Tumor Necrosis Factor-α Trimer Disassembly and Inactivation via Peptide-Small Molecule Synergy

Abstract

Aberrant signaling by tumor necrosis factor-α (TNFα) is associated with inflammatory diseases that can be treated with engineered proteins that inhibit binding of this cytokine to cell-surface receptors. Despite these clinical successes, there is considerable interest in the development of smaller antagonists of TNFα-receptor interactions. We describe a new 29-residue α/β-peptide, a molecule that contains 3 β-amino acid residues and 3 α-aminoisobutryic acid (Aib) residues, that displays potent inhibition of TNFα binding to TNFα receptor 1 (TNFR1) and rescues cells from TNFα-induced death. The complement of non-proteinogenic residues renders this α/β-peptide highly resistant to proteolysis, relative to all-α analogues. The mechanism of inhibitory action of the new 29-mer involves disruption of the trimeric TNFα quaternary structure, which prevents productive binding to TNFα receptors. Unexpectedly, we discovered that peptide-induced trimer disruption can be promoted by structurally diverse small molecules, including a detergent commonly used during selection procedures. The discovery of this synergistic effect provides a new context for understanding previous reports on peptidic antagonists of TNFα-receptor interactions and suggests new avenues for future efforts to block signaling via proteins with an active form that is oligomeric, including other members of the TNFα family.

Graphical Abstract

INTRODUCTION

The cytokine tumor necrosis factor-α (TNFα) has diverse biological activities, including promotion of inflammatory responses, regulation of immune function, and induction of apoptosis.^1–4 Misregulated TNFα signaling is associated with diseases such as rheumatoid arthritis and inflammatory bowel disease, which can be treated by inhibitors of the interaction between TNFα and its receptors (TNFRs).^{4, 5} TNFα antagonists in clinical use include etanercept (soluble TNFR2-Fc fusion protein), infliximab and adalimumab (two anti-TNFα antibodies).^{4, 5} These protein drugs bind to the bioactive form of TNFα, a non-covalent homotrimer, and occlude the receptor-binding site.⁶

Despite their clinical utility, large protein drugs targeting TNFα suffer from several disadvantages, including complex and costly production and low tissue penetration. To overcome these limitations, many groups have sought TNFα antagonists that are smaller than the current protein drugs. Approaches have included selections from libraries based on mini-protein scaffolds, such as those derived from Z domains,⁷ DARPins⁸ and ubiquitin,⁹ or from libraries based on even smaller linear or cyclic peptide scaffolds.^10–13 The non-peptide small molecule SPD304 (MW = 548) is a modest inhibitor of TNFα-TNFR1 interaction (IC₅₀ = 22 μM, ELISA),¹⁴ but the mechanism of inhibition by this compound has been a subject of debate.^{15, 16}

Recent efforts have identified medium-length peptides that display significant TNFα antagonism. Phage display selections by Luzi et al.¹² led to a series of peptide inhibitors containing internal chemical crosslinks. Fluorescence polarization data suggested K_d = 5 nM for binding to TNFα of a fluorophore-bearing derivative of the bicyclic 19-mer ACPPCLWQVLCGGSGSGSG in which the three Cys side chains are crosslinked by reaction with 1,3,5-trisbromomethylbenzene. The 19-mer itself inhibited TNFα-induced death of L929 murine fibrosarcoma cells, although activity depended on experimental conditions. Physical characterization indicated that this peptide induces dissociation of the TNFα trimer, a process that requires many hours to reach completion. A co-crystal structure showed the 19-mer peptide bound to a TNFα dimer.

Complementary approaches to polypeptide inhibitors of TNFα signaling have been based on the Z domain mini-protein scaffold, which adopts a three-helix-bundle tertiary structure.¹⁷ Libraries generated by randomizing side chains presented from the composite surface displayed by two of the three helices have been screened to identify Z domain variants that bind to TNFα.⁷ We subsequently showed that a 58-mer Z domain identified in this way (Z-TNFα, Figure 1) could be rationally converted to a 38-mer “mini-Z domain” containing an internal disulfide and several unnatural subunits; these modification were made with the goal of increasing resistance to proteolytic degradation.¹⁸ The most effective α/β-peptide (α/β-TNFα−2, Figure 1) contained five β-amino acid residues and two non-proteinogenic α-amino acid residues (Aib). This 38-mer miniZ α/β-peptide was comparable to the original 58-mer Z domain in potency for inhibiting binding of TNFα to TNFR1 in ELISA experiments with purified proteins.¹⁸

Figure 1.

(A) Primary sequences of three-helix α-peptide Z-TNFα and two-helix α- and α/β-peptide derivatives. [Helix3] = SQSANLLAEAKKLNDAQAPK. Non-proteinogenic residues are indicated by colored circles. Each cysteine is engaged in an intramolecular disulfide bond. (Note that Z-TNFα and α/β-TNFα−2 are previously reported compounds; they are named as in the previous study.¹⁸ New peptides reported in this study are given labels 1, 2, and 3.) (B) Structures of a generic α-residue, the Aib residue (green), a generic β³-residue (teal), and the cyclic β-residues ACPC (X) and APC (Z) (orange).

Here we describe a new 29-mer α/β-peptide that is comparable to the previous 38-mer and the original full-length Z domain in its ability to inhibit TNFα-TNFR1 association, but is far superior to these longer peptides in resistance to proteolysis. The new 29-mer serves as a potent antagonist of TNFα-induced death of mouse fibrosarcoma cells under specific assay conditions based on the ability of this peptide to disrupt the trimeric TNFα quaternary structure. Our studies reveal that diverse small molecules can increase the kinetics of the peptide-mediated trimer disruption process. This finding suggests the possibility of future designs in which trimer-destabilizing and dimer- or monomer-stabilizing moieties are combined in a single agent.

RESULTS AND DISCUSSION

Minimization of a peptide antagonist containing unnatural amino acid residues

The previously reported 38-mer α/β-TNFα−2 (Figure 1) inhibits the association of purified TNFα with a purified soluble form of TNFR1 with logIC₅₀ = −8.4 ± 0.4 M (IC₅₀ = 4 nM), as measured via ELISA (Table 1, Figures S1–S2). α/β-TNFα−2 was derived from the 58-residue Z-TNFα, which was discovered by Jonsson et al. via phage display, with selection based on binding to TNFα.⁷ α/β-TNFα−2 lacks the C-terminal helix of Z-TNFα and contains instead a disulfide intended to stabilize a helix-loop-helix tertiary structure that retains the TNFα-recognition surface of Z-TNFα.^19–21 In our ELISA, Z-TNFα displays logIC₅₀ = −8.41 ± 0.09 M (IC₅₀ = 3.9 nM) for inhibition of TNFα-TNFR1 binding. α-Peptide 1, which corresponds to the first two helices of Z-TNFα with five helix-stabilizing substitutions²⁰ but lacks a disulfide, is a very weak inhibitor of TNFα-TNFR1 association (logIC₅₀ = −6.1 ± 0.1 M, IC₅₀ = 900 nM). The analogous α-peptide that contains the disulfide, 2 (logIC₅₀ = −8.1 ± 0.4 M, IC₅₀ = 9 nM), is comparable to Z-TNFα as an inhibitor. The sites of unnatural residue incorporation that generated α/β-TNFα−2 from α-peptide 2 were chosen with the intent of avoiding changes in the phage-optimized surface⁷ that presumably engages TNFα. α/β-TNFα−2 matches both 2 and Z-TNFα in strong inhibition of TNFα-TNFR1 association.

Table 1.

LogIC₅₀ values for selected α- and α/β-peptides tested for inhibiting TNFα-TNFR1 interaction in ELISA, where 2 nM TNFα and peptides are pre-incubated with 0.05% TWEEN20 for 3 hours at room temperature (RT) or pre-incubated at 37 °C for 24 hours. LogIC₅₀ values are given as the mean ± standard deviation (SD) for at least two independent experiments. For each measurement, experimentally determined SD values were rounded to one significant digit, and associated mean value was then rounded to the same decimal place. Half-life values represent the susceptibility of 40 μM α- or α/β-peptide to cleavage by 10 μg/mL proteinase K in TBS, pH 7.5 at RT.

Peptide	TNFα-TNFR1 inhibition by ELISA, logIC50 (M)
	+ 0.05% TWEEN20 (3 h, RT)	No additive (24 h, 37 °C)	Proteinase K t1/2 (min.)
Z-TNFα	−8.41 ± 0.09	−8.74 ± 0.06	3.3
1	−6.1 ± 0.1	-	-
2	−8.1 ± 0.4	-	0.19
α/β-TNFα−2	−8.4 ± 0.4	-	57
3	−8.02 ± 0.05	−7.9 ± 0.2	400

One of our primary motivations for incorporating non-proteinogenic residues into α/β-TNFα−2 is to increase resistance to proteolytic degradation relative to comparable α-peptides, but our previous study¹⁸ did not evaluate proteolytic susceptibilities of this family of compounds. Here, we used proteinase K, an aggressive enzyme that cuts at many positions within a typical polypeptide,²² to evaluate peptide susceptibility to degradation. Under our experimental conditions, the half-life of Z-TNFα in the presence of proteinase K was 3.3 min (Table 1, Figure S3), which shows that the native three-helix bundle tertiary structure offers little protection against proteolysis. α-Peptide 2 was even more susceptible to degradation (t_1/2 = 0.19 min). α/β-TNFα−2 displayed t_1/2 = 57 min, which represents a 17-fold improvement relative to the full-length Z-TNFα and a 300-fold improvement relative to the analogue comprised entirely of proteinogenic residues (2).

We sought to develop new inhibitors that have decreased size and enhanced resistance to proteolysis relative to α/β-TNFα−2 while maintaining efficacy in blocking TNFα-TNFR1 association. Our iterative approach to this goal is outlined in the Supporting Information (See Note 1, Figures S1–S2, and Table S1).²³ The effort culminated in the discovery of α/β-peptide 3, which contains 29 residues, 9 fewer than α/β-TNFα−2. This new α/β-peptide is half the length of the parent peptide Z-TNFα. Six of the residues in 3 are non-proteinogenic, three Aib and three β residues. α/β-Peptide 3 (logIC₅₀ = −8.02 ± 0.05 M, IC₅₀ = 9.6 nM) is similar to α/β-TNFα−2 and Z-TNFα in potency as an inhibitor of TNFα-TNFR1 binding, but 3 is significantly more resistant to the action of proteinase K (t_1/2 = 400 min; 7-fold improvement relative to α/β-TNFα−2, 120-fold improvement relative to Z-TNFα, and 2100-fold improvement relative to two-helix α-peptide 2).

Peptide-mediated inhibition of TNFα-TNFR1 is very sensitive to assay conditions

In cell culture, the addition of exogenous TNFα initiates apoptosis in many cell types, and inhibiting TNFα binding to cell-surface receptors can rescue cells from TNFα-mediated death.^{10, 11, 24, 25} Based on our ELISA results, we expected that Z-TNFα and our two-helix α/β-peptide analogues would inhibit TNFα-mediated cell death of WEHI-13VAR mouse fibrosarcoma cells.²⁶ However, in preliminary studies, neither Z-TNFα nor α/β-TNFα−2 could rescue cells in our assay.

The unexpected outcome of our initial cell-based assays caused us to return to the ELISA format for careful evaluation of assay conditions. These studies revealed that potent inhibition of TNFα-TNFR1 association by Z-TNFα, α/β-TNFα−2 or α/β-peptide 3 required one of two conditions: (1) the presence of TWEEN20 (0.05% (w/v)); or (2) prolonged co-incubation of TNFα with the inhibitor peptide at 37 ^oC prior to ELISA (Figures 2, S2, and S4). IC₅₀ values measured under either of these two conditions were similar (Table 1). TWEEN20 (Figure 3) is a common component of ELISA buffers; this nonionic surfactant is widely used to suppress non-specific binding interactions in ELISAs, affinity-based selections and other biomolecular procedures that involve protein recognition. We observed that a relatively brief (3 hr) co-incubation of inhibitor and TNFα in the presence of TWEEN20 resulted in ELISA data shown in Table 1; however, no inhibition of TNFα-TNFR1 association was detected via ELISA after a peptide-TNFα co-incubation of similar duration in the absence of TWEEN20 (Figure 2A). On the other hand, peptide-TNFα co-incubation for 24 hr in the absence of TWEEN20 led to ELISA results that matched those obtained after short co-incubation in the presence of TWEEN20 (Figure 2B,C).

Figure 2.

Representative TNFα-TNFR1 competition ELISA dose-response data under the indicated conditions. For panel A, each data point represents a measurement from an individual well on the plate, and the curve represents fit to a 3-parameter competition model. For panels B and C, each point represents the mean ± SEM of duplicate wells on the plate, and curves represent fits to a 3-parameter competition model. RT = room temperature.

Figure 3.

Chemical structures of small molecules examined in this study.^{14, 30}

Hypothesis: surfactants can facilitate inhibition of TNFα-TNFR1 association by enhancing peptide-mediated TNFα trimer disassembly

The observations summarized above led us to hypothesize that inhibition of TNFα-TNFR1 association by Z-TNFα or one of the α/β-peptides we developed is subject to a substantial kinetic barrier in the absence of TWEEN20. A kinetic barrier could arise if the mechanism of peptide inhibition involves destabilizing the trimeric form of TNFα, which is required for signal transduction, by binding preferentially to a monomeric or dimeric form of TNFα. Dissociation of the TNFα trimer is known to be intrinsically slow,²⁷ which would explain why prolonged co-incubation of peptide and TNFα is required for inhibition in the absence of TWEEN20. Poiesi et al. demonstrated that TWEEN20 enhances the rate of TNFα dissociation;²⁷ therefore, we hypothesize that TWEEN20 alters TNFα trimer structural dynamics in a manner that promotes peptide action.

The synergistic, rate-enhancing role proposed for TWEEN20 and our compounds raised an important question: how could phage selection have identified Z-TNFα when the reported experimental protocol⁷ involved exposure of pooled phage to TNFα for only one hour? We noticed that the reported selections were conducted in the presence of 0.05% TWEEN20, which could have induced peptide-mediated TNFα trimer disruption during the selection process. Several other Z-domain-derived peptides that bind to TNFα have been identified via phage display with TWEEN20 included in the buffer used for these selections.^{7, 28} We designed, synthesized and evaluated two-helix α-peptides based on several such examples^{7, 28, 29} (Figures S5 and S6), and in each case inhibition of TNFα-TNFR1 association was observed via ELISA after a short pre-incubation of peptide and TNFα only if TWEEN20 was included (See Note 2 in the Supporting Information).

In order to identify molecular features required for the proposed synergistic effect on peptide-mediated inhibition of TNFα-TNFR1 association, we examined a small set of surfactants with at least partial structural similarity to TWEEN20 (Figure 3).³⁰ TWEEN20 itself contains a linear alkyl unit (dodecanoyl group) as the hydrophobic portion and a branched oligo-ethylene glycol unit as the hydrophilic portion. A related detergent, TWEEN80, has the same branched headgroup with a longer hydrocarbon unit that contains one double bond. TNFα-TNFR1 inhibition ELISA demonstrated that a short pre-incubation of Z-TNFα with TNFα in the presence of TWEEN80 led to potent inhibition of TNFα-TNFR1 association, although the maximal level of inhibition was somewhat reduced compared to TWEEN20 (Figure S7B). Brij L4, comprised of a dodecyl hydrophobic unit and a linear tetra-ethyleneglycol polar group, enabled peptide-mediated TNFα-TNFR1 disruption as well (Figure S7C). Note that the TWEEN20, TWEEN80 and Brij L4 surfactants were tested above their critical micelle concentrations.^{31, 32} In contrast, Pluronic F68, a surfactant comprised exclusively of oligo-ethylene glycol and oligo-propylene glycol segments with a much higher critical micelle concentration (>10% at room temperature),³³ displayed very little ability to synergize with peptides to disrupt the TNFα-TNFR1 interaction at 0.05–0.5% (Figure S7C).

To gain further insight on the role of TWEEN20 in peptide-mediated inhibition of TNFα-TNFR1 association, we evaluated the TNFα oligomerization state in the absence and presence of an inhibitor by size-exclusion chromatography (SEC) using a UPLC instrument (Figures 4 and 5). When 1 μM (50 μg/mL) TNFα trimer was dissolved in PBS, the SEC chromatogram showed only a single peak (Figure 4A), which we attribute to the trimeric form. Compared to the proteins in a standard mixture, the TNFα trimer runs at an apparent molecule weight (MW) of ~30–40 kDa, consistent with prior observations for TNFα by SEC analysis.^{9, 34–36} No change in this chromatogram was observed over a one hour incubation in PBS at room temperature. When the solution included 0.05% TWEEN20, the TNFα trimer peak again did not change over one hour at room temperature (Figure 4B), which suggests that the surfactant does not directly induce TNFα dissociation. Co-incubation of TNFα and α/β-peptide 3 for one hour at room temperature led to a slight decrease in area for the trimer peak coinciding with the emergence of a small peak corresponding to a smaller species with apparent MW <17 kDa (Figure 4C), which we tentatively assign as the TNFα monomer. However, after this one-hour incubation, TNFα remained largely trimeric. When the effects of α/β-peptide 3 on TNFα in the absence of TWEEN20 were monitored over a longer period at elevated temperature (37 °C), a gradual increase in the extent of TNFα trimer dissociation was observed (Figure 5A, Figure S8). The TNFα trimer was barely detectable after 14 hours under these conditions. However, extended incubation of TNFα in the absence of α/β-peptide 3 at 37 °C did not lead to a loss of trimer (Figure 5A, Figure S8).

Figure 4.

Oligomerization of TNFα in solution, as judged by SEC (representative chromatograms). (A-D) Trimeric TNFα (1 μM) was incubated in PBS containing 3% DMSO and the additive(s) indicated at the top of each panel, for the indicated time period at room temperature. Elution times for proteins that serve as molecular weight standards are indicated by vertical dotted lines.

Figure 5.

(A) α/β-Peptide 3 acts slowly without additives at 37 °C to dissociate the TNFα trimer (1 μM), as determined by SEC. More rapid dissociation of the TNFα trimer is observed at room temperature when α/β-peptide 3 is combined with (B) 0.05% TWEEN20 or (C) 10 μM SPD304. Note that the horizontal axis of part A differs from the horizontal axes of parts B and C. Peaks attributed to the TNFα trimer in the SEC chromatogram (220 nm) are normalized to the “PBS only” control at the “1 min” time point. All solutions contained 3% DMSO. Data points represent the mean ± SEM of duplicate experiments, and curves represent fit to an exponential decay model.

In contrast to other conditions monitored by SEC, co-incubation of TNFα and α/β-peptide 3 in the presence of TWEEN20 caused a rapid change in oligomerization state (Figure 4D, Figure 5B). After one hour at room temperature under these conditions, the TNFα was trimer was no longer detected, and conversion to the putative monomer was complete. Collectively, these observations are consistent with our hypothesis that TWEEN20 acts to promote peptide-mediated dissociation of the TNFα trimer. We further hypothesize that TWEEN20 binds to the TNFα trimer in a manner that facilitates subsequent binding of a peptide inhibitor, which then leads to trimer dissociation.

The small molecule SPD304 is a weak inhibitor of TNFα-induced cytotoxicity,¹⁴ but the mechanism of inhibitory action has been a subject of debate.^{15, 16} Hofmann et al. recently showed that SPD304 binds to the TNFα trimer in a manner that enhances structural dynamics but does not cause trimer dissociation.¹⁶ This group proposed that inhibition of TNFα signaling arises from increased trimer dynamics caused by SPD304 binding. In contrast, earlier results were interpreted to suggest that SPD304 induces TNFα trimer dissociation when TNFα is present at low concentrations.¹⁴ The conclusions of Hofmann et al.¹⁶ led us to hypothesize that SPD304 might promote peptide-mediated trimer dissociation; we tested this hypothesis via ELISA measurements.

SPD304 alone is a modest inhibitor of TNFα-TNFR1 association: ~30% of TNFα binding to TNFR1 was blocked by 100 μM SPD304 (Figure S9). However, when Z-TNFα was pre-incubated with 3 μM SPD304 (no TWEEN20), this mixture blocked up to ~50% of TNFα-TNFR1 association. At 20 μM SPD304, pre-incubation enabled Z-TNFα to inhibit TNFα-TNFR1 association almost entirely when the Z-TNFα concentration was greater than ~20 nM (Figures S9 and S10). Comparable observations were made with α/β-TNFα−2 and α/β-peptide 3 (Figure S10). Note that SPD304 was reported to have a critical aggregation concentration (CAC) of 10 μM as measured by an optical sensor assay.¹⁵ IC₅₀ values measured via competition ELISA after pre-incubation of peptide and TNFα with 20 μM SPD304 were similar to those measured after pre-incubation of peptide and TNFα with 0.05% TWEEN20 (Table 1, Figure S10).

SEC data obtained via UPLC provided additional insight on the role of SPD304 in altering peptide-TNFα dynamics (Figure 5C, Figure S11). These data indicate that SPD304 behaves similarly to TWEEN20 (Figure 4 and Figure 5B) in that neither compound directly causes TNFα trimer dissociation. However, in the presence of either SPD304 or TWEEN20, α/β-peptide 3 induces rapid trimer dissociation. We therefore conclude that SPD304 enhances peptide-mediated TNFα trimer disassembly. The parallel in behavior between SPD304 and TWEEN20 is intriguing because the structures of these two molecules are very different (Figure 3). We note that both of these molecules were used at concentrations at which they are known to aggregate. ^{15, 31} Despite their structural differences, both SPD304 and TWEEN20 appear to be able to bind to the TNFα trimer in a manner that alters the trimer’s structure, perhaps via enhanced internal dynamics,¹⁶ and ultimately lowers the energetic barrier to subunit dissociation, a form that is stabilized by bound peptide.

Peptide-mediated rescue of cells from TNFα-induced apoptosis is very sensitive to assay conditions

Based on the mechanism proposed above, we predicted that Z-TNFα, α/β-TNFα−2 and α/β-peptide 3 would individually be competent to inhibit TNFα-induced cytotoxicity if the peptide and TNFα were combined for one hour in the presence of TWEEN20 prior to exposure to cells. To test this prediction, we pre-incubated each of the three peptides with TNFα in PBS containing 0.05% TWEEN20, conditions virtually identical to those used in the ELISA experiments, and then diluted each solution ~1000-fold so that the concentration of TNFα was in the active range for cytotoxicity (2 pM, 0.08 ng/mL), while the TWEEN20 concentration was negligible (0.00005%, which is much lower than the concentration of TWEEN20 that is cytotoxic³⁷). Under these experimental conditions, each peptide, Z-TNFα, α/β-TNFα−2 or α/β-peptide 3, inhibited TNFα-mediated death of WEHI-13VAR cells; the potencies of the three peptides were comparable (Figure 6, Table 2, Figure S12). Experiments with a one-hour pre-incubation of peptide with TNFα in PBS containing 0.05% TWEEN20 showed that peptides could rescue ~75–90% cell viability. An experiment with three-hour pre-incubation of peptide with TNFα in PBS containing 0.05% TWEEN20 showed nearly full rescue for 3 and α/β-TNFα−2 (Figure S12G, H). No rescue was observed, however, when Z-TNFα or a related peptide was pre-incubated with TNFα for one hour in the absence of TWEEN20 (Figure 6A). In addition, pre-incubation of an unrelated α/β-peptide (α/β-VEGF-1, a previously reported Z domain-derived peptide selective for VEGF¹⁸) with TNFα in PBS containing 0.05% TWEEN20 caused no inhibition. Together, these experiments suggest that the inactivation of TNFα achieved during the short pre-incubation with inhibitor and 0.05% TWEEN20, which presumably results from TNFα trimer dissociation mediated by specific peptides, is not reversed upon dilution.

Figure 6.

Evaluation of potential TNFα antagonists for the ability to inhibit TNFα-mediated death of WEHI-13VAR cells (A) without TWEEN20 pre-incubation, (B) after pre-incubation with 0.05% TWEEN20 for 1 hour at room temperature, or (C) after pre-incubation with 20 μM SPD304 for 1 hour at room temperature. Cell viability was determined by CellTiter-Glo assay. Data points represent mean ± SEM for n = 2 wells (for inhibitors) or n = 6 wells (for controls), and the curves represent fit to a 3-parameter dose-response model. The horizontal axis represents peptide concentration during pre-incubation with 2 nM TNFα, before 1000-fold dilution and addition to cells.

Table 2.

LogIC₅₀ values for selected α- and α/β-peptides tested for inhibiting TNFα-mediated cell death under various pre-incubation conditions, before diluting 1000-fold and exposure to WEHI-13VAR cells (TNFα concentration during pre-incubation = 2 nM; final TNFα concentration = 2 pM). LogIC₅₀ values for inhibition of TNFα-mediated cell death are given as the mean ± standard deviation (SD) for at least three independent experiments. LogIC₅₀ values represent concentrations during pre-incubation, before dilution and addition to cells. For each measurement, experimentally determined SD values were rounded to one significant digit, and associated mean value was then rounded to the same decimal place. RT = room temperature.

Peptide	Inhibition of TNFα-induced cell death, logIC50 (M)
	No additive (1 h, RT)	+ 0.05% TWEEN20 (1 h, RT)	+ 20 μM SPD304 (1 h, RT)	No additive (24 h, 37 °C)
Z-TNFα	>−5	−8.7 ± 0.1	−8.70 ± 0.09	−7.85 ± 0.05
α/β-TNFα−2		−.49 ± 0.05	−8.54 ± 0.02
3	>−5	−8.47 ± 0.06	−8.4 ± 0.1	−7.1 ± 0.1
α/β-VEGF-1		>−5*	>−5*

^*Compounds showed slight activity at the highest concentration tested.

An analogous set of studies was conducted to determine whether SPD304 could facilitate inhibition of TNFα signaling by Z-TNFα, α/β-TNFα−2 or α/β-peptide 3 (Figure 6, Table 2; Figure S13). Pre-incubation of TNFα with 20 μM SPD304 alone did not substantially block TNFα-mediated cell death. In contrast, pre-incubating TNFα with Z-TNFα, α/β-TNFα−2 or α/β-peptide 3 and 20 μM SPD304 for one hour fully blocked cell death induced by TNFα in a manner that depended on peptide dose. Because these samples were diluted 1000-fold prior to exposure to cells, the final SPD304 concentration during the cytotoxicity assay was 20 nM, which is far below concentrations that have been reported to inhibit TNFα-induced cell death.^{12, 38} The IC₅₀ value for each of the three peptides determined from the SPD304-mediated cell-rescue experiments was comparable to the IC₅₀ value determined from cell-rescue experiments involving TWEEN20 (Table 2).

Peptide-based inhibition of TNFα-induced cell death could be achieved in the absence of either TWEEN20 or SPD304 by pre-incubating peptide with TNFα for 24 hours at 37 °C (Table 2; Figure S14). Z-TNFα and α/β-peptide 3 both rescued WEHI-13VAR cells from TNFα-induced death in these experiments, although the extent of rescue varied among the peptides. These observations are presumably explained by the ability of an inhibitory peptide to induce TNFα trimer disruption without a trimer-destabilizing agent, if sufficient time is allowed (Figure 5A). The IC₅₀ values for these peptides were roughly 10-fold larger in this cell-based assay than in the competition ELISA. In the absence of either TWEEN20 or SPD304, the maximum cell rescue obtained was ~55–75%. It is not clear why the maximum cell rescue was lower under these conditions relative to studies conducted in the presence of TWEEN20 or SPD304.

CONCLUSIONS

We have identified a new and relatively short α/β-peptide, 3, that can inhibit TNFα-mediated cytotoxicity. In our cell-based assay, involving WEHI-13VAR mouse fibrosarcoma cells and a final TNFα concentration of 2 pM (0.08 ng/mL), α/β-peptide 3 inhibits cytokine-induced cell death with logIC₅₀ = −7.1 ± 0.1 M (IC₅₀ = 80 nM) after 24-hour pre-incubation of 3 and TNFα. This inhibitory activity compares favorably with that reported for a 19-residue bicyclic peptide in inhibiting cytokine-induced death of L929 mouse fibroblast cells after 24-hour pre-incubation of the peptide and TNFα (IC₅₀ = 4 μM).¹² α/β-Peptide 3 contains 27 residues, 6 of which are nonproteinogenic. The unnatural components provide substantial resistance to proteolytic degradation (over 2100-fold relative to the analogous α-peptide 2 and 120-fold over parent three-helix peptide Z-TNFα).

α/β-Peptide 3 and its predecessors Z-TNFα and α/β-TNFα−2 appear to inhibit TNFα signaling by inducing dissociation of the trimeric form of the cytokine; the trimeric quaternary structure is required for activation of cell-surface receptors. Our ELISA experiments demonstrated that this trimeric disruption was able to severely limit the ability of TNFα to bind to a soluble form of TNFR1, one of two major receptors for TNFα. ^1–5 We did not evaluate whether these peptides block the interaction of TNFα with its other major receptor, TNFR2, but we predict that trimer-disruption would prevent interaction of TNFα with this receptor. The mechanism of action for the peptides reported here differs from that of engineered proteins that are used clinically to inhibit pathological TNFα signaling. A trimer-disrupting mechanism was previously proposed for the small molecule SPD304,¹⁴ but more recent data have undermined this hypothesis.^{15, 16} In contrast, Luzi et al. have provided strong evidence that a bicyclic peptide inhibitor acts by disrupting the TNFα trimer,¹² as we propose for α/β-peptide 3. Our data are consistent with a mechanism in which Z-TNFα and its derivatives bind to a monomeric or dimeric form of TNFα and prevent reassembly to the active trimer. In the absence of additive, the spontaneous dissociation and reassociation of trimer is slow at the concentrations studied.²⁷ However, the presence of small molecules such as TWEEN20 or SPD-304 increases the rate of trimer dissociation,²⁷ which would facilitate peptide binding. Alternatively, TWEEN20 or SPD-304 may increase the conformational dynamics of the TNFα trimer,¹⁶ exposing the peptide binding site and promoting peptide-induced dissociation.

Our data show that diverse chemotypes can enhance the rate of peptide-induced TNFα disruption. This synergistic effect is manifested by surfactants that contain linear hydrocarbon units, such as TWEEN20, as well as by SPD304. The synergy previously observed between SPD304 and a bicyclic peptide inhibitor of TNFα signaling was explained by invoking “SPD304’s ability to destabilize TNFα trimers to the dimeric form, which are bound and stabilized by complexation” with the bicyclic peptide.¹² However, more recent work has shown that SPD304 does not induce TNFα dissociation under the conditions employed,¹⁶ a conclusion that is supported by our SEC data; therefore, the previous mechanistic hypothesis cannot be extended to explain our findings. We speculate that the rate-enhancing effect documented here for TWEEN20 and SPD304 in the dissociative inhibition of TNFα signaling by Z-TNFα, α/β-TNFα−2 and α/β-peptide 3 is operative as well in the previously reported synergy between SPD304 and the bicyclic peptide inhibitor. We note that the selection steps of the phage display process used to identify the bicyclic peptide were conducted in the presence of TWEEN20.¹² TWEEN20-dependent binding has been noted for a few phage-derived miniproteins that target TNFα,^{8, 9} although the mechanism in these cases was either not carefully examined or differed from that described here.

The unanticipated synergistic mechanism we have proposed for surfactants and SPD304 may be relevant to other systems. Protein-binding polypeptides are commonly identified via phage display and related methods, and the selections are often performed in the presence of surfactants. Protein targets that can be engaged/inactivated via quaternary structure disruption, including other cytokines in the TNFα family³⁹ and other proteins that require a specific oligomer state for activity,^{40, 41} may be particularly prone to such effects. The mechanism proposed here raises the possibility that the inhibitory activity of a molecule such as α/β-peptide 3 might be augmented by covalent attachment of a quaternary structure-destabilizing domain. Such a compound may prove effective at disrupting TNFα-TNFR interactions in living systems, and may be the starting point for novel therapeutics acting through this mechanism. Exploration of this design strategy is underway.

METHODS

Additional details on experimental methods can be found in the Supporting Information.

Peptide synthesis and purification

The preparation of α- and α/β-peptides was carried out as previously described.¹⁸ After purification, final peptide purity was assessed by analytical-scale reverse-phase HPLC, and peptide identity was confirmed by MALDI-TOF-MS using a Bruker ULTRAFLEX-III (Figure S15).

TNFα-TNFR1 competition ELISA

The binding of α- and α/β-peptides to TNFα to block the TNFα-TNFR1 interaction was assessed by competition ELISA as previously described.¹⁸ See Supporting Information for full details.

Monitoring the TNFα oligomerization state by SEC

For each experiment, the time of the addition of 3 (or DMSO only for control experiments) was recorded, and 5 μL of the resulting solution was injected onto a Protein BEH SEC column (Waters, 186006505) on a Waters Acquity UPLC H Class instrument, with PBS, pH 7.4 as the mobile phase running at 0.4 mL/min; the chromatogram was monitored over 10 minutes. Each solution was injected repeatedly over the course of either 60 minutes at room temperature or 960 minutes at 37 °C. Each time course for each condition was repeated twice. For analysis, the area for the peak corresponding to the TNFα trimer (as determined by injection of TNFα in PBS alone and from MW standards) was monitored at 220 nm. We calculated the “Apparent trimer concentration (%)” for a given chromatogram by dividing trimer peak area at that time point by the average of the trimer peak areas in the “PBS only” condition at the 1 min time point. Within each set of injections that make up a time course, individual runs were aligned on the vertical axis by aligning UV signal at the 2.5 min time point for each injection. See Supporting Information for full details.

WEHI-13VAR cell viability assays

WEHI-13VAR mouse fibrosarcoma cells (ATCC, CRL-2148) were cultured using RPMI-1640 (RPMI) supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin (P/S) at 37 ˚C, 5% CO₂. For a given experiment, cells were trypsinized and reseeded into sterile tissue culture-treated, clear bottom, white-walled 96-well plates at a cell density of 8,000 cells/well in 100 μL of RPMI + 10% FBS + P/S.

For all experiments, after the pre-incubation, solutions were vortexed briefly and diluted 50-fold in RPMI, then 7-fold in RPMI with 4.4 ng/mL Actinomycin D. 50 μL aliquots of these solutions were added to the appropriate wells (bringing total volume to 150 μL per well, final TNFα concentration = 2 pM (0.08 ng/mL)). Plates were incubated at 37˚C, 5% CO₂ for 16–18 hours to allow TNFα to induce cell death. Following incubation, plates were removed from the incubator and allowed to cool to room temperature for 30 minutes. 150 μL of CellTiter-Glo reagent (Promega, G7571) was added to each well, the plate was agitated on an automated shaker for 2 minutes, and luminescence was read on a BioTek Synergy 2 microplate reader. See Supporting Information for full details.