Here we utilized an unbiased peptidomics approach to identify the endogenous substrates of the peptidase Prep in the mouse CNS (spinal cord, brain and hypothalamus). Earlier work aimed at identifying Prep substrates relied on radioimmunoassays and immunohistochemistry to measure changes in specific peptide levels as a function of Prep activity (47
). For example, substance P, α-melanocyte-stimulating hormone, thyrotrophin-releasing hormone and arginine-vasopressin, are all elevated upon Prep inhibition in several discrete regions of the rat brain, as measured by immunohistochemical methods (16
). Mass spectrometry based peptidomics methods improve on these immunological methods because they are not restricted to known peptides, do not require associated antibodies, and are not subject to the cross-reactivity encountered by some antibodies (49
In our studies, we used S17092 (17
), which we confirmed as a selective Prep inhibitor ( and Supporting Information
), to block Prep activity in the CNS. Comparison of the tissue peptidomes from S17092 treated mice to animals treated with vehicle revealed a number of peptide substrates for the enzyme including the known substrate substance P (14
). Surprisingly, there was no substrate overlap with a prior peptidomics study aimed at identifying Prep substrates in the CNS of rats. This could result from methodological differences such as inhibitor, model organism, duration of inhibition or quantitation strategy. In total, we found changes in 23 distinct peptides, with 15 peptides elevated in the S17092 treated samples (substrates) and 8 peptides elevated in the vehicle sample (products) ().
An overview of these peptides revealed that 9 peptides (~40%) contain post-translational modifications (PTMs) and 8 of these peptides (~35%) are protein fragments derived from regions of proteins known as PRDs. Three of the product peptides in the vehicle sample are shorter fragments of substrates found in the S17092 treated samples. For example, we detect the preproenkephalin peptide, Penk(123–133), in the vehicle treated sample, while we see the corresponding Prep substrate, Penk(114–133), elevated in the S17092 treated samples. We refer to these combinations where we are able to detect a substrate and its corresponding product as a “substrate-product pair”. This provides us with the best evidence that a peptide is an endogenous substrate.
Prep substrates also included bioactive peptides, such as substance P. More generally, the measurement of changes in substance P levels upon Prep inhibition validates our methodology by demonstrating our ability to detect known Prep substrates. Furthermore, novel interactions between Prep and bioactive peptides were also detected. Specifically, the novel Prep substrate Penk(114–133), a fragment of the proenkephalin gene, has recently been shown to stimulate glutamate release when introduced into the brains of rodents (50
). The determination of biological connections between Prep and bioactive peptides, such as Penk(114–133), are valuable because they highlight potential links between Prep and cellular and physiological functions, such as glutamate release in vivo
In addition, the thymosin β4 peptides, Tmsb4(8–22) and Tmsb4(8–25), are precursors of the bioactive peptide Ac-SDKP, which is produced by Prep cleavage of the thymosin β4 (51
). The Ac-SDKP peptide has been implicated in a number of biological processes, including fibrosis, proliferation, and angiogenesis (53
). While thymosin β4 is expressed in the brain, the Ac-SDKP peptide has mostly been studied in the periphery, specifically the cardiovascular system. Thus, the finding of thymosin β4 as a Prep substrate in the nervous system suggests that the Ac-SDKP peptide, and the thymosin β4-Prep-Ac-SDKP pathway in general, should also be studied for functions in the nervous system. This pathway could indicate new roles for Prep in the nervous system.
Next, we wanted to confirm that some of the S17092 elevated peptides were Prep substrates. We specifically targeted peptides that did not fit the typical mold of Prep substrates, which are usually short (<15 residues) with a single proline (12
), because we reasoned that looking at qualitatively different substrates provided the best opportunity to learn something new about the enzyme. Enzyme kinetics with recombinant Prep confirmed that the CGRP(20–37) is a Prep substrate (). In an elegant series of studies, Gorrão and colleagues designed fluorescence resonance energy transfer (FRET) Prep substrates to study factors such as length dependence and subsite specificity of Prep (41
). An exact comparison of results to this work is difficult due to differences in reaction conditions; however, it is clear that CGRP(20–27) is at least two orders of magnitude slower that a 17-residue peptide used by Gorrão and coworkers.
As mentioned, in vitro results are sometimes poor predictors of in vivo substrates, but this comparison does suggest that CGRP(20–37) is probably not a preferred physiological substrate of Prep. Our in vivo data supports this as well, since many of the better substrates return to baseline levels after 4 hours, but CGRP(20–37) is still elevated (). This suggests that the Prep is cleaving better substrates faster than it is cleaving CGRP(20–37) in vivo. Additionally, such poor substrates can sometimes also act as competitive inhibitors, which we demonstrated with CGRP(1–37). While it is possible that CGRP(20–37) is also acting as an inhibitor it is probably not occuring in vivo since CGRP(20–37) levels are up at 4 hours while many of the preferred substrates (e.g., substance P) return to baseline levels indicating a return of Prep activity.
Moreover, these FRET studies (41
), as well as others (57
), have demonstrated that longer peptides are poorer Prep substrates. For example, the aforementioned 17-residue FRET peptide was the worst substrate in a series of peptides ranging from 7 to 17 amino acids. We found similar results when comparing CGRP(1–37) and dsCGRP(1–37), a diserine mutant lacking the loop region found at the N-terminus of CGRP(1–37), to CGRP(20–37), even though all these peptides contain the exact same 18 amino acid cleavage site. These kinetic experiments confirmed the preference of Prep for shorter substrates, and demonstrated that Prep selectivity might allow the enzyme to control CGRP(20–37) without disturbing the full-length CGRP(1–37) (), even though CGRP(20–37) is likely not a preferred endogenous Prep substrate.
Furthermore, experiments with a Prep mutant, PrepD35A-K196A, support the importance of inter domain interactions in providing access to the Prep active site (11
), which is partly responsible for the length dependence seen with Prep (). Since the structures of these peptides are predominantly random coil (Supporting Information
), and the mutant still shows a preference for shorter peptides, these results also suggest the additional possibility that longer peptides are interacting with the enzyme at an undefined exosite. Moreover, since CGRP(1–37) is a poor Prep substrate which still binds to the enzyme, we tested this peptide as a Prep inhibitor and showed that it could inhibit the cleavage of substance P with an IC50 of 2 μM. This provides a mechanism to explain the previous observation that CGRP can enhance substance P activity (44
) (). Without estimating intracellular concentrations of CGRP it is impossible to assess whether this process is physiologically relevant, but this might provide an explanation for in vitro
experiments with exogenously added CGRP (45
). More generally, this experiment highlights the potential interplay between peptides and peptidases in the complex milieu of the cellular environment, providing another factor to consider when studying the mechanism of action of certain bioactive peptides.
One of the most striking contrasts between our results and the previous peptidomics analysis was the number of Prep regulated peptides we identified that contain more than one proline. More specifically, 11 out of 15 (~ 70%) of the substrates we identified had at least two prolines in the sequence, and 7 out of 15 endogenous substrates had more than two prolines, many of them derived from the PRDs of their precursor proteins (). PRDs are found in a number of proteins, including many scaffolding proteins, and mediate protein-protein interactions with SH3 proteins to assemble functional protein complexes (58
). For example, the PRD of dynamin participates in protein-protein interactions between microtubules, Grb2, amphiphysin, and dynamin (i.e., dynamin-dynamin interactions) (58
). Within a PRD there are usually shorter sequences that mediate protein-protein interactions, including the consensus SH3 binding motif, XPXXP (61
). Interestingly, we find an XPXXP motif in a number of the Prep regulated peptides, including Vgf(490–507), Dnm1(207–216), and Syn(673–682), which raises the possibility that these peptides might influence protein-protein interactions, through dominant-negative interactions.
We tested some of the substrates with multiple proline residues as substrates for Prep, including a pair of PRPs. One goal in these experiments was to gain evidence to support or refute an observation from our peptidomics data that suggests Prep shows a preference for individual prolines (). For example, despite having multiple proline residues we only detect a single product for Vgf(490–507) in our vehicle treated peptidomics data, Vgf(490–502), suggesting that this peptide is preferentially cleaved at one of eight proline residues. By contrast, Penk(123–133) is also a Prep product but this peptide contains no prolines indicating that both prolines are cleaved. We tested four peptides in these experiments as substrates for Prep: Penk(114–133), Ac-Hsp12a(2–23), Vgf(24–39), and Vgf(490–507). Based on our in vivo results we would predict that the Ac-Hsp12a(2–23), which contains two prolines, and Vgf(490–507), a PRP, would only be preferentially be cleaved at a single site. Our in vitro experiments corroborated these predications since Prep cleavage occurred predominantly at a single proline residue for both these peptides. By contrast we found that the other two peptides, Penk(114–133) and Vgf(24–39), were processed to similar degrees at multiple sites to generate multiple fragments. Therefore, Prep can show a preference between proline residues in certain peptides, supporting the hypothesis developed from our peptidomics profiling data.
A cursory look at the peptide sequences that were preferentially cleaved at a single site to those peptides that were cleaved at multiple sites did not reveal any sequence elements that might provide a clue into the underlying molecular basis of this selectivity. For example, the two Vgf substrates, Vgf(24–39) and Vgf(490–507), both contain a PPP site, but the Vgf(24–39) peptide is cut at this site while the Vgf(490–507) peptide seems to be refractory to cleavage at this site. It is known that Prep can show some specificity based on charge near the cut site, preferring positively charged peptides at neutral pH (56
). This specificity might provide an explanation for some of these differences, but we do not have enough substrates to identify such a pattern in our data. Furthermore, CD measurements indicated that most of the peptides are predominantly random coil in solution, so structural differences between the peptides cannot contribute strongly to these differences.
In conclusion, our peptidomics results have provided new insights into the biochemistry and biology of Prep in the nervous system. Prep regulated peptides included known substrates such as substance P and thymosin β4, which validates our profiling results, as well as novel substrates, including PRPs. Biochemical experiments tested and confirmed a subset of S17092 elevated peptides as substrates of Prep. These experiments also provided insights into the biochemistry of the enzyme, revealing an unappreciated specificity that enables Prep to preferentially cleave at a single proline in peptides that contain multiple prolines. In addition to these biochemical insights, the discovery of bioactive peptides, or bioactive peptide precursors, regulated by Prep facilitates the development of new hypotheses about the physiological function of Prep. For example, elevated levels of thymosin β4 precursor protein indicates that Prep is controlling the production of the Ac-SDKP peptide in the nervous system, which would link Prep to the regulation of diverse physiological processes such as angiogenesis and proliferation in the CNS. More generally, these results highlight the value of peptidomics approaches in obtaining biochemical and biological insights into the functions of mammalian peptidases.