A pathway sensor for genome-wide screens of intracellular proteolytic cleavage

A new system based on non-conventional secretion of the luciferase from Gaussia princeps (GLUC) can be used to detect intracellular proteolysis in vivo.


Background
Advances in automation and the availability of genomic sequence information have led to the development of sophisticated cell-based assays for high-throughput screening of functional phenotypes [1]. Most cell-based assays rely on fluorescent or luminescent reporters such as green fluorescent protein (GFP), secreted alkaline phosphatase (SEAP) or Photinus luciferase. Secreted luciferases offer many advantages over cellular reporter enzymes as they can be nondestructively harvested from cellular supernatants over time. Several secreted luciferases have been reported, from the marine copepods Gaussia princeps [2], and Metridia longa [3], the ostracod Vargula hilgendorfii [4], the decapod shrimp Oplophorus gracilirostris [5] and the ostracod crustacean Cypridina noctiluca [6]. In addition, intracellular luciferases, such as from the sea pansy Renilla reniformis, can be engineered to be secreted and stable in the extra-cellular milieu [7].
A cDNA encoding G. princeps luciferase (GLUC) activity has recently been isolated and found to direct the synthesis of a 19.9 kDa protein that has utility as a bioluminescent reporter [2]. GLUC can be used to monitor in vivo processes and can be easily harvested from biological fluids such as blood or urine [8]. Assays based on GLUC activity have been used to study, among other topics, processing through the secretory pathway [9], the strength of signal peptides [10], endoplasmic reticulum (ER) stress [11], DNA hybridization [12], and protein-protein interaction using complementary fragments derived from the enzyme [13]. By deletion of the signal peptide, a GLUC mutant has been engineered for monitoring in vivo gene expression; very low bioluminescence was detected in cell culture superanatants upon expression of this construct [2]. However, overall bioluminescence of this construct was greatly reduced compared to wild-type GLUC [2]. It has been noted that GLUC is secreted when fused to the ER retention signal KDEL, which has been attributed to changes in the protein conformation or processing in the ER and Golgi [2].
We have generated a GLUC variant that is secreted in the absence of a signal peptide. We present here a cell-based assay for the detection of general protease activity based on inducible luciferase secretion. GLUC can be anchored in cells by fusion to β-actin. Insertion of protease cleavage sites in a linker between β-actin and GLUC allows monitoring the cleavage of short peptides, as well as cleavage of native fulllength proteins of any sequence inserted. We present GLUCbased reporter systems for monitoring apoptosis and autophagy and describe applications of this reporter in genome-wide screening approaches.

Results
In the course of attempts to develop a GLUC reporter that is retained in cells and released after addition of a specific stimulus, we deleted the signal peptide to generate dNGLUC. Surprisingly, this deletion did not abolish the accumulation of GLUC activity in the supernatant (SN) of transiently transfected 293ET cells. Although the proportion of dNGLUC in SN was reduced to 30.5% compared to 96.7% of total GLUC activity, the overall activity was still very high (Table 1). By contrast, when dNGLUC was fused to the carboxyl terminus of β-actin, less than 1.5% of GLUC activity was detected in SN (Table 1), and the relative light unit values observed were close to background (not shown).
Most extracellular proteins are secreted from cells by transport through a secretory pathway that requires translocation of the nascent polypeptide from the ribosome to the lumen of the ER, followed by vesicular transport through the Golgi and subsequent compartments [14]. Initiation of secretion by this pathway is mediated by a hydrophobic amino-terminal signal sequence [14]. Some proteins, however, lack an amino-terminal signal peptide and are secreted by a mechanism that is insensitive to treatment with inhibitors of ER/Golgi trafficking such as Brefeldin A [15,16]. To further characterize the mechanism of secretion of dNGLUC, we treated 293ET cells expressing dNGLUC with drugs known to interfere with secretory pathways. Cells expressing dNGLUC were exposed to 7 μM Monensin, 10 μg/ml Brefeldin A or 5 μg/ml MG132 and the activity accumulating over 4 h at 37°C was determined ( Figure 1a). For comparison, we also measured the activity of SEAP, which is secreted by a classical signal peptide ( Figure 1b). We found that treatment with Monensin and Brefeldin A reduced secretion of both dNGLUC (by 75% and 82%, respectively; Figure 1a) and SEAP (by 88% and 90%, respectively; Figure 1b), while MG132, an inhibitor of the proteasome, reduced secretion by 32%. Since Monensin and Brefeldin A interfere with transport pathways originating from the Golgi apparatus, we propose that dNGLUC is secreted by a mechanism involving the secretory pathway. To confirm this hypothesis, we performed co-localization studies of dNGLUC and the Golgi marker protein Golgin-67. GFPtagged dNGLUC was localized in the cytoplasm. In addition, we observed co-localization of GFP-dNGLUC and DsRedtagged Golgin-67 at a perinuclear site (Figure 1c). Co-localization with Golgi-resident proteins is consistent with the view that secretion of dNGLUC requires ER/Golgi trafficking. As far as we are aware, the dNGLUC secretory pathway is presently the first example of Brefeldin A-sensitive non-conventional secretion.
The intracellular retention of Actin-dNGLUC opened the possibility of developing an assay for release of dNGLUC by protease-mediated cleavage. We inserted two tandem repeats of the caspase 9 protease consensus site DEVDG and a Flag tag between β-actin and dNGLUC to generate two candidate caspase sensors, Actin-DEVDG2-flagdNGLUC (DEVDG2F) and Actin-flagDEVDG2-dNGLUC (FDEVDG2) (Figure 2a). In cotransfection experiments in which variable amounts of caspase 8 or caspase 9 expression plasmid were co-delivered with a fixed amount of dNGLUC expression plasmid, dNGLUC activity was released into SN in a dose-dependent manner with increasing amounts of expression plasmid (Figure 2b). Luciferase activity in the SN was 4.8-fold, 9.3-fold and 32.8-fold higher in the presence of 10 ng, 100 ng and 1,000 ng of caspase 8, respectively, compared to control cells transfected with vector alone. Expression of caspase 9 resulted in 2.8-fold and 12.6-fold increases of released luciferase activity from FDEVDG in the presence of 100 ng and 1,000 ng of caspase 9, respectively (Figure 2b). We did not observe release of dNGLUC from cells expressing Actin-dNGLUC in the presence of caspase 8 or 9, confirming the specificity of cleavage for the DEVDG motif. In order to visualize the cleavage products, cell lysates from 293ET cells expressing FDEVDG2 were resolved by SDS-PAGE and immuno-blotted with anti-Flag antibodies. A band was detected at a size of 62 kDa, which corresponds to the calculated molecular size (61.5 kDa) of FDEVDG2 (Figure 2c, upper panel). In the presence of caspase 9, full-length FDEVDG2 was not detected, but a band at 46 kDa appeared, consistent with the removal of dNGLUC. Probing of the same blot with an antibody raised against GLUC identified a major band at 62 kDa that disappeared upon expression of caspase 9 (Figure 2c, lower panel).
In order to identify novel proteins that might induce caspasemediated cleavage, we performed a functional screen using the Origene Trueclone™ expression vector collection. We cotransfected 96-well plates with single cDNA expression vectors and DEVDG2F and measured luciferase activity in SN and cellular lysates in triplicate plates. To normalize for cellu-   (c) lar expression and cell numbers, we determined the ratio of luciferase activity in SN over cellular lysates from the same 96-well plate. Three wells on each plate were transfected with DEVDG2F only to determine the level of background secretion. In Table 2, we summarize genes that showed more than a three-fold increase in GLUC activity released from cells expressing DEVDG2F compared to cells transfected with reporter only. The candidates found include known inducers of apoptosis, such as BAK, FADD, BAD and caspase 8, in partial validation of the approach to identify regulators of caspase activation. In addition, we identified the novel genes for ASPH, PIR121, PERP and TBC1D10A, which induced 14.2-, 12.1-, 10.4-and 5.5-fold increases in GLUC activity in SN from DEVDG2F cells, respectively ( Table 2). TBC1D10A is a member of the Tre/Bub2/Cdc16 (TBC) family that exhibits GTPase activating protein (GAP) activity and, thus, is an interesting candidate gene in the context of apoptotic signaling. Since DEVDG2F harbors additional aspartate residues within the Flag peptide sequence that might serve as cleavage target sites, we also generated a construct with three DEVDGrepeats without a Flag tag, Actin-DEVDG3-dNGLUC (DEVDG3). In addition, we generated a variant reporter in which the DEVDG-motif was replaced with a DEVAG motif that is not a substrate for caspases. TBC1D10A was co-transfected with Actin-dNGLUC, DEVDG3 or DEVAG2F and the release of GLUC into SN was measured. Caspase 9 induced a 4.1-fold and TBC1D10A a 4.3-fold increase in extra-cellular GLUC activity compared to GFP, but did not release dNGLUC from Actin-dNGLUC or DEVAG2F (Figure 3a). These results are consistent with the view that the cleavage promoted by caspase 9 and TBC1D10A is specific to the caspase cleavage site introduced in the reporter substrate.
In the course of the screen we also identified genes that induce release of dNGLUC from Actin-dNGLUC. Co-expression of the serine peptidase HTRA4 with ActindNGLUC or DEVDG2F yielded a 201.5-fold increase of GLUC activity in SN from cells expressing Actin-dNGLUC and a 110.8-fold increase from DEVDG2F, indicating that the caspase cleavage site is not required for liberation of luciferase activity ( Figure  3b). Similarly, another family member, HTRA3, induced a 177.1-fold and 89.1-fold increase in GLUC activity in SN for Actin-dNGLUC and DEVDG2F, respectively. Caspases 8 and 9 induced a 9.5-fold and 15.0-fold increase of GLUC activity for DEVDG2F, but had no effect on Actin-dNGLUC. In accordance with previous reports that have identified HTRA2-mediated cleavage of β-actin by mass spectroscopy [17], these data support the view that HTRA3 and 4 cleave within the β-actin sequence. We therefore conclude that our assay also allows the detection of full-length protein cleavage under physiological conditions.
To further explore the suitability of GLUC fusions for detection of native protein cleavage, we inserted the open reading frame of hMAP1LC3 (LC3), a marker of autophagy, between β-actin and dNGLUC. Autophagy is a tightly regulated cellu-lar response to starvation that results in degradation of subcellular organelles. LC3 is cleaved during autophagy at the carboxyl terminus by the cellular protease hATG4B; the cleaved form is found associated with autophagosomes [18]. Amino acid starvation or treatment with rapamycin is sufficient to induce autophagy and LC3 cleavage. Upon treatment with 200 nM rapamycin, we detected a 7.2-fold increase in GLUC activity in SN in cells expressing Actin-LC3-dNGLUC but not Actin-dNGLUC (Figure 4a). In addition, co-expression of the cellular protease ATG4B, but not ATG4A or GFP, resulted in a 26.6-fold increase in extra-cellular luciferase activity (Figure 4b). Activity of ATG4B was confirmed by immunoblotting of transfected GFP-LC3. In the presence of ATG4B, the cleaved form of GFP-LC3 was visualized at 43 kDa and significantly increased in intensity compared to the full-length 45 kDa form, which was not evident in cells transfected with ATG4A ( Figure 4c). In order to confirm cleavage of Actin-LC3-dNGLUC by ATG4B, we resolved whole cell lysates by SDS-PAGE and immuno-blotting using an antibody raised against dNGLUC as a probe. In the absence of ATG4B, we detected the full-length construct Actin-LC3-dNGLUC at 83 kDa and a smaller band at 23 kDa (Figure 4d). In cells cotransfected with an ATG4B expression plasmid, the full-length product at 83 kDa disappeared; whereas in cells cotransfected with a GFP expression plasmid, the 83 kDa product was readily apparent (Figure 4d). To visualize the secreted cleavage product, we treated cells with Brefeldin A for 6 hours prior to cell lysis. In the setting of ATG4B coexpression in Brefeldin A-treated cells, the band corresponding to Actin-LC3-dNGLUC at 83 kDa is not seen and the 23 kDa band corresponding to the dNGLUC cleavage product has increased intensity, consistent with the view that Actin-LC3-dNGLUC is cleaved by ATG4B.
To study the role of endogenous proteins in the autophagy pathway, we used small hairpin RNA (shRNA)-mediated knockdown of candidate signaling molecules. We first identified a shRNA sequence targeting human ATG4B (sh4B) that significantly reduced the expression of a GFP-ATG4B construct by 80% of detected GFP mean fluorescence intensity in 293ET cells (Figure 5a). The dNGLUC activity in the SN of 293ET cells cotransfected with Actin-LC3-dNGLUC and sh4B was reduced by 40% compared to cells transfected with Actin-LC3-dNGLUC alone, indicating that endogenous ATG4B contributes to the observed luciferase release (Figure 5b). Next, we tested the effect of shRNA-mediated knockdown of AKT1, an upstream kinase that activates the key inhibitor of autophagy, mTOR [19]. Knockdown of AKT1 resulted in an increase of dNGLUC release from cells expressing Actin-LC3-dNGLUC compared to vector control in transient transfection as well as in a stable 293ET cell line expressing Actin-LC3-dNGLUC transfected with shAKT1 (Figure 5c,d). The magnitude of the effect of shRNA knockdown was similar to that observed following inhibition of mTOR with rapamycin ( Figure 4a). Ketteler et al. R64.5 Genome Biology 2008, 9:R64 Design of a GLUC-based caspase sensor  Casp8: GLUC activity in SN GLUC activity in SN

Non-conventional secretion of Gaussia luciferase
Protein secretion in most cells is mediated by signal sequences that target the nascent polypeptide chain of the elongating translation product to a secretory pore in the ER [14]. Within the ER and the subsequent compartments of the Golgi apparatus, folding and post-translational modifications take place, and the mature, modified polypeptide is released into the extracellular space. A number of secreted proteins that do not utilize the ER membrane translocation machinery, such as fibroblast growth factor, coagulation factor XIII and interleukin-1β are secreted by a non-conventional secretory pathway [16]. Different mechanisms for non-conventional secretion have been proposed [16], including lysosomal secretion for interleukin-1β [20], a plasma resident transporter for fibroblast growth factor 2 [21] and cell injury for coagulation factor XIII [22]. Two prevalent features of nonconventional secretion are the absence of a signal peptide and insensitivity to Brefeldin A [15]. The precise mechanism of secretion is still poorly understood and the underlying molecular signals remain to be elucidated.
The luciferase release assay reported here relies on a nonconventional secretion of dNGLUC that is inhibited by Monensin and Brefeldin A. Monensin inhibits acidification of terminal compartments thought to lie immediately prior to extracellular release, whereas Brefeldin A inhibits ER-to-Golgi transport. The amino-terminal amino acid sequence of the deleted luciferase studied here does not fulfill the accepted criteria for a signal peptide [23]. Because secretion is sensitive to treatment with Brefeldin A, we conclude that a previously unarticulated mechanism is responsible for the translocation of the polypeptide into the ER and/or Golgi. The molecular basis of this translocation, and subsequent passage through the terminal secretory apparatus, is presently under investigation. A Golgi-resident protein, GRASP, has been identified that is required for a non-conventional secretory pathway in Dictyostelium discoideum [24] and Drosophila melanogaster [25], and that is a candidate for mediating Brefeldin A-sensitive secretion of dNGLUC. Identification of GLUC mutants that are retained inside cells may help to identify the mechanism of non-conventional secretion.

A novel protease assay
The present assay system has several advantages over existing systems for measuring protease activity. Currently, protease cleavage sites can be inferred from comparison of primary sequences. The physiological relevance of predicted cleavage sites in a particular protein then can be assessed by further experimentation. Target motifs can be identified by analysis of protease action on peptide libraries, such as phage  [26]. The identification of protein cleavage in the context of live cells can be achieved by mass spectroscopic analysis of cleavage products [27], but requires a complex experimental setup and is not amenable to high-throughput approaches. Other cell-based protease assays rely on generation of a fluorogenic substrate upon cleavage, but these assays are not genetically encoded, thus limiting their applicability in vivo. Some in vivo protease assays have been developed that exploit the properties of fluorescence resonance energy transfer (FRET) [28]; in these, the protease-mediated separation of a donor and acceptor fluorophore results in changes of the ratio of fluorescence intensities at different wavelengths [29]. A major advantage of FRET-based methods is their ability to provide information about the sub-cellu-A GLUC-based sensor to monitor autophagy lar localization of protease activity. However, FRET-based assays are frequently not highly sensitive, require a carefully characterized cohort of control samples in a single experiment and typically demand advanced instrumentation.
To date there has been little use of FRET in genomic screening applications. In contrast, the assay system described here non-invasively measures protein cleavage over time in the context of the complex physiology of intact living cells, is compatible with high-throughput screening methodologies, and can be designed to monitor protease function with high specificity. The luciferase release system can detect cleavage of short peptides as well as cleavage of full-length proteins. Evaluation of actin-specific or non-specific screening hits can be identified and eliminated by secondary screening with a luciferase fusion bearing a mutated version of the protease cleavage motif to be investigated. It has previously been established that GLUC secretion is proportional to cell number [2]. Differences in cell number as well as variation in transcription and translation rate can be assessed by determining ratios of extracellular luciferase to cellular activity. We recommend that the optimal harvest and collection times be assessed in pilot studies. For instance, extensive cell death results in reduced reporter production, and in the case of the apoptosis sensor used here, best results were seen when the cultures were assayed 24-32 hours after initiation of apoptosis. The assay is highly reproducible following transient transfection, and can also be used in cell lines stably transfected with the reporter if desired. Both transfected and endogenous protease activities are easily detected with this system. The transfer of a reporter enzyme across cell membranes constitutes an unexpected assay principle that adds a flexible, broadly applicable approach to current cell-based multi-color and multienzyme assays.

Applications of the protease sensor to study β-actin cleavage, apoptosis and autophagy
Cleavage of Actin-dNGLUC by HtrA3 and 4 suggests that members of the HtrA family of heat shock proteases, which are known to have significant functions in protein folding and apoptosis, may have the general property of cleaving actin in a manner that eliminates its ability to form insoluble fibers.
Recently, a proteomic approach based on mass spectroscopic identification of cleavage products was undertaken to identify HTRA2 substrates [17]. Major cleavage products included βactin and tubulin alpha/beta and it was suggested that HTRA2 regulates apoptosis at the level of the cytoskeleton [17]. Although β-actin has been reported as a substrate for a number of caspases, including caspase 3 [30], we have not observed release of dNGLUC from Actin-dNGLUC in response to caspase 3, 8 or 9, suggesting either that cleavage did not occur, or that it did not impair the ability of β-actin to anchor dNGLUC in the cell. In contrast to observations on cell-free extracts, cleavage of β-actin by caspases has not been detected in intact cells [31].
In a functional screen using the caspase sensor, we have identified the TBC family member TBC1D10A as an inducer of DEVDG-mediated cleavage. The TBC family of proteins exhibit GAP activity towards small GTPases of the Rab family [32]. TBC1D10A has recently been identified as a GAP for Rab27A, suggesting a role in melanocyte transport and secretion [33]. In addition, TBC1D10A binds to a complex of EBP50 with Ezrin and ARF6-GTP to regulate microvillus structure [34]. Based on these data, TBC1D10A has been proposed as a regulator of protein trafficking in cells.
Recently, a genome-wide screen for cell death effectors identified another family member, TBC1D10C, as an inducer of apoptosis [35]. In agreement with this observation, our findings confirm a role for TBC1D10A as an effector of protein cleavage.
Autophagy is an essential cellular process for the degradation of proteins and organelles that has been associated with neurogenerative diseases, cancer and infection [36]. Although autophagy is currently widely investigated, the systematic identification of molecular events in autophagy has been hampered by the lack of suitable assays. Current assays to study autophagy measure the accumulation of autophagic vacuoles by staining with fluorescent dyes such as monodansylcadaverine [37], or the sequestration of radioactive sugars or enzymes such as lactate dehydrogenase [38]. However, these assays are difficult to quantify due to the presence of background levels of autophagic vacuoles or non-specific staining. Recently, immuno-blotting of hMAP1LC3 cleavage products, and GFPhMAP1LC3 translocation to autophagosomes [18] have been proposed as specific assays for autophagy. However, since the cleavage product of hMAP1LC3 is itself degraded by autophagy, interpretation of these assays requires additional controls [39]. The assay presented here is a simple, easily implemented, quantitative assay that measures induction of autophagy without destruction of the cell being studied. As such, we anticipate it will be useful to many investigators in their studies of this enigmatic process.

Conclusion
It has been estimated that the human genome contains more than 500 proteases [40], most of which are poorly characterized. The luciferase secretion assay described here can be used to identify protease regulatory pathways as well as protease targets. The actions of nongenomic proteases, such as the HIV or HCV proteases or Anthrax lethal factor can be easily assessed by inserting the appropriate peptide target sequence in an actin-peptide-dNGLUC reporter construct.
The finding that Gaussia luciferase is capable of exiting the cell by a non-conventional secretion pathway is unusual in itself, and provides a tool to explore aspects of non-conventional secretion. Regulated non-conventional secretion of an enzymatic reporter has not been previously demonstrated to our knowledge, and affords several advantages over existing methods for analysis of intracellular cleavage events.
Of particular interest is the process of autophagy. Autophagy is a highly regulated process that appears to provide additional energy to cells under conditions of starvation. Autophagy has been suggested to play roles in the prevention and progression of cancers [41]. The precise role that autophagy plays in these settings is not well understood, and high interest is currently directed toward understanding the contribution of autophagy to tumor growth. Large-scale screening approaches to identify regulators of autophagy to date have not been reported, possibly due to the absence of suitable screening assays. Analysis of autophagy is presently based on qualitative ultramorphological analyses, immunoblotting, or translocation of GFPLC3. Such assays can be nonquantitative, laborious and subject to multiple confounding factors [39]. The analysis system described here facilitates insight into the regulation of autophagy and enables large scale shRNA knockdown and expression screening approaches.

Transfection
293ET cells were cultured in DMEM (supplemented with 10% calf serum plus iron, 0.25 μg/ml gentamycin and 50 μM βmercaptoethanol) and transfected using calcium phosphate precipitation as described elsewhere [43]. The Origene True-clone™ cDNA library consisting of approximately 12.000 human expression cDNAs arrayed in 96-well plates were transfected by TransFectin (BioRad, Hercules, CA, USA) along with a GFP expression construct in 293ET cells and screened for morphological changes by fluorescence microscopy (RK and BS, unpublished). Clones displaying signs of cell death were selected for transfection with Actin-DEVDG2F-dNGLUC. Supernatants were harvested after 24-32 h for luciferase analysis. Inhibitors of non-conventional secretion (7 μM Monensin, 10 μg/ml Brefeldin A, 5 μg/ml MG132; all from Sigma) were added 24 h after transfection and medium was collected over a 4 h time period.

Generation of stable 293ET cell line
Actin-dN and Actin-LC3-dN were subcloned into pMOWS [43] and co-transfected in 293ET cells with expression plasmids for VSV-G and retroviral gag-pol. The medium was changed after 24 h and virus supernatant was harvested and filtered through 0.45 μm filters 48 h after transfection. Untransfected 293ET cells were incubated with retroviral supernatant supplemented with 8 μg/ml polybrene; 48 h later, transduced 293ET cells were selected with puromycin at a concentration of 0.3 μg/ml.