TGFβ signaling is a therapeutic target for many pathological conditions such as fibrosis, cancer and ocular disorders (Nakamura et al., 2004
; Breitkopf et al., 2005
; Saunier and Akhurst, 2006
). There are several approaches to TGFβ inhibition including antisense oligonucleotides, small molecule kinase inhibitors, neutralizing antibodies and soluble receptors. Representatives of each type of inhibitor are at various stages of development and each type of inhibitor has its own advantages and disadvantages. The rationale for TGFβ inhibition and the design and efficacy of the various TGFβ inhibitors has been reviewed (Yingling et al., 2004
; Akhurst, 2006
Several obstacles have been encountered in the development of effective TGFβ inhibitors such as delivery to the target location and specificity for the target. In addition, any or all of the three TGFβ isoforms may be expressed by a given cell in a specific context. Although small molecule kinase inhibitors have the advantage of high-throughput screening, their synthesis could be costly. Another disadvantage of kinase inhibitors is the possible side effects and toxicity, in part, due to their non-specific inhibition of other kinases (Foringer et al., 2005
; Park et al., 2006
). TGFβ RI (ALK5) kinase inhibitors typically inhibit ALK4 and ALK7 with high affinity as well. TGFβ RI and ALK4 are identical at the ATP binding site (http://www.touchbriefings.com/pdf/2287/boriack-sjodin.pdf
; Drug Discovery, 2006—September 2006, Use of ALK4 as a Surrogate Kinase for TGFβ RI).
Since excessive TGFβ signaling can lead to pathological conditions, but some TGFβ signaling is required for normal development and homeostasis, a better strategy may be to inhibit excess TGFβ extracellularly. Evidence for this approach has been shown in animal models of human carcinomas and fibrotic disease by utilizing a soluble RIII to block TGFβ-induced cell migration, invasion, angiogenesis, metastasis and firbrosis (Bandyopadhyay et al., 1999
, 2002a, 2002b, 2005
; Liu et al., 2002
; Dong et al., 2007
). Furthermore, in a mouse model of metastatic breast cancer, transgenic mice expressing the Fc-RII showed suppressed metastasis with no apparent side effects after lifetime exposure to this TGFβ antagonist (Yang et al., 2002
). This suggests that soluble TGFβ antagonists can reduce excessive TGFβ extracellularly without affecting normal TGFβ function required to maintain healthy tissue homeostasis.
Fc-RII is one of the most potent receptor-based inhibitors of TGFβ1 and β3; however, it does not inhibit TGFβ2 (Komesli et al., 1998
). Some cancers are known to specifically overexpress TGFβ2 such as malignant gliomas (Schlingensiepen et al., 2006
) and prostate cancer cells (Dallas et al., 2005
), whereas other types of cancer are known to overexpress all three TGFβ isoforms including breast and gastric cancer (Reiss, 1999
). All three TGFβ isoforms were also shown to mediate renal fibrogenesis, and it was suggested that blockade of all three TGFβ isoforms may yield the best therapeutic results (Yu et al., 2003
). Therefore, in addition to effectively inhibiting excess TGFβ extracellularly, it is also necessary to effectively inhibit all three isoforms in certain diseases where all three isoforms are implicated in disease progression.
Receptor-based inhibitors are relatively easy and cost-efficient to produce with the added possibility of scale-up. It was previously found that the antagonistic potency of soluble receptors is inversely related to the rate of dissociation from TGFβ (De Crescenzo et al., 2001
). Dimerization of the Fc portion of an antibody fused to the RII extracellular domain (Fc-RII) dramatically increased antagonistic potency. Homo and heterodimerization of TGFβ receptor extracellular domains through the introduction of coiled-coils was shown to enhance antagonistic potency (De Crescenzo et al., 2004
). Using this system, an RII extracellular domain homodimer was generated along with an RII extracellular domain and RIII membrane proximal domain heterodimer. It was found that the dimers had better antagonistic potency for TGFβ1 than the corresponding monomers.
Our study describes the design, production and characterization of a fusion protein containing both the RII receptor extracellular domain and the RIII receptor endoglin domain. Incorporating both domains into one protein has the advantages of blocking the activity of all three isoforms of TGFβ with high affinity and for equal delivery of both domains to the same physiological or pathological location. The production of the chimeric receptor is relatively simple and cost-effective. TGFβ ligands are differentially expressed and bind heterologously to RII and RIII. In cell-based affinity labeling experiments, RII was previously shown to bind with similar affinity to both TGFβ1 and β3 (Kd
~40 pM), but with much lower affinity to TGFβ2 (Kd
~500 pM) (Cheifetz et al., 1990
). While TGFβ1, β2 and β3 share 70–80% sequence identity, the lower affinity of RII for TGFβ2 was found to lie in three key residues on TGFβ2 (De Crescenzo et al., 2006
). When these residues were substituted to match the residues present in TGFβ1 and β3, binding affinity of RII to this variant of TGFβ2 was restored to that seen with TGFβ1 and β3. Although the type III receptor serves limited signaling function in the canonical TGFβ pathway, it has been shown to enhance the binding of TGFβ, particularly TGFβ2, to RII (Ohta et al., 1987
; Cheifetz et al., 1990
; Lopez-Casillas et al., 1993
). Its soluble extracellular domain binds to all three isoforms of TGFβ, but with highest affinity to TGFβ2, followed by TGFβ3 and then TGFβ1 (Vilchis-Landeros et al., 2001
). The extracellular domain of RIII is composed of two distinct domains. The membrane proximal domain shows sequence similarity to uromodulin (U-domain) and the membrane distal domain shows sequence similarity to endoglin (E-domain). Mutagenesis analysis has shown that both domains can bind independently to TGFβ (Fukushima et al., 1993
; Lopez-Casillas et al., 1994
; Pepin et al., 1994
). Competition assays have shown that, similar to wild-type RIII, both subdomains bind with higher affinity to TGFβ2 than TGFβ1 (Esparza-Lopez et al., 2001
). The same study showed that while only the E-domain enhanced TGFβ binding to RII, both domains enhanced intracellular signaling mediated by TGFβ2. Therefore, while RII preferentially binds to TGFβ1 and β3, RIII binds preferentially to TGFβ2. Theoretically, joining the two proteins together should generate a single chimeric protein that could bind all three isoforms of TGFβ with high affinity. RIII has also been shown to bind inhibin A. This would be an undesirable characteristic as the main goal is to bind only TGFβ. Although RIII contains two separate TGFβ-binding domains, there is no apparent difference in the ability of either domain to bind TGFβ. However, the inhibin A binding ability has been shown to reside exclusively in the U-domain (Esparza-Lopez et al., 2001
; Wiater et al., 2006
). Thus, utilization of the E-domain would be more favorable than the U-domain or wild-type RIII for a chimeric receptor. Therefore, we designed a fusion protein consisting of an N-terminal BGE
and C-terminal soluble RII.
The expression and purification methods described in this study were used to rapidly produce sufficient quantities of BGERII to determine its biological activity and feasibility as a pan-TGFβ inhibitor. The purification and refolding steps, which may be further optimized, proved to be adequate for the production and characterization of this novel protein. Our study demonstrates that this novel fusion protein can be expressed to a moderate level in a prokaryotic expression system and that active protein can be purified and recovered by dilution into non-denaturing buffer and through air oxidation. BGERII binds with high affinity to all three isoforms of TGFβ and is a more effective inhibitor of TGFβ signaling than equimolar concentrations of soluble RII or RIII. BGERII binds with higher affinity to TGFβ1 and β2 compared with RIII alone and with higher affinity to TGFβ1, β2 and β3 compared with RII. One might conclude that this is the result of adding an extra binding domain in the receptor which would effectively double the relative concentration. However, our results suggest that is not the case. Because RII does not bind significantly to TGFβ2, one would expect that BGERII should have the same binding affinity to TGFβ2 as RIII or even more closely to BGE. But our data show that BGERII binds TGFβ2 with higher affinity than RIII or BGE. Furthermore, BGERII is a potent TGFβ2 antagonist whereas BGE and soluble RII show little or no antagonistic activity against TGFβ2. Therefore, this is not just an effect of adding another mole of receptor. At least one possible explanation for this is an avidity effect, wherein BGERII simultaneously contacts TGFβ at two separate sites and that the binding energy provided by such simultaneous contacts is additive.
BGERII contains 16 cysteines which participate in the formation of eight intramolecular disulfide bonds in the native conformation. This fact severely hampers a critical step in the purification of any recombinant protein overexpressed in a bacterial system: refolding to the native conformation. We found that after refolding BGERII exists as a heterogenous mixture of monomer and soluble disulfide-linked multimers and that only the monomer contains high activity. Monomeric BGERII was shown to have higher TGFβ1 and β3 inhibitory activity compared with the two most potent TGFβ inhibitors, Fc-RII and the pan-specific anti-TGFβ antibody, with the additional advantage over Fc-RII of being able to inhibit TGFβ2, albeit to a lesser degree than the pan-specific antibody.
RII has some advantages when compared with Fc-RII and the pan-specific antibody. These two large-molecule inhibitors exist as disulfide-linked complexes in their native form. The multimeric conformation of these two inhibitors is likely the basis for their very high activity, but also makes them quite large. Being a smaller molecule with high activity may give BGE
RII a therapeutic advantage with respect to its accessibility to target tissues. Another advantage of BGE
RII is our ability to produce it in a bacterial expression system which is much faster, has the potential for scale-up and would be relatively more cost-effective compared with the mammalian cell systems used to produce the other two inhibitors. Additionally, BGE
RII can be engineered using native host sequences, thus reducing a potential immune response. On the other hand, further studies are needed to determine the stability, bioavailability and clearance of BGE
RII in vivo
and whether it will induce more unwanted side effects for being a more potent TGFβ inhibitor in comparison with other TGFβ inhibitors. Recently, RIII has been shown to also bind bone morphogenetic proteins (BMPs) and to enhance BMP activity (Kirkbride et al., 2008
). Because TβRII does not bind BMP and BGE
alone does not antagonize TGFβ activity, it is not expected that BGERII will inhibit BMP signaling. Nonetheless, further investigation is needed to address this issue.
This report is meant to introduce BGERII as a novel and potent pan-TGFβ inhibitor that was successfully produced by bacterial expression and has the potential to be developed further. Ongoing work is focusing on improving purification and refolding conditions to increase the yield of monomeric BGERII. Additional experiments are being designed to evaluate the efficacy of BGERII as a TGFβ inhibitor as well as its tolerability in vivo.