Macrocyclic θ-defensins suppress tumor necrosis factor-α (TNF-α) shedding by inhibition of TNF-α converting enzyme

Theta-defensins (θ-defensins) are macrocyclic peptides expressed exclusively in granulocytes and selected epithelia of Old World monkeys. They contribute to anti-pathogen host defense responses by directly killing a diverse range of microbes. Of note, θ-defensins also modulate microbe-induced inflammation by affecting the production of soluble tumor necrosis factor (sTNF) and other proinflammatory cytokines. Here, we report that natural rhesus macaque θ-defensin (RTD) isoforms regulate


INTRODUCTION
Mammalian defensins are host defense peptides composed of three structural families, designated as α-, β-or θ-defensins. The three defensin families are genetically related, but are distinguished by amino acid chain length and distinctive tridisulfide motifs that are characteristic of each family (1)(2)(3)(4).
Isolation and characterization of each of the defensin families was the result of studies aimed at identifying antimicrobial substances in granulocytes or epithelial cells implicated in host defense (5,6). Among defensins, θ-defensins are unique in that they are 18-amino acid macrocyclic peptides known to be expressed only in cells of Old World monkeys and are the only cyclic polypeptides known in animals (7).
Multiple θ-defensin isoforms are expressed in neutrophils and monocytes of rhesus monkeys (six; (8)) and olive baboons (ten; (9)), each of which is a macrocyclic octadecapeptide that includes six disulfide linked cysteines (Fig. 1). The sequence diversity of θ-defensins results from the binary combinations of gene-encoded nonapeptides that are ligated head-to-tail to form the θ-defensin backbone ( Fig. 1) (10). Humans and great apes lack θ-defensins due to stop codon interruptions in the signal peptide coding regions of the respective θ-defensin precursors (11).
For example, the prototype θ-defensin RTD-1 reduced lethality in murine models of polymicrobial sepsis, E. coli bacteremia, and SARS coronavirus infection, and the therapeutic effects in each model were associated with significant reductions in tissue proinflammatory cytokine and chemokine levels (14,15). Studies in the mouse SARS coronavirus model strongly implicated hostdirected anti-inflammatory effects of RTD-1 because the peptide had no direct antiviral activity (15). RTD-1 was also effective in reducing pulmonary pathology in murine models of endotoxic lung injury (16) and cystic fibrosis (17) by moderating inflammatory responses. Also, RTD-1 arrested joint inflammation in a rat model of rheumatoid arthritis (RA), pristane-induced arthritis (PIA), an autoimmune disease characterized by dysregulated pro-inflammatory cytokines and erosive joint changes similar to those associated with RA (18).
Parenteral administration of RTD-1 to rats with established PIA rapidly induced arrest of disease progression and resolution of arthritis that correlated with significant reductions in proinflammatory cytokines in joint tissues (in press). sTNF is produced when pro-TNF, a type-II transmembrane protein, is cleaved at the cell surface by TACE (ADAM17) (19)(20)(21). TACE is a membrane-anchored zinc-metalloprotease and is responsible for "shedding" the ectodomain of TNF and many additional cytokines, growth factors, receptors, and adhesion molecules (22,23). Dysregulated TACE activity has been associated with disruption of cytokine homeostasis, elevating levels of TNF in chronic and acute inflammatory diseases including RA, sepsis, and colitis (24)(25)(26)(27)(28)(29) as well as cancer progression (22,30). Inhibition of TACE activity with broad-spectrum metalloprotease inhibitors prevents TNF release from cell surfaces, suppressing levels of sTNF (31-33). In a previous study on the kinetics of RTD-1 inhibition of TNF release by E. colistimulated leukocytes we found that that suppression of sTNF occurred rapidly upon peptide addition to the bacteria-blood incubation mixture (14). We hypothesized that the blockade of sTNF release is mediated by inhibition of proTNF ectodomain shedding.
Thus, we performed studies to evaluate the effects of RTD isoforms on the TNF convertase TACE and the related sheddase, ADAM10.

RESULTS
RTD-1 suppresses TNF release by human blood leukocytes stimulated by diverse TLR agonists. In a previous study we showed that RTD-1 suppressed the release of several proinflammatory cytokines, including TNF, IL-1α, IL-1β, IL-6, IL-8, CCL2, CCL3, and CCL4 by human buffy coat leukocytes stimulated with agonists of TLRs 2, 4, 5, and 8 (14). TNF release was markedly suppressed irrespective of the stimulus (14). In the current study we analyzed the effects of 5 or 15 µM RTD-1 on TNF secretion by human blood leukocytes stimulated with an expanded panel of TLR ligands including agonists of TLRs 1/2, 2, 4, 5, 2/6, 8, and 9. In the absence of peptide, stimulated cells secreted 30-1100 pg/ml of TNF (Fig. 2). Consistent with our previous report, RTD-1 dose-dependently reduced TNF release by agonists for TLRs 2, 4, 5, and 8 and also had similar effects on cells stimulated with ligands for TLRs 1/2, 2/6, and 9 (Fig. 2). These findings, and the rapid blockade of TNF release observed when blood leukocytes were stimulated with E. coli cells in the presence of RTD-1, suggested that the peptide regulates proteolytic release of TNF.

RTD-1 inhibits TNF release by THP-1 cells but does not affect downstream signaling of sTNF in colonic epithelial cells.
We previously showed that RTD-1 dose-dependently suppressed TNF release by LPS stimulated THP-1 monocytes (14). To determine whether RTD-1 pretreatment of THP-1 cells blocked LPS-induced TNF secretion, cells were incubated for 60 min with 5 µM RTD-1 or vehicle, washed, and stimulated with LPS in the presence or absence of 5 µM RTD-1. As shown in Figure 3A, suppression of LPS-induced TNF secretion only occurred if RTD-1 was present when the cells were stimulated, as washout of the peptide prior to LPS stimulation had no effect on TNF release. This result was not due to neutralization of LPS as we showed previously that RTD-1 does not block the endotoxic properties of LPS (14).
Cellular responses to soluble TNF are mediated by paracrine and autocrine signaling, resulting in further TNF release and activation of other downstream inflammatory effects (34). To determine whether RTD-1 affects downstream TNF signaling, we tested whether the peptide alters the response of HT-29 cells to TNF. HT-29 cells express TNF receptor I (TNFRI) and release IL-8 in response to sTNF (35). Consistent with previous reports, TNF-stimulated HT-29 cells induced robust IL-8 release (Fig. 3B). This was unaffected by the co-incubation with RTD-1 (Fig.  3B), and RTD-1 alone had no effect on IL-8 release by HT-29 cells (data not shown). Thus, RTD-1 does not affect TNF signaling by direct neutralization of sTNF nor does the peptide appear to interfere with sTNF/TNFRI signaling in HT-29 cells.
Inhibition of TACE by θ-defensins. RTD-1 blockade of TNF release by human leukocytes stimulated with E. coli cells is extremely rapid (14), and the peptide down-regulates TNF release by leukocytes irrespective of the stimulating TLR ligand (Fig. 2). Based on these findings, we hypothesized that RTD-1 inhibits TNF release by inhibition of its mobilization from the cell surface by its principal convertase, TACE (ADAM17). RTD-1 dose-dependently inhibited recombinant human TACE (rhTACE) cleavage of its fluorogenic substrate (Fig. 4A) with an IC50 = 0.11 µM +/-0.04. Of note, the inhibition by RTD-1 did not require pre-incubation of the peptide and the TACE enzyme. As discussed later, we found that RTD-1 rapidly bound to and inhibited the enzyme. The sigmoidal inhibitory response of RTD-1 was approximately 10-fold less potent than that of marimastat (MRM), a small molecule inhibitor of zinc matrix metalloproteinases (36) (Fig. 4A). A single chain opening in the RTD-1 backbone, producing the S7 analog ( Fig. 1), reduced inhibitory potency >99% (Fig. 4A). We also tested the TACE inhibitory activities of human αdefensins (human neutrophil peptides, HNPs 1-4) that have proinflammatory and/or immune activating properties (37-43). Neither HNP-2 nor HNP-4 inhibited TACE (Fig. 4A). We then evaluated the effects of RTD-1, S7, marimastat, and human α-defensins for their inhibition of TNF release by LPS stimulated THP-1 cells. None of the HNPs, nor S7 was effective in blocking TNF release, whereas RTD-1 and marimastat dosedependently inhibited sTNF release (Fig. 4B), consistent with the TACE inhibitory activity of each compound (Fig. 4A).
In a previous study, we showed that natural RTD isoforms (RTDs 1-5) varied markedly in their inhibition of TNF release by stimulated blood leukocytes or THP-1 monocytes (14). Therefore we evaluated RTDs 1-5 for their relative inhibition of rhTACE.
Each peptide dosedependently inhibited TACE, but with IC50 values that ranged from 50 -265 nM, giving an inhibitory hierarchy of RTD-2 > RTD-5 > RTD-4 > RTD-1 > RTD-3 (  Table 1). These peptides were then evaluated for their inhibition of sTNF release by LPS-stimulated THP-1 cells. The hierarchy of dose-dependent inhibition of sTNF release ( Fig.  4D) was the same as that obtained for TACE inhibition by RTDs 1-5. Indeed, the TACE IC50 values of the RTDs were highly correlated with their TNF-release IC50 values in assays using LPSstimulated THP-1 cells or E. coli-stimulated blood leukocytes (Fig. 5).

-Defensins inhibit TACE activities of THP-1 and HT-29 cells.
To test whether the effects of θ-defensins on rhTACE are replicated with TACE-expressing cells, we incubated RTDs 1-5 with THP-1 macrophages and HT-29 epithelial cells in the presence of a fluorogenic substrate that mimics the cleavage site of membrane-bound TNF, allowing for continuous, real-time measurement of TACE activity in live cells (44,45). After a brief lag, substrate conversion in RTD-free controls was linear for 60 min. RTDs dose-dependently inhibited substrate hydrolysis by both THP-1 macrophages and HT-29 cells (Fig.  6). Nearly maximum inhibition of THP-1 cell TACE was achieved with 5 µM of RTDs 1, 3, and 5 while RTDs 2 and 4 were slightly less inhibitory at this concentration (Fig. 6A). The absence of complete inhibition at the highest RTD levels tested is similar to findings reported for the small molecule TACE inhibitors GM6001 and BB94 (45).
We assessed the effect of RTDs on ADAM10, a TACE-related metalloprotease, by monitoring processing of AP-BTC. Ionomycin stimulation of COS7 cells increased ADAM10 cleavage of AP-BTC by 10-20-fold more than unstimulated cells (Fig. 7C).
RTD-1 dosedependently reduced ionomycin-stimulated ADAM10 activity, but acyclic S7 had little inhibitory activity. HNP-4 enhanced ADAM10 activity by an unknown mechanism that was not investigated further. The effect of θ-defensins on ADAM10 was analyzed further by testing for inhibition rADAM10 by RTDs 1-5. While each RTD inhibited ADAM10 as a function of peptide concentration, (Fig. 7D), the inhibitory potency was lower, and the isoform-specific hierarchy differed, compared to inhibition of RTDs of rTACE and cellular TACE expressed by THP-1 cells (Figs. 4C and 5).
RTD-1 is a fast-binding non-competitive inhibitor of TACE. As noted above, the hypothesis that RTDs inhibit TACE was based in part on the exceedingly rapid blockade of sTNF release by E.
coli-stimulated leukocytes exposed to RTD-1 (14). To further study the kinetics of inhibition, we analyzed the temporal effect of RTD-1 addition to a steady state reaction of fluorogenic substrate cleavage by rTACE. As shown in Fig. 8A, addition of RTD-1 very rapidly inhibited enzyme activity and the degree of blockade was dosedependent. These findings are consistent with rapid binding of RTD-1 to TACE. Additionally, a large molar ratio of RTD-1 (0.05-25 µM) to TACE (2 nM) is required for proteolytic inhibition, indicating that RTD-1 is not a tight-binding inhibitor.
Michaelis-Menten kinetics of RTD-1 inhibition of rhTACE was determined by measuring substrate conversion in reactions with varied substrate and peptide concentrations (Fig.  8B). RTD-1 inhibition of TACE could not be overcome by increasing concentrations of substrate, thus Vmax decreased with increasing RTD-1 concentrations while Km remained unaffected, kinetics consistent with classical noncompetitive inhibition. Best fit modeling of enzyme inhibition kinetics was performed (see Methods) disclosing that RTD-1 is indeed a noncompetitive TACE inhibitor which likely binds to an exosite that modifies interaction of the enzyme with proTNF and related soluble substrates. As the Ki value for a non-competitive inhibitor is the same as its IC50 value, the values reported in Table  1 reflect both the measured IC50 and apparent Ki values for each peptide.

DISCUSSION
ADAMs are key regulators of growth factor signaling, development, tumor progression and inflammation (22,50,51), and among known ADAMs TACE plays a central role in both EGFR signaling and proinflammatory pathways (52). TACE-mediated shedding of EGFR ligands is required for normal organogenesis (53)(54)(55)(56)(57). TACE is also critical in generating sTNF (19)(20)(21), the ectodomain of proTNF, as well as the production of other proinflammatory cytokines and their receptors (58). Regulation of TACE activation is rapid and reversible and recent studies have demonstrated the important role of two membrane proteins, inactive Rhomboid 1 and 2, which selectively regulate activation of TACE in different mouse tissues (29,(59)(60)(61). The current study discloses a previously unknown mechanism of sTNF regulation wherein macrocyclic θdefensins allosterically inhibit TACE. To our knowledge, θ-defensins are the only known endogenous inhibitors of TACE sheddase activity other than tissue inhibitor of matrix metalloproteinases 3 (TIMP3), which has been shown to play a critical role in regulating proTNF shedding (62)(63)(64)(65).
Enzymology experiments demonstrate that θ-defensin inhibition of TACE is rapid, non-tight binding, reversible, and non-competitive. Inhibition of TACE by RTDs explains how RTD-1 is able to suppress sTNF release by blood leukocytes stimulated by an array of TLR agonists (Fig. 2). Moreover, the correlation between potency of θ-defensin isoforms in blocking TACE enzymatic activity and inhibition of sTNF release supports the conclusion that θ-defensins regulate sTNF production by TACE inhibition (Fig. 5). This was further validated using cell-based assays that showed that RTDs inhibit TACE and ADAM10 shedding of the ectodomains of their respective membrane bound substrates.
θ-defensins are expressed at high levels in neutrophils and monocytes of rhesus macaques (8), olive baboons (9), and African green monkeys (unpublished data). Despite the high degree of sequence conservation among RTDs 1-5 (identity in 11 of 18 residue positions), the TACE inhibitory potency of RTDs varied by more than 5fold. Of note, the hierarchy of TACE inhibition of RTDs 1-5 strongly correlated with the suppression of TNF release by each isoform from stimulated THP-1 monocytes and whole blood leukocytes.
θ-defensins are readily measurable in plasma of baboons with experimental E. coli bacteremia, and are released by granulocytes challenged with E. coli in vitro (66) Moreover, RTDs are present in saliva of healthy macaques (unpublished data). In clinical sepsis, high levels of human neutrophil α-defensins are released into the circulation (67). In baboons, θ-defensins are secreted by blood granulocytes stimulated with E. coli in vitro, and θ-defensins are rapidly released into baboon plasma following experimental E. coli bacteremic challenge (66). We hypothesize that θdefensins are endogenous regulators of TACE, ADAM10, and potentially other matrix metalloproteases in Old World monkeys. Since αdefensins stimulate proinflammatory pathways (38)(39)(40)(41)(42)(43) and lack inhibitory activities against TACE or ADAM10, TACE inhibition by θdefensins may underlie the differential susceptibility of humans and other great apes to endotoxin, compared to endotoxin resistance of Old World monkeys that express θ-defensins (68,69).
RTD-1 markedly suppressed mRNA expression and release of TNF, IL-1β, and IL-8 by LPS-stimulated THP-1 macrophages (70). Suppression of transcriptional activation was associated with suppression of NF-κB and MAPK signaling with a concomitant increase in pAkt which negatively regulates these pathways (70). Thus θ-defensins are pleiotropic effectors of host defense and inflammation.
However, unlike their acyclic counterparts, θ-defensins also function as antiinflammatory molecules that moderate proinflammatory stimuli at both transcriptional and post-translational levels.
Given the importance of TACE as a regulator of sTNF, numerous laboratories have sought to identify selective TACE inhibitors for pharmacologic regulation of the enzyme's sheddase activity to antagonize TNF expression in diseases such as rheumatoid arthritis, inflammatory bowel diseases, diabetes, and sepsis. Small molecules evaluated for this purpose have included succinates, hydroxamates, sulfonamides, γ-lactams, β-benzamido compounds, benzothiadiazepine inhibitors, and zinc chelators (31, [71][72][73][74], each designed to competitively target the zinc-containing catalytic site (31). To date none of these programs has produced an approved TACE-targeted drug due to lack of efficacy and/or off target toxicities (27,31,75,76).
Perhaps molecules with structural and functional features embodied in θ-defensins may provide a new avenue for development of TACE inhibitors for treatment of TNF-driven diseases. purified (>98%) (4), and human neutrophil αdefensins 1-4 (HNPs; > 98%) were purified from human buffy coat leukocytes as described previously (77). Peptides were dissolved in 0.01% (v/v) acetic acid or HPLC-grade water, and peptide concentrations were confirmed by coupled liquid chromatography and mass spectrometry (LC-MS) using previously quantified standards.
TNF in cell-free supernatants was determined by ELISA (Life Technologies). For studies using peripheral blood leukocytes, EDTAanticoagulated blood was obtained from healthy adult volunteers, and buffy coat leukocytes were harvested by centrifugation at 200 × g. Cells were washed twice with 3-5 ml of RPMI, counted with a hemocytometer, and suspended at 5×10 5 cells/ml in RPMI + 5% human EDTA plasma. Leukocytes were incubated for 4 h with the indicated concentrations of TLR ligands or 100 CFU/ml E. coli #UCI 9021, and TNF in clarified supernatants was quantified by ELISA.
Half-maximal inhibition concentrations (IC50s) were obtained from fitted curves.
Enzyme kinetics. RTD-1 (0 -960 nM) was incubated with 2.5, 5, 10, and 20 µM Mca-PLAQAV-Dpa-RSSSR-NH2 (320ex/405em) in the presence of 25 ng/ml rTACE and enzyme activity was measured every 30 sec for 90 min. Initial velocity (Vo) was plotted as a function of substrate concentration (Mean +/-SD, n = 3), subjected to Michaelis-Menten kinetics analyses which were compared with best-fit models for competitive, non-competitive, uncompetitive, and mixed modeling inhibition using GraphPad Prism. Inhibition of rTACE by RTD-1 was also determined by addition of 0 -25 µM RTD-1 to steady-state (Vs) TACE-Mca-PLAQAV-Dpa-RSSSR-NH2 reactions in which substrate conversion was monitored for 10 min prior to addition of peptide, and then for 25 min after the addition of the indicated concentration of RTD-1.       . Substrate conversion was monitored continuously for 60 min at 37 °C and plotted as a function of percent product formation rate relative to peptide-free controls. Results are shown as means ± SD of two independent experiments containing 2-3 technical repeats each.

FIGURE 7.
RTDs inhibit cellular TACE and ADAM10 sheddase activities. A-B, RTDs, but not acyclic S7 or α-defensin HNP-4, suppress TACE-mediated ectodomain shedding of TGFα in COS7 cells. Data are expressed as fold TGFα shedding relative to constitutive TGFα release. C, RTD-1, but not S7 or HNP-4, inhibits ADAM10 dependent cleavage of BTC in COS7 cells. Data are expressed as fold BTC shedding relative to constitutive BTC release. Data in panels A-C represent means ± SD of a representative experiment containing 3 technical repeats. D, RTDs inhibit rADAM10 proteolysis of its fluorogenic substrate. Enzyme reactions were performed for 30 min at 37 °C with 1 nM ADAM10 and 10 µM substrate (R&D Systems ES010). Enyzme inhibition is expressed as percent change in substrate conversion rate relative to peptide-free controls. Data represent means ± SD of 2 independent experiments containing 2 technical repeats each.