Abstract
Multiplexed Assays of Variant Effect (MAVEs) are a powerful tool for assessing the effects of variation in proteins to understand sequence-function relationships, inform biological structures, and infer pathogenicity in clinical settings. We wished to use MAVEs to study the functional impact of all possible missense variants on coagulation factor IX (FIX). Variation in the F9 gene can cause decreased levels of functional FIX, leading to the bleeding disorder hemophilia B. However, genes that can be studied with MAVEs were limited to those that encode intracellular or membrane-bound proteins. This is because each variant protein must be physically connected to the genome of the cell of origin to allow functional protein selection and subsequent DNA sequencing to measure individual variant effects. However, FIX is a secreted plasma protein that does not remain physically associated with the cell that expresses it, making it incompatible with current MAVE methods. We hypothesized that cell surface display, as has been used in yeast and phage, could be adapted to establish the connection between FIX protein and F9 DNA variation needed for MAVEs in mammalian cells.
Thus, we developed a FIX mammalian cell surface display system in HEK-293 cells by fusing FIX protein to a single pass transmembrane domain with a C-terminal linker. We measured FIX expression in our mammalian cell-surface display system using flow cytometry and multiple monoclonal, fluorescent anti-FIX antibodies to different FIX domains. We introduced a library of nearly all ~10,000 FIX missense variants into this system to profile each variant's effect on FIX function. Antibody-stained cells were then sorted into quartile bins by FACS. Variants in each bin were then deeply sequenced and scored based on their frequency within each bin. Scores were then used to assign functional significance to each FIX variant.
To detect variants that impact FIX secretion, we used both antibodies to both FIX heavy chain and a strep II epitope engineered into the C-terminal linker. With the anti-FIX heavy chain antibody, we found that approximately 36.2% of variants were lowly secreted and that secretion scores alone correlated well with FIX antigen data reported in the Factor IX (F9) Gene Variant Database. We further found that proline and cysteine substitutions had strong negative effects on FIX expression. The anti-strep II antibody yielded similar results, except that cysteine substitution was generally better-tolerated. We concluded that the FIX heavy chain antibody binding is particularly sensitive to loss of function due to cysteine substitutions.
FIX post-translational modification, particularly γ-carboxylation of the GLA domain, is critical for normal FIX function. A γ-carboxylation-sensitive anti-FIX monoclonal antibody can detect properly γ-carboxylated FIX. In our system, binding of this γ-carboxylation-sensitive antibody was lost when cells were grown in the presence of warfarin, indicating that our displayed FIX is γ-carboxylated. We used this carboxylation-sensitive antibody to detect FIX variants associated with reduced FIX carboxylation-sensitive antibody binding relative to secretion. As expected, the majority of variants with disproportionately low γ-carboxylation were located within the propeptide and GLA domains.
Thus, we developed a novel membrane-tethered mammalian display system to study the functional impact of all possible missense variation in the secreted coagulation protein FIX. Our results provide direct functional evidence for loss of function in a high proportion (39.6%) of all possible FIX missense variants, either by decreasing secretion or by loss of γ-carboxylation.
Disclosures
Zapp:Adaptive Biotechnologies: Ended employment in the past 24 months; Shield T3: Ended employment in the past 24 months. Johnsen:CSL Behring: Consultancy, Honoraria; Takeda: Consultancy, Honoraria; Octapharma: Consultancy, Honoraria, Research Funding. Fowler:MAZE Therapeutics: Consultancy.
Author notes
Asterisk with author names denotes non-ASH members.
