NOVEL APPLICATIONS OF BISPECIFIC DART PROTEINS - Fusion Protein Technologies for Biopharmaceuticals: Applications and Challenges - page 523

V H -A) can either be expressed as separate chains (diabody)
or combined in a single polypeptide that folds back on itself
(single-chain diabody) [12,14]. With appropriate linker
selection, one can also drive the latter to a dimeric tetra-
valent structure (tandem diabody) [15,16].
Bispecific Ig-like structures, that are monovalent for each
antigen-binding domain, can also be constructed recombi-
nantly in a number of ways. Several recombinant strategies
that promote heterodimer formation and deter homodimeriza-
tion of the two Fc-bearing chains have been developed by
modifying interface sequences in CH3 of the respective chains.
These include (1) a knobs-into-holes strategy that creates a
protuberance on one strand by replacing small amino acid side
chains with larger one, and a structurally complementary cavity
on the other strand by replacing large side chains with smaller
ones; (2) strand-exchange engineered domain (SEED) of CH3
composed of alternating segments of human IgA and IgG
sequences; and (3) reversing the electrostatic interactions on
the respective CH3 sequences [17-19] by swapping acidic for
basic amino acids and vice-versa. However, it remains difficult
to preferentially assemble different antigen-binding domains
on opposite sides of the molecule in a homogeneous manner.
This can be achieved by selecting two antibodies that use a
common light chain; by expressing a scFv on one chain and a
normal Fab on the other; or by a “cross-over” strategy in the Fab
region of one arm. One “cross-over” approach is to replace the
C k of one antibodywith CH1 on the light chain and likewise the
CH1withtheC k on the heavy chain [20].
Ig-like structures that are bivalent for each antigen can
simply involve a scFv appended to the C-terminus of an IgG,
but also can be made by fusing a second set of variable
regions onto the existing variable regions (V L -B onto V L -A
and V H -B onto V H -A) of the respective IgG light and heavy
chains with flexible intervening linker sequences, yielding a
dual-variable domain Ig (DVD-Ig) [21].
Bivalent bispecific antibodies can also be generated by
linking two independent target-binding peptides to a scaf-
fold antibody through chemical conjugation. One such
bispecific antibody, designed to block angiogenesis through
combined inhibition of VEGF and Ang2 is currently under-
going clinical evaluation [22].

The novel bispecific antibody platform called dual-
affinity retargeting (DART 1 ) is based on covalently link-
ing Fv regions of two distinct antibodies. DART 1 proteins
arrange the V H and V L domains specific for different
antigens in a bispecific diabody arrangement with the
order V L (A)-V H (B) and V L (B)-V H (A). The association
of the two chains is stabilized through covalent car-
boxy-terminal disulfide linkage. DART 1 proteins have
been focused mainly on two areas: (1) recruiting inhibitory
receptors to activating receptors on the same cell to
negatively modulate cell activation [23] and (2) recruiting
effector cells to tumor cell targets for redirected cytolysis
of the tumor cells [24,25].
DART 1 proteins are easily expressed in mammalian
systems at high levels. During secretion, the two chains
associate and become disulfide bonded in an intracellular
compartment to form the heterodimeric bispecific unit.
Preferential formation of the heterodimeric form can be
promoted by the introduction of native Ig sequences
upstream from the hinge of the C k and CH1 regions
[24] or by the use of oppositely charged coil sequences
on the respective chains [26]. These “E-coil” and “K-coil”
sequences strongly favor heterodimer over homodimer
formation, greatly simplifying downstream processing
by minimizing the formation of unwanted homodimeric
species.
For downstream processing of DART 1 proteins, both
protein A and nonprotein A based schemes have been success-
fully developed. The former requires that one or both of the Fv
sequences in the DART 1 be able to bind to protein A through
the use of human V H 3 family sequence(s) (or an antibody
sequence fromanother species, determined empirically to bind
protein A) [27]. While not absolutely required, protein A
binding is preferable in that the downstream steps can be
adapted from established antibody purification processes
that are standard in the industry. Nonprotein A based purifica-
tion of DART 1 proteins for research use can be accomplished
using an affinitymatrix composed of one of the target antigens
immobilized to Sepharose, or using a specific antibody that
recognizes the associated E-coil/K-coil structure.
The remainder of this chapter focuses on the application
of DART 1 technology for oncology and autoimmune dis-
eases and introduces a Universal DART 1 (U-DART)
platform; a technique for rapid screening of monoclonal
antibodies to prioritize them for subsequent development
into DART 1 molecules.

36.2 DART 1 PROTEINS

Fragment-based and Ig-like strategies often both rely on
noncovalently linked scFv or diabody structures to maintain
the integrity of the binding unit. As the association between
the V L and V H is quite variable, and there is always equili-
brium between the associated and nonassociated forms, the
stability of these molecules in solution and under conditions
of use have been highly unpredictable. This is probably the
major reason that bispecific antibodies have not, as a class,
moved further in the clinic and product development.

36.3 APPLICATION OF DART 1 TO CROSS-LINK
INHIBITORY AND ACTIVATING RECEPTORS

Bispecific antibody-based approaches to autoimmune dis-
ease control have largely aimed at dual blockade of potential
disease mediators. For instance, simultaneous blockade of