To achieve a fast regulated cellular response resulting in the release of a protein, we decided to mimic the release of insulin from beta cells where the protein of interest is pre-formed and present in the cell in secretory granules. In contrast to the specialized storage and release mechanism of insulin from beta cells we wanted to develop a more general and modular solution by making use of components already existing in different types of cells. Additionally, there should be minimal leakage from the protein depot in the uninduced state and after induction secretion from the cell should be fast.

Many proteins that reside on the membrane or in the lumen of the ER contain short peptide signals. Proteins present in the lumen of the ER contain a KDEL C-terminal sequence (Lys-Asp-Glu-Leu) while type I transmembrane (TM) proteins contain a dilysine (KKXX) motif on their C-terminus (cytosolic side) Munro1987, Jackson1990, Stornaiuolo2003 . . The mechanism that allows these proteins to stay in the ER is more retrieval than retention. However, we decided to use the term retention for description of this process. ER luminal proteins interact with the KDEL receptor, a transmembrane ER resident protein. The cytosolic part of the KDEL receptor interacts with coat proteins I (COP I) which coat vesicles and are responsible for transporting proteins from the cis end of the Golgi apparatus (cis-GA) back to the ER. The KKXX motif present on type I TM proteins can directly interact with the COP I for retrieval Stornaiuolo2003, Letourneur1994 .

Our idea was that if we proteolytically remove the retention signal, the protein of interest would no longer be retrieved back to the ER and could be secreted from the cell. To achieve this we designed two types of secretory reporters, one type based on the luminal retention using KDEL sequence and the other based on the transmembrane retention with a KKMP sequence. In each case, the retention sequence was preceded by a TEVp cleavage site to allow for inducible secretion, which could be replaced by any other peptide target of our orthogonal protease set.

 Secretion from the ER lumen

To achieve and detect the inducible secretion from the ER lumen, we created two reporter constructs with a cleavable KDEL sequence targeted to the ER lumen: SEAP^KDEL and TagRFP^KDEL. Those proteins contained a protease target motif between the reporter domain and the KDEL domain, aimed to enable protein secretion after the proteolytic cleavage. We used a TEVp variant (erTEVp) for all of our experiments with luminal retention.

When the TagRFP^KDEL reporter ( 1 A) was expressed in the ER without an active erTEVp we confirmed its localization in the ER with confocal microscopy ( 1 B). Additionally, we could not detect any TagRFP in the cell medium with Western blotting. When erTEVp was present and active in the ER, the KDEL sequence was removed from the reporter and the protein was secreted from the cell, which we detected with Western blot ( 1 C), demonstrating that proteolytic activity in the ER can regulate protein secretion.

Using SEAP^KDEL we were able to confirm that the reporter is not present in the cell medium without coexpression of erTEVp. When erTEVp was cotransfected with the reporter, we detected a large increase in enzymatic activity in the medium ( 2 ).

 Secretion from the ER membrane

The second approach to regulate protein secretion from the ER by protease was to used KKMP ER retention peptide linked to the transmembrane protein with a protease target motif on the cytoplasmic side, N-terminal to the KKMP peptide. A transmembrane (TM) domain from the B-cell receptor (Bba_K157010) was used for the integration of target proteins in the ER membrane. Similar as described above, two reporter constructs with SEAP and TagRFP (SEAP:TM^KKMP and TagRFP:TM^KKMP) were designed and the constructs also contained a signal sequence at their N-terminus and a proteolytically cleavable ER retention signal at their C-terminus. In case of the transmembrane targeted reporters we used the KKMP retention signal preceded by 3 copies of the TEVp cleavage site on the cytosolic side of the membrane.

Additionally, either one or four furin cleavage sites were inserted between the protein of interest on the luminal side of the ER, which enable cleavage of the reporter protein from the membrane, but this could occur only after the KKMP had been removed and the protein could enter the trans-GA. Furin is a native cellular endoprotease that is active only in the trans-GA.Henrich2003 This allowed us to design our constructs so that they are cleaved off of the membrane without any modified scar sequences attached to them.

Localization of the TagRFP:TM^KKMP reporter was confirmed by the confocal microscopy. We used a control reporter without the KKMP retention signal (TagRFP:TM) which we detected both on the ER and the plasma membrane ( 3 A). In case of the present KKMP retention signal, the reporter was detected only on the ER ( 3 B). When TagRFP:TM^KKMP was coexpressed with TEVp, localization of the reporter was similar to the localization of the positive control (TagRFP:TM) on the plasma membrane and the ER ( 3 C).

A band with a slightly larger apparent size than the expected size of TagRFP (28 kDa) was detected by western blotting in cells transfected with TagRFP:TM. We showed that the unexpected difference in size was due to glycosylation, as we detected the protein at the expected size after deglycosylation of the medium sample with N-glycosidase F. We were unable to detect a corresponding band in the medium of cells transfected with TagRFP:TM^KKMP in the absence of the protease.

Together, these results confirm that localization and secretion of the protein reporter with the transmembrane domain depends on the presence and proteolysis of the KKMP retention signal and that proteolysis can be used to induce secretion of already synthesized protein.

Finally, we cotransfected cells with SEAP:TM^KKMP and rapamycin-inducible split TEVp. We detected increased levels of the SEAP enzymatic activity in the medium of cells stimulated with rapamycin, which was dose dependent with respect to the amount of the transfected reporter-coding plasmid ( 4 ). These results confirm that secretion of a target protein can be made inducible by an externally supplied signal, processed through our split protease system.