Background

Targeting signals allow the correct intracellular distribution of newly synthesized proteins. Proteins that are transported into their target compartment in a folded conformation often show complex signals that are displayed on the surface of their native, folded structure. These signals are difficult to recognize in the primary sequence since they only can be deciphered when the fold of the native proteins is known.

In contrast, proteins that are threaded through translocases in the membrane of their target compartments in an unfolded conformation typically employ N-terminal signals. This organization allows that protein translocation can commence before translation is completed (Wickner and Schekman, 2005). N-terminal targeting sequences include the signal sequences found on secretory proteins, the transit peptides of chloroplast proteins, the MTSs of mitochondrial proteins and the leader peptides that target bacterial proteins into or across the inner membrane.

Prediction algorithms were developed that recognize and interpret these N-terminal targeting sequences (Juncker et al., 2009). These algorithms can either search for defined characteristic features (such as helicity, hydrophobicity, charge distribution or the presence of specific residues) or they are based on self-learning algorithms that were trained with test sets of proteins of known cellular localizations. One of the latter programs is TargetP which uses the primary sequence of a given protein (in FASTA or text format) to come up with a ‘probability’ score that predicts how likely it is that this sequence starts with a functional matrix targeting sequence (Emanuelsson et al., 2000; Emanuelsson et al., 2007). In the mitochondrial community, TargetP values of larger than 0.6 are typically interpreted as evidence that a given protein is targeted into the mitochondrial matrix.

Recent studies suggested that the important targeting information is not confined to the N-terminal MTS. Obviously, mature parts of proteins are critical for the translocation efficiency in vivo (Yamamoto et al., 2009; Chatzi et al., 2017; Backes et al., 2018). It is known for a long time that strongly folded domains can impede translocation (Eilers and Schatz, 1986; Harner et al., 2011; Schneider, 2018). Hence the unfolding probabilities of cytosolic domains of translocation intermediates strongly influence the velocity and yield of the translocation reaction (Wilcox et al., 2005; Yagawa et al., 2010). On the mitochondrial surface, protein unfolding is supported by cytosolic chaperones and by surface receptors, in particular by Tom70 (Hines et al., 1990; Söllner et al., 1990). The cytosolic domain of Tom70 represents a tetratricopeptide repeat structure which it shares with many co-chaperones of the Hsp90 and Hsp70 chaperone systems (Melin et al., 2015). Indeed, Tom70 cooperates with Hsp90 (in mammals) and Hsp70 (in mammals and fungi) in protein translocation across the outer membrane (Young et al., 2003; Bhangoo et al., 2007; Hoseini et al., 2016; Zanphorlin et al., 2016). It was recently shown that internal sequence stretches that share the typical biochemical properties of mitochondrial presequences support the binding to Tom70 and facilitate protein translocation into mitochondria (Backes et al., 2018, Hansen et al., 2018). Due to their similarity to mitochondrial targeting signals, these internal MTS-like sequences (iMTS-Ls) can be found using the same algorithms that are predicting mitochondrial presequences. The only necessary change one has to make is to relieve the restriction of analyzing only the N-terminus of a given protein, but iterate through the complete amino acid sequence. In this protocol, we explain how TargetP (or a similar tool) can be used to detect and analyze these iMTS-Ls protein sequences of interest. This will be helpful to analyze the features of mature regions of mitochondrial proteins and to estimate their import competence using a simple in silico test. We provide step-by-step instructions for three alternatives that contrast the different settings in data analysis. Depending on linking and scientific background, the detection of iMTS-Ls can be performed using either spreadsheet software (e.g., Microsoft Excel), a language and environment for statistical computing (R) or a general-purpose programming language (F#). While dedicated analysis platforms are the common choice in the field there is an advent for easy and intuitive general-purpose languages to facilitate the integration of the data analysis tasks into production tools and web apps or smooth the linkage of processing with high computational demand and statistical analysis.