Secretion Stimulates Intramembrane Proteolysis of a Secretory Granule Membrane Enzyme*

Regulated intramembrane proteolysis, a highly conserved process employed by diverse regulatory pathways, can release soluble fragments that directly or indirectly modulate gene expression. In this study we used pharmacological tools to identify peptidylglycine α-amidating monooxygenase (PAM), a type I secretory granule membrane protein, as a γ-secretase substrate. PAM, an essential enzyme, catalyzes the final step in the synthesis of the majority of neuropeptides that control metabolic homeostasis. Mass spectroscopy was most consistent with the presence of multiple closely spaced NH2 termini, suggesting that cleavage occurred near the middle of the PAM transmembrane domain. The luminal domains of PAM must undergo a series of prohormone convertase or α-secretase-mediated cleavages before the remaining transmembrane domain/cytosolic domain fragment can undergo a γ-secretase-like cleavage. Cleavage by γ-secretase generates a soluble fragment of the cytosolic domain (sf-CD) that is known to localize to the nucleus. Although PAM sf-CD is unstable in AtT-20 corticotroph tumor cells, it is readily detected in primary rat anterior pituitary cells. PAM isoform expression, which is tissue-specific and developmentally regulated, affects the efficiency with which sf-CD is produced. sf-CD levels are also modulated by the phosphorylation status of the cytosolic domain and by the ability of the cytosolic domain to interact with cytosolic proteins. sf-CD is produced by primary rat anterior pituitary cells in response to secretogogue, suggesting that sf-CD acts as a signaling molecule relaying information about secretion from the secretory granule to the nucleus.

initiates many of the signaling cascades involved in these pathways. The i-CLiP family includes metalloproteases (S2P), rhomboid serine proteases, signal peptide peptidase aspartyl proteases, and ␥-secretase aspartyl protease complexes (1,2). For example, signaling pathways that regulate lipid homeostasis modulate cleavage of SREBP by S2P, releasing its transcription factor domain. The unfolded protein response involves S2P-mediated cleavage of ATF6 and nuclear translocation of its N-terminal transcription factor region (3). Many aspects of development involve cleavage of Notch by ␥-secretase, generating a cytosolic domain fragment that translocates to the nucleus, where it participates in the regulation of gene transcription (4). The ␥-secretase-mediated cleavage of amyloid precursor protein (APP) generates amyloidogenic fragments along with a cytosolic fragment whose role is under investigation (5).
Secretory granules in the regulated exocytic pathway play an essential role in maintaining homeostasis. Their soluble content includes chromogranins, prohormone convertases, neuropeptides, small molecules like ATP, and various metals (6 -8). Little is known about how secretory granule function is integrated into overall cell metabolism. The majority of neuropeptides require an essential post-translational modification, ␣-amidation, to be active (9). This modification, one of the final steps in the conversion of preprohormones into active product peptides, requires peptidylglycine ␣-amidating monooxygenase (PAM), a type I membrane protein. PAM is dependent on ascorbic acid, copper, and molecular oxygen, and mice lacking PAM do not survive past mid-gestation (10). After exocytosis, membrane PAM is subject to endocytosis followed by regulated re-entry into granules or by degradation (11,12).
The identification of cytosolic proteins that interact with PAM and examination of the effects of overexpressing PAM suggested that it might serve as an indicator of secretory granule status (13,14). The cytosolic domain of PAM interacts with Rho GDP/ GTP exchange factors (Kalirin and Trio), a Ser/Thr protein kinase (Uhmk1, also known as P-CIP2 or KIS), and a Ras-association domain family (RASSF) member (Rassf9, also known as P-CIP1). The intrinsically unstructured, multiply phosphorylated cytosolic domain of PAM is essential for trafficking but not for catalysis. We recently identified a soluble fragment of the PAM cytosolic domain (sf-CD) in the nuclei of anterior pituitary endocrine cells and AtT-20 corticotroph tumor cells (15). In AtT-20 cells, sf-CD was generated in response to prolonged secretogogue stimulation and was subject to rapid proteasomal degradation. Based on its solubility and apparent mass, we suggested that sf-CD was generated by intramembrane cleavage.
In this study we demonstrate that juxtamembrane cleavage of PAM by an ␣-secretase-like enzyme or by prohormone convertases must occur before a ␥-secretase-like enzyme can cleave within its transmembrane domain, producing sf-CD. Mutations in the cytosolic domain of PAM-1 modulate the generation of sf-CD from the intact protein. Mass spectroscopic analysis of sf-CD produced from Myc-TMD-CD is most consistent with the presence of multiple N termini located 7-12 amino acids from the cytosolic domain of PAM. Levels of sf-CD in anterior pituitary endocrine cells rose after repeated stimulation and depletion of secretory granule stores.
Primary Anterior Pituitary Cultures-Rat anterior pituitary cultures were prepared from adult male and female rats as described (12). Cells were plated in 12-well plastic dishes coated with protamine and NuSerum (0.5 pituitary/well) in/Dulbecco's modified Eagle's medium/F-12 (Mediatech) containing 25 mM Hepes, 10% NuSerum, 10% fetal bovine serum, and penicillin/streptomycin. For the next 2 days, the cells were maintained in complete serum free medium (Dulbecco's modified Eagle's medium/F-12, 25 mM Hepes, pH 7.4, penicillin/streptomycin/ insulin/transferrin/selenium, 1 mg/ml fatty acid-free bovine serum albumin) and cytosine arabinoside (10 M). Cells were maintained for an additional day in complete serum free medium without cytosine arabinoside before harvesting.
Treatment with Inhibitors and Preparation of Extracts-AtT-20 cell lines and primary pituitary cells were maintained in complete serum-free medium with 0.02 mg/ml bovine serum albumin. Unless indicated otherwise, inhibitors were prepared as stock solutions in DMSO and were diluted at least 1000-fold into culture medium. Both 2 M MG132 (Enzo Life Sciences; 2 mM stock) and 5 M lactacystin (Enzo Life Sciences; 5 mM stock in water) were used as proteasomal inhibitors. Other inhibitors used included 10 M 3,4 dichloroisocoumarin (EMD Chemicals; 10 mM stock), 5 M benzyloxycarbonyl-VLL-CHO (EMD Chemicals; 10 mM stock), 10 M GM6001 (EMD Chemicals; 10 mM stock), 10 M DAPT (Biomol; 10 mM stock), 10 M MDL-2810 (Biomol), and 1 M L685458 (Tocris; 1 mM stock). Treatment times are specified in each figure legend; in general, specific protease inhibitors were present during an overnight incubation. For AtT-20 cells, this was followed by a 5-h incubation that included that same inhibitor along with MG132. Control cells were treated with 0.1% DMSO overnight followed by DMSO along with MG132 for an additional 5 h. Doses were selected based on the literature, with visual assessment to monitor cell viability. GM6001 and DAPT were tested on AtT-20 PAM-1 cells at 5, 10, and 20 M; even the highest dose had no morphological effect on the cells. DAPT blocked sf-CD formation even at 5 M; processing of the PAM luminal domains in the regulated secretory pathway was unaffected. The lowest dose of GM6001 (5 M) did not block formation of soluble 100-kDa PAM (PAMs) (20,21), and 20 M GM6001 had a greater effect than 10 M (data not shown).
AtT-20 cells and pituitary cells were extracted into ice-cold 20 mM NaTES, 10 mM mannitol, 1% Triton X-100 (Pierce SurfActs), pH 7.4, containing protease inhibitors (0.3 mg/ml PMSF, 50 g/ml lima bean trypsin inhibitor, 2 g/ml leupeptin, 16 g/ml benzamidine, and 2 g/ml pepstatin); lysates were allowed to tumble for 30 min at 4°C and were then clarified by centrifugation at 10,000 ϫ g for 10 min. For pituitary samples, cell extracts and spent media were concentrated by overnight precipitation with 80% acetone at Ϫ20°C. After centrifugation at 10,000 ϫ g for 30 min, supernatants were removed; air-dried pellets were suspended in Laemmli sample buffer for SDS-PAGE.
Western Blotting-Samples were subjected to SDS-PAGE and Western blot analysis as described (12). Antigen-antibody complexes were detected using horseradish peroxidase-conjugated secondary antibody and Super Signal West Pico chemiluminescent substrate (Pierce). An affinity-purified rabbit polyclonal C-Stop antibody raised against the final 12 residues of PAM was used to analyze sf-CD levels (15). ␤III-Tubulin was visualized with a monoclonal antibody (Covance), and APP was visualized with a polyclonal APP C-terminal-specific antibody (Zymed Laboratories Inc.). Membranes were cut above the 25-kDa marker, and the top and bottom pieces were incubated separately with the same stock of C-Stop and secondary antibody (15). Western blots were quantified using GeneTools software and non-saturated images (Syngene). All experiments were repeated at least twice, and representative gels are shown and quantified.
Immunoprecipitation and Mass Spectrometry-Myc-TMD-CD cells or wild type AtT-20 cells were grown in 15 confluent 150-mm dishes each and treated with lactacystin for 5 h in complete serumfree medium containing 0.02 mg/ml bovine serum albumin. Cells were extracted in 20 mM NaTES, 10 mM mannitol containing protease inhibitors, passed through a 25-gauge needle, and centrifuged at 1000 ϫ g for 5 min to remove cell debris. The cytosolic fraction was prepared by centrifuging the lysate at 430,000 ϫ g for 15 min. Immunoprecipitation was performed using equal amounts of the two lysates and equal amounts of C-Stop antibody (20 g of affinity-purified antibody for ϳ20 mg of cytosolic protein). Binding to antibody was carried out at 4°C overnight followed by binding to Protein A beads (250 l of beads/20 mg of cytosolic protein) for 2 h with tumbling. Beads were washed with phosphate-buffered saline, pH 7.4, and finally with ice-cold water. The proteins bound to the beads were eluted with ice-cold 0.4% TFA in 30% acetonitrile at 4°C and lyophilized.
Comparative Matrix-assisted Laser Desorption Ionization Mass Spectrometry (MALDI-MS) Analysis-Lyophilized aliquots of immunoprecipitated control and sf-CD were dissolved in 10 l of 70% formic acid and then diluted with 20 l of 0.1% TFA. To ensure the removal of all salts, the samples were desalted on a C4 (Millipore) ZipTip. The eluted, desalted sample was dried in a vacuum centrifuge before dissolving in 2 l of MALDI matrix ((␣-cyano-4-hydroxy cinnamic acid matrix (3.0 mg/ml in 0.05% TFA, 50% acetonitrile)); 1 l was loaded onto the target plate. MALDI-MS was performed in linear mode over the mass range of m/z 1,000 -18,000 on an Applied Biosystems (AB) 4800 MALDI-Tof/Tof mass spectrometer. MS spectra were overlaid for comparative analysis.
Protein Digestion-Before LC-MS/MS analysis, aliquots of the immunoprecipitated control and sf-CD samples were dissolved in 20 l of freshly made 8 M urea, 0.4 M ammonium bicarbonate; disulfide bonds were reduced by adding 2 l of 45 mM dithiothreitol and incubating at 37°C for 20 min followed by alkylation using 2 l of 100 mM iodoacetamide and incubation at ambient temperature in the dark for 20 min. The urea concentration was reduced to 2 M by the addition of water, and lysyl endopeptidase (Wako Chemicals) was added at a 1:10 weight to weight ratio. Digestion proceeded for 16 h at 37°C.
Titanium Dioxide Enrichment-After digestion, the samples were acidified with 0.5% TFA, 50% acetonitrile. Top Tips (Glygen Corp.) were prepared by washing 3 times with 40 l of 100% acetonitrile followed by 0.2 M sodium phosphate, pH 7.0, and 0.5% TFA, 50% acetonitrile. Washes were spun through into a microcentrifuge tube at 2000 rpm for 1 min. The acidified digest supernatant was then loaded into the TopTip and spun at 1000 rpm for 1 min and then at 3000 rpm for 2 min. The Top Tip was next washed with 40 l of 0.5% TFA, 50% acetonitrile, and the spin was repeated. The flowthrough from these washes was saved and analyzed by LC-MS/MS as below. Phosphopeptides were eluted from the TopTip with 3 ϫ 30 l of 28% ammonium hydroxide. Both the flow-through and eluted fractions were dried in a vacuum centrifuge and re-dried from 40 l of water. Samples were dissolved in 3 l of 70% formic acid, vortexed, diluted with 7 l 50 mM sodium phosphate, spun, and transferred to LC-MS/MS vials where 5 l was injected.
LC-MS/MS on the LTQ Orbitrap-The LTQ Orbitrap (Thermo Scientific) was equipped with a Waters nanoACQUITY UPLC system, with a Waters Symmetry C18 180-m ϫ 20-mm trap column and a 1.7-m, 75-m ϫ 250-mm nanoACQUITY TM UPLC TM column (35°C). Peptide trapping was done at 15 l/min, 99% buffer A (100% water, 0.1% formic acid) for 1 min. Peptide separation was performed at 300 nl/min with buffer A (100% water, 0.1% formic acid) and buffer B (100% acetonitrile, 0.075% formic acid). A linear gradient (51 min) was run with 5% buffer B at initial conditions, 50% B at 50 min, and 85% B at 51 min. MS was acquired in the Orbitrap part of the instrument (400 -2000 m/z) using 1 microscan and a maximum inject time of 900 ms in the LTQ part of the instrument followed by six data-dependent MS 2 acquisitions. For phosphopeptide analysis, Multistage Activation was used, triggering on the Ϫ98.0, Ϫ49.0, and Ϫ32.7 atomic mass unit neutral loss fragment ions.
Data Base Searching-All MS/MS spectra were searched inhouse using the Mascot algorithm (version 2.2.06) for uninterpreted MS/MS spectrum after using the Mascot Distiller program to generate Mascot compatible files. Carboxamidomethylated cysteine, possible methionine, and tryptophan oxidation and phosphoserine, threonine, and tyrosine were considered along with a peptide tolerance of ϩ20 ppm, MS/MS fragment tolerance of ϩ0.6 Da, and peptide charges of ϩ2 or ϩ3. Normal and decoy data base searches were run.

sf-CD Production from Myc-TMD-CD Is Blocked by ␥-Secre-
tase Inhibitors-The apparent molecular weight and solubility of sf-CD suggest that it is generated through intramembrane proteolysis (15). Because AtT-20 cells stably expressing Myc epitope-tagged PAM TMD-CD (Fig. 1A) produce sf-CD, we FIGURE 1. Production of sf-CD from Myc-TMD-CD is blocked by ␥-secretase inhibitors. A, topology of Myc-TMD-CD and sf-CD is illustrated; the signal sequence that precedes the Myc epitope has been removed. The epitope recognized by the C-Stop antibody is illustrated; the arrow indicates postulated intramembrane cleavage. B, AtT-20 cells expressing Myc-TMD-CD were treated with the indicated protease inhibitor for 16 h and with the same inhibitor plus MG132, a proteasomal inhibitor, for an additional 5 h. Duplicate cell extracts were subjected to Western blot analysis using the C-Stop antibody; ␤III tubulin was visualized as a loading control. C, control and DAPT lysates from B were analyzed using antibody specific for the C-terminal fragment of APP (CTF). D, triplicate wells of AtT-20 cells expressing Myc-TMD-CD were treated with L685458, another ␥-secretase inhibitor, and analyzed as described in B. All experiments were repeated three times with duplicates or triplicates, and representative gels are shown and quantified. Data were analyzed using a two-tailed t test assuming unequal variance; error bars indicate S.D.; *, p Ͻ 0.05. used these cells to identify the protease involved in the final step of sf-CD production. AtT-20 cells expressing Myc-TMD-CD were treated with inhibitors of calpain (MDL-28170) (22), ␤-secretase (benzyloxycarbonyl-VLL-CHO) (23), ␣-secretase (GM6001) (24), and ␥-secretase (DAPT) (25). The morphology of cells treated with inhibitors was indistinguishable from that of control cells. Given the rapid turnover of sf-CD observed in previous experiments (15), MG132, a proteasomal inhibitor, was included during the final 5 h of the incubation with each protease inhibitor. Control cells were treated with vehicle (DMSO) and MG132. Lysates were probed with C-Stop antibody, which is specific for the C terminus of PAM (Fig. 1A).
Only the ␥-secretase inhibitor DAPT resulted in decreased levels of sf-CD and increased levels of intact Myc-TMD-CD (Fig. 1B). The calpain, ␣-secretase, and ␤-secretase inhibitors were without effect. AtT-20 cells express APP, which is a ␥-secretase substrate (26). We, therefore, probed lysates from control and DAPT-treated cells with an APP C-terminal-specific antibody that detects full-length APP isoforms (APP 695,751,770) and the ϳ11-kDa C-terminal fragment that serves as a ␥-secretase substrate. Lysates from DAPT-treated cells accumulated the C-terminal fragment (Fig. 1C), as observed for Myc-TMD-CD (Fig. 1B); the 5-kDa APP intracellular domain fragment, which would be equivalent to sf-CD, could not be detected in control AtT-20 cells.
To verify an essential role for ␥-secretase, a different inhibitor, L685458 (27), was tested using the same treatment paradigm (Fig. 1D). Levels of sf-CD were reduced in L685458-treated cells and intact Myc-TMD-CD accumulated, although the effect was less pronounced than that of DAPT.
Processing of Luminal Domains of PAM-1 Affects sf-CD Generation-Unlike Myc-TMD-CD, PAM-1 has a large luminal domain that is subject to tissue-specific endoproteolytic cleavage both in secretory granules (Granule pathway) and in cells lacking secretory granules (Non-granule pathway) ( Fig. 2A) (21). The prohormone convertases that cleave pairs of basic amino acids in prohormones also cleave PAM at pairs of basic amino acids, releasing its soluble enzymatic domains, PHM and PAL, and generating a TMD-CD fragment that resembles Myc-TMD-CD ( Fig. 2A). In atrial myocytes and in transfected kidney cells, which do not express prohormone convertases, a luminal domain cleavage that occurs near the transmembrane domain releases soluble 100-kDa PAMs. Studies on ␥-secretase have consistently revealed a need to remove much of the luminal domain of a potential substrate to allow access of the target transmembrane domain to its active site (28).
To begin to explore the role of various endoproteases in the generation of sf-CD from PAM-1, AtT-20 cells stably expressing PAM-1 were treated with a matrix metalloproteinase inhibitor (GM6001), a rhomboid inhibitor (DCI, 3,4 dichloroisocoumarin), a ␥-secretase inhibitor (DAPT) or vehicle (Con) (Fig.  2B). Levels of PAM-1 and PALm were not altered by any of these treatments. Production of sf-CD was blocked in the presence of the matrix metalloproteinase inhibitor or the ␥-secretase inhibitor, but the 19-and 22-kDa intermediates that presumably serve as precursors to sf-CD accumulated only in the presence of the ␥-secretase inhibitor. As for Myc-TMD-CD, the ␤-secretase inhibitor benzyloxycarbonyl-VLL-CHO had no effect on the cleavage of PAM-1 (data not shown).
To better evaluate the effect of each protease inhibitor on PAM cleavage, we examined the PAM proteins released into the medium under basal conditions (Fig. 2C). Although 45-kDa PHM and 50-kDa PAL are generated by prohormone convertase cleavage, 100-kDa PAMs is not detected in secretory granules and is thought to be formed by a cleavage that occurs at a different subcellular location (11,16). Levels of 100-kDa PAMs were reduced by incubation with GM6001, the matrix metalloproteinase inhibitor, but not by incubation with the rhomboid or ␥-secretase inhibitor (Fig. 2C). The ability of GM6001 to block this cleavage would limit production of 19-kDa TMD-CD, indirectly reducing the ability of the cells to produce sf-CD.
To verify the role of ␥-secretase in sf-CD production from PAM-1, an additional ␥-secretase inhibitor, L685458, was tested (Fig. 2D). Like DAPT, L685458 treatment resulted in a decrease in sf-CD levels. As observed for Myc-TMD-CD, L685458 was less effective than DAPT at blocking creation of sf-CD, and no accumulation of the 19-or 22-24-kDa TMD-CD intermediate was observed.
Lactacystin Increases sf-CD Accumulation in AtT-20 Cells-In AtT-20 cells expressing Myc-TMD-CD or PAM-1, sf-CD is barely detectable in the absence of the proteasomal inhibitor MG132, a peptidyl aldehyde that also inhibits calpain 1 (15,29). Because the proteasome seemed to play a key role in sf-CD metabolism, we tested another inhibitor. We chose lactacystin, a ␤-lactone that binds covalently but also inhibits cathepsins A and B (29,30). We first examined AtT-20 cells expressing Myc-TMD-CD and treated them for 5 h with lactacystin or MG132 (Fig. 3A). Levels of sf-CD rose after treatment with either inhibitor, but the effect of lactacystin was more pronounced. Data were quantified, making it clear that both drugs increased the total signal observed with the C-Stop antibody to a similar extent (Fig. 3B). Lactacystin had a greater effect on the level of sf-CD than did MG132 (Fig. 3, A and B).
The effects of lactacystin and MG132 on AtT-20 cells expressing PAM-1 were also compared (Fig. 3C). In the absence of proteasomal inhibitor, sf-CD is not detectable (15). sf-CD levels were higher in lactacystin-treated cells than in MG132treated cells, whereas the levels of 19-kDa intermediate were higher in MG132-treated cells than in lactacystin-treated cells. AtT-20 cells expressing Myc-TMD-CD and PAM-1 responded in a similar manner to these two drugs. Levels of the 22-24-kDa intermediates, which are thought to arise from prohormone convertase-mediated cleavage of PAM-1, were indistinguishable in MG132 and lactacystin-treated cells, consistent with the idea that the 19-and 22-24-kDa intermediates are metabolized through distinct pathways.
Different Isoforms of PAM Yield Different Amounts of sf-CD-Access of substrates like SREBP and Notch to the relevant i-CLiP requires juxtamembrane endoproteolytic cleavage of their luminal domain (2,4,31). In secretory granules, the luminal domains of PAM are subject to prohormone convertase-mediated cleavage ( Fig. 2A); cleavage of PAM-1 at the Lys-Lys sequence in the linker region separating PHM from PAL generates PALm (Fig.  4A). This linker region is absent from PAM-2, the other major splice variant of PAM (32). Prohormone convertase mediated cleavage at the Lys-Lys site after PAL leaves a non-catalytic fragment of PAM that is similar to Myc-TMD-CD, which includes nine residues of the stalk region (Fig. 4A). To determine whether luminal cleavage is required for the generation of sf-CD, AtT-20 lines stably expressing these four proteins were treated with MG-132 and subjected to analysis with the C-Stop antibody (Fig. 4B). Levels of intact protein, 19/22-kDa intermediates, and sf-CD were quantified.
In Myc-TMD-CD cell lysates, sf-CD accounted for ϳ25% as much of the C-Stop signal as did intact Myc-TMD-CD (Fig. 4, B and C). In contrast, very little sf-CD was present in PALm cell lysates. While sf-CD accounted for about 10% as much of the C-Stop signal as intact PAM-1, sf-CD was less prevalent in PAM-2 cells (Fig. 4, B and C). Conversion of PALm, PAM-2, and PAM-1 into sf-CD presumably involves trafficking of the newly synthesized protein through multiple subcellular compartments and the actions of multiple proteases. To separate production of potential i-CLiP substrate from production of sf-CD, we compared the level of 19/22-kDa intermediate present at steady state to the level of intact protein (open bars, Fig.  4D); this step is bypassed in Myc-TMD-CD cells. PAM-1 was more efficiently converted into intermediates than was PAM-2 or PALm. It is not clear whether this difference reflects a difference in protein structure or a difference in protein trafficking.
When levels of sf-CD in each cell line were compared with levels of 19/22-kDa intermediate (or intact Myc-TMD-CD), a very different answer was obtained (Fig. 4D). Regardless of the PAM protein expressed, the steady state level of sf-CD was ϳ5-fold less than the steady state level of 19/22-kDa interme- diate. Although very little 19/22-kDa intermediate was detected in PALm cells, the small amount present was converted into sf-CD as efficiently as it was in PAM-1, PAM-2, or Myc-TMD-CD cells. Two different aspects of PAM processing appear to govern sf-CD production, the amount of intermediate produced from the intact protein and access of these intermediates to the enzyme that produces sf-CD.
sf-CD Levels Are Modulated by Properties of the Cytosolic Domain-PAM cytosolic domain, an unstructured 86-amino acid region, is multiply phosphorylated (15) and interacts with several cytosolic proteins including Kalirin and Trio, GDP/ GTP exchange factors for small GTP-binding proteins of the Rho subfamily, and Uhmk1 (KIS, PCIP-2), a Ser/Thr kinase (14). We analyzed sf-CD production from PAM-1 mutated at sites known to abrogate its interaction with Uhmk1 or with both Uhmk1 and Kalirin and at specific phosphorylation sites. Mutation of Phe 929 -Phe 930 to Ala-Ala eliminates interaction with Uhmk1, whereas mutation of Lys 919 to Arg eliminates interaction with both Uhmk1 and Kalirin (Fig. 5A). Phosphomimetic mutations at Ser 937 or Thr 946 -Ser 949 were previously shown to alter endocytic trafficking (33,34) and block sf-CD generation (15). Additional phosphorylation sites were mutated in PAM-1/GFP 3P and 5P to analyze the effects of multiple phosphorylation.
Levels of 19/22-kDa intermediates and sf-CD produced from each mutant protein were evaluated by Western blot (Fig. 5B) and quantified as described above (Fig. 5C) Levels of 19/22-kDa intermediates and sf-CD did not differ significantly in PAM-1/GFP and PAM-1 cells (Fig. 5C). Simultaneous phosphomimetic mutations at Ser 932 , Ser 937 , and Ser 945 (3P) or at these three sites along with Thr 946 and Ser 949 (5P) had little effect on levels of the 19/22-kDa intermediates. However, levels of sf-CD were very low in PAM-1/GFP 5P lysates versus PAM-1/GFP 3P lysates. Because PAM-1 with a single phosphomimetic mutation at Ser 937 yielded only low steady state levels of sf-CD (15), phosphorylation seems to affect sf-CD levels in a combinatorial fashion.
Anterior Pituitary sf-CD Is Produced in Response to Secretogogue-We turned to primary cultures of rat anterior pituitary cells to test the ability of secretogogue to stimulate sf-CD production. Both PAM-1 and PAM-2 are expressed endogenously in the pituitary and are subjected to cell typespecific cleavage (Fig. 6A) (35); the C-Stop cross-reactive bands at 24 -26 kDa represent TMD-CD intermediates that could be derived from either the Granule or the Non-granule pathway. Under basal conditions, in the absence of proteasomal inhibitors, sf-CD was readily detected in primary anterior pituitary cells (Fig. 6A); in contrast, in AtT-20 cells, sf-CD is barely detectable in the absence of proteasomal inhibitors (15).
To stimulate each of the different pituitary cell types and cause granule depletion, cultures were exposed alternately over a 5.5-h period to BaCl 2 , to mimic elevated intracellular calcium, and phorbol myristate acetate, to stimulate protein kinase C (12). Analysis of media from the first BaCl 2 sample verified that secretion was stimulated (Fig. 6A, bottom panel). Cells harvested at the end of the depletion paradigm had reduced levels of intact PAM-1 and PAM-2, TMD-CD intermediates, and sf-CD (Fig. 6A). Bands were quantified and analyzed as for PAM-1 AtT-20 cells (Fig. 6B); the ratio of Intermediates to intact PAM declined with depletion, whereas the ratio of sf-CD to intermediates rose. This shift suggests more efficient conversion of intermediates into sf-CD or increased stability of sf-CD in depleted cells (Fig. 6B). A similar response to secretogogue stimulation was observed previously in PAM-1 AtT-20 cells (15).
We next asked whether a single exposure to secretogogue could alter levels of sf-CD. Primary pituitary cells were exposed to medium containing both BaCl 2 and phorbol myristate acetate for 1 h. Cells and media were collected and analyzed as above. Robust stimulation of secretion was verified by examining PHM released into the medium (Fig. 6C, bottom panel). Analysis of cell extracts with C-Stop antibody revealed decreased levels of intact PAM and intermediates; sf-CD was readily detected in both samples. Although trending in the same direction, the ratio of sf-CD to intermediates was not significantly increased after only 1 h of secretogogue stimulation (Fig. 6D).
Mass Spectrometric Analysis of sf-CD Start Sites-We used Myc-TMD-CD and wild type AtT-20 cells to try to identify the N terminus of sf-CD. sf-CD was immunoisolated from Myc-TMD-CD cell cytosol; wild type AtT-20 cells processed in parallel served as a control. Both cells types were treated with lactacystin to improve the yield of sf-CD (Fig. 7A). Immunoprecipitation efficiency was verified by immunoblotting with the C-Stop antibody (Fig. 7B). Along with the major sf-CD band, a slightly smaller minor band was detected in the immunoprecipitate by the C-Stop antibody (arrow). No cross-reactivity was observed in the wild type AtT-20 sample (Fig. 7B). Aliquots of the immunoisolated protein were prepared for LC-MS/MS analysis after digestion with endopeptidase Lys-C. LC-MS/MS was performed on a LTQ Orbitrap on unenriched fractions as well as after phosphopeptide enrichment using TiO 2 tips with both the bound and unbound fractions analyzed (Fig. 7C). Peptides derived from the cytosolic domain of PAM were identified only in the sample prepared from Myc-TMD-CD cells (Fig. 7C). MS/MS data for four peptides covering 52 residues of the cytosolic domain along with one site of phosphorylation are presented in supplemental Figs. S1-S4. No PAM peptides were identified in the wild type control sample.
The sf-CD sample immunoisolated from the Myc-TMD-CD cell cytosol and the corresponding fraction prepared from wild type cells were also subjected to linear MALDI-Tof analysis to determine the intact mass of sf-CD. The fact that the C-Stop antibody requires the intact C terminus of PAM-1 (Ser 976 ) makes it possible to deduce the N terminus based upon the mass of intact sf-CD. The entire eluate was analyzed, meaning that sf-CD was a minor component of the sample being analyzed. The sample prepared from wild type cells provided a perfect control, allowing identification of masses present in the Myc-TMD-CD sample and absent from the wild type control sample.
Peaks were aligned based on their centroid masses at 80% peak height (used to eliminate peak tailing); the region shown (m/z 9,100 -11,200) was the only region of the spectrum in which peak areas from the Myc-TMD-CD cells were greater than the corresponding control cell peak areas ( Fig. 8 and supplemental Table 1). Five putative sf-CD-related peaks were identified in this manner (Fig. 8). To adequately match peptides terminating with Ser 976 to the masses observed, we allowed phosphorylation (ϩ80); P-Ser 949 was identified in peptides prepared from Myc-TMD-CD cells (Fig. 7C) and multiple phosphorylation sites were previously identified in the cytosolic domain of PAM-1 (15). Oxidation of Met and Trp was not observed in these samples, so this modification was not considered. Peak #1, with a mass of 9396.3, cannot include any of the transmembrane domain and is best fit to a monophosphorylated peptide extending from Trp 892 to Ser 976 . Peaks #2-5 can each be fit to non-phosphorylated, monophosphorylated, or diphosphorylated peptides that begin with a residue in the transmembrane domain and terminate with Ser 976 . sf-CD masses beginning 12, 10, 9, and 7 residues before the stop transfer signal ( 891 RWKK) were tentatively identified by this method. Cleavage at Ile 878 -Pro, Val 880 -Leu, Leu 881 -Val, and Leu 883 -Leu bonds would yield the products observed. Several other ␥-secretase substrates are cleaved at multiple sites (4),  22-24 kDa), and sf-CD were quantified; the ratios of intermediates to intact PAM-1 and sf-CD to intermediates were calculated. Mutations affected production of intermediates and sf-CD differently. Experiments were repeated at least twice, and representative gels are shown and quantified. Data were analyzed using a two-tailed t test assuming unequal variance; error bars indicate S.D. *, p Ͻ 0.05. although cleavage followed by trimming from the N terminus could also account for the products observed.

DISCUSSION
i-CLiP Cleavage of PAM-Regulated intramembrane proteolysis is an ancient process known to occur in prokaryotes like Enterococcus faecalis and Bacillus subtilis as well as eukaryotes (2). Levels of sf-CD in Myc-TMD-CD cells were reduced in the presence of two ␥-secretase inhibitors, DAPT and L685458. DAPT treatment resulted in the accumulation of intact Myc-TMD-CD, whereas L685458 did not. L685458 is a transition state analog, whereas DAPT binds to a different site in the C-terminal fragment of presenilin1, perhaps contributing to their different effects on Myc-TMD-CD metabolism (36 -38). Consistent with identification of ␥-secretase as the enzyme cleaving the PAM transmembrane domain, array analysis revealed expression of Presenilin1 and Nicastrin, two of the components of the ␥-secretase complex, in AtT-20 cells (40).
As previous attempts to sequence immunoprecipitated sf-CD by Edman degradation failed, we utilized MALDI-MS coupled with the fact that the C-Stop antibody cannot recognize PAM fragments lacking Ser 976 to suggest four intramembrane N-terminal start sites, Pro 879 , Leu 881 , Val 882 , and Leu 884 , for sf-CD. It is not yet clear whether these fragments are actual ␥-secretase cleavage products or the products of subsequent trimming. Well studied ␥-secretase substrates like Notch and APP are often cleaved at multiple sites (4,5). Cleavage of Notch generates the ␥-secretase substrate NEXT, which is cleaved at multiple sites in its transmembrane domain, generating Notch intracellular domain (NICD). Once released from the membrane, NICD can translocate to the nucleus and regulate gene expression. The most stable NICD fragment includes four transmembrane domain residues; slightly shorter fragments are less stable and thought to play a lesser role in Notch signaling (4). Although sf-CD accumulates in the nucleus, the N-terminal product of the ␥-secretase cleavage of PAM TMD-CD has yet to be identified.
Regulation of PAM sf-CD Levels-In both primary anterior pituitary cells and PAM-1 AtT-20 cells, levels of sf-CD increased in response to secretogogue stimulation, providing support for the hypothesis that this soluble fragment of PAM plays a role in signaling from secretory granules to the nucleus. The steady state level of sf-CD reflects many factors, including cleavages within the luminal domain of PAM, cytosolic domain FIGURE 6. Endogenous sf-CD increases in response to exocytosis. A, duplicate wells of primary anterior pituitary cells were kept under basal conditions (Con) or treated alternately with BaCl 2 (2 mM) and phorbol myristate acetate (1 M) using a paradigm known to deplete cells of hormone (Depleted) (12). Medium collected in the first 1.5 h of BaCl 2 stimulation or under basal conditions was probed with the Exon 16 antibody, revealing stimulated secretion of 45-kDa PHM. Cell lysates were probed with C-Stop and ␤III-Tubulin antibodies. B, intact PAM (PAM-1 ϩ PAM-2), intermediates, and sf-CD were quantified, and the indicated ratios were calculated. C, duplicate wells of primary anterior pituitary cells were kept under basal conditions or exposed simultaneously to BaCl 2 and phorbol myristate acetate for 1 h. Medium was probed with the Exon 16 antibody to verify secretion. Cell extracts were analyzed as described above. D, data were quantified as described for B. Experiments were repeated three times with duplicates; representative gels are shown and quantified. Data were analyzed using a two-tailed t test assuming unequal variance; error bars indicate S.D. *, p Ͻ 0.05. phosphorylation, the subcellular localization of PAM and of ␥-secretase, the activity of the ␥-secretase complex, and the turnover of sf-CD (Fig. 9).
As for other ␥-secretase substrates, removal of the luminal, catalytic domains of PAM is essential before ␥-secretase cleavage can occur. Tissue-specific and developmentally regulated alternative splicing yields two major isoforms of membrane PAM, PAM-1 and PAM-2 (32); PAM-2 lacks the exon that encodes a protease-sensitive linker region separating the two catalytic domains of PAM. In AtT-20 cells, PAM-1 is converted into sf-CD more efficiently than PAM-2. Splice variants of the receptor-tyrosine kinase ErbB4 that include a juxtamembrane ␣-secretase cleavage site can be cleaved by ␥-secretase, whereas splice variants lacking this site cannot (38).
PAM-1 is subject to prohormone convertase-mediated cleavage in the regulated secretory pathway and to juxtamembrane ␣-secretase-mediated cleavage elsewhere in the cell (Fig.  9). Cleavage of PAM-1 or PAM-2 at the Lys-Lys 822 site that follows the catalytic core of PAL leaves a 44-residue stalk region attached to the transmembrane domain, presumably corresponding to the 22-24-kDa protein visualized with the C-Stop antibody. The ␣-secretase cleavage site has not yet been iden-tified but is substantially closer to the transmembrane domain (Fig.  9). ADAM17/TACE or ADAM10 metalloprotease cleavage of Notch leaves a 12-residue stalk attached to its transmembrane domain. Whether the product created by prohormone convertase cleavage of PAM requires further trimming before it can access the ␥-secretase complex or is cleaved differently by ␥-secretase is not yet clear.
The unstructured cytosolic domain of PAM is subject to multisite phosphorylation and is known to interact with several cytosolic proteins (14,15). Phosphomimetic mutants of PAM differed in their ability to generate sf-CD, as did mutants unable to interact with the Rho guanine nucleotide exchange factors Kalirin and Trio. Both differences in subcellular localization and conformational differences could contribute to the observed alterations in sf-CD levels. The active ␥-secretase complex may be primarily localized to the plasma membrane and to the endocytic compartment (39). Because PAM-1 that reaches the plasma membrane is rapidly endocytosed, it could come into contact with ␥-secretase at multiple sites (Fig. 9). PAM-1 mutants unable to enter the intralumenal vesicles of the multivesicular body (e.g. PAM-1/TS/DD) might have limited access to ␥-secretase (11). A role for Kalirin in restraining sf-CD production is suggested by the fact that PAM-1/K919R, which cannot interact with Kalirin or Uhmk1, yields low levels of TMD-CD intermediate and relatively high levels of sf-CD. Processing of a PAM-1 mutant that can interact with Kalirin but not with Uhmk1 (FF/AA) was indistinguishable from that of PAM-1.
The turnover of sf-CD differs substantially in primary pituitary cells and AtT-20 cells. sf-CD is readily detectable in anterior pituitary cell lysates in the absence of proteasomal inhibitors. In PAM-1 AtT-20 cells, the addition of MG132 or lactacystin is essential to visualize sf-CD. In Myc-TMD-CD cells, treatment with lactacystin or MG132 produced a similar increase in total cross-reactive material, but lactacystin had a greater effect on sf-CD, whereas MG132 had a greater effect on Myc-TMD-CD. A similar difference in the effects of the two inhibitors was observed in PAM-1 AtT-20 cells, where MG132 brought about a selective increase in levels of the 19-kDa TMD-CD intermediate, the presumed product of ␣-secretase cleavage, without altering levels of the larger TMD-CD fragments derived from prohormone convertase-mediated cleavage of PAM-1. These differences suggest that TMD-CD frag- The mass range shown contained the only unique masses (indicated by mass) in the sf-CD sample. Centroid masses were measured at 80% peak height to minimize peak tailing effects; sharper m/z peaks have higher mass accuracies than m/z peaks with shoulders. The corresponding region from the WT sample was identified so that peak areas could be compared (supplemental Table 1). Peaks #1-5 were enriched or present only in Myc-TMD-CD cells, indicating that they arose from Myc-TMD-CD. Peaks #1-5 had masses consistent with cleavage at the indicated sites in or near the transmembrane domain (green letters) of PAM, coupled with serine phosphorylation as indicated.
ments generated in secretory granules are distinguishable from TMD-CD fragments generated elsewhere in the cell (Fig. 9).
During endocytosis, PAM is known to enter the intraluminal vesicles of the multivesicular body (11). In addition, a significant amount of the PAM-1 biotinylated on the cell surface undergoes a juxtamembrane cleavage that releases 100-kDa PAM. Cleavage of this type would generate a TMD-CD fragment that could serve as a ␥-secretase substrate. The C-terminal fragment of APP accumulates in the intraluminal vesicles of multivesicular bodies and is released in exosomes when the multivesicular bodies fuse with the plasma membrane. Inhibition of ␥-secretase increases exosomal release of APP C-terminal fragment (27). sf-CD formation from PAM TMD-CD may also involve cleavages that occur in multivesicular bodies. Exosomes play an important signaling role in several systems including the immune system.
Signaling from Secretory Granules-Unlike synaptic vesicles, peptide-containing secretory granules cannot be refilled and reused. For a pituitary cell or a neuron to maintain its supply of secretory granules, there must be a means of reporting granule use. Several observations support the conclusion that iCliPmediated production of sf-CD plays a role in a regulatory loop of this type (Fig. 9). First, production of sf-CD is triggered by stimulated secretion. Second, sf-CD accumulates in the nucleus. Its ability to do so is modulated by phosphorylation. sf-CD with phosphomimetic mutations at its Uhmk1 site (TS/ DD) is less well localized to the nucleus; co-expression of sf-CD and active Uhmk1 eliminates nuclear localization of sf-CD. Third, array analysis revealed altered expression of a subset of transcripts upon induction of PAM-1 expression in AtT-20 cells. Aquaporin 1 (Aqp1), a water channel known to be important for secretory granule biogenesis, was among the transcripts upregulated by PAM-1 (40). Transient expression of Myc-tagged PAM cytosolic domain (Myc-CD) in AtT-20 cells produced a similar increase in Aqp1 protein levels (40).
Regulated intramembrane proteolysis of SREBP by S2P is triggered by low cholesterol levels and initiates a collection of responses to counteract this deficit. Regulated intramembrane proteolysis of ATF6 by S2P is triggered when an excess of unfolded proteins removes the BiP bound to ATF6, initiating responses that reduce the level of unfolded proteins in the endoplasmic reticulum. For Notch, the binding of ligands like Delta and Jagged (4) triggers the ectodomain shedding that must occur before ␥-secretase can release the NICD from NEXT. The events that trigger ␥-secretase-mediated cleavage of the APP and generate APP intracellular domain are not well understood.
The cytosolic fragments generated by the i-CLiP cleavage of SREBP, ATF6, Notch, and APP translocate to the nucleus and regulate gene expression. The basic helix-loop-helix transcription factors generated from SREBP and ATF6 affect gene expression directly. NICD, the product of Notch cleavage, associates with CSL, a DNA-binding protein, and activates transcription of specific genes. APP intracellular domain, the product of APP cleavage, is thought to associate with the adaptor protein Fe65 and the histone acetyltransferase Tip60 and can drive transcription in an artificial reporter assay. PAM sf-CD is also thought to regulate gene expression, although the mechanism is not yet clear. Microarray analysis identified a subset of genes up-regulated upon expression of a tetracycline inducible PAM-1 cell line (40). One of the up-regulated genes, Aqp1, was also found to be up-regulated upon overexpression of Myc epitope-tagged PAM cytosolic domain alone. Further analysis found Aqp1 to be down-regulated in PAM heterozygous mice, indicating a direct role of PAM cytosolic domain in regulating Aqp1 levels in cells (40). Aqp1, a water channel, has been shown to be important for the formation of secretory granules (6) (Fig.  9). PAM-sf-CD generated by regulated intramembrane proteolysis in response to secretion may participate in the control of secretory granule biogenesis. FIGURE 9. PAM sf-CD signaling pathway. Secretory granules contain neuropeptides and processing enzymes like the prohormone convertases, carboxypeptidase E, and PAM. Granules release their soluble content during regulated exocytosis. PAM, a secretory granule membrane protein, reaches the plasma membrane and can be recycled and reused. In the granules, PAM is cleaved by prohormone convertases, generating a TMD-CD fragment with a longer stalk region. When PAM is cleaved by an ␣-secretase-like enzyme on the plasma membrane or in the endocytic pathway, a TMD-CD fragment with a shorter stalk is generated. These TMD-CD fragments can be degraded or cleaved by a ␥-secretase-like enzyme, possibly in the endocytic pathway, generating sf-CD. sf-CD is produced in response to stimulated exocytosis and is more stable in primary pituitary cells than in AtT-20 cells. sf-CD accumulates in the nucleus in a phosphorylation-dependent manner, modulating the expression of genes like Aqp1, which is involved in secretory granule biogenesis.