We present a comprehensive study of putative SPs in the genome of L. plantarum WCFS1 for which SignalP predicted a unique cleavage site for signal peptidase I. The results provide genome-wide insight into SP functionality, new tools (vectors) for secretion of proteins using homologous SPs, and increased insight into the predictability of SP functionality on the basis of sequence only.
82% (p < 0.05) of the 76 tested SPs led to secretion of NucA. While this result may be taken to confirm that the 62 L. plantarum
proteins containing these SPs indeed are secreted, it does not imply that the remaining 14 SPs do not function at all and that their cognate proteins are not secreted. SP functionality depends on which protein is being secreted [41
], meaning that SPs that do not work for NucA may function when coupled to another protein (and, in principle, vice versa
). Furthermore, in some cases prediction of the signal peptidase cleavage site may have been wrong, despite the unanimous prediction by the two Signal P algorithms (see also below). Indeed, comparison of the sequences of some of the seemingly non-functional SPs (see additional file 1
) with what is known about cleavage site sequences (illustrated by the sequence logos in additional file 2
) show that alternative cleavage sites are possible in some of these SPs. The detected levels of extracellular NucA varied by three orders of magnitude. Since the only difference between the constructs is the SP, the large differences in secretion capacity are due to variation in the SP, directly or indirectly. To try to unravel the causes of these variations we set up additional experiments and looked closer into the properties of the SPs.
Real-time PCR studies of cultures containing different constructs did not reveal significant differences in mRNA levels. This indicates that the large variation in secretion capacity observed for these constructs is not due to differences in transcription levels. This is an expected result, since the constructs contain identical transcription initiation and termination signals. Thus, the variation in secretion capacities must be governed by (inter-related) post-transcriptional factors such as secondary structure of mRNA, codon usage and translation efficiency, the interaction between the precursor protein and the translocation machinery, the efficiency of the signal peptidase for the SP in question, the rate of (non-desirable) intracellular and (desirable) extracellular folding, and possible interactions between the secreted protein and the bacterial cell wall [41
Although the Western blot of Figure provides only limited quantitative insight, the data do suggest that all L. plantarum
transformants produced approximately equal amounts of NucA, meaning that all transformants experienced approximately equal "protein loads". The data show a (rough) correlation between translocation efficiency and the levels of secreted protein (Figure ). One possible cause of variation in secretion efficiency is variation in the efficiency of SP processing. However, in their genome-wide study of B. subtilis
SPs Brockmeier et al. [42
] showed that the rate of precursor processing had limited effects on levels of extracellular reporter protein. Assuming a similar situation in L. plantarum
, differences in the efficiency of the translocation process itself remain as the main cause of the variation in extracellular NucA levels.
Mutagenesis studies have confirmed that secretion levels in Gram-positive bacteria are not only affected by variation in the SPs [46
] but also by variation in the N-terminal part of the mature protein [21
]. Le Loir et al. [21
] showed that negative charge in the N-terminal part of the secreted protein was beneficial for secretion. The NucA variants in the present study varied only with respect to residues +1 and +2 and we did not observe correlations between the character of these residues and secretion performance of the SP. The very efficient Lp_3050 sequence has a basic residue (Lys) at position +2 which is unexpected on the basis of the conclusions drawn by Le Loir et al. [21
]. Taking into account the above considerations, it is likely that the variation in the secretion of NucA observed in this study is caused by the variation of the SP only and its effect on the interaction between the precursor and the translocation machinery.
The translocation process is a complex process which involves many interactions that are affected by the characteristics of both the SP and the protein. It is conceivable, that SPs are evolutionary adapted to their cognate protein to ensure efficient and controlled secretion. The importance of the protein part is clearly shown in both the present study and a previous genome-wide study on SPs from B. subtilis
], which show that the efficiency of many SPs depends on the reporter protein. Thus, high secretion efficiency requires an optimal combination between the SP and the target protein. Recent studies suggest that SP function may be much more complex than previously thought, and may direct surface proteins to different subcellular locations [49
]. Clearly, such underlying complexities in SP functionality, will weaken correlations between SP sequence properties and secretion levels.
Several studies have shown that changes in hydrophobicity of the H-domain can affect the secretion capacity [47
] and this is indeed one of the correlations that we discovered in the present genome-wide study. However, in a study of 148 SPs from B. subtilis
] no such correlation was found. In the present study, we also identified a clear correlation between a predicted transmembrane helix by the programme TMHMM and high secretion capacity. On the basis of our results, running TMHMM seems one of the best ways to select SPs that are likely to perform well, and this analysis should thus be performed next to SignalP. In addition, the length of the H+C domain should also be taken into account when selecting an SP. It is interesting to note that SPs from proteins that are thought to be anchored to the cell wall tend to perform less well than other SPs. It is conceivable that these proteins do not require high translocation efficiencies, since they are not meant to be actively secreted to the surrounding media and therefore may be produced at lower levels than released proteins.
In this study, we have based the prediction of signal peptides on the original analysis of the L. plantarum
genome as described by Kleerebezem et al [2
] and we have used SignalP 3.0 to check and predict the cleavage sites. Clearly, the annotation of the L. plantarum
genome will evolve as bioinformatic tools evolve and today's annotation, e.g. with respect to the subcellular localization of proteins, will differ from the one published in 2003. The most accurate prediction of extracellular protein localization in L. plantarum
WCFS1 is found in the Secretome database [18
. Another prediction tool is the newly developed Locate P [54
] that combines existing predictors and produces genome-wide predictions for the subcellular locations of bacterial proteins in a fully automated manner. Predictions based on both methods/databases for the 78 proteins relevant for this study are included in additional file 1
and show several differences. For example, all but one (Lp_1524) of the selected SPs are predicted to be cleaved by SPaseI in the Secretome database, whereas Locate P predicts such cleavage only for 63 of the SPs. The present set with experimental data may be used to evaluate prediction quality and, hopefully, to improve prediction methods. Our data show that the SPs of several proteins predicted to be N-terminally anchored by LocateP lead to efficient secretion of NucA, meaning that they are cleaved by SPaseI as predicted by SignalP and according to the prediction in the Secretome database. Likewise, several proteins predicted to be multi-membrane proteins according to Locate P contain SPs that are quite efficient for NucA secretion.
To test the general performance of the SPs we replaced NucA with AmyA in selected constructs. When produced at levels applied in this study, AmyA seems to be difficult to handle for L. plantarum
. Secretion efficiencies were below, often far below, 100% for all constructs. Table shows that the AmyA constructs lead to highly variable overall production levels, creating a complicating variable that was less prominent in the studies with NucA. Previous studies have shown that overexpressed amylase can be difficult to handle for B. subtilis
and induce stress reactions [55
]. Table also shows that high production levels of AmyA correlate with low secretion efficiencies, suggesting that the translocation machinery is overloaded. In addition to slow or blocked translocation, secretion stress may cause intracellular or extracellular proteolytic degradation [41
]. Proteolytic degradation was not analyzed because of the lack of a suitable antibody for AmyA. The stress caused by AmyA expression is illustrated by cells harbouring the pLp_2940sAmy construct that leads to high levels of AmyA production. These cells showed impaired growth (data not shown), cell lysis and a change in morphology (Figure ). Lp_2940 did not perform very well for NucA (rank 41) and it does not have the properties that are typical for SPs that work well with NucA (see additional file 1
). It is possible that the combination of a high production level with an unfavourable SP stressed the cells to the extent that lysis occurred. All in all, our observations with AmyA indicate that this protein is not a suitable reporter to search for characteristics in SP-sequences that correlate to secretion capacity.