Chemical crosslinking and mass spectrometry to elucidate the topology of integral membrane proteins

Abstract

Here we made an attempt to obtain partial structural information on the topology of multispan integral membrane proteins of yeast by isolating organellar membranes, removing peripheral membrane proteins at pH 11.5 and introducing chemical crosslinks between vicinal amino acids either using homo- or hetero-bifunctional crosslinkers. Proteins were digested with specific proteases and the products analysed by mass spectrometry. Dedicated software tools were used together with filtering steps optimized to remove false positive crosslinks. In proteins of known structure, crosslinks were found only between loops residing on the same side of the membrane. As may be expected, crosslinks were mainly found in very abundant proteins. Our approach seems to hold to promise to yield low resolution topological information for naturally very abundant or strongly overexpressed proteins with relatively little effort. Here, we report novel XL-MS-based topology data for 17 integral membrane proteins (Akr1p, Fks1p, Gas1p, Ggc1p, Gpt2p, Ifa38p, Ist2p, Lag1p, Pet9p, Pma1p, Por1p, Sct1p, Sec61p, Slc1p, Spf1p, Vph1p, Ybt1p).

Fig 1. Experimental workflow and chemical structure and properties of the crosslinking agents employed.

A, schematic representation of the experimental workflow used for growing cells, preparing microsomes, carbonate wash of membranes, chemical crosslinking of proteins and preparation of peptides for MS analysis. B, chemical structure and reactivity of the crosslinkers employed BS3 (bis[sulfosuccinimidyl] suberate) and EDC (1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride.

Fig 2. Molecular mass and scores of XLs found by pLink for Pma1p and BSA.

Distributions of scores and calculated molecular masses of XLs generated by pLink by scanning through all mgf files from BS3 experiments (A-D) or EDC experiments (E-H) for Pma1p and BSA, a protein that is not present in the sample, are plotted as calculated in S7 Table. The cut off lines used to filter all pLink XLs (stippled red lines) indicate a minimal molecular mass of XL of 1’500 Da and scores of ≤ 7.5 x 10⁻⁴ for BS3-XLs, or ≤ 4 x 10⁻⁴ for EDC-XLs.

Fig 3. Frequency of XLs found by pLink for a given protein is correlated with its copy number.

The number of XLs for 18 arbitrary chosen MSPs listed in S4 Table is plotted as a function of the copy number of the respective MSP (A) or its amino acid length (B). In panel C, the number of BS3- and EDC-generated XLs for these 18 proteins are plotted separately, as a function of copy number. In panel D, the distance in the primary sequence between crosslinked amino acids in 160 XLs found in 27 proteins of S14 Table are plotted.

Fig 4. BS3 and EDC crosslinks mapped onto the structural models of Pma1p, Por1p and Sec61p.

Structures of Pma1p (A, B), Por1p (C-F), and Sec61p (G, H) were homology modeled by HHPRED [15] using plasma membrane H⁺-ATPase from Neurospora crassa (1mhs_A) [16], the voltage-dependent anion channel VDAC1 from Mus musculus (4c69_X) [17], and Sec61 from Canis lupus/Bos Taurus (3jc2_1) [18] as template. Structural models were visualized by PyMOL and the position of crosslinks connecting the Cα atoms of amino acids were added manually based on the experimental data from pLink.

Fig 5. BS3 crosslinks mapped onto the topological model of Fks1p, the catalytic subunit of 1,3-beta-D-glucan synthase.

A, Arch model of Fks1p indicating the positions of BS3 (red) XLs visualized using xVis. B, topology model of Fks1p proposed using the crosslinking data to validate a topology prediction visualized by Protter [19].