|Home | About | Journals | Submit | Contact Us | Français|
Proteolysis plays a key role in regulating the levels and activity of peptide hormones. Characterization of the proteolytic pathways that cleave peptide hormones is of basic interest and can, in some cases, spur the development of novel therapeutics. The lack, however, of an efficient approach to identify endogenous fragments of peptide hormones has hindered the elucidation of these proteolytic pathways. Here, we apply a mass spectrometry (MS)-based peptidomics approach to characterize the intestinal fragments of peptide histidine isoleucine (PHI), a hormone that promotes glucose-stimulated insulin secretion (GSIS). Our approach reveals a proteolytic pathway in the intestine that truncates PHI at its C-terminus to produce a PHI fragment that is inactive in a GSIS assay—a result that provides a potential mechanism of PHI regulation in vivo. Differences between these in vivo peptidomics studies and in vitro lysate experiments, which showed N- and C-terminal processing of PHI, underscore the effectiveness of this approach to discover physiologically relevant proteolytic pathways. Moreover, integrating this peptidomics approach with bioassays (i.e. GSIS) provides a general strategy to reveal proteolytic pathways that may regulate the activity of peptide hormones.
Peptide hormones control a number of vital physiological processes (1, 2) and the dysregulation of these signaling pathways can lead to prevalent diseases such as diabetes (3). Understanding the molecular pathways that regulate peptide hormones can therefore provide insight into disease mechanisms and new opportunities for therapeutic intervention. Proteolysis of the hormones angiotensin (4, 5) and glucagon like peptide 1 (GLP-1) (6, 7) modulate the levels and activities of these peptides. Additionally, both these proteolytic pathways have been targeted in the development of drugs (4–6) by using peptidase inhibitors to regulate endogenous hormone levels. The characterization of the proteolytic pathways that cleave other peptide hormones would identify key regulatory mechanisms and may eventually lead to the development of novel therapeutics.
We present a general peptidomics-based approach that relies on measurements of endogenous fragments of peptide hormones to elucidate the proteolytic pathways that cleave these peptides in tissues. We develop this approach by investigating the proteolytic pathways that process the intestinal peptide hormone peptide histidine isoleucine-27 (PHI(1–27)) (8). PHI(1–27) and vasoactive intestinal peptide (VIP) are both products of the VIP gene (9), which is expressed in the gut and other tissues. PHI(1–27) was originally discovered from porcine intestine using an innovative biochemical strategy that detected peptides with C-terminal amides (8, 10), a hallmark of peptides derived from the secretory pathway. Since its discovery PHI(1–27) has been linked to a number of biological functions including prolactin secretion (11), glucose stimulated insulin secretion (GSIS) (12), and the regulation of physiological glucose tolerance (9), which led to our interest in the regulation of this hormone.
Our peptidomics approach revealed that proteolysis of PHI(1–27) in the intestine occurs through C-terminal processing. This result differed from initial experiments with intestinal lysates, which identified proteolytic pathways that truncate PHI(1–27) at its N- and C-termini, to demonstrate the difficulty in using in vitro proteolysis experiments to predict physiologically relevant pathways. Through the integration of peptidomics studies with bioassays we can identify the potential impact of proteolysis on the bioactivity of a peptide. In this case, C-terminal proteolysis of PHI(1–27) abrogates the activity of this peptide hormone in a glucose-stimulated insulin secretion (GSIS) assay (13), which suggests that the intestinal proteolytic pathway discovered through peptidomics may be involved in the inactivation of PHI(1–27) in vivo. These results demonstrate the ability of peptidomics to discover relevant proteolytic pathways involved in the processing, and possibly the regulation, of peptide hormones.
Wild type (strain C57BL/6, DPP4+/+) mice used in this study were either purchased (Jackson Labs, Bar Harbor, ME) or taken from a breeding colony. The DPP4−/− mice used in this study (a generous gift from Dr. Didier Marguet) have previously been described (7) and are on a C57BL/6 background. Mice in these studies were not littermates from het × het crosses, but were obtained from separate colonies of DPP4−/− and DPP4+/+ mice. Animals were kept on a 12-h light, 12-h dark schedule and fed ad libitum. For intestinal tissue collection, animals were euthanized with CO2, their tissue dissected, flash frozen in liquid N2, and stored at −80 °C. All animal care and use procedures were in strict accordance with the standing committee on the use of animals in research and teaching at Harvard University and the National Institute of Health guidelines for the humane treatment of laboratory animals.
The first three inches of the mouse intestine extending from the stomach (duodenum) were dounce-homogenized in 1 × phosphate-buffered saline (PBS) and the homogenates were then centrifuged at 1,000 × g at 4 °C for 5 min to remove cellular debris. The supernatant from this spin was transferred to thick-walled centrifuge tube and then centrifuged at 100,000 ×g at 4 °C for 45 min using a TLX ultracentrifuge (Beckman Instruments). The resulting supernatant was transferred to a new eppendorf tube and used as the soluble fraction. The pellet was washed twice with 1 × PBS and then suspended in 100 µL of 1 × PBS, which was used as the membrane fraction. The protein concentrations of these lysates were determined using a Bradford assay. All lysates were diluted using 1 × PBS to a final concentration of 1 mg/mL for all subsequent experiments.
The concentrations of the peptide stocks were determined by amino acid analysis (Microchemistry Facility, Harvard University). Peptides (100 µM) were incubated with soluble and membrane intestinal lysates (1 mg/mL) at 37 °C for 15 min. These reactions were then quenched using 8 M GndHCl and desalted using a Zip-Tip (Millipore) prior to LC-MS or MALDI analysis.
Lysate samples were analyzed using a nano flow LC (Nano LC-2D, Eksigent Technologies) system coupled to a linear ion trap mass spectrometer (LTQ, ThermoFinnigan). The analytical column (Self-pack picofrit column, 75µm ID, New Objective) was packed 15 cm with 3µm C18 (Magic C18 AQ 200A 3U, Michrom Bioresources Inc). The trap column was obtained pre-packed from New Objective Inc. (Integrafrit sample trap, C18 5µm, 100µm column ID). The samples were trapped at an isocratic flow rate of 2µl/min for 10 minutes and eluted at a flow rate of 300 nl/min via a mobile phase gradient of 5 – 40% B in 145 min (mobile phase A: 0.1% formic acid in water, mobile phase B: 0.1% formic acid in acetonitrile). The peptides were detected in the positive mode and the LC/MS/MS experiment was performed in a data-dependent acquisition (DDA) mode. The mass range for data acquisition was set from m/z 400–1600. The MS/MS data were collected for the most intense peak from the MS chromatogram. Dynamic exclusion was set for 30s, exclusion size list was set to 200 and the normalized collision energy for CID was 35%. The capillary spray voltage was set to 2.5 kV. Peptide identification was performed using SEQUEST with differential modification of methionine to its sulfoxide. The uniprotmus_frc.fasta mouse database, concatenated to a reversed decoy database, served to estimate a false discovery rate (FDR). Peptides were accepted within 1 Da of the expected mass, meeting a series of custom filters on ScoreFinal (Sf), −10 Log P, and charge state that attained an average peptide FDR of < 2% across data sets. Manual inspection of spectra, FDR calculation, and protein inference were performed in Proteomics Browser Suite 2.23 (ThermoFisher Scientific). The peak areas served to quantify the different peptides generated during the incubation with the soluble and membrane lysates.
MALDI spectra of the peptides were obtained with a Waters MALDI micro MX™ (Waters MS Technologies, Milford, MA) operated in reflectron positive mode. Operating conditions were: source voltage, 12 kV; matrix suppression delay, 800 amu; pulse voltage, 1.9 kV; reflectron voltage, 5.2 kV; laser firing rate, 5 Hz. Instrument control, spectral acquisition and processing were performed with MassLynx software (version 4.1). Samples were desalted by a Zip-Tip, mixed with a solution of peptides (0.5 µL) on the stainless steel MALDI target together with 0.5 µL of matrix (saturated α-cyano-4-hydroxycinnamic acid in 50% acetonitrile) and dried under ambient conditions. The peptide identification were also confirmed by LC-tandem MS experiments as described above.
The protease inhibitors used were e-64 (L-trans-Epoxysuccinyl-leucylamideo(4-guanidino)butane) (10 µM) and iodoacetamide (1 mM) for cysteine proteases; EDTA (ethylenediaminetetraacetic acid, 1 mM), o-phenanthroline (1 mM) as general metalloprotease inhibitors; captopril (1 mM) and enalapril (1 mM) as ACE inhibitors; LAF-237 (1 µM) as a DPP4 inhibitor; phosphoamidon (10 µM) as neprylsin inhibitor; pepstatin A (10 µM) for aspartyl proteases; PMSF (phenylmethylsulfonyl fluoride, 1 mM) and DFP (diisopropyl fluorophosphonate, 1 mM) as a general serine protease inhibitor. PHI(1–27) (100 mM) was added along with these inhibitors to membrane lysate (1 mg/mL, diluted with 1× PBS) in a total volume of 40 µL. The incubation was allowed to proceed at 37 °C for 15 min prior to the addition of 40 µL of 8 M GndHCl to quench the reaction. Samples were then desalted using a Zip-Tip and subsequently analyzed by MALDI (qualitative) and LC-MS (quantitive) assay.
Tissue peptide isolation and fractionation followed the recently developed protocol (14). Briefly, frozen intestines were placed in 500 µL of water and boiled for 15 minutes to inactivate any residual proteolytic activity prior to tissue homogenization. The aqueous fraction was separated and saved, and the tissue was dounce-homogenized in ice-cold 0.25% aqueous acetic acid. The aqueous fraction and the homogenate were combined and centrifuged at 20,000 × g for 20 min at 4 °C. The supernatant was then sent through a C18 Sep Pak cartridge (HLB 1cc; 30 mg, Oasis), which was washed thoroughly with water (10 mL) to desalt the sample. The peptides were then eluted with 1 mL of 70:30 H2O/ACN and concentrated under vacuum using a speedvac prior to fractionation by strong cation exchange (SCX).
SCX was performed using a PolySULFOETHYL A™ column (200 × 2.1mm, 5 µm, 300 Å; PolyLC INC.) connected to an Agilent Technologies 1200 series LC. All runs were operated at 0.3 mL/min. The SCX buffers consisted of: A) 7 mM KH2PO4, pH 2.6, 25% ACN (vol/vol); B) 40 mM KCl, 7 mM KH2PO4, pH 2.6, 25% ACN (vol/vol); C) 100 mM KCl, 7 mM KH2PO4, pH 2.6, 25% ACN (vol/vol); D) 600 mM KCl, 7 mM KH2PO4, pH 2.6, 25% ACN (vol/vol). Prior to the SCX runs, all samples (N = 4) were dissolved in 100 µL buffer A (100 µL injections). A step-gradient was applied that included 60 min with Buffer A, 40 min with Buffer B, 40 min with Buffer C, and 40 min with Buffer D, with 20 min transitions between the different buffer conditions. Fractions were collected separately for each of the different buffer conditions (e.g., a buffer A fraction, a buffer B fraction, and so on). All four salt fractions (A, B, C, and D) were then applied to a C18 Sep Pak cartridge, washed with water (25 mL) to desalt the samples, and then eluted with 1 mL of 70:30 H2O/ACN and concentrated using a speed vac. The peptides were dissolved in 0.1% aqueous formic acid (75 mg tissue/40 µL), normalized according to the original tissue weight, prior to LC-MS analysis.
Fractionated intestinal samples were analyzed using a nano flow LC (Nano LC-2D, Eksigent Technologies) system coupled to a linear ion trap mass spectrometer (LTQ, ThermoFinnigan). The analytical column (Self-pack picofrit column, 75εm ID, New Objective) was packed 15 cm with 3µm C18 (Magic C18 AQ 200A 3U, Michrom Bioresources Inc). The trap column was obtained pre-packed from New Objective Inc. (Integrafrit sample trap, C18 5µm, 100µm column ID). The samples were trapped at an isocratic flow rate of 2µl/min for 10 minutes and eluted at a flow rate of 300 nl/min via a mobile phase gradient of 5 – 40% B in 227 min (mobile phase A: 0.1% formic acid in water, mobile phase B: 0.1% formic acid in acetonitrile). The peptides were detected in the positive mode and the mass range for data acquisition was set from m/z 400–1600. The data were collected in Full MS mode (N = 4) for quantitation and then in Top 6 MS2 mode (N = 2) for sequencing (14). For the Top 6 analysis dynamic exclusion was set for 30s, exclusion size list was set to 200 and the normalized collision energy for CID was 35%. The capillary spray voltage was set to 2.5 kV. We utilized an algorithm written in-house that reveals related MSMS spectra, (MuQuest, Harvard Proteomics Browser Suite). To analyze the intestinal proteome to search for PHI(1–27) fragments we used the MSMS data from the in vitro lysate experiments. MuQuest is then applied to compare the in vitro MSMS spectra with those of the in vivo data set to determine which in vitro MSMS spectra (i.e. which peptides) are present in the in vivo samples. The output files are filtered based on charge state, mass to charge values, and statistical scores.
Automated (PS3™ Protein Technology, Inc.) FMOC solid-phase peptide synthesis (SPPS) was carried out with Rink Amide MBHA resin for PHI(3–27) and PHI(1–27), and with an aspartic acid loaded Wang resin for PHI(1–22). Cleavage from the resin was accomplished using a standard FMOC cleavage conditions (TFA, water, mercaptoethanol) and successful synthesis was assessed by MALDI-TOF (Waters) measurements of the crude peptide mixture. The crude peptides were purified by RP-HPLC (Shimadzu) using a C18 column (150 mm × 20 mm, 10 µm particle size, Higgins Analytical). Mobile phase A consisted of 99% H2O, 1% acetonitrile, and 0.1% TFA and mobile phase B consisted of 90% acetonitrile, 10% H2O, and 0.07% TFA. The HPLC gradient proceeded from 20–50% B over 40 min. HPLC fractions were analyzed by MALDI-TOF (Waters) using a-cyano-4-hydroxycinnamic acid as the matrix to isolate fractions containing the pure peptide, which were then combined and lyophilized.
A plasmid containing the mouse DPP4 gene was obtained from Open Biosystems and the extracellular domain (residues 37–760) was subcloned into the EcoRI and XhoI site of a modified pFastBac vector (Invitrogen). The final construct contains a baculovirus gp64 signal peptide followed by a His6 tag fused to the coding sequence corresponding to residues 37–760 of DPP4. Recombinant baculovirus was generated by transposition using the Bac-to-Bac system (Invitrogen). Mouse recombinant DPP4 was expressed using a baculovirus expression system with Sf9 cells in liquid culture and purified by Ni(II) –affinity chromatography.
PHI(1–27) (300 µM) was incubated with recombinant DPP4 (200 nM) in PBS. The degradation products were monitored by MALDI at 30 min, 1 h, and 2h time intervals.
PHI(1–27) (300 µM) was incubated with DPP4+/+ (0.1 mg/mL) and DPP4−/− (0.1 mg/mL) plasma diluted in PBS. The degradation products were monitored by MALDI at 30 min, 1 h, and 2h time intervals.
Insulin release in primary mouse islets was examined using the batch release method as previously described4. Peptides (GLP1, PHI1–27, PHI3–27, and PHI1–22) were directly added at 100 nM final concentration to 1 ml of Krebs assay buffer (13) containing basal (1.67 mM) or stimulatory (16.7 mM) glucose concentrations. Islets were incubated for one hour in the Krebs assay buffer at 37 °C. The supernatant was then collected for insulin measurement using an insulin ELISA kit (Crystal Chem Inc., IL, USA) according to the manufacturer’s instructions for non-serum/plasma samples using half the amount of sample than suggested. The insulin concentration was calculated by using KaleidaGraph 4.0 software (Synergy Software, PA, USA) with 4-parameter sigmoidal logistic curve fitting.
This assay was performed as a service with Shanghai ChemPartner Co., Ltd. Breifly, 4 × 104 cells from a reporter cell line (GLP-1R/CRE-HEK293, clone 9–5) were plated into a 96-well plate and then incubated overnight in a CO2 incubator at 37 °C. In parallel, a 10-point serial dilution of PHI(1–27) and GLP-1 were prepared from a DMSO stock with concentrations ranging from 0.2–2 × 10−10 mM, and 0.5 uL of these stock solutions were transferred to the cell plates. The cells and peptides were then incubated for 6 hours in a CO2 incubator at 37 °C. Steady-Glo assay Kit (from Promega, Cat. No.E2520, Lot. No.256974) was then prepared according to the manufacturer’s instructions and 100 uL of the Steady-Glo reagent was added to each well and then mixed for 5 minutes to induce cell lysis. The luminescence of each well was then read using a Flexstation 3 with an integration time of 500 ms. The data are plotted by normalizing the luminescence response and fitted to the dose-response logistic equation using KaleidaGraph 4.0 to generate an EC50 curve.
In vitro experiments with synthetic peptide hormones and recombinant enzymes are commonly used in an attempt to identify a proteolytic pathway that cleaves the peptide hormone (16, 17). This approach relies on using the known substrate preferences of an enzyme to identify candidate substrates from the pool of known peptide hormones. For example, the hormone arginine-vasopressin (AVP), which contains a single proline residue, was tested as a prolyl endopeptidase (Prep) substrate (Figure 1A). AVP is an excellent in vitro Prep substrate (17), but subsequent tissue measurements indicate that AVP is not a Prep substrate in vivo (18). This example highlights the shortcomings of in vitro assays that neglect important features of physiological regulation (e.g. competitive pathways and localization effects) when trying to predict endogenous proteolytic pathways.
To create improved approaches in vitro and in vivo experiments were combined to elucidate physiologically relevant proteolytic pathways (19). This integrated approach relies on in vitro lysate experiments to generate a list of candidate peptide fragments, followed by immunoassays (i.e., ELISA, RIA, and EIA) of tissues to test whether these fragments are present in vivo (Figure 1B). The presence of a peptide fragment in tissues is then used to infer the physiological relevance of a particular proteolytic pathway. For example, incubation of the bioactive form of the insulinotropic hormone GLP-1, GLP-1(7–36), with human plasma produced an inactive N-terminal truncated fragment, GLP-1(9–36) (19, 20). Next, an effective, albeit labor-intensive, strategy that combines HPLC fractionation of plasma with a radioimmunoassay (RIA) of the resulting fractions (19) was used to confirm the physiological relevance of this proteolytic pathway by characterizing GLP-1(9–36) as the major endogenous GLP-1 fragment (19) (Figure 1B).
The determination of relevant proteolytic pathways and their impact on the bioactivity of a peptide hormone is a pivotal step in defining a role for proteolysis in peptide hormone regulation. With GLP 1(7–36), for example, subsequent experiments, which relied on the prior identification of the endogenous proteolytic pathway, established dipeptidyl peptidase 4 (DPP4) as the enzyme that cleaves and inactivates this hormone in vivo (7). Moreover, this finding prompted the development of DPP4 inhibitors as new class of anti-diabetic drugs (6). While useful at defining these pathways through the identification of peptide hormone fragments, targeted immunoassays suffer from shortcomings, including the need for specialized reagents (antibodies) (19), cross-reactivity (21), and can only be used to confirm, not discover, endogenous peptides. These challenges, in particular the need for antibodies, limit the rate at which peptide hormone proteolysis can be studied and indicate the necessity for a better approach to detect physiologically relevant bioactive peptide fragments.
We reasoned that a liquid chromatography-mass spectrometry (LC-MS)-based peptidomics method (14, 22, 23) could be used to identify fragments of peptide hormones in tissues and circumvent some of the challenges associated with immunoassays (Figure 1C). In particular, peptidomics does not require specialized reagents (i.e. species specific antibodies) and will not suffer from cross-reactivity due to the specificity of mass spectrometry. Our peptidomics-based approach, which combines in vitro lysate experiments with in vivo peptidomics, should provide a straightforward, general, and rapid method for the discovery of the proteolytic pathways that cleave peptide hormones in vivo. We develop and test this strategy through investigation of the intestinal proteolysis of peptide histidine isoleucine (PHI(1–27)) (8, 10), a peptide hormone with insulinotropic activity (12) similar to that of GLP-1(7–36).
Full-length PHI(1–27) was incubated with the soluble and membrane fractions derived from intestinal lysates to identify where the majority of PHI(1−27)-proteolytic activity is located. After incubation of synthetic PHI(1–27) with lysates for 15 minutes, the remaining amount of PHI(1–27) in each sample was quantified using the ion intensity of the resulting PHI(1–27) peak in the LC-MS chromatogram. This experiment revealed that the majority of PHI(1–27) proteolytic activity is associated with the membrane fraction (Supporting Information). This finding is consistent with the idea that peptide hormones, such as PHI(1–27), are primarily processed extracellularly after release from secretory vesicles (24).
Next, these samples were analyzed by mass spectrometry to identify the peptide fragments that were being generated. A matrix-assisted laser desorption/ ionization (MALDI)-MS spectrum of these membrane lysate samples revealed a number of new ions upon incubation of PHI(1–27) with membrane lysates. These ions corresponded to the major PHI(1–27) fragments being produced, and included PHI(3–27), PHI(1–22), and PHI(3–22) (Figure 2A). All of these structural assignments were confirmed by analysis of these samples by tandem LC-MS experiments (LC-MSMS).
These particular PHI(1–27) fragments provide a wealth of information about the proteolytic pathways that are active in these intestinal tissue lysates. For example, PHI(3–27) indicates that N-terminal proteolysis of PHI(1–27) occurs through the removal of a dipeptide, which is similar to what is seen for GLP-1 in plasma. Similarly, C-terminal processing of PHI(1–27) is also taking place to generate the PHI(1–22) fragment. These experiments also reveal a fragment resulting from the N- and C-terminal processing PHI(3–22). Together the data from these in vitro intestinal lysate experiments result in a model for PHI(1–27) proteolysis that degrades this peptide at both the N- and C-terminus to afford the PHI(3–27) and PHI(1–22), respectively (Figure 2B).
Class selective protease inhibitors (e.g., metallo-, cysteine, serine, and aspartyl) were added to these lysate experiments to inhibit the production of PHI(1–27) fragments (Supporting Information). After incubation, the impact of these inhibitors on PHI(1–27) processing was determined by LC-MS analysis of the samples and quantitation of the relevant ions. These experiments reveal that the serine protease inhibitors PMSF and DFP strongly inhibited the production of PHI(3–27), which indicates that N-terminal processing is being carried out by a serine protease (Figure 2C). Similarly, the metalloprotease inhibitor 1,10-phenanthroline impacted C-terminal processing of PHI(1–27) by reducing the production of PHI(1–22) (Figure 2D).
Next, we tried to identify candidate peptidases that process PHI(1–27) in these intestinal lysates by looking for enzymes that cut between amino acids 2–3 and 22–23 of the PHI(1–27) sequence. A search of the MEROPS database (25) using the cleavage sites we identified on PHI(1–27) and the enzyme class inhibitor suggested the matrix metalloproteases (MMPs) and neprilysin (Nep) as candidate C-terminal processing enzymes, and dipeptidyl peptidases as a potential N-terminal cleaving enzymes. Prior work investigating the in vitro breakdown of peptide histidine methionine (1–27) (PHM(1–27)), the human homolog of PHI(1–27), in serum implicated DPP4 in the proteolytic processing of this peptide (20). To assess the relative contributions, if any, of these enzymes we utilized chemical inhibitors and tissues from animals lacking these enzymes.
We found no impact of the MMP (26) inhibitor (GM6001) and Nep (27) inhibitor (phosphoramidon) on C-terminal processing of PHI(1–27) by measuring PHI(1–22) production in lysates treated with these inhibitors (Supporting Information). These experiments highlight the difficulty in trying to use the known substrate preferences of enzymes to predict enzyme-substrate relationships. Considering that there are over 500 proteases in the mammalian genome (28) and that many of these enzymes need to be characterized, the identification of these enzyme-substrate relationships will continue to be a challenge.
In some cases, however, especially when the cleavage site is very distinctive (e.g. N-terminal dipeptide truncation) the processing of a peptide can immediately reveal a candidate enzyme. Indeed, the specific DPP4 inhibitor, LAF-237 (29) reduced the production of PHI(3–27) by greater than a 100-fold, to indicate that DPP4 was responsible for N-terminal PHI(1–27) processing (Figure 2D). Furthermore, this conclusion garnered additional support from intestinal lysate experiments with tissues from DPP4 null (DPP4−/−) mice (7), which also showed greatly reduced production (>100-fold) of PHI(3–27) (Figure 2D). Together these in vitro lysate experiments predict that PHI(1–27) undergoes N- and C-terminal processing in the intestine, and that the N-terminal processing pathway is mediated by DPP4.
As mentioned, in vitro assays are often unreliable in predicting cleavage in vivo because important aspects of endogenous regulation can be disrupted during lysate preparation, such as spatial organization and competitive pathways. In prior work, immunoassays were developed to identify these physiologically relevant fragments of peptide hormones. However, the shortcomings of immunoassays, including the need for specialized reagents and unwanted cross reactivity (21), prompted us to implement a state-of-the-art LC-MS-based peptidomics platform (14, 22, 23) to characterize fragments of PHI(1–27) in the mouse intestine.
The processing of these samples is identical to our previous peptidomics experiments (14, 22, 23). The key step in this workflow is the heating of tissues prior to homogenization, which inactivates all proteolytic activity to maintain sample integrity for the subsequent LC-MS analysis. Indeed, an assay of DPP4 activity after tissue homogenization shows that this enzyme is completely inactive (14). We analyzed the LC-MS data to identify PHI(1–27) and its fragments in the intestinal peptidome of wildtype (WT) mice. To simplify this complex sample, the intestinal peptidome was subjected to multidimensional fractionation. Specifically, an offline strong cation exchange (SCX) step was applied to the sample prior to LC-MS and yielded four salt fractions of reduced complexity (14). These samples were then analyzed by LC-MSMS, which provided the dataset that were searched for PHI(1–27) fragments.
Our initial attempts at identifying PHI(1–27) fragments in this dataset using standard proteomics analysis software (Sequest) (30) were unsuccessful and we did not see a single PHI(1–27) fragment. During a proteomics experiment every protein generates multiple peptides that can be used to detect the protein, and the identification of multiple peptides per protein is also used as a measure of confidence in the protein identification. One of the challenges faced in all peptidomics experiments is that each peptide results in a single MSMS spectrum that can be used to identify the peptide. As a result, the standard proteomics analysis software are not optimized for peptidomics and we found that direct determination of PHI(1–27) and its fragments using the standard proteomics parameters for Sequest was not efficient. While lowering some of the thresholds optimized for proteomics (in particular the total intensity count) generates additional MSMS data that can be searched for the identification of PHI(1–27) peptides, these changes increase the time it takes for data analysis so much that it is no longer feasible. Additionally, even when these parameters are changed it still results in poor identification of the additional MSMS spectra due to a higher frequency of MSMS spectra with low information content.
To circumvent this limitation, we utilized an algorithm written in-house that reveals related MSMS spectra, (MuQuest, Harvard Proteomics Browser Suite). MuQuest compares MSMS spectra between different data sets (or within the same data set) to identify the same or related peptides in two (or more) different samples based on similar or related MSMS spectra. It is based on the principle that MSMS spectra derived from peptides of identical or shared sequence, independent of charge state or modifications, should have a subset of identical product ions. If one data set has a series of known peptides (i.e. standards) then it becomes an easy search for those peptides in a second sample using MuQuest. To analyze the intestinal proteome to search for PHI(1–27) fragments we used the MSMS data from the in vitro lysate experiments. Sequest identified PHI(1–27) and its fragments PHI(1–22) and PHI(3–27) with high statistical confidence in this in vitro sample to ensure that we were looking for the correct peptides. MuQuest is then applied to compare the in vitro MSMS spectra with those of the in vivo data set to determine which in vitro MSMS spectra (i.e. which peptides) are present in the in vivo samples. The output files are filtered based on charge state, mass to charge values, and statistical scores.
The analysis of the intestinal peptidome samples revealed the presence of PHI(1–27) and PHI(1–22) (both +5 charged peptides at pH < 3 found in the 600 mM KCl salt fraction) as the predominant PHI(1–27) fragments (Figure 3A). PHI(1–22) was one of the major fragments generated during the in vitro intestinal lysate experiments, and the detection of this peptide in vivo indicates that a C-terminal pathway for processing PHI(1–27) is active in the intestine. More generally, this example also highlights the utility of MuQuest in peptidomics analysis of complex biological samples, and indicates that this should be the standard approach to these experiments.
Interestingly, these experiments did not detect PHI(3–27) in the intestine, one of the other major in vitro fragments that results from DPP4 processing of PHI(1–27) (Figure 2A). We confirmed that the absence of a detectable PHI(3–27) ion was not due to the inability of this peptide to ionize efficiently by measuring the ionization of synthetic PHI(1–27), PHI(1–22), and PHI(3–27) (2 nmoles each) spiked into intestinal lysate samples. All three peptides produced peaks of comparable intensity that indicates that we would have detected PHI(3–27) if it was naturally occurring in the intestine (Supporting Information). The absence of a PHI(3–27) fragment in these samples demonstrates the importance of making in vivo measurements when elucidating physiologically relevant proteolytic pathways. Moreover, these experiments highlight the value of peptidomics experiments since we did not have to develop a specific assay to measure PHI(1–27), but simply relied on our standard peptidomics protocol (14) combined with an optimized data analysis approach using MuQuest (Harvard Proteomics Browser Suite, v3.2).
The absence of PHI(3–27) indicates that DPP4 does not process PHI(1–27) in the intestine. This differs from the conclusion reached from in vitro lysate experiments, which clearly showed a role for DPP4 in the N-terminal processing of PHI(1–27) to produce PHI(3–27). This is an important finding because it highlights the necessity for in vivo measurements to truly understand the physiological pathways that breakdown a peptide hormone. In our previous comparative peptidomics studies DPP4 substrates were discovered by identifying peptides that were elevated in DPP4−/− mice (14, 23). To confirm that DPP4 is not regulating PHI(1–27) we quantitated PHI(1–27) and PHI(1–22) levels in the intestines from wildtype (DPP4+/+) and DPP4−/− mice using our comparative peptidomics approach (14). If DPP4 is processing PHI(1–27) or PHI(1–22) levels in vivo the prediction would be that PHI(1–27) or PHI(1–22) would be elevated in the absence of DPP4. According to these comparative peptidomics experiments, which showed similar levels of PHI(1–27) and PHI(1–22) in DPP4+/+ and DPP4−/−, DPP4 does not regulate intestinal levels of PHI(1–27) (Figure 3B).
In total, these in vivo peptidomics experiments establish that PHI(1–27) undergoes C-terminal processing to produce PHI(1–22) (Figure 3C). The data also indicates that PHI(3–27), the product of a DPP4-mediated pathway, is not produced in vivo, which differs from the conclusion obtained from in vitro experiments. As mentioned, this discordance highlights the importance of complementing in vitro experiments with in vivo analysis, since in vitro experiments cannot account for important aspects of in vivo biochemistry.
Based on the in vitro results with tissue lysates we were surprised by the absence of PHI(3–27), the DPP4 cleavage product, in vivo. The discordance between in vitro measurements and in vivo analysis could be explained by a number of factors. First, DPP4 might not be able to access PHI(1–27) due to differential localization. Specifically, DPP4, which is found on the extracellular portion of the plasma membrane (31), might not have access to PHI(1–27) if a majority of this peptide is stored in secretory vesicles. Another possibility is that DPP4 is regulated by a posttranslational modification (PTM) or protein-protein interaction (PPI) that would impact DPP4 activity; however, DPP4 is not known to have any PTMs that influence the activity of this enzyme and known DPP4-PPIs are not thought to influence the activity or localization of this peptidase (32).
Another possibility we considered is that substrate competition might be playing a role in the substrate selectivity we are seeing. We previously identified a number of proline-containing peptides as DPP4 substrantes in the intestine through comparative peptidomics (14). Specifically, we found 22 peptides in total that were elevated in the intestine of DPP4−/− mice, which indicates that they are DPP4 substrates. Of these 22 candidate substrates only a single peptide has a penultimate alanine (i.e. H2N-Xaa-Ala-peptide) while the remainder all are proline-containing peptides (i.e. H2N-Xaa-Pro-peptide). Importantly, other penultimate alanine-containing peptides could be found in this sample, but these were unchanged between DPP4+/+ and DPP4−/− mice. This is consistent with in vitro experiments that show that proline-containing peptides are superior DPP4 substrates (23, 33).
The observation that DPP4 is cleaving a number of proline-containing substrates in the intestine suggests that the difference between our in vitro observations and our in vivo measurements might be due to differences in the available substrate pool. The preparation of lysates necessarily removes all endogenous peptide substrates. When PHI(1–27) is added in these in vitro lysate experiments it is clearly processed by DPP4 to remove a dipeptide from the N-terminus of this hormone. In this experiment the only substrate available to DPP4 is exogenously added PHI(1–27). In vivo there is no evidence for DPP4 processing of PHI(1–27), either through the detection of PHI(3–27) or the regulation of PHI(1–27) or PHI(1–22) levels when comparing DPP4+/+ and DPP4−/− intestinal tissue. We hypothesize that the presence of a large number of proline-containing peptides in the intestine leads to a situation where substrates compete for the DPP4 active site. As a result, this competition leads to preferential cleavage of proline-containing peptides, which are the best substrates of DPP4 (34), instead of alanine-containing peptides (Figure 4A).
To test this possibility we incubated PHI(1–27) with membrane lysates and added increasing concentrations of a proline-containing DPP4 substrate meprinβ (21–41) (mepβ(21–41)), which is a bona fide DPP4 substrate in the kidney(14, 23) (Table 1). The addition of mepβ(21–41) inhibits DPP4 processing of PHI(1–27) in a dose-dependent manner to demonstrate that proline-containing substrates may slow the cleavage of alanine-containing substrates in a competitive manner in tissue lysates (Figure 4B). Admittedly, we do not know the ratio of proline-to-alanine containing substrates in vivo, however, this experiment in conjunction with our in vivo comparative peptidomics experiments supports a model where proline-containing peptides may be inhbiting the processing of the alanine-containing peptides. Moreover, these results are also supported by measurements in the kidney where we used SCX and peptidomics to identify 67 proline-containing DPP4 substrates but not a single alanine-containing peptide substrate was found (14, 23). More generally, this experiment suggests another possibility for differences between in vitro and in vivo experiments—the presence of competitive substrates—that impact our ability to predict endogenous substrates from in vitro experiments.
Based on the data obtained from the in vivo peptidomics experiments, where we detect PHI(1–22) as a physiologically relevant fragment of PHI(1–27), we conclude that the proteolytic pathway that processes PHI(1–27) relies on C-terminal degradation (Figure 3C). We wondered whether this pathway could influence any of the known bioactivities of this hormone. PHI(1–27) has been shown to stimulate prolactin and insulin secretion (11, 12, 35). To assess the functional impact of the newly discovered intestinal pathway for C-terminal processing of PHI(1–27), we tested the PHI fragments PHI(1–27), and PHI(1–22) in a glucose-stimulated insulin secretion (GSIS) assay (13) using mouse pancreatic islets. GLP-1 has been reported to stimulate insulin secretion in these experiments and we used this peptide a positive control (36).
Mouse pancreatic islets were freshly isolated from whole pancreases for these experiments (13). These islets were then incubated with glucose and GLP-1, PHI(1–27), or PHI(1–22). The conditioned media from these islets were then isolated and insulin levels were measured. In these experiments, PHI(1–27) is able to promote insulin secretion to demonstrate the insulinotropic activity of this hormone (Figure 5A). Importantly, PHI(1–22) is inactive in this GSIS assay at the same dose as PHI(1–27) to indicate that the proteolytic processing that we see in vivo can regulate the biological activity of this peptide hormone (Figure 5A). Thus, this newly discovered C-terminal degradation pathway that process PHI(1–27) in vivo attenuates the bioactivity of this peptide (Figure 5B).
Interestingly, we tested and found that PHI(1–27) does not activate the GLP-1 receptor (GLP-1R) (Supporting Information), despite a large degree of N-terminal homology between GLP-1(7–36) and PHI(1–27) (Table 1). The inactivity of PHI(1–27) with the GLP-1R indicates that PHI(1–27) and GLP-1 would operate differently in a biological setting and suggests that subsequent efforts to define the peptidase(s) that regulate PHI(1–27) would be of benefit in studying the regulation and activity of this hormone, because it might lead to the discovery of an alternative pathway for regulating physiological insulin levels.
This work demonstrates the utility of peptidomics as a general method for elucidating the endogenous pathways involved in the proteolysis of peptide hormones. We discovered an intestinal proteolytic pathway that converts PHI(1–27) into PHI(1–22) and, in the process, abolishes the activity of this peptide in a GSIS assay. Obtaining this information required an optimized computational workflow that relied on MuQuest for the discovery of peptide fragments in complex biological matrices. Since we detected some, but not all, of the same peptides we saw in vitro, we believe that our method can also be improved to increase the sensitivity and quantitation by using multiple reaction monitoring (MRM) methods (37). Compared with other standard immunoassay approaches, peptidomics offers some distinct advantages that accelerate the rate at which endogenous proteolytic fragments of bioactive peptides can be discovered. Furthermore, integrating this peptidomics strategy with bioassays, such as GSIS, will enrich our understanding of bioactive peptide function and regulation. Looking forward, the use of this peptidomics strategy with additional bioactive peptides, and improved methods for characterizing the peptidases that mediate these proteolytic pathways, will enrich our understanding of proteolytic regulation of bioactive peptides and might eventually provide new targets for drug development. In this case, we will begin efforts to identify the enzyme(s) responsible for the C-terminal processing of PHI(1–27) in the intestine, which may eventually lead to a better understanding of the regulation and bioactivity of this peptide hormone in vivo.
We thank Horatio R. Thomas for technical assistance and David R. Liu for his constructive insight.
This work was supported by the Mary Fieser Postdoctoral Fellowship (A.D.T.), Searle Scholar Award (A.S.), Burroughs Wellcome Fund Career Award in the Biomedical Sciences (N.N.D. and A.S.), National Institutes of Health Grant 5 R01 DK078081 (N.N.D.), and National Institutes of Health Grants 1DP2OD002374 (A.S.).
SUPPORTING INFORMATION PARAGRAPH
Bar-graph showing the increased degradation of PHI(1–27) by membrane fraction (Figure S1). Bargraph showing the inhibition of c-terminal degradation of PHI(1–27) by phenanthroline (Figure S2). Intestinal peptidomics demonstrate the ability to detect PHI(1–27) and PHI(1–22), but not PHI(3–27), in vivo (Figure S3). GLP-1 receptor assay showing that GLP-1 but not PHI(1–27) is a GLP-1 receptor agonist (Figure S4). Supplemental materials may be accessed free of charge online at http://pubs.acs.org.