products. Both Toll-like receptors and members of the interleukin-1 receptor superfamily activate NF-κB through the recruitment of MyD88 as part of their signal transduction 39 . In addition, there is also evidence that NF-κB can be activated by chemokine signaling 40 .
The CTLL-2 and A20 cell lines express on their surface a large variety of receptors of immunological importance and should therefore be applicable for many different bioactivity assays. If a researcher is interested in the use of a different cell line, the viral transduction system described in Fig. 1 should enable the rapid preparation of the corresponding reporter system. The intracellular expression of mCherry was found to be useful for sorting of positively transduced cells by flow cytometry, while the secreted luciferase rapid a facile readout of the reporter activity.  In vivo the F8 antibody binds to the extradomain A (EDA) of fibronectin that is present in the tumor-associated neovasculature. This is thought to lead to clustering of the cytokine fused to the F8 antibody and therefore to enhance activation of target cells such as cytotoxic T cells. To mimic this situation in vitro, microwell plates were coated with the extradomain A of fibronectin. The activity of the immunocytokine when clustered on EDA and without clustering was compared for different formats. www.nature.com/scientificreports www.nature.com/scientificreports/ In summary, we have generated new universal reporter systems for the facile measurement of biological activity of various types of pro-inflammatory mediators and engineered cytokine products. We anticipate that the newly-developed cell lines and vector may find a broad applicability in biological, pharmaceutical and immunological research.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.