BTK autoinhibition analyzed by high-throughput swaps of SH2 domains

BTK, a Tec-family tyrosine kinase, resembles the Src and Abl kinases in that an SH2-SH3 module regulates the activity of the kinase domain, principally through an inhibitory interaction between the SH3 and kinase domains. In Src-family kinases, phosphorylation of a C-terminal tail latches the SH2 domain onto the kinase domain, stabilizing the inhibitory conformation of the SH3 domain; in Abl, interaction between the kinase domain and a myristoyl group on the N-terminal segment provides a similar latching function. The structure of autoinhibited BTK resembles that of the Src and Abl kinases, but BTK lacks an obvious SH2-kinase latch. To assess the role of the SH2 domain in autoinhibition of BTK, we generated hundreds of chimeric BTK molecules in which the native SH2 domain is replaced by other SH2 domains. We measured the fitness of these chimeric proteins using a high-throughput assay in T and B cells. Surprisingly, many SH2 domains increased fitness when substituted into BTK. Analysis of one set of chimeric proteins indicates that the increase in fitness stems from the ability of the substituted SH2 domains to disrupt BTK autoinhibition while maintaining phosphotyrosine targeting. Thus, although BTK lacks a specialized latch, distributed interactions between the SH2 and kinase domains stabilize the autoinhibitory conformation of BTK. While phosphotyrosine recognition can be conferred on BTK by evolutionarily distant SH2 domains, autoinhibition requires specific interactions with the kinase domain that arose through evolutionary refinement of the regulatory mechanism, and is less easily be mimicked by heterologous SH2 domains.

Methods
Quantification of fitness from sequencing data was performed as in Eisen et al., Sci Signal. 2024. Briefly, Fastq files from MiSeq runs were aligned to the Fasta files containing the full sequences of each variant using Kallisto (Bray, Nat. Biotech. 2016) to generate read counts for each variant. A read cutoff of 50 reads was applied to the input libraries such that any variant not passing this threshold was discarded. Next, the unnormalized scores were calculated by dividing the number of reads in the sorted dataset by the number of reads in the input dataset and taking the log10. These unnormalized scores were normalized by subtracting the mean of the wild-type fitness scores. The SH2-domain library included 22 synonymous wild type sequences that were generated by randomly choosing 5 codons and substituting them for randomly chosen synonymous counterparts. Sequences that introduced additional BsaI restriction sites were avoided. In the helix I library, 44 synonymous wild type sequences were included. Fitness scores were calculated by subtracting the mean of these synonymous sequences. Code to generate saturation-mutagenesis sequences and to analyze RNA-seq libraries was written using R and Python and is available on Github (https://github.com/timeisen/MutagenesisPlotCode).