Structure reveals function of the dual variable domain immunoglobulin (DVD-Ig™) molecule

The bispecific tetravalent immunoglobulin known as the dual variable domain immunoglobulin or DVD-Ig™ molecule was first described by Wu, et al. in 2007.¹ Like a conventional IgG, the DVD-Ig™ molecule is composed of two heavy chains and two light chains. Unlike IgG, however, both heavy and light chains of a DVD-Ig™ molecule contain an additional variable domain (VD) connected via a linker sequence at the N-termini of the VH and VL of an existing monoclonal antibody (mAb). Thus, when the heavy and the light chains combine, the resulting DVD-Ig™ molecule contains four antigen recognition sites (Fig. 1). The outermost or N-terminal variable domain is termed VD1 and the innermost variable domain is termed VD2; the VD2 is proximal to the C-terminal CH1 or CL. We and others have previously reported that DVD-Ig™ molecules can be manufactured and purified to homogeneity in large quantities, have pharmacological properties similar to those of a conventional IgG₁, and show in vivo efficacy in multiple mouse models.^1,2

Theoretically, the structurally novel DVD-Ig™ format design can impose certain structural and functional constraints on VD2, the inner variable domain. In a conventional mAb, the complementarity-determining regions (CDRs) of a VD are surface exposed with no N-terminal constraints, no limitations as to the size and location (soluble or cell surface) of the target antigen, and no constraints as to conformational changes that might occur upon target binding (VD stabilization). In the DVD-Ig™ molecule, the juxtaposition of VD1 to VD2 via linkers could potentially occlude VD2 CDRs, limit VD2 rotational flexibility, impose limits on target size and location, or impose constraints on VD2 conformational changes (stabilization) upon target binding. We and others, however, have observed that affinity at the inner antigen binding site (VD2) may be somewhat dependant on the VD1/VD2 pair combination (amino acid sequences), the VD1/VD2 orientation, and linker selection.^2-4 The choice of linker length between the VD1 and VD2 (e.g., either both short linkers [S-S], both long linkers [L-L], or one short and one long linker [S-L or L-S] as shown in Table 1), can affect the affinity of the inner VD2 domain.⁴ In addition, the antigen affinity at the outer antigen binding domain (VD1) is often nearly equal to the parent antibody in the DVD-Ig™ format. Both VD1 and VD2 of a DVD-Ig™ molecule can successfully target soluble and cell surface antigens and can bind antigens simultaneously and with full occupancy.¹ Here, we report the structure of the interleukin (IL)12-IL18 DVD-Ig™ Fab with IL18 bound at VD2. The structure provides the basis for understanding how the DVD-Ig™ molecule binds two different antigens simultaneously and lays the foundation for hypothesis-driven design of new DVD-Ig™ molecules with adjustable binding properties.

An initial view of the DVD-Ig™ DFab structure immediately reveals how the DVD-Ig™ molecule functions to bind two different antigens on each DFab simultaneously (Fig. 1B and C). With IL18 bound to VD2 (the inner variable domain), the outer VD1 rests entirely on top of the heavy chain of the inner variable domain. This orientation of the outer variable domain positions the CDRs of VD1 for binding the second antigen approximately 85° from the inner antigen CDRs (Fig. 2A), leaving ample room for binding the outer antigen (IL12). The structure helps explain how the binding affinity for both antigens in this DVD-Ig™ molecule remains essentially unchanged from the parent antibodies [K_D (pM) mAbs: IL12 = 120, IL18 = 140; IL12-IL18 DVD-Ig™: IL12 = 130, IL18 = 160].¹ Inspection of the structure from the top of the DVD-Ig™ molecule reveals no apparent contact between IL18 and the outer VD1(Fig. 2B), and no significant interactions between the linkers and the inner VD2 CDRs or the outer VD1. Viewing the structure from the side and looking underneath the outer VD1 (Fig. 2C), there appears to be a surface complementarity between the inner surface of the outer VD1 and the outer surface of the VD2 heavy chain, and there is an approximate 10° twist to the outer VD with respect to the inner VD. When the outer VD1 rests upon the VD2 heavy chain surface, it results in an overall additional buried surface area of 1745Ų.⁵ The surface complementarity statistic, Sc = 0.67 for the interface between the outer VD2 and the inner VD1 heavy chain indicates considerable surface complementarity, but also suggests that the interface can be significantly improved.⁶ Adjusting the linker sequences may help situate the outer variable domain, VD1, in a more favorable position where the surface complementarity statistic and the overall buried surface area would be greater.

In the context of the DVD-Ig™ format, we are able to identify critical amino acid interactions as well as water mediated hydrogen bonds that are responsible for producing the (140pM) K_D for the IL18 parental antibody. It is worth noting that, until now, we have been unsuccessful at producing crystals of this particular IL18 Fab (derived from the parental IgG) complexed with IL18 at a sufficient quality for X-ray crystal structure determination. Some of the critical interactions made by the IL18 Fv to IL18 are shown along with a representative portion of the electron density map in Figure 2D. The surface complementarity statistic is Sc = 0.73 for the interface between IL18 and the IL12-IL18 DFab, which is about average for antibody antigen complexes, and is coincidentally identical to the previously reported value for the IL18/125–2H Fab interface^6,7 that recognizes an entirely different epitope. Comparing the K_D of the parent anti-IL18 antibody (K_D = 140pM) to the K_D of the IL12-IL18 DVD-Ig™ for IL18 (K_D = 160pM), it is clear that there is little loss of affinity for the binding of IL18 at the inner antigen binding site (VD2) in the DVD-Ig™ architecture. Additionally, superposition of the parental IL18 Fab (2VXV)⁶ onto the IL18 Fab portion of the IL12-IL18 DFab reveals a low main chain rmsd = 0.334 Å.

The ‘short-short’ linkers (Table 1) were used for the construction of IL12-IL18 DVD-Ig™ and are composed of sequences taken proximal to the elbow regions of a conventional IgG₁. They directly connect the C-terminus of the outer IL12 VD1 to the N-terminus of the inner IL18 VD2 of the molecule, which results in a tetravalent, dual specific, molecule with full activity for both antigens compared with the parental IgGs. The light chain linker sequence is TVAAP and the heavy chain linker sequence is ASTKGP.¹ We see complete electron density for the light chain linker, but we only see main chain density for the heavy chain linker because it is more solvent exposed. The outer variable domain can be imagined as a swinging bucket that can move over the inner antigen binding site from side to side with the linkers acting as flexible handles connecting it to the mAb Fab portion below. This outer VD “bucket” concept suggests that we may affect the outer VD movement by adjusting linker lengths and thereby modulate inner domain binding.

Having two linkers to the outer variable domain in the DVD-Ig™ format affords additional flexibility when engineering DVD-Ig™ molecules. By adjusting the length and sequence of the linkers, the position of the outer variable domain may be altered, making it possible to tune the affinity of the inner antigen site. A cartoon schematic illustrating how varying the linker length can reposition the outer VD to improve interactions with the inner VD is shown in Figure 3A. We have seen a significant loss of affinity for vascular endothelial growth factor (VEGF) when it is expressed at the inner VD2 with SS linkers (Table S2). To investigate the effect of linker length and outer VD sequence pairing on inner VD antigen affinity, we constructed a matrix of DVD-Ig™ proteins pairing anti-VEGF VD with four different paratopes of an alternate therapeutic target of 54.7 kDa that we refer to as AgA. These DVD-Ig™ molecules were made with four different linker sets, short-short (SS), short-long (SL), long-short (LS), and long-long (LL), as described in Table S2. The VEGF binding kinetics for this series of DVD-Ig™ proteins suggest that the affinity for the inner antigen can be modulated by adjusting linker length, and that it is possible to screen for desired linker length combinations. The binding kinetics (Fig. 3B) reveal a linker length dependent reduction of VEGF affinity (K_D) for the inner VD that is primarily the result of reduced on-rate (k_a) for short linkers, relative to the reference IgG. Furthermore, the binding kinetics show that the K_D can be modulated by adjusting linker length and by using different outer VD sequences. Overall, we can always identify a DVD-Ig™ molecule retaining binding affinities to both inner and outer variable domains through a process of selecting the best inner/outer variable domain combination, optimizing the orientation of the two variable domains and adjusting the linker length and type. The ability to modulate the affinity of one antigen with respect to the other may be an important feature for any dual specific binding protein when dosing two paratopes simultaneously. Although it is not necessary for the proper function of a DVD-Ig™ molecule, in certain applications one or both of the linkers can be particularly engineered to be enzymatically cleaved to enhance or activate antigen binding at the inner antigen site. Cleavage of one linker allows outer VD1 additional rotational freedom, and greater accessibility to the inner VD2 antigen binding site. This feature may enable tissue targeting of the DVD-Ig™ molecule with the outer variable domain while masking the paratope of the inner domain until the DVD-Ig™ molecule is “activated” at the targeted site by tissue specific proteases. We have recently shown that a DVD-Ig™ molecule that substantially lost antigen affinity at the inner VD2 binding site can regain binding affinity to near that of the parent antibody while still maintaining the binding functionality of the outer VD1 after selective cleavage of one of the linkers.³

The recent approval of the first bispecific therapeutic mAb (catumaxomab) in the EU,⁸ emerging clinical data,⁹ and entry of several bispecifics into clinical trials has renewed interest in design and development of bi- and multi-specific biologics. In addition, bispecific antibodies offer novel opportunities not possible with mAbs, such as recruitment of specific immune cell populations to kill tumors,^8,9 targeting biologics across the blood brain barrier,¹⁰ and delivering therapeutic agents, including stem cells, to specific sites.^11,12 Challenges in developing therapeutic bispecifics, however, remain, e.g., molecular stability, pharmacokinetics, manufacturability of individual bispecific formats. Of the more than 30 bispecific formats described to date, there are no crystal structures reported for any other bispecific format that reveals both paratopes with an antigen bound. This fact speaks to the robust nature of the DVD-Ig™ format and the ability to produce these proteins in large quantities in a highly purified and stable form. The DVD-Ig™ format is a novel immunotherapeutic platform that exploits the benefits of the immunoglobulin fold through optimization of its various components while engineering in the necessary flexibility to bind an additional antigen. This platform can be utilized to bind a combination of both soluble or cell surface expressed antigens simultaneously. The information contained within the structure of the complex of the DFab with IL18 has proven to be particularly useful for understanding antigen binding properties of other DVD-Ig™ protein pairs and for designing future DVD-Ig™ molecules. There is obviously a need for additional structural studies with a variety of antigens at both the inner and outer variable domain positions in order to more fully comprehend the dynamic nature of the DVD-Ig™, in particular how the various states of antigen occupancy affect VD positioning and how the different antibody frameworks interact with one another and reposition upon antigen binding. It is expected that this structure and future structures will facilitate the design of a new generation of dual specific therapeutics.