Application of a chemical probe to detect neutrophil elastase activation during inflammatory bowel disease

Neutrophil elastase is a serine protease that has been implicated in the pathogenesis of inflammatory bowel disease. Due to post-translational control of its activation and high expression of its inhibitors in the gut, measurements of total expression poorly reflect the pool of active, functional neutrophil elastase. Fluorogenic substrate probes have been used to measure neutrophil elastase activity, though these tools lack specificity and traceability. PK105 is a recently described fluorescent activity-based probe, which binds to neutrophil elastase in an activity-dependent manner. The irreversible nature of this probe allows for accurate identification of its targets in complex protein mixtures. We describe the reactivity profile of PK105b, a new analogue of PK105, against recombinant serine proteases and in tissue extracts from healthy mice and from models of inflammation induced by oral cancer and Legionella pneumophila infection. We apply PK105b to measure neutrophil elastase activation in an acute model of experimental colitis. Neutrophil elastase activity is detected in inflamed, but not healthy, colons. We corroborate this finding in mucosal biopsies from patients with ulcerative colitis. Thus, PK105b facilitates detection of neutrophil elastase activity in tissue lysates, and we have applied it to demonstrate that this protease is unequivocally activated during colitis.


Results
Selectivity and potency of PK105b against purified serine proteases. We synthesized an analogue of our previously published PK105 probe 16 , referred to as PK105b, in which the polyethylene glycol (PEG) linker was omitted and sulfoCy5 was used in place of the unsulfonated version (Fig. 1). This change allowed for a more direct comparison to our Cy5-V-DPP probe 13 , which included sulfoCy5 and no PEG. The only difference between the two probes compared in this study is the specificity region: Cy5-V-DPP contains a single P1 valine residue, while PK105b contains a tetrapeptide consisting of non-natural amino acids (Nle(OBzl)-Met(O)2-Oic-Abu) (Fig. 1).
We tested the reactivity of the PK105b against recombinant serine proteases, and compared its potency with Cy5-V-DPP. After incubation of proteases with increasing concentrations of probes for 30 minutes, the proteins were resolved by SDS-PAGE and binding was detected by in-gel fluorescence. Both probes clearly labeled NE and PR-3 in a concentration-dependent manner ( Fig. 2A), though PK105b was more potent than Cy5-V-DPP. PK105b also clearly labeled trypsin, another serine protease, while trypsin binding by Cy5-V-DPP was negligible ( Fig. 2A).
Using a commercially available fluorogenic substrate probe, AAPV-AMC, we compared the ability of the two probes to inhibit recombinant NE. Like Sivelestat, a commonly used NE inhibitor, PK105b immediately inhibited rNE activity at 0.1 µM while Cy5-V-DPP only partially inhibited rNE activity at 1 and 10 µM (Fig. 2B). These results indicate that PK105b binds to rNE more rapidly and more potently than Cy5-V-DPP. As Cy5-V-DPP and Selectivity and potency of PK105b in normal tissue lysates. We compared the ability of PK105b and Cy5-V-DPP to detect protease activity in lysates prepared from mouse bone marrow. Two proteases were labeled by PK105b (Fig. 2C), and the 25-kDa protein was confirmed to be NE by immunoprecipitation with an NE-specific antibody (Fig. 2D). NE was specifically labeled by Cy5-V-DPP, but with a much lower potency than PK105b. We also examined the reactivity of the probes in lysates prepared from mouse pancreas, a tissue rich in (C,E) Lysates from mouse bone marrow or pancreas were incubated with increasing concentrations of Cy5-V-DPP or PK105b and binding was assessed by in-gel fluorescence. (D,F) PK105b-labeled lysates were immunoprecipitated with antibodies for neutrophil elastase, pancreatic elastase (PE) or trypsin 3 (Try3), as indicated. In (F), gain settings for pulldowns were enhanced in order to observe faint bands (right panel). Note: Cy5-V-DPP and PK105b labeling are depicted at equal gain settings, set independently for each sample type.
www.nature.com/scientificreports www.nature.com/scientificreports/ serine proteases. With PK105b, we observed strong labeling of 25-kDa proteins that was much more apparent than labeling by Cy5-V-DPP (Fig. 2E). Immunoprecipitation of PK105b-labeled lysates revealed that the target proteases consisted of a combination of NE, pancreatic elastase (PE), and trypsin 3 (Try3, also known as PRSS3 or mesotrypsin; Fig. 2F). Thus, PK105b exhibits a dramatic improvement in labeling efficiency in protein lysates over Cy5-V-DPP. In addition to the improved binding kinetics demonstrated above, the improved labeling profile of PK105b may also be due in part to enhanced stability of the compound afforded by the non-natural specificity region, which may be less prone to degradation than Cy5-V-DPP.

Validation of PK105b in inflamed tissues.
To determine the effectiveness of PK105b to measure NE activation in inflamed tissues, we used a mouse model of Legionella pneumophila infection. L. pneumophila infection may result in Legionnaire's disease, a common cause of community or hospital-acquired pneumonia, and is associated with high neutrophil infiltration in the lung 18,19 . PK105b labeling was significantly increased in lysates prepared from infected lung tissues compared to uninfected lungs (Fig. 3A,C). The identity of the major 25-kDa species was confirmed to be NE by immunoblotting (Fig. 3B,D,E) and immunoprecipitation ( Fig. 3F) with an NE-specific antibody. We also confirmed that PK105b binding was mediated by the DPP warhead and the specificity region, as the labelling could be competed by pre-treatment with PK101 14,15 , a biotinylated (non-fluorescent) analogue of PK105 (Fig. 3G). As neutrophils are the predominant source of NE during L. pneumophila infection, we also examined PK105b labelling in neutrophils from infected-lungs, which were sorted by flow cytometry at >97% purity (Fig. S1). Within this population of cells, we observed specific labeling of NE by PK105b (Fig. 3H).
We next determined the utility of PK105b to detect NE activation in a cancer setting, which is also rich in neutrophils [20][21][22] . Specifically, we utilized a mouse orthotopic xenograft model of oral squamous cell carcinoma in which human cancer cells (HSC-3) were injected into the tongue 23 . In this context, we observed clear labeling of a 25-KDa species in tumor tissues, but not normal tongue tissues (Fig. 4A,C). This species coincided with the size of mature NE as determined by immunoblot (Fig. 4B,D,E) and immunoprecipitation ( Fig. 4F) with an NE-specific antibody. Several other high-molecular weight species were abundantly labeled by PK105b in these lysates ( Fig. 4A), but they have not yet been identified. Nonetheless, they are likely to be binding to PK105b through the DPP and specificity region, and not Cy5, as binding could largely be competed with PK101, the non-fluorescent PK105 analogue (Fig. 4G).

Application of PK105b to measure NE activation in experimental colitis. Having validated
PK105b in inflamed mouse tissues, we next applied this probe to investigate NE activation during acute experimental colitis induced by trinitrobenzenesulfonate (TNBS). As expected, mice exhibited loose stools, delayed defecation, weight loss (Fig. S2A), and colon shortening (Fig. S2B). We also observed damage to the mucosa by histological evaluation, as well as edema and inflammatory infiltrate (Fig. S2C). We analyzed colon lysates for NE activation by PK105b labeling and measurement of in-gel fluorescence. In proximal colons from healthy and inflamed mice, we observed little PK105b labeling. By contrast, in the distal region of inflamed colons, which is most affected in the TNBS model, we observed clear labeling of a 25-kDa protein (Fig. 5A). This band was virtually absent in distal colons of healthy mice that received vehicle instead of TNBS (Fig. 5A). We confirmed the identity of the protease to be NE by immunoprecipitation with an NE-specific antibody (Fig. 5D). All of the PK105b bands could be competed by PK101, the non-fluorescent PK105 analogue (Fig. 5E).
We also transferred the fluorescent gels to nitrocellulose membranes in order to immunoblot the samples for total NE expression. In all proximal colons and in healthy distal colons, we observed bands at 37 and 25 kDa (Fig. 5B). In the TNBS-treated distal colons, a new band appeared just below the 25-kDa protein. Only the lower species was labeled by PK105b, as revealed by overlay of the Cy5 fluorescence ( Fig. 5C) and immunoprecipitation (Fig. 5D). To verify that the appearance of this smaller NE species was not an artefact of probe labeling, we immunoblotted inflamed distal colon samples in the presence and absence of PK105b. The smaller species was detected regardless of the presence of PK105b (Fig. 5F). Taken together, these data suggest that NE is subject to trimming in inflamed regions of the colon that permits its activation and thus its reaction with the PK105b probe.
For comparison, we also tested our previous NE probe, Cy5-V-DPP, in distal colon lysates and labeling of the 25-kDa species was barely distinguishable from the background (Fig. 5G). Thus, PK105b is clearly superior to Cy5-V-DPP for its ability to detect NE activity in tissue lysates. Both probes exhibit binding to several species in the 50-75-kDa range (Fig. 5A), and future proteomics assays will be required to determine their identity. Furthermore, we investigated secreted proteases found in the lumen of the colon (either luminal flush or in fecal pellets) with PK105b (Fig. 6A). In both samples, we observed two labeled proteases at 25 kDa. Immunoprecipitation confirmed low levels of NE in these samples, with pancreatic elastase and trypsin 3 being the predominant species (Fig. 6C). Nonetheless, NE activity could be clearly delineated by PK105b in lysates from colon tissues.
Application of PK105b to measure NE activation in mucosal biopsies from IBD patients. To translate our findings in mouse colitis to human disease, we examined PK105b labeling in mucosal biopsies from patients ( Table 1). As in mice, we observed a significant increase in labeling in samples from patients with active UC compared to healthy individuals brought in for routine colonoscopy screening (Fig. 7A,D). In contrast to mice, where we observed a single 25-kDa species at labeled by PK105b, three species were labeled in human mucosal lysates, with the smallest form having the most activity. The banding pattern resembled that which was observed with recombinant human NE ( Fig. 2A and in refs 24,25 ). We confirmed these bands to be NE by immunoprecipitation with an NE-specific antibody (Fig. 7G), and they could be competed with PK101 (Fig. 7H).
Furthermore, when the same samples were immunoblotted for total NE expression, we observed NE bands in the healthy tissue at 37 and 25 kDa (Fig. 7B). UC tissues, however, displayed an additional doublet that was smaller than the 25-kDa species. The most active species, as indicated by PK105b labeling, corresponded to these www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
While PK105 was selective for NE in purified neutrophils, its ability to label NE in tissue lysates, which contain many cell types and proteases, had not been established. We tested PK105b, a close analogue of PK105, in bone marrow, pancreas, inflamed lungs, oral cancer, and human and mouse colitis tissues and fecal samples. Due to the covalent nature of PK105b, we could verify that NE was clearly targeted by this probe. We also observed that PK105b exhibited cross-reactivity with other serine proteases such as pancreatic elastase, proteinase-3, trypsin 3 www.nature.com/scientificreports www.nature.com/scientificreports/ and other unknown proteases. This cross-reactivity was dependent on the tissues examined, with greater selectivity observed in purified lung neutrophils, lung tissue, and human colon biopsies and more cross-reactivity in pancreas, fecal samples, and cancer tissue. We confirmed that all PK105b binding was mediated by the DPP warhead in combination with the specificity region, and not due to non-specific binding via the Cy5 fluorophore, as all labeling could be blocked by pre-treatment with PK101, a non-fluorescent analogue of PK105. www.nature.com/scientificreports www.nature.com/scientificreports/ While PK105b clearly has limitations, it is advantageous over traditional substrate probes such as AAPV-p-nitroanilide and BODIPY-FL-elastin, as these tools do not bind their targets covalently and the extent of their cross-reactivity in various contexts cannot readily be determined. Additionally, PK105b was advantageous over our previous diphenylphosphonate probe, Cy5-V-DPP 13 . In competition assays with the fluorogenic substrate, AAPV-AMC, PK105b reacted with rNE much more rapidly and potently than Cy5-V-DPP. This is most likely mediated by the improved non-natural specificity region in PK105b, which may bind more tightly within the NE active site. PK105b may also have enhanced sensitivity in tissue lysates due to increased stability, as its non-natural peptide sequence may be more resistant to degradation by other proteases than valine. PK105b is therefore one of the most effective tools available to study NE activation.
We applied PK105b to investigate NE activation during mouse colitis. In inflamed colons, we observed labeling of active NE, which was significantly increased compared to control tissue. Probe-labeled NE corresponded to the most mature species of NE detected by immunoblotting, which was not present in healthy tissue. We corroborated these findings in human colitis using mucosal biopsies from patients with UC, unequivocally demonstrating for the first time that NE is activated in IBD. Thus, PK105b is well suited for detection of both mouse and human NE during IBD and will be a valuable resource for future investigation of NE function. Furthermore, it may have utility as a diagnostic agent for IBD and for validating target engagement and efficacy of neutrophil elastase inhibitors in preclinical development as therapeutic agents.

Experimental Procedures
Probe synthesis and characterization. Refer to Supplemental Materials for information on the synthesis of PK105b. Cy5-V-DPP 13 and PK101 14 were synthesized in house as described previously.
Recombinant protease labeling. Recombinant proteases (500 ng) were diluted in 20 µl of phosphate-buffered saline (PBS): human neutrophil elastase (Elastin Products Company), porcine pancreatic trypsin type II-S (beta trypsin; Sigma), and human proteinase-3 (Sigma). PK105b or Cy5-V-DPP (0, 0.1, 0.5 or 1 µM) was added from a 100x DMSO stock, and reaction was carried out at 37 °C for 30 minutes. Proteins were solubilized in 4x sample buffer (40% glycerol, 200 mM Tris-Cl [pH 6.8], 8% SDS, 0.04% bromophenol blue, 5% beta-mercaptoethanol), boiled and resolved on a 15% SDS-PAGE gel under reducing conditions. Probe labeling was detected by scanning the gel for Cy5 fluorescence on a Typhoon 5 flatbed laser scanner (GE Healthcare). Detailed protocols for ABP application are available in ref. 26 . Ex vivo tissue labeling. Bone marrow was obtained by flushing tibias and femurs from healthy C57BL/6J mice with PBS. Cells were washed and resuspended in PBS prior to sonication on ice. Tissues were lysed by sonication on ice in PBS (10 µl/mg tissue), and supernatants were cleared by centrifugation at 21,000 g for 10 min at 4 °C. Total protein (60 µg, as measured by BCA assay, Pierce) was aliquoted in a total volume of 20 µl PBS, and probe labeling and SDS-PAGE was carried out as above. Where indicated, PK101 (10 µM) was added for 10 minutes prior to the addition of PK105b.
Western blotting. Fluorescent gels were transferred to nitrocellulose membranes and blotted using the Turbo Blot system (BioRad). Membranes were blocked using Li-Cor Odyssey blocking buffer diluted by 50% with PBS containing 0.05% Tween 20. Sheep anti-mouse neutrophil elastase/ELA2 (1:1000; R&D AF4517) was incubated overnight at 4 °C. Secondary antibody (goat-IR800, 1:5000; LiCor) was incubated for one hour at room temperature. Binding was detected by scanning with the IRLong filter on the Typhoon 5.