Anti-interleukin-6 therapy through application of a monogenic protein inhibitor via gene delivery

Anti-cytokine therapies have substantially improved the treatment of inflammatory and autoimmune diseases. Cytokine-targeting drugs are usually biologics such as antibodies or other engineered proteins. Production of biologics, however, is complex and intricate and therefore expensive which might limit therapeutic application. To overcome this limitation we developed a strategy that involves the design of an optimized, monogenic cytokine inhibitor and the protein producing capacity of the host. Here, we engineered and characterized a receptor fusion protein, mIL-6-RFP-Fc, for the inhibition of interleukin-6 (IL-6), a well-established target in anti-cytokine therapy. Upon application in mice mIL-6-RFP-Fc inhibited IL-6-induced activation of the transcription factor STAT3 and ERK1/2 kinases in liver and kidney. mIL-6-RFP-Fc is encoded by a single gene and therefore most relevant for gene transfer approaches. Gene transfer through hydrodynamic plasmid delivery in mice resulted in hepatic production and secretion of mIL-6-RFP-Fc into the blood in considerable amounts, blocked hepatic acute phase protein synthesis and improved kidney function in an ischemia and reperfusion injury model. Our study establishes receptor fusion proteins as promising agents in anti-cytokine therapies through gene therapeutic approaches for future targeted and cost-effective treatments. The strategy described here is applicable for many cytokines involved in inflammatory and other diseases.


Effective inhibition of IL-6 in vitro and in vivo with recombinant mIL-6-RFP-Fc.
To characterize the bioactivity of mIL-6-RFP-Fc, muHepa cells were stimulated with murine IL-6 that had been preincubated with purified mIL-6-RFP-Fc in molar ratios ranging from 0.2 to 10. As a read-out of IL-6 activity tyrosine phosphorylation of STAT3 was analyzed (Fig. 2c). Complete inhibition of IL-6 was already achieved at about equimolar concentrations of mIL-6-RFP-Fc and IL-6. In the next experiment, mIL-6-RFP-Fc and IL-6 were not preincubated, but cells were pretreated with mIL-6-RFP-Fc (Fig. 2d). Subsequent addition of IL-6 did not elicit phosphorylation of STAT3 over a prolonged period of time indicating that even in direct competition with the cell surface receptors mIL-6-RFP-Fc neutralizes IL-6 immediately and completely. Activities of the original mIL-6-RFP 15 and the novel mIL-6-RFP-Fc were compared on Ba/F3 cells stably expressing mIL-6Rα and mgp130 (Ba/F3-mgp130/mIL-6Rα). IL-6-dependent proliferation of Ba/F3-mgp130/mIL-6Rα cells was inhibited in a concentration-dependent manner by both inhibitors. However, mIL-6-RFP-Fc's inhibitory activity was one order of magnitude higher compared to the non-optimized version of mIL-6-RFP ( Fig. 2e; IC50 mIL-6-RFP-Fc: 3.3 × 10 −5 μ M, IC50 mIL-6-RFP: 3.8 × 10 −4 μ M) which is in agreement with the higher avidity of mIL-6-RFP-Fc as proposed in our model (Fig. 1c) and with the increase of the activity of the IL-6 trans-signaling inhibitor sgp130 upon Fc fusion 17 .
To establish its bioavailability in a mammalian system, mIL-6-RFP-Fc was administered to mice as 1 μg intravenously (i.v.), 1.5 μg intraperitoneally (i.p.) and 1.5 μg subcutaneously (s.c.) and recovered via blood sampling. For i.v. injection a lower amount was chosen because a higher recovery was expected upon direct administration to the blood stream. As determined by Western blot analysis of serum samples, the i.v. and i.p. routes gave rise to the highest systemic levels with the latter resulting in prolonged serum detectability with no signs of major degradation of the fusion protein (Fig. 3a). S.c. application did not result in significant plasma levels, which might be explained by the known slow-onset and prolonged pharmacokinetics of this route. It should be noted that a non-optimized, non-Fc-containing variant of mIL-6-RFP 15 was successfully applied s.c. for local IL-6 antagonism in a cutaneous tumor model 18 . To further characterize the pharmacokinetics of mIL-6-RFP-Fc, different amounts, i.e. 1.5, 4, and 40 μg were injected i.p., followed by repeated blood sampling for quantitative serum ELISA directed against the epitope tags. At the highest amount (40 μg) applied in one single administration, mIL-6-RFP-Fc could  with D1-D3 of murine IL-6Rα for high-affinity binding of murine IL-6, the Fc-fragment (CH2-CH3) including the hinge region of mIgG2a which has been mutated to reduce antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) 50 followed by three V5 and three HA epitopes (3V5-3HA) for sensitive detection by immunofluorescence and reliable quantification by ELISA. Arrows indicate binding of IL-6 to mIL-6-RFP-Fc with the dashed line indicating binding to a second mIL-6-RFP-Fc resulting in the formation of a complex in analogy to the hexameric IL-6 receptor complex (shown in (c)). Secretion of mIL-6-RFP-Fc is driven by the signal sequence of preprotrypsin (not shown) as established previously for mIL-6RFP 15 . (b) Purified mIL-6-RFP-Fc was analyzed by 7.5% SDS/PAGE. Purity and identity was determined by staining with Coomassie brilliant blue and Western blotting, respectively, under non-reducing and reducing conditions. The molecular mass of mIL-6-RFP-Fc was determined by calculating the R f of five marker proteins and inserting the R f of mIL-6-RFP-Fc monomer and dimer into the equation for the linear regression. (c) Schematic representation (left panel) and structural model (right panel) of the (IL-6) 2 (mIL-6-RFP-Fc) 2 inhibitory complex. The structural model is based on the crystal structures of the human hexameric IL-6 signaling complex (PDB: 1P9M), human IL-6Rα (PDB: 1N26), and the hinge region, CH2 domain and CH3 domain of murine IgG2A (PDB: 1IGT).
Scientific RepoRts | 5:14685 | DOi: 10.1038/srep14685 be detected in serum for up to 48 h (Fig. 3b, left diagram). Normalization of the different animals' serum mIL-6-RFP-Fc levels on their respective 2 h value, allowed for statistical analysis and determination of a time constant (t 1/2 = 3.5 h) across the amounts of protein spiked (Fig. 3b, right diagram). The observed half-life is identical with the initial clearance determined for an IL-2-Fc fusion protein harbouring the same Fc-fragment (mIgG2a) 19 .
To determine the efficacy of mIL-6-RFP-Fc in mice, we established a model of IL-6-stimulated STAT3 and ERK1/2 phosphorylation in liver and kidney tissues. In analogy to the in vitro experiments, mIL-6-RFP-Fc was either pre-incubated with IL-6 ( Fig. 4a) or the proteins were administered sequentially (Fig. 4b). In both scenarios mIL-6-RFP-Fc significantly reduced IL-6 downstream signaling. Pretreatment with 40 μg (70× molar excess) of mIL-6-RFP-Fc i.p. for 45 min attenuated STAT3 and ERK1/2 phosphorylation by 12 and 63% in the liver and by 80 and 50% in the kidney, respectively. Phosphorylation was assessed by the ratio of phosphorylated versus total STAT3 or ERK1/2 (Fig. 4b). Variations of total STAT3 levels might be due to the fact that activated STAT3 induces its own gene expression 20,21 . Collectively, these data established mIL-6-RFP-Fc as a potent therapeutic agent for the inhibition of IL-6 in vitro and in vivo.
Effective gene delivery and inhibitory capacity of mIL-6-RFP-Fc in mice and use in a preclinical model of acute kidney injury. Anti-cytokine therapy through application of recombinant protein has been shown to be effective but protein production is expensive and tedious. We argued that a gene therapeutic approach, where small amounts of a delivered plasmid generate substantial protein levels by the recipient himself, might be much more efficient. mIL-6-RFP-Fc is encoded by a single gene and therefore ideally suited for gene delivery.
In order to achieve high systemic levels of circulating mIL-6-RFP-Fc we used the method of hydrodynamic plasmid delivery via the tail vein, which mainly results in transfection of the liver and to a much lower extent of the kidney and other organs (1:100-1000) 22 . The phosphoenolpyruvate carboxykinase (PEPCK) promoter was chosen because it is active in the liver and generates high systemic levels of soluble proteins 23,24 . Following gene delivery, mIL-6-RFP-Fc was detected in liver tissue (Fig. 5a) and serum (Fig. 5b). The concentration of expressed protein in the liver correlated well with the amount of plasmid used for transfection (i.e. 10 and 37 μg/animal, respectively). Using the latter amount, a mIL-6-RFP-Fc serum level at 24 hrs of 1.3 ± 1.5 μg/ml (mean ± SD; n = 4) was achieved, thus corresponding to the 2 h

. Inhibition of IL-6 dependent STAT3 and ERK phosphorylation by mIL-6-RFP-Fc in vivo.
(a) Western blot analysis of mouse liver lysates prepared 15 min following intravenous administration of either PBS, IL-6 (100 ng) or preincubated IL-6/mIL-6-RFP-Fc-complexes (100 ng/3.5 μg, corresponding to a molar ratio of 1:5, incubation for 30 min). (b) Representative Western blot of mouse liver and kidney lysates prepared 15 min following intravenous administration of either PBS or IL-6 (100 ng). Mice were pretreated for 45 min with PBS or mIL-6-RFP-Fc i.p. at dosages of 4, 12 or 40 μg, corresponding to molar mIL-6-RFP-Fc (RFP): IL-6 ratios as indicated. Phosphorylated STAT3 (α-pSTAT3), phosphorylated ERK1/2 (α-pERK1/2), total STAT3 (α-STAT3) and total ERK1/2 (α-ERK1/2) were detected. GAPDH and HSP70 served as loading controls. Statistical evaluation of experiments are depicted as bar charts. Data shown are means ± SD, *p ≤ 0.05, one-sided t-test (n = 2 for PBS i.v.-stimulated groups; n = 5 for PBS i.p./IL-6 i.v.; n = 3 for RFP 7x i.p./IL-6 i.v.; n = 3 for RFP 20x i.p./IL-6 i.v.; n = 4 for RFP 70x i.p./IL-6 i.v.). Because of variations in STAT3 expression levels, pSTAT3 was normalized relative to total STAT3. Similarly, pERK1/2 was normalized relative to total ERK1/2. and 12-24 h levels upon singular i.p. spiking of 4 and 40 μg of mIL-6-RFP-Fc, respectively (Fig. 3b). Subsequently, 37 μg of plasmid/animal were used for hydrodynamic transfection throughout all experiments. Analysis of liver sections of mice that were simultaneously transfected with plasmid encoding mIL-6-RFP-Fc and GFP revealed that a single cell either expresses the fusion protein or GFP but rarely both (Fig. 5c). This finding emphasizes the advantage of a single gene construct such as receptor fusion proteins over proteins encoded by multiple genes such as antibodies for gene transfer approaches. The bioactivity of mIL-6-RFP-Fc administered through gene transfer could be directly confirmed by the suppression of hepatic acute phase protein production which had been induced by the invasive transfection procedure. Expression of mRNA for serum amyloid A1 (SAA1) was significantly reduced and expression of mRNA for α2-macroglobulin (A2M) showed a trend of a reduction (Fig. 5d). Both proteins are known to be governed by IL-6 25 . A trend for slightly increased IL-6 mRNA levels in the mIL-6-RFP-Fc treated group might reflect compensatory IL-6 production (Fig. 5d). The fact that SAA1 is not completely suppressed in this model might be potentially explained by other SAA1-inducing cytokines such as IL-1 and TNF being released following hydrodynamic transfection.
To assess the usefulness of the hydrodynamic transfection approach for disease models, the pharmacokinetics of our RFP were evaluated in more detail. mIL-6-RFP-Fc was detected in serum for up to one week after gene transfer, again with no signs of degradation. Average serum levels of 1.6 μg/ml (± 1.6 SD; n = 4) mIL-6-RFP-Fc were achieved for at least 48 h (Fig. 5e), indicating continuous synthesis as opposed to singular spike and turnover (Fig. 3b).
To assess the therapeutic potential of mIL-6-RFP-Fc produced via gene delivery in a relevant disease model we applied a model of acute kidney injury (AKI) by transient bilateral renal ischemia followed by reperfusion for 24 hours (I/R). This well-established model of tubular cell necrosis 26 has been previously demonstrated to depend critically on IL-6 27 . Injury is significantly reduced in IL-6 −/− mice and can also be alleviated by the adoptive transfer of IL-6 −/− bone marrow or by treatment with IL-6 neutralizing antibodies, resulting in improved renal serum and histological parameters 28,29 . Our experimental design is outlined in Fig. 6a. Mice with detectable mIL-6-RFP-Fc serum levels (0.25-4.5 μg/ml; mean 1.5 μg/ ml ± SD 1.5) at 20 h post hydrodynamic transfection or equally treated mice having received control vector were subjected to bilateral ischemia of 33 minutes duration at 24 h and sacrificed at 48 h for further organ analysis. Successful transfection and biological activity of mIL-6-RFP-Fc was confirmed by suppressed hepatic SAA1 and A2M mRNAs, which also correlated with the 20 h serum level of mIL-6-RFP-Fc (Fig. 6b,c). In line with the previous studies 28, 29 , we observed similar beneficial effects on renal endpoints. Serum creatinine, intrarenal mRNA synthesis of the acute kidney injury marker neutrophil gelatinase-associated lipocalin (NGAL)/lipocalin-2 (Lcn2) 30,31 and histologic tubular injury scoring were significantly reduced following transfection with mIL-6-RFP-Fc (Fig. 6d-f). These data are also in line with another report in which IL-6 −/− mice were protected in the initial phase of a mercury-induced proximal tubular injury model 32 . It must be noted that the renal functional and renal histologic endpoints of the mIL-6-RFP-Fc group exhibited a comparatively wide variance (Fig. 6d,f). This is in agreement with published data on this complex acute kidney injury model 28  Collectively, the consistent reduction of the hepatic acute phase response in both the hepatic transfection and the I/R model (Figs 5 and 6, respectively) and the significant -though variant -benefit of renal endpoints in I/R (Fig. 6) establish endogenous expression of mIL-6-RFP-Fc via gene transfer as an effective treatment in murine disease models.

Discussion
There is an ever-growing requirement for biologics to treat inflammatory diseases and cancer. Production of biologics involves expression of recombinant proteins in cell culture and subsequent purification under standardized conditions. The entire process is time-consuming, intricate and costly. Consequently, increasing prescription rates of biologics will become a heavy burden for health care systems 33 . Therefore, alternative cost-effective strategies that circumvent recombinant protein production are highly desired. A potential solution of the problem is to use the protein-producing capability of the host and to deliver the genetic information instead of the protein itself. This strategy of therapeutic gene delivery would work best for biologics encoded by a single gene and requires optimization of the protein, the encoding cDNA and reliable gene transfer methods.
Here, we have introduced mIL-6-RFP-Fc as a receptor fusion protein for the inhibition of IL-6 that has been optimized for application through gene delivery in mice. IL-6 is a well-established therapeutic target in rheumatic disease, other inflammatory diseases and cancer [34][35][36][37][38] . Starting with our previously described mIL-6-RFP 15 , we added an Fc-fragment for several purposes. Cysteine residues within the Fc-fragment enforce dimerization of mIL-6-RFP-Fc through disulfide bond formation favouring the formation of an inhibitory complex consisting of a mIL-6-RFP-Fc dimer that binds two IL-6 molecules. This stoichiometry is analogous to the high-affinity hexameric receptor complex consisting of two molecules of each IL-6, sIL-6Rα and gp130 that has been verified by X-ray crystallography 16 . Earlier work demonstrated that IL-6-RFP indeed forms complexes similar to the native IL-6 receptor complex 39 . The extraordinary inhibitory potency of mIL-6-RFP-Fc is demonstrated by its capacity to inhibit IL-6-mediated proliferation of Ba/F3-mgp130/mIL-6Rα cells with a tenfold higher potency than the original mIL-6-RFP.
Furthermore, the Fc-fragment of IgG interacts with the neonatal Fc receptor (FcRn) that mediates recycling of pinocytosed protein, thus increasing the plasma half-life of the fusion protein 40 . Plasma half-life can be further increased through mutations within the Fc-fragment that affect the Fc-FcRn interaction 41 . Finally, when mIL-6-RFP-Fc is produced as a recombinant protein the Fc-fragment allows  Serum creatinine was assessed 5 days prior to gene delivery and at sacrifice. All data are from n = 8 animals/ group, shown as means (horizontal bars) ± SD (vertical bars) and single values (symbols), *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 one-sided t-test. (b,c) qRT-PCR of acute phase mRNAs from liver of sacrificed mice as calculated relative (x-fold) to mean of healthy control mice using housekeeper GAPDH and the Δ Δ CT method 51 . Expression of both acute phase proteins by mIL-6-RFP-Fc transfection is significantly reduced (red triangles) in comparison to control vector (black dots). Nonlinear regression analysis revealed that suppression of SAA1 and A2M correlates with serum mIL-6-RFP-Fc levels. (d) Increased serum creatinine following I/R is significantly reduced by prior transfection with mIL-6-RFP-Fc. Verification of t-test results by ANOVA with Tuckey multiple group analysis revealed an overall p ≤ 0.0001 and an identical significance level of ≤ 0.01 for I/R control versus mIL-6-RFP-Fc. (e) Transcription of the tubular injury marker NGAL in both kidneys as assessed using housekeeper GAPDH and the Δ Δ CT method 51 , indicating the significant mitigation of renal injury upon hydrodynamic transfection of mIL-6-RFP-Fc (red triangles) compared to empty vector (black dots). (f) Representative renal tissue sections, original magnifications, ×200. Dotted lines, representative renal tubules. Normal healthy kidney showing homogeneous basophilic and eosinophilic staining of the cytoplasm and of the apical brush border membrane, respectively. Following I/R, tubular dilatation/atrophy (arrowheads), loss of the brush border (arrow), tubular necrosis (asterisk) and leukocyte infiltration (L) ensue, the degree of which is mitigated following transfection with mIL-6-RFP-Fc. Far right: statistical analysis of tissue injury score for the right kidney (left kidney not assessed). Red triangles, mIL-6-RFP-Fc; black dots, empty vector. See methods for details. convenient purification through affinity chromatography with immobilized Protein A or G. mIL-6-RFP-Fc can be reliably quantified with an ELISA directed against the two triple tags (3V5-3HA) that follow the Fc-fragment. By simple cloning procedures, the Fc-3V5-3HA module can be easily transferred to other receptor fusion proteins so that different proteins can be quantified by the same assay.
Interestingly, the expression level of mIL-6-RFP-Fc increased through optimization of the cDNA sequence that affects codon usage and all sequences that could interfere with stability and translation of the mRNA. An analogous human hIL-6-RFP-Fc for clinical trials should therefore be optimized the same way. A great advantage of an endogenously produced therapeutic protein is that the glycosylation is entirely host-like which cannot be achieved by recombinant protein production. This feature is expected to dramatically reduce immunogenicity which often is a result of aberrant glycosylation 42 .
The experimental method of hydrodynamic transfection 22 allowed us to effectively characterize the biological action of mIL-6-RFP-Fc in vivo but obviously this mode of gene delivery will not be suited for the treatment of humans. For future therapeutic applications in patients, an analogous hIL-6-RFP-Fc designed for delivery through safe virus-based transfection methods such as integrating lentiviral vectors or non-integrating adeno-associated virus-derived vectors will be required and is under current development 43 . Similarly, transfer vectors can be designed that control expression of mIL-6-RFP-Fc through organ-or tissue-specific promoters so that the consequences of local IL-6 inhibition can be studied.
A large variance within the IL-6 knockout groups in the I/R model was also observed in another study (at least twofold of the wildtype control groups) 28 . To better understand this phenomenon a careful correlation analysis of the 20 h serum mIL-6-RFP-Fc levels and the associated renal outcomes was undertaken. While the animal with the highest serum level of 4.5 μg/ml exhibited the lowest serum creatinine level (0.21 mg/dl) and the lowest average NGAL expression level (right kidney: 40-fold, left kidney 156-fold of normal) no clear correlation between these outcome markers and the mIL-6-RFP-Fc level was found in the other animals (data not shown). On the other hand, a clear dose-response effect was observed in the inhibition of the hepatic acute phase response (Fig. 6b,c).
Intriguingly, in the hepatic stress response at 24 h following hydrodynamics based transfection (Fig. 5d) acute phase transcriptional suppression by mIL-6-RFP-Fc is significant but incomplete while in the I/R model it is total with some animals ranging below the level of healthy controls (Fig. 6b,c). Indeed, as discussed above, 24 h post hydrodynamics stress non-IL-6 driven SAA1 transcription (IL-1, TNF) might be the case, while for the I/R model it has been elegantly established that IL-6 is mainly derived from the kidney contributing to high systemic levels 28 . Renal IL-1 and TNF release is a feature of I/R but not in the setting of IL-6 KO or antagonism 29 . Collectively, this could explain the discrepancy between acute phase transcriptional suppression in hepatic stress and I/R.
The approach of gene delivery of a monogenic inhibitor is not confined to IL-6 and can be exploited to target other cytokines that signal through heteromeric cytokine receptors. For instance, through replacement of the IL-6Rα moiety by the IL-11Rα, a mIL-11-RFP-Fc can be generated that is expected to potently block IL-11, another cytokine of the IL-6 family 6 . IL-11 has been recently identified as a dominant cytokine during gastrointestinal tumorigenesis 44 .
Taken together, to our knowledge this is the first report on the successful gene delivery of an IL-6 inhibitor that is encoded by a single gene. Very recently, the well-established anti-IL-6-Rα antibody tocilizumab has been demonstrated to be effectively deliverable by a gene therapeutic approach 45 . However this antibody required two separate plasmids, given the nature of antibody assembly from two chains. This is simplified by the approach presented herein. IL-6 antagonism, as pioneered by the clinical use of tocilizumab and siltuximab 46 is a promising approach for the treatment of a variety of autoimmune disorders. In our view, gene therapeutic treatment options for IL-6 antagonism that are effective and economic may be of substantial benefit for patients and society in the future.
Quantitative mRNA analysis. RNA isolation and cDNA synthesis from tissue were performed using standard columns (Qiagen, Hilden, Germany) and random primers (Roche), respectively. Real-time quantitative PCR was carried out using qPCR Core Kit for SYBR Green (Eurogentec, Liege, Belgium) and an ABI Prism 7300 sequence detector (Life Technologies, Carlsbad, CA). Data were normalized using glyceraldehyde-3-phosphate dehydrogenase (Gapdh) as an internal control and calculated using the Δ Δ CT-method. The following primer sequences were used: Gapdh, serum amyloid A1 (Saa1), and a2-macroglobulin (A2m) as described elsewhere 21  Statistical analysis. Unpaired one-sided student's t-test with Welch's correction when appropriate was performed at all instances when not otherwise specified to compare control-and RFP-treated groups as indicated using GraphPad Prism version 6.0b for Mac OS X, GraphPad Software (La Jolla, CA). The one-sided test approach was chosen since our hypothesis H1 was one-sided, i.e., mIL6-RFP-Fc inhibits IL-6 or is beneficial in the renal ischemia-reperfusion model compared to control.