A diverse family of thermostable cytochrome P450s created by recombination of stabilizing fragments

Abstract

Thermostable enzymes combine catalytic specificity with the toughness required to withstand industrial reaction conditions. Stabilized enzymes also provide robust starting points for evolutionary improvement of other protein properties. We recently created a library of at least 2,300 new active chimeras of the biotechnologically important cytochrome P450 enzymes. Here we show that a chimera's thermostability can be predicted from the additive contributions of its sequence fragments. Based on these predictions, we constructed a family of 44 novel thermostable P450s with half-lives of inactivation at 57 degrees C up to 108 times that of the most stable parent. Although they differ by as many as 99 amino acids from any known P450, the stable sequences are catalytically active. Among the novel functions they exhibit is the ability to produce drug metabolites. This chimeric P450 family provides a unique ensemble for biotechnological applications and for studying sequence-stability-function relationships.

Navigating the protein fitness landscape with Gaussian processes

“…The most stable chimera, EXPc5 (Dataset S3), had a T 50 of 69.7°C. EXPc5 is 8.7°C more stable than CYP102A1 variants that have been engineered using directed evolution (31) and 5.3°C more stable than previously identified thermostable chimeric P450s (16). EXPc5 differs from this previously published most-stable chimera by 23 mutations.…”

“…This P450 sequence-stability dataset was modeled in our previous work, using a linear regression model that associated weights to individual sequence fragments (16). The fragment-based regression model also worked well (cross-validated r = 0.90, MAD = 2.0°C) and was used to predict the sequences of new, highly stable chimeric P450s.…”

“…The first column shows the thermostabilities of the three parent cytochrome P450s. The next two columns show the results from a large sampling of a P450 recombination library, followed by sequences that were predicted to be stabilized using a fragment-based regression model (16). The next four columns (UCBr1-4) show four rounds of batch-mode upper confidence bound sequence optimization, providing a diverse sampling of thermostabilized sequences.…”

“…Chimeric protein libraries are particularly desirable training sets because they uniformly sample a massive combinatorial space of mutations. In addition, the sequences within chimera libraries have a high probability of functioning (35) and display significant functional diversity (16,36).…”

“…All chimeric cytochrome P450 genes were constructed from fragments of previously published chimeric P450s, which were originally constructed from the heme domains of CYP102A1, CYP102A2, and CYP102A3 (15)(16)(17). Single-and double-crossover chimeric genes were assembled using overlap extension PCR and cloned into pCWori (P450-specific vector) (40) or pET22b expression vectors containing a C-terminal 6×His tag.…”