Results
AlphaFold2 models help solve crystal structure of the inner membrane reductase FoxB (CASP: T1058) by molecular replacement – by IJ and HT.
From email to the CASP Prediction Center: The model you sent me (from the leading group) worked for molecular replacement and we finally solved the structure by MR-SAD. I am still astonished that the human expert model worked, while none of the server models we tried did (as they were rather similar). Henning Tidow
Brief description of the target
Most microorganisms rely on the bioavailability of iron for their survival. Due to the low solubility of ferric iron, they often use secreted siderophores for the chelation and uptake of iron. In Gram-negative bacteria, siderophores are usually taken up by TonB-dependent transporters (TBDTs) located in the bacterial outer membrane. The route of ferric-siderophores across the inner membrane (IM) is less straightforward and differs across many bacterial species and siderophore chemistries. Ferric-siderophore complexes are either recognized by the dedicated periplasmic-binding proteins for delivery to IM transporters for uptake into the cytoplasm or the iron is released from the ferric-siderophore complexes by a reduction mechanism. The Gram-negative bacterium Pseudomonas aeruginosa (an opportunistic human pathogen) is able to take up Fe-siderophore complexes called ferrioxamines via a dedicated TBDT FoxA in an act of siderophore piracy16. For several years we also worked towards the structure determination of FoxB, another protein of unknown function located in the same operon as FoxA. With the help of the AlphaFold2 model generated in the course of the CASP14 competition, we were able to determine the structure of FoxB. It possesses a novel fold with the transmembrane domain harboring two heme molecules indicating a role as inner membrane reductase involved in Fe-siderophore uptake and processing17.
Workflow of how an AlphaFold2 model helped to solve the structure
Native FoxB crystals obtained in decyl-maltopyranoside (DM) diffracted to approximately 5 Å resolution on average. Most of the crystals belonged to the P212121space group. All crystals were obtained in 30% PEG 600, 0.1 M BICINE pH 9, 0.1 M ZnSO4. Use of a lipid-like peptide (LLP7) as additive allowed us to collect several datasets extending to 3.4-3.5 Å18.
All molecular replacement attempts using distant homologs and homology models thereof failed. We acquired Se-Met anomalous data to 3.5 Å resolution, with anomalous signal to 4.5 Å as well as anomalous data at the Fe edge with anomalous signal to 4 Å, as we knew from spectroscopic characterization that FoxB most likely contained at least one heme group. Combining all anomalous data provided some experimental phases and allowed partial model building. A single FoxB was present in the asymmetric unit. However, phasing power was only sufficient to build approximately 60-70% of the backbone structure (Fig. 1). Although two heme groups could be successfully placed, further tracing of the protein backbone and confident sequence assignment was prevented by the low number of Met residues (5/382) and low resolution of the datasets. Lysozyme fusion at the N-terminus also resulted in crystals diffracting to approximately 4.5 Å.