Fig 1.
Approach to engineering Fn3 proteins to recognize tumor biomarker MSLN for diagnostic and therapeutic applications.
(A) The tenth domain of human fibronectin type III (Fn3) (PDB 1TTG) is a highly stable protein structure with three loops (BC, DE, and FG) broadly tolerant of mutation to confer novel binding properties. Structure was rendered in PyMOL. (B) We employed a previously developed hydrophilic Fn3 yeast surface display library [56] that incorporates a range of loop lengths and biased amino acid composition to mimic the diversity of naturally occurring antibody complementarity-determining regions. (C) Fn3 proteins that bind cell surface protein MSLN have numerous potential clinical applications, such as through diagnostic imaging, internalization for drug delivery, and metastatic reduction by blocking MSLN-MUC16 interactions. Stars represent conjugated imaging or therapeutic molecules.
Fig 2.
Directed evolution of a naïve yeast surface display library yielded Fn3 variants that bind soluble MSLN.
We started with a naïve yeast surface display library with 2.8 x 109 variants of the Fn3 non-antibody scaffold. The library was sorted for full-length protein expression, detected by an antibody to a terminal c-myc epitope tag, and binding to MSLN using MACS and FACS. Red polygon indicates example cell population collected for further enrichment and analysis. Additional diversity was introduced into the enriched library through a single round of mutagenic PCR and sorting of this second generation library resulted in further enrichment for MSLN binding variants. A double-negative population of yeast cells is characteristic of yeast surface display.
Fig 3.
Yeast displayed Fn3 variants 1.4.1 and 2.4.1 bound specifically to tumor biomarker MSLN.
(A) Two dominant Fn3 variants, 1.4.1 and 2.4.1, were recovered from a first generation and second generation Fn3 library, respectively. (B) Individual clones and Fn3 WT were displayed on the surface of yeast and incubated with a range of concentrations of soluble MSLN. Experimental triplicate data were collected, and the dissociation constant is reported as the mean and standard deviation of the KD values calculated for each replicate. A representative binding curve is shown for each variant. (C) Individual clones were displayed on the surface of yeast and incubated with a range of concentrations of a biotinylated, Fc fragment. Experimental triplicate data were collected. A representative curve is shown for each variant.
Fig 4.
Production and characterization of selected Fn3 variants.
Engineered Fn3 clones 1.4.1 and 2.4.1 were expressed in bacteria with a C-terminal hexahistidine tag and a short peptide tag containing GKSK residues for later bioconjugation chemistry. (A) Fn3 protein 1.4.1 was purified by nickel affinity chromatography followed by SEC, demonstrating desired product with retention time of ~ 42 min. (B) Fn3 protein 2.4.1 was purified by nickel affinity chromatography followed by HPLC, demonstrating desired product with retention time of ~30 min. (C) Proteins were purified to high purity > 99% as analyzed by SDS-PAGE. Yields of Fn3 protein production were routinely ~ 10 mg/L.
Fig 5.
Engineered Fn3 protein variants bound cancer cells expressing MSLN.
A431/H9 cells, epidermoid carcinoma cells transfected to express high levels of MSLN, and MCF-7 cells, breast cancer cells lacking surface MSLN, were used in all binding assays. (A) Analysis by flow cytometry confirms MSLN presence on the surface of A431/H9 cells as detected by an anti-MLSN antibody. The MCF-7 cell line does not express MSLN. (B) Fn3 variants 1.4.1 and 2.4.1 were isolated and binding to MSLN was measured using equilibrium binding assays. A431/H9 and MCF-7 cells were incubated with a range of concentrations of soluble fluorescently labeled 1.4.1 or 2.4.1. The assays were performed in experimental triplicate. Data from each replicate were fit to a sigmoidal curve, and a KD value was calculated for each replicate. The KD is reported as the mean +/- standard deviation. A representative binding curve of each clone for both cell lines is shown.
Fig 6.
Engineered Fn3 protein variants 1.4.1 and 2.4.1 localized to early endosomes upon binding MSLN.
Analysis by (A) flow cytometry and (B) imaging flow cytometry confirms MSLN presence on the surface of KB-3-1 cells compared to the MSLN-negative MCF-7 cells, as detected by an anti-MSLN antibody. (C, D) KB-3-1 cells (top) internalize AF488-labeled 1.4.1 (C) and AF488-labeled 2.4.1 (D), while MCF-7 cells show no specific binding or internalization (C bottom, D bottom). Endosomes are detected by an AF647-conjuated antibody recognizing the EAA1 early endosomal marker. Yellow in the merged images indicate co-localization between AF488-1.4.1 or AF488-2.4.1 anti-MSLN engineered proteins (green) and EEA1 (red). Original magnification 40X. Co-localization is quantified by the Bright Detail Similarity (BDS) metric, with values near 1 indicating co-localization. KB-3-1 BDS = 1.31 and 0.919 for AF488-1.4.1 and AF488-2.4.1, respectively. BDS values are not quantifiable for the negative control cell line, due to insufficient fraction of negative control cell population staining for binding or internalization of engineered protein variants.