BS3 and its ability to cross-link proteins of all mitochondrial subcompartments 79

Two commonly used and commercially available cross-linkers are disuccinimidyl suberate (DSS) and its sulfonated derivative bis(sulfosuccinimidyl) suberate (BS3)¹⁸³. Both of them are non-cleavable, homobifunctional N-hydroxysuccinimide (NHS) esters that preferably react with the ε-amino group of lysine residues and protein N-termini. Both cross-linkers possess a spacer arm that, after reaction, covers a distance of 11.4 Å. The difference between both compounds is the charge in BS3 that is introduced by the sulfo-group.

Therefore, BS3 is charged, hydrophilic, and, hence, water-soluble. DSS is uncharged, hydrophobic, and therefore only soluble in organic solvents. Consequently, DSS is supposed to be membrane-permeable, while BS3 is not. When cross-linking mitochondria, an organelle with two lipid bilayers, membrane-permeability of the cross-linking ingredient is a key parameter. The functionality of DSS for cross-linking human mitochondria was demonstrated recently¹⁰¹. However, the use of DSS comes along with a major drawback.

Since organic compounds such as DMSO are necessary for the solubilization of DSS (or, in general, for all membrane-permeable cross-linkers), an interference of such solvents with the membrane system has to be considered, at least when using higher concentrations of the cross-linker. In general, DMSO tends to induce water pores in membranes and increases their permeability³⁰³. By applying transmission electron microscopy, Yuanet al.²⁷⁶ could show that mitochondria in cultured astrocytes treated with 1 % DMSO within a time period of 24 h underwent swelling. An increase to 5 % DMSO resulted in even more severe damages such as disruption and loss of cristae. While

0.5-4.2 BS3 and its ability to cross-link proteins of all mitochondrial subcompartments

1.5 % DMSO are widely accepted especially in cell cultures addressing medical problems³⁰⁴, higher concentrations seem to be problematic for membrane integrity. To catch also low abundant proteins and their interactions, a high cross-linker concentration is crucial²¹¹. Therefore, using a water-soluble cross-linker for sensitive membrane systems might be the better choice. Based on this information, BS3 was tested in this study for its suitability to cross-link mitochondria, although it is supposed to be membrane-impermeable.

Surprisingly, BS3 showed similar cross-linking efficiency as DSS (with 1 % DMSO) in all mitochondrial subcompartments (see subsection 3.1.1). This raises the question, how a negatively charged 570 Da molecule can pass through two functional membrane bilayers. The outer mitochondrial membrane might act as a sieve with free permeability for molecules smaller than 1-2 kDa due to Por1³⁰⁵. Por1, also known as the voltage-dependent anion channel (VDAC), is a highly abundant protein within the outer mitochondrial membrane, forming a pore with a diameter of approximately 2-3 nm. In its open form this channel is selective to anions³⁰⁶. More recent studies suggested a permeability of Por1 for hydrophilic small molecules of up to 5 kDa³⁰⁷ (also for ATP as modeled for murine VDAC³⁰⁸) and even linear DNA might use Por1 as transporter in plants³⁰⁹. Therefore, Por1 might be one entry gate to pass through the outer mitochondrial membrane. Additionally, besides the translocase of outer membrane (TOM) complex that is responsible for protein import into mitochondria, three other outer membrane transporters have been identified recently, namely Ayr1, Omc7 and Omc8³¹⁰. Ayr1 is a NADPH-dependent channel whereas the latter two are anion channels whose substrates are still unknown. In theory, also these channels might aid BS3 import.

The inner mitochondrial membrane forms a closed system with specific carrier proteins what makes it much more difficult for the charged BS3 molecule to pass through this particular membrane. One of the most abundant carrier proteins is Pet9, the ADP/ATP exchange carrier³¹¹. Since this protein carries ADP and ATP, which are both charged and hydrophilic molecules, Pet9 might aid the transport of BS3 in an unspecific manner. In the past, it was assumed that Pet9 is the pore forming unit of the mitochondrial megachannel (MMC), also known as permeability transition pore (PTP)^312,313. More recent studies propose that rather the F1F0 ATP synthase than Pet9 is the pore forming unit^314,315. The PTP is Ca²⁺- and reactive oxygen species (ROS)-dependent and forms, similar to Por1, a 3 nm large pore, allowing for molecules up to 1.5 kDa to pass^316-318. The PTP is also responsible for mitochondrial swelling³¹⁹ that might have been induced by the slightly hypotonic cross-linking buffer used in this thesis. However, channels and pores might also

freezing. Additionally, the cross-linking reaction was performed for 1 h at room temperature in a slightly hypotonic buffer that might have increased the damage enabling the cross-linker to reach every mitochondrial subcompartment. The membrane integrity of mitochondria could be tested, for example, by recording electron microscopy (EM) images of mitochondria after incubation in cross-linking buffer with and without cross-linker.

4.3. Benefits and drawbacks of restricted databases for cross-linked peptides searches

Cross-linked peptides were identified by database search with the software pLink 1^255,256. It identifies peptides cross-linked with a non-cleavable cross-linker such as BS3 and DSS, i.e. the covalent bond formed between two amino acids will not be cleaved during the fragmentation in the gas phase of the mass spectrometer (see section 4.8). This poses a huge computational challenge, since one spectrum does not only contain information of fragment ions of one peptide – as is the case in common proteomics – but information of fragment ions of two peptides, linked by the cross-linker. Consequently, spectra displaying cross-links are chimeric and, hence, more complex. A major bottleneck in cross-linking experiments is therefore the computational analysis of cross-linked peptides spectra¹⁸⁰. In contrast to a conventional database search of linear peptides, the search space of possible peptide combinations increases quadratically (in fact: (n²+n)/2, withn being the number of tryptic peptides). This is termed the n²-problem^180,220 and makes the analysis computationally expensive. In XL-MS studies that investigate single proteins or protein complexes containing only a few proteins, the n²-problem in database search can be neglected¹⁸⁰. However, the n²-problem becomes unsurmountable in the elucidation of large-scale protein-protein interactions. The analysis of 50 cross-linked proteins inflates the search space as much as a search for linear peptides of the whole human proteome²²⁰. The higher the number of proteins within the database, the higher the possibility of identifying false positives. To counteract this, in this thesis restricted databases were generated for the identification of cross-linked peptides, reducing the search space to the 400 most abundant proteins (proteins that are included in the respective databases are listed in Supplementary Table 6 provided on a CD-ROM attached to the hardcopy version of this thesis). Three benefits come along with the use of restricted databases: (i) Database searches with pLink 1 were still possible in a reasonable time frame. (ii) Smaller databases decrease the number of false positive identifications¹⁹⁵. (iii) Cross-linking reactions occur mostly within and between high abundant proteins^211,320 that are covered by the used databases in this thesis. Interactions between low abundant proteins are therefore anyway difficult to detect. A similar observation was made by Liuet al.¹⁰⁰ who cross-linked murine

4.4 Protein-protein cross-links in mitochondria derived from yeast grown on either