4 Discussion
In vitro display platforms have been employed extensively over
the past 3 decades for protein discovery and
characterization.7 Within the past 10 years, whole
cell selection technologies have empowered selection approaches against
MP targets by simulating the natural context of the protein relevant for
ligand recognition.18 In this study, we build upon
these whole cell selection technologies by presenting and characterizing
a novel platform that we call biofloating to interrogate protein
interactions within the context of a yeast/mammalian cell system. In
contrast with previous yeast/mammalian cell interaction systems, our
approach enables incubation of yeast with mammalian cells in suspension
and utilizes quantitative flow cytometry analysis, without requirement
for the incorporation of genetic fusions into either the yeast or
mammalian cells. Hence, this versatile interaction system allows for
compatibility with existing yeast display infrastructure. Moreover, we
found that the biofloating platform demonstrated superior sensitivity
compared to biopanning, both in terms of kinetics and equilibrium
interactions investigated in this study. In particular, dramatic
differences in binding behavior were evident in the context of low
target binding affinity or reduced target expression (avidity). Of note,
the biofloating approach detected high, medium, and low affinity
interactions between yeast-displayed scFvs and target antigen-expressing
mammalian cells, whereas biopanning detected only high and medium
affinity interactions, and with significantly weaker potency (Figure 3).
In addition, the biofloating method detected scFv-displaying yeast
interactions with dense and medium density antigen-expressing mammalian
cells (Figure 5), whereas biopanning was only able to detect
interactions with antigen-dense cells, and with significantly weaker
sensitivity.
Interestingly, the biofloating platform achieved maximum yeast/mammalian
cell binding virtually instantaneously in the case of high and medium
affinity interactions, whereas the biopanning platform exhibited a more
gradual rise in binding (Figure 2). The biofloating approach also led to
immediate saturation of dense and medium density target-expressing cell
lines, whereas the biopanning strategy effected a time-dependent rise in
binding for only the dense cell line (Figure 4). The significant
differences in on-rate kinetics between the two platforms likely results
from distinct spatial arrangements and mixing capacities between the
2-dimensional biopanning and 3-dimensional biofloating systems. In the
biopanning platform, yeast gradually settle on top of the mammalian cell
monolayer, thus manifesting a diffusive limitation in achieving
cell-cell binding. In contrast, in the biofloating platform, yeast cells
are well mixed in suspension with mammalian cells, thereby eliminating
any diffusive barriers.
A major advantage of the biofloating platform is integration with flow
cytometry, which enables precise quantification of binding dynamics.
This gives us access to important molecular parameters that define
yeast/mammalian cell interactions, such as the number of yeast bound per
mammalian cell and the percentage of mammalian cells that interact with
yeast. In addition, the biofloating platform interrogates protein
interactions fully in suspension, which can be exploited for in
vitro evolutionary schemes. To this end, we demonstrated enrichment of
yeast displaying a specific scFv from a library of scFv-displaying yeast
using a fully suspension cell-based MACS selection approach. This work
builds upon a recently reported MACS-based selection
platform20 to achieve specific enrichment of a clone
spiked into a naïve scFv library and to elucidate the dynamics of
enrichment over of multiple rounds of evolution. We showed that this
suspension cell-based approach led to robust enrichment of yeast
specific for the mammalian cell-expressed target antigen, and this
enrichment was more rapid than for an analogous adherent cell-based
selection approach (Figure 6). Importantly, substantial enrichment
occurred in the first round using the suspension cell-based approach.
This observation is significant since MACS is most useful in early
rounds to debulk large libraries, compared to alternative approaches
such as FACS, which take prohibitively lengthy amounts of time to screen
highly diverse libraries. Indeed, the throughput of our approach enabled
rapid screening of a 109 member scFv library
incorporating a 10-fold excess of yeast cells. We ultimately envision
integration our methodologies into a hybrid MACS/FACS selection
workflow. Initial rounds of MACS could be implemented to accommodate
large library sizes, followed by subsequent rounds using FACS for
superior control over the affinities of the enriched proteins. The
96-well plate biofloating format we have developed could then be
implemented for high-throughput screening of individual clones from the
evolved library to monitor selection progress and identify lead clones.
This post-selection screening would be logistically challenging and
significantly less sensitive if implemented using the biopanning format,
underscoring the advantage for our novel biofloating platform.
Furthermore, although FACS was not used in this study, our
characterization of yeast/mammalian cell interactions via flow cytometry
offers useful insight that could inform the design of FACS-based
selections in future implementations. An important limitation to keep in
mind is that specific scFv-expressing yeast were spiked into the naïve
yeast scFv library at a ratio of 1:1000. Representation of a binding
clone within the library would likely be much lower; thus, additional
rounds of selection could be required to achieve adequate enrichment. In
addition, these selections were carried out using a high affinity
anti-PD-L1 scFv, which may not be present in a naïve library. Indeed,
Lown et al found that enrichment of yeast displaying a target-specific
fibronectin using MACS-based selections against target-expressing
mammalian cells was more challenging for fibronectin clones with lower
affinity.20 Thus, the selection process may require
optimization to enable enrichment of lower affinity clones.
Antibody discovery against MPs has immense potential to advance
scientific research and therapeutic development. The ability to study
protein interactions and perform selections against MPs in their native
conformations in the context of the cell membrane empowers the
characterization and manipulation of complex MPs such as GPCRs and ion
channels, for which the native antigen display has been historically
difficult. Our novel biofloating platform enables quantitative
investigation of MP interactions in a fully suspension cell-based system
that is compatible with existing yeast display infrastructure. This work
also paves the way for development and optimization of new selection
technologies that will advance MP-focused therapeutic design.