RESURGENCE OF BISPECIFIC ANTIBODIES - Fusion Protein Technologies for Biopharmaceuticals: Applications and Challenges - page 518

epitopes may allow simultaneous receptor binding by the
antibody, while others only allow binding of the two recep-
tors in a mutually exclusive manner. It is believed that
formats at or below 100 kDa are more suitable for target
tissue penetration than IgG-like fusion protein of 150-
250 kDa [52]. A smaller size may translate into a higher
biological activity when it comes to receptor binding on
target tissue, but may not matter for neutralizing ligands in
the blood stream or interstitial fluid.
For engagement of immune effector cells by bispecific
antibodies, it will greatly matter which kind of immune cells
shall be linked to target cells for mediating redirected lysis.
For engagement of NK cells, presence of a Fc g 1 part is a
frequently selected option because of the positive clinical
experience with mostly ADCC-mediating antibodies ritux-
imab and CAMPATH-1. Point mutations in the CH2 domain
or manipulations avoiding the addition of fucose residues to
the complex carbohydrate moiety in the CH2 domain have
been shown to greatly enhance ADCC [51,86]. Moreover,
Fc g parts can engage immune cells other than NK cells, such
as Fc g R I (CD64)-expressing macrophages, or Fc g RII
(CD32)-expressing granulocytes. The use of anti-CD16A
or anti-CD64 antibodies for cytotoxic immune cell engage-
ment has been preferred in several instances over Fc g 1
domains, as done by particular Tandabs and DARTs, and
by MDX-210. Arguments include that this way the inhibi-
tory Fc g R IIB (CD32B) will not be stimulated and that
excess IgG in serum will not compete for drug binding to the
Fc g receptor.
For engagement of T cells, different design premises of
bispecific antibodies may apply as for engagement of NK
and other immune cells. Owing to the superior cytotoxic
potential of T cells, T-cell engagement by bispecific anti-
bodies holds the promise of treating cancer by completely
eradicating target-expressing cancer cells. Owing to an
enormously potent signal-amplifying cascade, T cells are
highly responsive to minute stimulation of their TCR com-
plex. In fact, single digit numbers of MHC/peptide com-
plexes can fully activate the lytic potential of effector
memory T cells [87, 90]. T cells do not bear Fc g receptors
and can therefore not be directly engaged by Fc g parts of
regular IgG antibodies. Therefore, most T cell-engaging
bispecifics target an invariant sequence of the TCR complex.
The flip side is that by potential binding to every T cell,
such bispecific antibodies can broadly activate the circulat-
ing T-cell compartment, which can cause serious cytokine-
related adverse events. These were observed on first infusion
of anti-CD28 IgG4 antibody TGN1412 to healthy volunteers
[87], and are naturally triggered upon infection with bacteria
secreting “superantigens,” which can cross-link the TCR
b -chain with MHC class II molecules [88]. Likewise, the
bivalently CD3-binding murine mAb OKT-3 (Orthoclone 1 )
can upon first infusion cause a cytokine release syndrome,
which is followed by TCR down-modulation, T-cell anergy,

and mutual redirected T-cell lysis (fratericide). The latter
three reactions are clinically exploited for the prevention of
acute transplant rejection. T-cell engagement and activation
by bispecific antibodies has thus to be subtle, well con-
trolled, and ideally not directly cross-linking the TCR
complex. T-cell activation should rather occur in a highly
conditional manner, only as long as target cells are present,
and not trigger significant TCR down-modulation and T-cell
anergy. This may require a delicate tuning of the anti-TCR
binding arm for optimal binding affinity, epitope recognition
and conformational impact on the bound TCR complex.
Likewise, the binding affinity and epitope recognition of the
second arm of the bispecific antibody for target cells must be
optimal and well balanced with that for T-cell binding.
T cells engaged by BiTE antibodies can eliminate both
dividing and non-dividing cells bearing the target antigen.
This is in contrast to ADCs carrying an antiproliferative drug
payload, which will predominantly eliminate dividing cells.
Hence, great care has to be taken when selecting target
antigens for T-cell-engaging antibodies.
Fc g domains—as present in numerous bispecific anti-
body formats—bind to FcRn on endothelium and to other
Fc g receptors as found on most immune cells. As a conse-
quence, these functionally “trispecific” antibodies can tether
T cells to a plethora of Fc g receptor-positive cells and
eventually trigger this way T-cell activation and redirected
lysis of Fc g receptor-expressing normal cells. This potential
issue of trispecific antibody catumaxomab, which may
explain its low systemic tolerability, has obviously been
managed by both its intraperitoneal administration and
intrapatient dose escalation starting from very low micro-
gram doses [3].
A notorious challenge with BiTE antibodies is the lack of
appropriate animal models. This relates to the very limited
repertoire of available and suitable TCR-binding antibodies.
While xenograft studies in immunodeficient mice can still
use human T cells as effector cells, T cells of rodents or
macaques, as are typically used for preclinical safety studies,
will not be recognized by bispecific antibodies. Unless, a
species-reactive surrogate antibody is generated, the conse-
quences of T-cell activation and expression of a particular
target antigen on normal tissues can therefore not be
assessed prior to Phase I clinical testing in patients. As a
consequence, only minute doses delineated in vitro by the
“minimum-anticipated biological effect level” (MABEL)
approach can be used to define start doses for first-in-man
studies, which may then require long-lasting escalation
periods to reach active dose levels. Recently, a human/
primate cross-reactive anti-CD3 human scFv antibody has
been developed as a basis for constructing fully human/
primate cross-reactive BiTE antibodies. Exemplary anti-
EGFR BiTE antibodies were shown to cause the expected
target-dependent adverse events in a macaque model [89].
The opportunity to explore potential target-specific adverse