Successful screening of human antibody panels requires an understanding of the therapeutic design goals including specificity, functional activity, and affinity. Targeting multi-pass membrane proteins can pose challenges as they can be difficult to over-express in a biologically relevant manner to be used for assays. This talk will discuss strategy for antibody screening for multi-pass protein target using examples from therapeutic antibody discovery work. 

We recently developed µCellect, an on-cell phage display approach that recapitulates the complex in vivo binding environment to produce high-performance human antibodies. In a proof-of-concept screen against human Frizzled-7, a key ligand in the Wnt signaling pathway, antibodies with picomolar affinity were discovered in two rounds of selection, outperforming current best-in-class reagents. This approach, termed µCellect, is low-cost, high-throughput, and compatible with a wide variety of cell types.

In neuromyelitis optica patients, antibody identification against the aquaporin-4 (AQP4) membrane protein traditionally involves labor-intensive single B-cell sorting, cloning, and analysis. To accelerate patient-specific discovery, we compared two approaches to screen anti-AQP4 antibodies using yeast surface display libraries. We found that both cell-based biopanning and solubilized antigen FACS were effective to select for AQP4-binding clones. These established screening techniques will accelerate library-scale antibody discovery against AQP4 and other membrane proteins.

Many novel tumor-specific targets are intracellular proteins. To efficiently target them, we can assess which are loaded onto HLA receptors and in turn generate antibodies targeting these peptide-HLA complexes. These T cell receptor mimicking (TCRm) antibodies can be highly tumor-specific, however they can be very challenging to generate. My presentation will touch on these challenges and the methods we have used to overcome them

I will present a set of machine learning technologies for the de novo design of epitope specific antibodies to membrane protein targets. Our methods encompass structure-based design of CDRs for optimal epitope recognition and sequence-based generative models ensuring favorable developability properties. We show that we obtain nanomolar Fab binders to a specified novel epitope.

Unstructured regions comprise key functional epitopes within therapeutic proteins and cell signaling proteins. However, the generation of antibodies against unstructured epitopes remains challenging. Here, we will describe two case studies on the discovery of anti-peptide antibodies with novel specificities: strong yet paradoxical degenerate recognition and recognition of a two amino acid motif. We will describe the generation, characterization, and application of these antibodies to reveal new biological insights.

DNA immunization is effective in stimulating both innate and adaptive immunities to elicit high levels of antibody responses. The in vivo expressed proteins can maximally maintain the native structures and go through appropriate post-transcriptional modifications. DNA immunization has been used by us successfully in various host species to elicit monoclonal antibodies (mAbs) against a wide range of targets including those with complex structures such as G-protein-coupled receptors (GPCR).

Antigen design, expression, and purification remain a challenge when raising mAbs to GPCRs. Here we present a case story, using a GPCR purified in Lauryl Maltose Neopentyl Glycol (LMNG). The purified receptor was applied as a soluble antigen in an in vitro Antibody discovery campaign (Adimab). Optimization of antibody selection, as well as characterization of identified hits, will be presented.

Recombinant production of multi-span transmembrane proteins is frequently challenging due to their complex structure and low target stability. Selection of the appropriate model membrane system and protein engineering strategy can address some of these challenges. The antigen formats, i.e., virus-like-particles, nanodisc, or polymer-extracted membranes must be carefully considered in the application context. In this talk, I will compare several antigen formats and engineering strategies to produce a GPCR.

Compared to soluble proteins, membrane protein structures often suffer from lower resolutions, more missing atoms/residues, and ambiguous electron densities attributed to ligand or lipid/detergent molecules. Moreover, they sometimes deviate from the necessary membrane contexts. Here we will showcase some examples of structural and functional studies of membrane proteins that used molecular dynamics simulations to remedy the shortcomings of their incomplete structures.

Synthetic antibody libraries facilitate the rapid discovery of monoclonal antibodies for virtually any target. However, synthetic methods lack in vivo filtering for broad reactivity. To overcome this limitation, we created machine learning models that quantitatively assess synthetic camelid antibody fragment (‘nanobody’) polyreactivity and predict amino acid substitutions to decrease polyreactivity. Using our model, we decreased the polyreactivity of a GPCR targeting nanobody without compromising its functional properties.

Single-domain antibodies (nanobodies) offer signature advantages for targeting cell surface proteins (e.g. GPCRs). However, it remains difficult to identify nanobodies that directly activate receptor function. We have devised methodology to link nanobodies with GPCR ligands. We show that conjugation could improve the pharmacological properties of ligands for different GPCRs. For parathyroid hormone (PTH) receptor ligands, we demonstrate improvements in potency of up to 10,000-fold. We also provide evidence that this approach can be applied to small molecule ligands for adenosine binding GPCRs. These findings set the stage for extension of this technology to other cell surface receptors.

A multi-component co-receptor complex, which includes the essential signaling protein CD19 and the tetraspanin CD81, carefully regulates the strength of BCR signaling. We determined the cryo-EM structure of the CD19-CD81 complex, revealing that CD81 opens its ectodomain upon binding to CD19 to expose a hydrophobic CD19-binding surface and reorganizes its transmembrane helices to occlude a cholesterol-binding pocket present in the apoprotein. We then developed a platform to follow the dynamic sequence of events that occur inside a B cell in response to activation. We tracked CD19 signaling dynamics in a time window from seconds to hours after BCR stimulation, revealing the kinetics of recruitment of both previously known and novel mediators of BCR signaling.