Multiplex, multimodal mapping of variant functional effects in secreted proteins

Despite widespread advances in DNA sequencing in the past decade, the functional consequences of most rare genetic variants remain poorly understood, severely limiting our ability to connect variants to their consequences on protein function, identify biochemical mechanisms by which variation causes disease, and interpret variant pathogenicity. Multiplexed Assays of Variant Effect (MAVEs), which can measure the function of tens of thousands variants, are beginning to address this problem. However, existing MAVEs cannot be applied to the approximately 10% of human genes encoding secreted proteins, about a quarter of which are associated with disease. We developed a flexible and scalable human cell surface display method, Multiplexed Surface Tethering of Extracellular Proteins (MultiSTEP), that can simultaneously measure the functional effects of tens of thousands of variants in secreted proteins. We used MultiSTEP to study the consequences of missense variation in coagulation factor IX (FIX), a vitamin K-dependent plasma serine protease where variation can cause FIX deficiency and the bleeding disorder hemophilia B. We used a panel of antibodies to detect FIX secretion or FIX post-translational modification, measuring a total of 45,024 effects for 9,007 variants. 43.8% of all possible F9 missense variants impact FIX secretion, post-translational modification or both. We also identify new signals of functional constraint on secretion including within the signal peptide, folded domains, and for nearly all variants that caused gain or loss of cysteine. FIX secretion scores correlate strongly with FIX levels in patient plasma and also reveal that most F9 missense variants causing severe hemophilia do so by profoundly impacting secretion. We integrate the secretion and post-translational modification data to develop a F9 variant classifier that can identify loss of function variants with high specificity. We use the resulting classifications to reinterpret and upgrade 62 of 97 F9 variants of uncertain significance (VUS) in the MyLifeOurFuture hemophilia genotyping project to likely pathogenic. Lastly, we show that MultiSTEP can be applied to a wide variety of secreted proteins, ranging from small signaling proteins like insulin to large proteins like factor VIII. Thus, we establish a multiplexed, multimodal, and generalizable method for systematically assessing variant effects for secreted proteins at scale, paving the way for improved understanding of biochemical mechanisms of disease and clinical variant interpretation. Overall design: Three barcoded single amino acid variant libraries of FIX variants spanning its entire sequence were recombined into an engineered, polyclonal Freestyle 293F line containing a lentivirally-introduced Bxb1 recombination site. FIX was engineered to be displayed on the surface of Freestyle 293F cells after recombination by C-terminal fusion of a flexible linker, eptiope tag, and single pass transmembrane domain. Cells were then stained with antibodies directed towards either FIX protein or the epitope tag and sorted using flow cytometry for Alexa Fluor 647 or Alexa Fluor 488 expression. Genomic DNA was then extracted from sorted cells, the barcodes were amplified by PCR, and the barcodes were sequenced by Illumina sequencing. Barcodes were associated back to corresponding FIX variants by comparison with a barcode-variant map created by sequencing the variant library plasmids using PacBio HiFi long-read sequencing.