Up to now, various ATs have been used for the cell surface display of in total over thirty passenger domains . In the past, the versatility and universality of AT based cell surface display has been emphasized [6, 28]. Even more recently it was pointed out that this technology could be used for the extracellular accumulation of proteins with a diversity in size, structure and function . With this in mind, we set out to test the broad applicability of AT based cell surface display using the recently characterized autologous EstA AT  in P. stutzeri A15 as a test-case.
In the first experiments, an expression module based on the EstA β-barrel and α-helix was constructed. Whole cell proteinase K assays and esterase activity assays validated the expression module by showing the successful cell surface display of the original EstA passenger domain. Also cell surface display of a functional beta-lactamase as the passenger domain could be attained. So far, our results are in line with previous papers describing AT based cell surface display of esterases/lipases [18, 19, 29, 30] and a beta-lactamase  in E. coli. However, the same expression module failed to successfully display the majority of the fusion proteins with one of three different lactonases, AiiAsoil, AiiB or AttM, as passenger domain. Also when one of the fluorescent proteins mCherry, eGFP or yEVenus was used as passenger domain, no successful cell surface display could be realized. Most reports focusing on passenger domain structure have looked at formation of disulfide bonds within the passenger domain in the periplasm and the compatibility with surface display. Recent research seems to have reached a consensus on this point suggesting that only disulfide bonds between closely spaced cysteine residues in the passenger domain can be tolerated to achieve successful surface display [31, 32]. The crystal structures of the fluorescent passenger domains reveal a β-barrel fold without disulfide bonds [33–35]. The same is true for the crystal structure of the AiiA lactonase of Bacillus thuringiensis and on the basis of homology also for the lactonases used in this study. The only passenger domain that could successfully be surface displayed was a TEM β-lactamase. Interestingly, the crystal structure of the TEM-1 β-lactamase , contains one disulfide bond. The overall structure of this protein reveals a more globular structure which resembles most the structure of the original GDSL passenger domain. However, the globular structure is most likely not the main reason for surface display since β-lactamases have already been used as passenger domain for surface display using ATs with a completely different passenger domain structure [25, 38]. Whereas the surface display of β-lactamases is in line with our findings, the display of the mCherry protein at the cell surface using the Pet AT  and the mRFP1 protein using the YfaL AT  is in contrast with our findings. In both studies [7, 8] cell surface display was achieved in E. coli.
A minimal translocation unit for AT based cell surface display, consisting of the β-barrel domain and the α-helical linker domain, has been described for several ATs [39–43] including the EstA AT of P. putida. More recently however, this original view was debated by the finding that the cell surface expression of heterologous proteins was increased when the full-length IcsA AT was used rather than only the β-domain . This was supported by work on the Hbp AT where an intact β-stem was needed for efficient extracellular expression of the ESAT6 antigen .
Because of these reports and other publications in which a full-length EstA AT of P. aeruginosa was applied [19, 20] we designed full length EstA expression modules based on the EstA AT of P. stutzeri A15. Apart from the β-barrel domain and the α-helical linker domain also the original esterase passenger domain was present in these modules, once with and once without the catalytic serine residue. The presence of the original passenger domain did not substantially resolve the problems encountered previously with the β-barrel domain based construct.
Also, in various cases smaller sized bands could be visualized. They could be indicative of translation intermediates or breakdown products. In E. coli it has been shown that surface exposed proteins can be proteolytically cleaved by the presence of proteases. Therefore, E. coli strains lacking the outer membrane protease (OmpT), like UT5600, are often preferred as host .
A first possible explanation for the differences between our results and reports describing successful surface display is the use of a different host organism. However, an often addressed issue for autodisplay in E. coli, the use of an for the host organism autologous AT, is nonetheless fulfilled . Another reason could be the difference in viewpoint. Instead of focusing on the small fraction of passenger proteins that are proteinase K accessible and heat modifiable we concentrated on the major fraction of the proteins that were not secreted. The benchmark for successful surface display in our case was set by the data available for the surface display of the original EstA passenger domain using the β-barrel domain based construct. Others might have applied the term successful surface display in a less stringent way. Also, literature presumably has a bias towards the successful use of ATs for surface display. Despite the fact that the number of reports criticizing AT based surface display are limited, publications have reported the negative effect on membrane permeability  and on cell survival .
To end, it is important to mention that some reports are available which have addressed the nature of the passenger domain. The mechanism of translocation and secretion of passenger domains in the majority of these studies was only addressed by looking to the effects of introducing cysteine residues and the resulting disulfide bond formation. Some of the initial studies revealed the incompatibility of periplasmic disulfide bond formation with translocation of passenger domains [46–48]. These early reports were later contested by work describing the successful translocation of folded domains containing disulfide bonds [26, 49, 50]. The most recent studies on this topic however concluded that disulfide bond formation can only be tolerated if the cysteine residues are closely spaced [31, 32]. Using a different approach it was suggested that the folding of a MalE passenger domain fused to the IgA AT interfered with its translocation across the outer membrane . Also the introduction of the rigid and bulky calmodulin in the backbone of the Hbp AT has been shown to negatively affect secretion . Finally it was proposed that secretion efficiency is dependent on the folding properties of the passenger domain itself . Based on our results and the discussion above we suggest that more basic molecular knowledge of naturally occuring autotransporters is needed before such systems can be used for heterologous expression.