Multisite aggregation of p53 and implications for drug rescue

Significance

Destabilized mutants of the tumor suppressor p53 are inactivated by self-aggregation in a substantial number of tumors and may also coaggregate with and inactivate WT p53 and family members. We found in vitro that self-aggregation proceeded via a network of multiple aggregation-prone sites in p53, and inhibition of an individual site did not inhibit aggregation. Nevertheless, peptides designed to be complementary to various aggregation sequences and inhibit their polymerization can specifically kill cancer cells and be potential anticancer drugs. We found that those peptides can also function by p53-independent routes in cancer cell cultures, implying further therapeutic targets.

Abstract

Protein aggregation is involved in many diseases. Often, a unique aggregation-prone sequence polymerizes to form regular fibrils. Many oncogenic mutants of the tumor suppressor p53 rapidly aggregate but form amorphous fibrils. A peptide surrounding Ile254 is proposed to be the aggregation-driving sequence in cells. We identified several different aggregating sites from limited proteolysis of harvested aggregates and effects of mutations on kinetics and products of aggregation. We present a model whereby the amorphous nature of the aggregates results from multisite branching of polymerization after slow unfolding of the protein, which may be a common feature of aggregation of large proteins. Greatly lowering the aggregation propensity of any one single site, including the site of Ile254, by mutation did not inhibit aggregation in vitro because aggregation could still occur via the other sites. Inhibition of an individual site is, accordingly, potentially unable to prevent aggregation in vivo. However, cancer cells are specifically killed by peptides designed to inhibit the Ile254 sequence and further aggregation-driving sequences that we have found. Consistent with our proposed mechanism of aggregation, we found that such peptides did not inhibit aggregation of mutant p53 in vitro. The cytotoxicity was not eliminated by knockdown of p53 in 2D cancer cell cultures. The peptides caused rapid cell death, much faster than usually expected for p53-mediated transcription-dependent apoptosis. There may also be non-p53 targets for those peptides in cancer cells, such as p63, or the peptides may alter other interactions of partly denatured p53 with receptors.

The tumor suppressor p53 is inactivated by mutation in a substantial number of tumors (1–3). Some 30–40% of those oncogenic mutants are simply destabilized by mutations in its core domain. Those mutants are temperature-sensitive, having a WT structure at lower temperatures, but melt at close to body temperature or below and rapidly aggregate (4–6). Protein aggregation occurs in many diseases (7–9). The best mechanistically characterized examples involve the polymerization of aggregation-prone peptides (10, 11) or small proteins to give well-defined fibrils based on a regular repeat structure (12–18). The fibrillar aggregates have a characteristic cross-β X-ray diffraction pattern and bind such dyes as Congo Red and Thioflavine T (ThT) (16). The kinetics of aggregation usually follow a nucleation-growth mechanism, with very slow nucleation (19–21). WT p53 itself aggregates at body temperature (4–6, 22–24), and the oncogenic destabilized mutants aggregate even faster to give amorphous structures that display the characteristic diffraction pattern and bind those diagnostic dyes (23, 25), although under certain conditions, such as very high pressure, they will generate regular fibrils (23, 26). The mechanism of initiation of aggregation of p53 differs from the usually studied examples. Two molecules of the core domain of p53 extensively unfold (27) and then combine, on a much shorter time scale than in classical diseases of fibril formation. Mutant and WT protein form mixed aggregates in a cross-reaction and coaggregation process rather than by seeding by already aggregated molecules because the initiation of aggregation is relatively fast even in WT protein (27, 28).

The core domain of p53 is built around a β-sandwich, comprising two antiparallel β-sheets of four strands (S1, S3, S8, and S5) and five strands (S10, S9, S4, S7, and S6) that pack against each other across an extended hydrophobic core (29). The β-sandwich serves as a scaffold for formation of a DNA-binding interface by two large loops, L2 and L3, and a loop–sheet–helix motif, containing loop L1 (113–123), a short β-sheet composed of β-hairpin S2–S2′ and the C-terminal residues of the extended strand S10, and α-helix H2 (278–289) (30, 31). Xu et al. (32) have proposed residues 251–257 (ILTIITL) in β-strand S9 to be the key aggregation-prone sequence. The mutation I254R is reported to decrease greatly the aggregation propensity of the isolated peptide in model studies and to inhibit heterologous aggregation-promoting activities of p53 (32). In contrast, Vousden and coworkers (33) find the 251–257 region of p53 is not required for p63 or p73 binding in cancer cells. Neither the I254R mutation in WT p53 nor the deletion of 251–312 or 251–257 in p53R175H inhibits their binding to p63 or p73, and even p53I254R itself can promote invasion in a p53-null cancer cell line.

The inhibition of aggregation of p53 may be a route for rescuing mutant p53 in cancer cells (5). We initially tackled the rescue by stabilizing the folded conformation against unfolding, first by a generic approach using a peptide that binds to native p53 (34) and mutants (35) and then by targeting a specific mutant where the mutation forms a druggable cavity, Y220C. Its melting temperature is raised, and its aggregation is slowed in vitro by small mutant-specific molecules that bind in the cavity (36, 37) and rescue Y220C in cancer cell lines (25, 38, 39). Eisenberg and coworkers (40) have taken a different approach of inhibiting the aggregation process, per se, with a peptide that caps the specific exposed aggregation-prone sequence, centered on Ile254, rather than stabilizing the native structure.

A ϕ-value analysis reveals that many of the β-strands in the cores of two associating molecules of the core domain are substantially unfolded in the initiation of aggregation (27). We suspected that more than one of those strands could be involved in pairing during aggregation. We searched the whole p53 core domain sequence for additional aggregation-prone sites by both computer algorithms and limited proteolysis of the harvested aggregate to identify regions that are buried in the aggregate. We mutated those sites to inhibit their aggregation tendency and measured the kinetics of aggregation. Knocking out by mutation the aggregation propensity of Ile254 and other individual sites did not prevent aggregation from the native state in vitro. The initial aggregate did not elongate by forming regular repeating of a unique peptide as found in the most frequently studied examples, where well-ordered fibrils are formed, but a variety of alternative heterologous or homologous pairings of β-strands were involved.

Results

Sequences of p53 Involved in Aggregation.

WT p53 core domain.

We used limited proteolysis of harvested aggregates from WT p53 core domain (WTC) to analyze which parts of the structure are resistant to proteases after aggregation from the native state at 37 °C in vitro. The large clumps of WTC aggregate were highly protease-resistant. Treatment of native soluble WTC by proteinase K for 9 h gave only small peptides (Fig. S1A). In contrast, similar proteolysis of the aggregate harvested after incubation of WTC overnight at 37 °C still gave residual full-length protein and peptides larger than 5 kDa (Fig. S1B).

Fig. S1.

Proteolysis resistance of aggregate of WTC. Limited digestion of 0.45 mg/mL native p53 WTC (A) and aggregate of WTC from overnight aggregation at 37 °C using 20 μg/mL proteinase K at 20 °C (B). The lane for 0 h is from the same gel as the other time points.

To reveal the sequences protected in the initial stages of aggregation, we harvested the aggregate after 2 h of incubation at 37 °C, which is sufficient time for near-maximal binding of ThT but early in the maturation of the aggregate to large clumps. This aggregate was more readily proteolyzed than the aggregate from overnight incubation. After 4 h of digestion with 1:50 (wt/wt) p53/proteinase K, only a small fraction of the aggregate remained. Mass spectroscopy (MS) revealed that the main peaks lower than 3 kDa in the remaining aggregate were 108–114 (S1L1), 137–146 (S3), 146–160 (S4), 126–136 (S2S2′), and 248–257 (S9).

We then analyzed in more detail a time course of tryptic digestion of the aggregate formed after 2 h of incubation at 37 °C (identical results were found after 1-h and 3-h incubations). After digestion of the aggregate for defined times, we separated the remaining aggregate from the supernatant and analyzed both phases by MS. Peptides 249–267 (2,069 Da, S9), 214–248 (3,880 Da, S7S8), 182–196 (1,609 Da, H1S5), 193–213 (2,097 Da), and 203–213 (1,428 Da) were not present in the supernatant after 24 h of digestion with 1:100 (wt/wt) trypsin (Fig. 1A), although they were present in the main peaks in the digested native soluble WTC (Fig. 1B). Only after a further 24 h of digestion by 1:50 (wt/wt) trypsin did 2,069 Da, 1,609 Da, and 3,880 Da become three of the main peaks released into the supernatant (Fig. 1C). The cleavage pattern indicated that those peptides were buried in the aggregate or formed relatively inflexible structures that were initially inaccessible to trypsin. Thus, S5, S7, S8, and S9 were all involved in the aggregate of WTC.

Fig. 1.

Limited proteolysis of native WTC, WTC aggregate, native WTCI254D, and WTCI254D aggregate by trypsin. Mass spectra of supernatant of WTC aggregate after the initial 24-h digestion (A), native WTC after the initial 24-h digestion (B), supernatant of WTC aggregate after the second 24 h of digestion (C), and remaining aggregate of WTC aggregates after second 24 h of digestion (D). WTC aggregate was from a 2-h incubation at 37 °C. Native WTCI254D (E) and the remaining aggregate of WTCI254D aggregate (F) after a 6-h digestion.

After 48 h of digestion, the peptides that were much more abundant in the remaining aggregate (Fig. 1D) than in the supernatant (Fig. 1C) were 249–267 (2,069 Da, S9), 102–120 (2,090 Da, S1), 182–209 (3,169 Da, S5S6), 182–213 (3,688 Da, S5S6), 176–202 (3,037 Da, H1S5), 102–132 (3,356 Da, S1L1S2), and 92–132 (4,301 Da, S1L1S2). There were far fewer accessible tryptic cleavage sites in the WTC aggregate (Fig. 1D) than in the native protein (Fig. 1B), and most of the inaccessible cleavage sites (Fig. 1D) resided in the peptides in the insoluble aggregate. The much larger peak of S9 in the remaining aggregate was consistent with a previous finding (32) that this peptide is a major site for aggregation.

Peptides 249–267 (2,069 Da, S9) plus peptides 214–248 (3,880 Da, S7S8), 102–110 (1,079 Da, S1), 140–156 (1,855 Da, S3), 182–196 (1,609 Da, S5), and 165–174 (1,215 Da, L2) were present in both remaining aggregate and supernatant after 48 h of digestion. Their presence in both phases was not due to poor solubility because digestion of the same concentration of native WTC did not result in any aggregate during the digestion process and the intensity of their corresponding MS peaks in the supernatant increased with digestion. The presence in both phases was not caused by the aggregate dissolving and being digested in the supernatant: More than 97% of the aggregate remained insoluble after 48 h of incubation without protease.

The most likely reason for the presence of those peptides in the residual core structure and being released into solution is that the aggregate is heterogeneous: In some part of the aggregate, a particular sequence is highly buried, whereas it is only loosely buried in other parts. Further, there were multiple aggregation sites in addition to the aggregation site of S9.

WTCI254D.

The mutation I254D in the middle of the S9 aggregation-prone sequence greatly reduces its intrinsic propensity to aggregate. However, the mutated protein still forms an amorphous aggregate from the native state at 37 °C (Fig. S2) and rapidly forms a ThT-binding state and scattering particles (Fig. 2). The harvested aggregate from I254D was also protease-resistant. After 6 h of digestion by trypsin, most of the native soluble I254D was digested to peptides of M_r lower than about 2 kDa, except for peptide 214–248 (3,880 Da), which does not contain a tryptic cleavage site. In contrast, there was only partial digestion of the aggregate of I254D after 2 h of incubation at 37 °C, and the peptides in the remaining aggregate were mainly of M_r greater than 2 kDa. Major peaks that were enriched in the remaining aggregate relative to the supernatant were similar to those peaks from the digestion of WTC aggregate, with common peaks in the MS peaks in regions of 182–213 (3,688 Da, 3,169 Da; S5S6), 102–120 (4,301 Da, 3,356 Da, 2,090 Da; S1), and 176–202 (3,037 Da, H1S5) (Fig. 1 E and F). Further, minor peaks corresponding to peptides 268–282 (1,710 Da, S10H2), 157–181 (3,064 Da, 2,809 Da), 140–174 (3,983 Da), and 133–156 (2,677 Da, S2′S3) were also enriched in the remaining aggregate of WTCI254D. Digestion of the aggregate after overnight incubation also showed the same major and minor peaks enriched in the remaining aggregate as for proteolysis of the 2-h aggregate. These data provide further evidence that there are multiple aggregation-prone sequences and that S9 is not necessary for aggregation to occur.

Fig. S2.

Similar morphologies of WTC and WTCI254D aggregates. Transmission electron micrograph of WTC aggregate (A) and WTCI254D aggregate after aggregation at 37 °C for 6 h (B).

Fig. 2.

Aggregation of WTC, WTCI254D, and WTCI254R from the native state. Time course of aggregation of WTC (A) and WTCI254D (B) at 37 °C monitored by ThT binding, WTCI254D (C) and WTCI254R (E) at 30 °C by ThT binding, and WTCI254D (D) and WTCI254R (F) at 30 °C by light scattering.

Full-length p53.

Most of the fragments found in the aggregate of core domain were present in the aggregates of wild-type full-length p53 (WTFL) and WTFLG245S (Fig. 3 and Table 1). The first three main segments contributing to the aggregation process (249–267, 182–213, and 268–282) were also the highly aggregation-prone segments we found in the aggregation of denatured p53 mutants. Aggregation of the full-length protein was thus similar to aggregation of the core domain. Further, a small amount of the C-terminal fragment 373–393 (2,377 Da, 373–393; 2,594 Da, 373–395), which is known to bind to the core domain, was buried in the aggregation core.

Fig. 3.

Limited proteolysis of WTFL aggregate and WTFLG245S aggregate by trypsin. Mass spectra of supernatant of WTFL aggregate after digestion (A), the remaining aggregate of WTFL aggregates after digestion (B), supernatant of FLG245S aggregates after digestion (C), and the remaining aggregate of FLG245S aggregates after digestion for 76 h (D).

Table 1.

Regions protected against proteolytic degradation

Peptide regions enriched in remaining aggregate*				Effect on aggregation†	Region
WTC	WTCI254D	WTFL	FLG245S
249–267	—	249–267	249–267	+++++	S9
182–213	182–213	182–213	182–213	+++++	H1S5S6
—	268–282	268–282	268–282	+++++	S10H2
102–120	102–120	102–120	102–120	+++	S1L1
—	—	284–292	284–292		H2
140–174	140–174	140–174	140–174	++	S3S4L2
—	—	373–393	373–393		C-term
—	133–156	—	—		S2′S3

^*Aggregates from aggregation of the proteins at 37 °C were digested by trypsin for 48 h, 6 h, 76 h, and 76 h, respectively, for WTC, WTCI254D, WTFL and FLG245S at 20 °C. Peptides are shown in boldface: major peaks in the spectra. For WTFL and WTFLG245S, peptides are listed according to relative size of peaks from large to small in MS spectra of the remaining aggregate after digestion for 76 h.

^†Effect on aggregation is based on the aggregation rates of denatured aggregation-inhibition mutants in the corresponding region. Measured by initial rates: +++++ = 0.6–1%, ++++ = 1–2%, +++ = 5–10%, ++ = 20%, + = 30–50%.

Effects of Mutating Aggregation-Prone Sites in Denatured p53.

Using the programs Tango (41), Waltz (42), Aggrescan (43), FoldAmyloid (44), and Amylpred2 (45), we identified several aggregation-prone sequences (Table 2). We lowered the aggregation propensity of each sequence by introducing mutations that were modeled to be compatible with the crystal structure (46) and are predicted to be effective using the above programs. Mutation to Asp or Arg should be the most effective; however, because of steric constraints in the folded protein, these changes were not always possible, and so we made multiple, conservative mutations in each sequence instead. S5 was not predicted to be a site, but we mutated it because it was identified by limited proteolysis to be protected in the aggregate.

Table 2.

Aggregation from 3 M urea

Region*	Residue no.	Sequence	Mutation	Effect of mutation on aggregation kinetics†
S1&L1	109–113	FRLGF	F113A	+++
L1 & S2	121–127	SVTCTYS	C124D	+
S3	143–147	VQLWV	W146D	++
S4	158–164	VRAMAIYK	I162G	++
S5	194–203	LIRVEGNLRV	L194AV197AL201DV203A‡	+++++
S7	214–218	HSVVV	V217D	+
S8	230–238	TTIHYNYMC	T231DH233D	+++
S9	251–257	ILTIITL	I254R/D‡	+++++
S10	271–276	EVRVCA	V272AV274AC275S‡	++++

^*S5 was identified by limited proteolysis, whereas the others were predicted by computer algorithms.

^†Measured by initial rates: +++++ = 0.6–1%, ++++ = 1–2%, +++ = 5–10%, ++ = 20%, + = 30–50%.

^‡These mutants are in the WT construct.

We determined the kinetics of aggregation of the core domain of p53 from binding of ThT (Fig. 4A) after denaturation in 4 M urea at 0 °C and then diluting the denaturant to 3 M at 37 °C. The curve for aggregation of our standard mutant core domain Y220C (QCYC) (27) had an exponential fast phase. A double-logarithmic plot of the initial rate (measured against t²) vs. the concentration of [QCYC] (Fig. 4B) should give as its slope the number of molecules in the nucleus plus 2 for homogeneous nucleation (47). The slope was 5.8 ± 0.5, in contrast to 1–2 found for thermal aggregation of WT and mutants (27).

Fig. 4.

Aggregation of denatured p53 variants in 3 M urea, measured by binding of ThT. (A) Concentration dependence of QCYC. (B) Double-logarithmic plot of the initial rates from A (versus t²) versus concentration of QCYC. (C) Aggregation curves for mutations into the QCYC framework. (D) Aggregation curves for mutations into the WTC framework.

Mutations in S1 (F113A), S2 (C124D), S3 (W146D), S4 (I162G), S7 (V217D), and S8 (T231DH233D) (Fig. 4C) had small to significant lowering of rates (Table 2). Mutations in S5 (WTCL194AV197AL201DV203A), S9 (WTCI254D, WTCI254R), and S10 (WTCV272AV274AC275S) had very large effects (Fig. 4D and Table 2), and there was little, if any, light scattering after overnight incubation in 3 M urea. Accordingly, S9 and S5 are highly aggregation-prone regions, followed by S10, S8, and S1 regions. Many regions within the p53 sequence are very aggregation-prone and can contribute to aggregation of p53 if they are exposed.

Urea Dependence of Aggregation.

WTCI254D did not visibly aggregate after incubation for 24 h in 3 M urea. However, the rate increased greatly at lower urea concentrations, and a logarithmic plot of the initial rate against [urea] had a slope of −2. The denatured WT p53 core domain and mutant QCYC were much less sensitive to the concentration of urea, with a corresponding slope of only −0.69 (Fig. S3A). The effects of mutation on aggregation decreased with decreasing concentrations of urea as the rates greatly increased (Fig. S3A). The sensitivity of rate constants to the concentration of denaturant reflects the change in solvent-accessible surface area as a protein-folding reaction proceeds: The more the area buried, the greater is the slowing down of the rate. Mutation of I254 to D254 was the most sensitive, indicating that the rate-determining step in aggregation involving the region surrounding I254 buried more surface area than in the other regions probed.

Fig. S3.

Double-logarithmic plots of initial aggregation rate of denatured p53 mutants on concentration of urea (A), initial rates of light scattering (B), and ThT binding vs. concentration of WTCI254D (μM) (C). (V_o is measured from plots of signal vs. t²). Aggregation of denatured p53 mutants was measured at 37 °C from the denatured state. Aggregation of WTCI254D was measured at 30 °C from the native state. NA, not applicable.

Effects of Mutation on Aggregation from the Native State.

We analyzed, among other mutations of aggregation-prone sites, I254D and I254R in S9, which greatly affected aggregation from the denatured state and are reported to inhibit totally its gain of function in cells (32). The structure of WTC is destabilized by mutations of I254D and I254R at 10 °C, by 6.15 kcal/mol and 4.28 kcal/mol, respectively, and decreases in their melting temperatures (T_ms) to 37.8 ± 0.1 °C and 39.7 ± 0.1 °C from 45.8 ± 0.1 °C, respectively. Mutating I254R and I254D increased the aggregation rate of WTC (Fig. 2) but retained the apparent two-step sequential first-order kinetics (25) for the binding of ThT. At 37 °C, aggregation rate constants for WTC at 12 μM were 0.027 ± 0.001 min⁻¹ and 0.79 ± 0.16 min⁻¹, respectively, whereas the aggregation rate constants for 4 μM WTCI254D are 0.85 ± 0.02 min⁻¹ and 4.3 ± 0.5 min⁻¹. Even at 30 °C, the aggregation rate constants of WTCI254D (k₁ = 0.069 ± 0.001 min⁻¹, k₂ = 0.33 ± 0.01 min⁻¹) and WTCI254R (k₁ = 0.0255 ± 0.0001 min⁻¹, k₂ = 0.38 ± 0.02 min⁻¹) at 10 μM were still high. The amplitude of light scattering was not diminished by mutation of I254 (Fig. 2 D and F), with the values being typical of WTC and all other mutants, and the aggregate of WTCI254D had similar morphology to aggregates of WTC (Fig. S2).

The slopes of double-logarithmic plots of initial rates of light scattering and ThT binding vs. concentration of WTCI254D were 1.5 ± 0.2 and 2.1 ± 0.1, respectively (Fig. S3 B and C), which is typical of WTC and the majority of mutants (27), and greatly different from the value of 5.8 ± 0.5 (Fig. 4) for aggregation from the denatured state. The kinetics of aggregation from the native state was unchanged by mutation of the sites that greatly reduced aggregation rates from the denatured state, and the mechanism of aggregation from the native and denatured states was clearly quite different. The mutations I254D and I254R caused fast aggregation without changing the aggregation mechanism, suggesting that S9 is not crucial to aggregation from the native state at physiological temperature.

Self- and Cross-Aggregation of Peptides Derived from p53.

Self-aggregation of isolated peptides.

We examined by ThT fluorescence the aggregation properties of the isolated S9 and S5–6 peptides that contain the sequences of residues 249–258 and 193–206 in p53, respectively, with modifications of N-terminal acetylation and C-terminal amidation. S9 peptide contains more hydrophobic residues and self-aggregated relatively fast, but with a low fluorescence yield. S5–6 is more hydrophilic than S9 and aggregated far more slowly (Fig. 5A), but yielded a high ThT fluorescence intensity, similar to the p53 protein core domain (25). Although having a high predicted aggregation propensity, S10 peptide self-aggregated much more slowly than S9 and S5–6 (Fig. 5D). Nevertheless, mutations inhibiting its aggregation greatly decrease the aggregation propensity of denatured p53, suggesting that S10 coaggregated with other regions of p53 rather than self-aggregating. S10 reached intermediate levels of ThT fluorescence (Figs. 5D and 6A).

Fig. 5.

Aggregation of S9, S10, and S5–6 peptides and its inhibition in standard buffer of 25 mM KPi, 150 mM NaCl, 1 mM TCEP, and 22 μM ThT, unless indicated otherwise, at pH 7.2 and various concentrations of GdmCl. (A) Fifteen micromolar S9 peptide and 15 μM S5–6 peptide (+0.18 M GdmCl). (B) Inhibition by Anti197_1 and Anti197_2 of aggregation of 15 μM WT S5–6 peptide in buffer + 0.18 M GdmCl. (C) Anti254_1 on 15 μM S9 in buffer + 1.2 M GdmCl. (D) Anti272_1 on aggregation of S10 in buffer + 0.37 M GdmCl. (E) Anti254_1 and ReACp53 on 62 μM S9 peptide in buffer + 1.9 M GdmCl and 20 μM ThT. (F) Fifteen micromolar S9 peptide in buffer + 0.18 M GdmCl. (The peptides were dissolved in high concentrations of GdmCl to prevent self-aggregation, which led to varying concentrations of GdmCl on mixing and dilution.)

Fig. 6.

Cross-aggregation and seeding of S9, S10, and S5–6 peptides. Coaggregation of 15 μM S9 and 160 μM S10 (A), 15 μM S5–6 and 160 μM S10 (B), and 15 μM S5–6 and 15 μM S9 (C) in buffer of 25 mM KPi, 150 mM NaCl, 22 μM ThT, 1 mM TCEP, and 0.3 M GdmCl (pH 7.2) at 37 °C.

Cross-aggregation and seeding.

After 800 min, 160 μM S10 barely aggregated, whereas 15 μM S9 rapidly self-aggregated to produce a very low yield of ThT fluorescence (Fig. 6A). However, 15 μM S9 stimulated the aggregation of S10, giving a high yield of ThT fluorescence. This evidence is compelling for seeding and coaggregation of the two. There were indications that 160 μM S10 interacted with 15 μM S5–6 because its aggregation was somewhat inhibited and the fluorescence yield was higher than the sum of the two separate reactions (Fig. 6B). S9 slowed down the aggregation of S5–6, again suggesting some interaction between the two (Fig. 6C).

Inhibition of aggregation of isolated peptides.

We designed peptides targeting three main aggregation-prone sequences (251–257, 194–203, and 271–276) in p53 (Table 3) to inhibit aggregation of mutants or possibly also interfere with the interactions between those regions and other cellular proteins.

Table 3.

Sequences of designed peptides

Peptide name	Peptide sequence*	Target
Anti254_m	ILTIITL-NH2	S9
Anti197_m1	LIRVEGNLRV-NH2	S5–6
Anti197_m2	LIRVEGNLRV-NH2	S5-6
Anti272_m1	EVRVCA-NH2	S10
Anti272_m2	EVRVCA-NH2	S10
Anti254_1	LCLRRRPIARIIRLL-NH2	S9
Anti254_2	R10PILTPITLR-NH2	S9
Anti254_3	R10PIARIIRLL-NH2	S9
Anti197_1	R10GPHLIRREARYRAEYL	S5–6
Anti197_2	LCLRPHLIRREARYRAEY	S5–6
Anti272_1	LCLRPNSFEVRACA-NH2	S10
Anti272_2	LCLRPNSFERRVCA-NH2	S10

N-termini of the unmethylated peptides are acetylated except Anti272_1. R₁₀, 10 arginines.

^*Underlined residues are N-methylated.

We tested first the inhibition of the isolated peptide sequences from p53. Low concentrations of Anti197_1 and Anti197_2 inhibited the aggregation of S5–6 (Fig. 5B). Anti254_1 dramatically decreased the aggregation rate of S9 (Fig. 5C), and a much lower concentration of Anti272_1 also inhibited aggregation of S10 peptide (Fig. 5D). Among several aggregation-prone segments predicted by ZipperDB, the peptide ReAcp53 is reported to target (40) the S9 region. We compared peptides Anti254_1 and ReACp53. In the presence of a high concentration of GdmCl, neither Anti254_1 nor ReACp53 self-aggregated. Anti254_1 inhibited S9 aggregation more effectively than did ReACp53 (Fig. 5E). At a low concentration of GdmCl, Anti254_1 was also very effective in inhibiting aggregation of S9 peptide (Fig. 5F).

Peptides did not inhibit aggregation of p53 in vitro.

The designed peptides were not effective in inhibiting aggregation of p53 protein, although they were very effective in inhibiting aggregation of the isolated target p53 peptides (Fig. 5 C–F). Twenty micromolar Anti254_m only slightly, if at all, inhibited aggregation of 1 μM p53 Y220C core domain (Fig. 7 A and B). Twofold (Fig. 5C) or even 0.4-fold Anti254_1 (Fig. 5 E and F), relative to its target peptide, dramatically decreased the aggregation of the targeting peptide S9. Thirty-five micromolar Anti254_1 only slightly inhibited aggregation of 1 μM mutant R175H p53 core domain or a mixture of 1 μM mutant R175H p53 and 2 μM p73 (Fig. 7 E and F). Thirty-five–fold ReACp53 relative to p53 protein did not inhibit aggregation of 1 μM p53 R175H either (Fig. 7C). Twenty micromolar Anti197_1 could effectively inhibit aggregation of its target peptide 15 μM S5–6 (Fig. 5B), although neither 20 μM Anti197_1 itself nor together with Anti254_1 inhibited aggregation of 1 μM p53 Y220C core domain (Fig. 7D). The peptides were unable to decrease the initial aggregation rate of p53. The lack of inhibition is consistent with our findings that the rate-determining steps in aggregation of p53 are early unfolding events and that there are multiple aggregation sites after initial unfolding.

Fig. 7.

Peptides not effectively inhibiting aggregation of p53 proteins. Inhibition effect of Anti254_m on aggregation of 1 μM p53 Y220C mutant at 34 °C monitored by light scattering (A) and ThT binding (B). Thirty-five micromolar ReACp53 (C) and 35μM Anti254_1 (E) on aggregation of 1 μM p53R175H at 37 °C. (D) Anti197_1 and Anti254_1 on aggregation of 1 μM p53 Y220C. (F) Thirty-five micromolar Anti254_1 on aggregation of the mixture of 1 μM p53R175H and 2 μM p73 at 37 °C.

Cytotoxicity Toward Cancer Cells with Structural Mutants of p53.

We examined the effects of the peptides on cancer cell lines containing WT p53 or destabilized structural mutant p53Y220C, p53V143A, or R175H. All of the designed peptides targeting the three major regions had higher cytotoxicity toward the gastric cancer cell line NUGC3 (Y220C^+/+) than toward NUGC4 (WT^+/+) (Fig. 8). N-methylated peptides without a cell-penetrating sequence were significantly more cytotoxic toward NUGC3 than toward NUGC4 (Fig. 8A). The more hydrophobic peptide Anti254_m was more effective than the other two sequences, possibly because with negative charged residues and low hydrophobicity, Anti197_m and Anti272_m were less cell-penetrating.

Fig. 8.

Peptides selectively kill cancer cells harboring mutant p53. Viability was measured after treatments with peptides Anti254_m (A); Anti254_1 (B); Anti197_1 (C); Anti254_1, Anti197_1, and Anti197_2 (D); Anti254_2 (E); Anti197_2 (F); Anti272_1 (G); and Anti272_2 (H) on cancer cells harboring mutant p53 (NUGC3, MKN1, SKBR3) and cancer cells harboring WT p53 (NUGC4, MCF-7) for 24 h unless otherwise indicated.

Anti254_1 was cytotoxic at concentrations lower than 30 μM toward NUGC3 and MKN1 (V143A^+/+) (Fig. 8B), although being much less toxic toward NUGC4. Anti197 also effectively destroyed cancer cells harboring mutant p53 with either poly-arginine (Fig. 8C) or Xentry cell-penetrating sequence (Fig. 8E). Two consecutive 24-h treatments with 8 μM Anti197_1 were even more effective than 24-h treatment with 16 μM Anti197_1. Anti254_1, Anti197_1, and Anti197_2 showed higher cytotoxicity toward breast cancer cell lines SKBR3 (R175H^+/+) than toward MCF7 (WT^+/+) (Fig. 8D). Anti254_2, with a proline to interrupt the formation of the β-sheet, had much higher toxicity toward NUGC3 than toward NUGC4, although it was much less effective than Anti254_1 at concentrations higher than 20 μM. Peptides targeting region 272 also showed some cytotoxicity (Fig. 8 G and H). Anti272_1 is very sensitive to the presence of serum and is effective only at high concentrations, possibly because of its unmodified N-terminus (48).

We changed the cell-penetrating sequence of Anti254_1 to poly-arginine (Anti254_3) as in ReACp53 for better comparison. Anti254_3 was much more cytotoxic than ReACp53 toward cancer cells (Fig. 9). The IC₅₀s of Anti254_3 with NUGC3 and NUGC4 cells were 4.2 ± 0.1 μM and 6.5 ± 0.2 μM, respectively, whereas the IC₅₀s for ReACp53 were 17.4 ± 0.4 μM and 17.2 ± 0.6 μM, respectively. ReACp53 did not show selectivity between NUGC3 (Y220C^+/+) and NUGC4 (WT^+/+), whereas Anti254_3 demonstrated selectivity toward the mutant. The presence of 10% heat-inactivated serum in our experiments would have led to higher values of IC₅₀s than reported (40).

Fig. 9.

Relative cytotoxicity of Anti254_3 and ReACp53. Viability was measured after treatment with Anti254_3 (A) and ReAcp53 (B) for 48 h.

Synergistic and Additive Effects of Cytotoxic Peptides.

Pairwise synergy and additivity.

There were synergistic and additive effects between the peptides targeting the 254 region and peptides targeting the 197 and 272 regions. Fifty micromolar Anti254_m had minor toxicity to NUGC3, which was greatly enhanced by the addition of 5 μM or 10 μM Anti197_1 (Fig. 10A). Combination of Anti254_1 and Anti197_1 greatly enhanced the effect of either alone on both NUGC3 and MKN1 cell lines (Fig. 10B). The combination index (CI) is ∼1, based on the Chou–Talalay method (49), indicating that the effects were additive rather than synergistic. Combinations of Anti254_1 and Anti197_2 had much greater effects than the sums of the effects of each peptide on NUGC3 (Fig. 10C) and MKN1 (Fig. 10D) cells. The combination between 10 μM Anti254_1 and 50 μM Anti197_2 (CI is around 0.8) is synergistic, and the combination of 15 μM Anti254_1 and 25 μM Anti197_2 shows even stronger synergy (CI = 0.55).

Fig. 10.

Peptides targeting the 197 region enhance the effect of peptides targeting the 254 region. Anti197_1 enhances the effect of N-methylated peptide (A) and Anti254_1 (B). Anti197_2 enhances effect of Anti254_1 on NUGC3 (C) and MKN1 (D) cancer cells. Western blot of the combination effect of Anti254_1 and Anti197_2 on NUGC3 cells (E) and MKN1 (G) for different lengths of time and on MKN1 cells after treatment for 9 h (F).

Western blots showed that treatment of NUGC3 cells with Anti254_1 slightly up-regulated p21 levels, whereas incubation with Anti197_2 up-regulated p21 more so (Fig. 10 E and F). The combination of these two peptides consistently up-regulated the p21 level in both cell lines. This up-regulation of p21 was detectable within 3 h in MKN1 cells (Fig. 10G). Further, combination of Anti254_1 and Anti197_2 decreased levels of MDM2 in both MKN1 and NUGC3 cell lines. Levels of the apoptosis regulator Bax did not change in either combination or separate treatment of these two peptides. A decrease in p53 levels was detectable after 9 h of incubation with the combination of both peptides (Fig. 10 E and F), and levels of p53 dramatically decreased after treatment for 24 h (Fig. 10G), indicative of degradation of aggregated p53.

Anti272_1 had little effect on MKN1 cancer cells at concentrations lower than 400 μM, but it significantly enhanced the effect of Anti254_1 (Fig. S4A). Eighteen micromolar Anti254_1 by itself reduced the viability of MKN1 by 80%, whereas the combination of 50 μM Anti272_1 and 18 μM Anti254_1 decreased the viability to 46%, demonstrating synergism (CI = 0.88). However, further increasing the concentration of Anti272_1 undermined the synergy, indicating antagonism also exists. The antagonism may come from interactions between the two peptides. Synergy and its impairment at higher concentrations of Anti272_1 were also found at 21 μM Anti254_1. Anti272_1 had little effect on NUGC3 at concentrations lower than 200 μM, but there was enhancement of the effects of Anti254_1 (Fig. S4B).

Fig. S4.

Combination effect between Anti254_1 and peptides targeting the 272 region. Combination of Anti254_1 and Anti272_1 on killing MKN1 (A) and NUGC3 and NUGC4 (B) cancer cells. (C) Combination of Anti254_1 and Anti272_2 on killing MKN1 cells. Combination of Anti254_1, Anti272_2, and Anti197_2 on killing NUGC4 and NUGC3 (D) and MKN1 (E) cancer cell lines.

Anti272_2 at 200 μM had little effect on MKN1 cells, but greatly enhanced the effect of 14 μM Anti254_1 in a dose-dependent manner (Fig. S4C). At 200 μM, Anti272_2 had synergy with 14 μM Anti254_1 (CI = 0.8), inducing an 80% decrease in viability of MKN1 cells. Fifty micromolar Anti272_2 also enhanced the cytotoxicity of 15 μM Anti254_1 toward NUGC3 (Fig. S4D).

Triple-wise synergy and additivity.

Anti272_2 and Anti197_2 did not show synergistic or additive effects at the concentrations tested; however, combination of Anti254_1, Anti272_2, and Anti197_2 was more effective than a single peptide or combination of two (Fig. S4 D and E) with both NUGC3 and MKN1 cell lines. There was synergy of cytotoxicity of the three peptides in combination (CI = 0.8) toward MKN1 cancer cells. Combination of 100 μM Anti272_2 and 12 μM Anti254_1 decreased their viability to ∼80%. Addition of a further 50 μM Anti197_2 dramatically decreased the viability to ∼14% (Fig. S4E).

Peptides and Peptide Combinations Selectively Kill Cancer Cells.

Anti254_1 had far less effect on a cancer cell line harboring a contact mutant of p53 SW480 (R273H) and a cancer cell line with no p53 (H1299) (Fig. S5A) than on human cancer cell lines containing structural mutants of p53, NUGC3, and SKBR3 (Fig. 8 B and D). It was least effective on noncancerous human fibroblast cell line WI38 (Fig. S5A), demonstrating selective toxicity toward cancer cells. Anti272_2 at concentrations less than 400 μM was toxic to NUGC3 cells but showed no toxicity to WI38. It did not enhance the effect of Anti254_1 on WI38 (Fig. S5B) at concentrations that greatly enhanced the effect of Anti254_1 on NUGC3 cells (Fig. S4D). There was no synergistic effect of Anti254_1 and Anti272_2 or Anti197_2 on H1299 cells (Fig. S5C) at the concentrations at which there was synergy for cytotoxicity toward NUGC3 cells (Fig. 10C and Fig. S4D) and MKN1 cells (Fig. 10D). The combination of 15 μM Anti254_1 and 100 μM Anti272_2 was selectively toxic toward NUGC3 cells, although lacking toxicity toward normal WI38 fibroblast cells (Fig. S5D). The combination of 10 μM Anti254_1 and 50 μM Anti197_2 also had high cytotoxic selectivity toward cancer cells (Fig. S5D).

Fig. S5.

Effects of peptides and their combinations on a cancer cell line lacking p53 (H1299), a cancer cell line harboring a contact mutant of p53 (SW480), and a noncancerous fibroblast cell line (WI38). (A) Dose-dependent effect of Anti254_1 on viability of H1299, SW480, and WI38 cells. (B) Anti272_2 and its combination with 15 μM Anti254_1 had no effect on the viability of WI38. (C) Effect of combination of 15 μM Anti254_1 and 100 μM Anti272_2 and the combination of 10 μM Anti254_1 and 50 μM Anti197_2 on the viability of H1299 cells. (D) Combination of Anti254_1 and Anti197_2 or Anti272_2 induced a significant decrease in the viability of NUGC3 cells, but had no or a minor effect on WI38 cells.

Dependence of cytotoxicity of peptides on p53.

Although Anti254_1 had higher toxicity toward cancer cell lines containing structural mutants of p53 than toward cancer cell lines containing a contact mutant or lacking p53 (Fig. 8 and Fig. S5), it was still significantly toxic to the latter two cancer cell lines (Fig. S5A), indicating a major p53-independent effect of Anti254_1, especially at concentrations higher than 14 μM. A better control is to compare two closely isogenic cell lines. To check further the p53 dependency of Anti254_1, Anti197_1, and other peptides, we knocked down p53 using p53-targeted small interfering RNA (siRNA) and compared their viability with the same cells that had been treated with nontargeted siRNA (Fig. 11). Near-complete deletion of p53 (Fig. S6) did somewhat reduce the effect of Anti254_1 (Fig. 11 A–C) and Anti197_1 (Fig. 11D) on both sets of NUGC3 and MKN1 cell lines. However, there was still a major p53-independent effect of Anti254_1 and Anti197_1, especially at concentrations higher than 15 μM for Anti254_1, which is consistent with observations of the peptides on the other cell lacking p53 or harboring a contact mutant.

Fig. 11.

Effects of p53 knockdown by siRNA on cytotoxicity of peptides. After knockdown for 48 h, cancer cells were washed with PBS. Then, NUGC3 cells were incubated with Anti254_1 for 24 h (A); NUGC3 cells (B) and MKN1 cells (C) were incubated with Anti254_1 and ReACp53 for 48 h; NUGC3 and MKN1 cells were incubated with 12 μM Anti197_1 for 24 h and 48 h (D); NUGC3 cells were incubated with 100 μM Anti272_2, the combination of 15 μM Anti254_1 and 100 µM Anti272_2, and the combination of 15 μM Anti254_1 and 50 μM Anti197_2 for 24 h (E); NUGC3 and MKN1 cells were incubated with various concentrations of Anti197_2 for 24 h or 48 h (F); and NUGC3 and MKN1 cells were incubated with 30 μM Anti254_2 for 24 h (G).

Fig. S6.

Levels of p53 after knockdown by siRNA detected by Western blots. SI, small interfering.

There was significant p53-independent cytotoxicity of combinations of 15 μM Anti254_1 and 100 μM Anti272_2 or 50 μM Anti197_2 (Fig. 11E), as there was with Anti254_2 toward NUGC3 cells. The cytotoxicity toward MKN1 cells was completely independent of p53 (Fig. 11G). The transfection reagent lipofectamine and nontargeting siRNA weakened the effect of Anti197_2, making it ineffective at concentrations lower than 150 μM. At high concentrations of Anti197_2, knocking down p53 actually enhanced the cytotoxicity (Fig. 11F). Because treatment of the transfection reagent greatly affected the efficacy of Anti272_2, we did not test further its effect using siRNA. Similar to Anti254_1, ReACp53 also had a significant p53-independent effect (Fig. 11 B and C).

Peptides did not increase native p53 or reduce denatured p53 in cells.

To test further whether the peptides affected the status of p53, we treated NUGC3 cells with 12 μM Anti254_1 or 120 μM Anti197_2 for 6 h. In contrast to the reported effect of ReACp53 (40), immunostaining showed no dramatic change in the level of p53 in different states (Fig. 12). We detected a great decrease in the denatured p53 level, which is similar to the effect of ReACp53; however, this decrease only happened when the cell nuclei were shrinking, indicating they were dying (Fig. S7).

Fig. 12.

Anti254_1 treatment induces little change in the status of p53. NUGC3 cells were treated with 12 μM Anti254_1 for 6 h before fixation and immunostaining. Pab240 was used to detect unfolded p53, whereas Pab1620 was used to detect p53 in the native status. (Scale bar, 20 μm.)

Fig. S7.

Effect of Anti254_1 on cell nucleus, p53 status, and mitochondria. Anti254_1 caused cell nucleus condensation, little change of native p53 level (A), degradation of unfolded mutant p53 in NUGC3 cancer cells (B), and loss of mitochondria potential. Nuclei were stained with DAPI, and mitochondria were stained with MitoTracker Red CMxRos. NUGC3 cells were treated with 21 μM Anti254_1 or DMSO for 30 min. (Scale bar, 25 μm.)

mRNA levels of p53 target genes.

Consistent with our Western blot results (Fig. 10E), mRNA quantification by real-time PCR showed that treatment of NUGC3 cells with Anti254_1 slightly up-regulated levels of the cyclin-dependent kinase inhibitor p21, whereas incubation with Anti197_2 up-regulated p21 more so (Fig. 13). The p53 upregulated modulator of apoptosis (PUMA) was up-regulated to a very high level (Fig. 13). PUMA up-regulation normally will induce apoptosis; however, the PUMA up-regulation did not lead to cells dying through apoptosis. Anti197_2 induced caspase 3/7 activation to some extent, but no caspase 3/7 activation was detected for Anti254_1 treatment and carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone (z-VAD) could not inhibit the cytotoxicity of the peptides toward NUGC3 cancer cells (Fig. 14). Necrostatin-1 could not inhibit its cytotoxicity either (Fig. 14C), so Anti254_1-induced NUGC3 cancer cell death is not through necroptosis.

Fig. 13.

Effect of peptides on a subset of p53 transcriptional targets. mRNA levels of p53 target genes were measured by real-time PCR after a 4-h treatment. In NUGC3 cells, p21, Noxa, and Puma are all up-regulated, especially when treated with combined Anti254_1 and Anti197_1. The MDM2 level was not greatly affected.

Fig. 14.

Effect of Anti254_1 and Anti197_2 on caspase 3/7 activity. Caspase 3/7 activity was induced by treatment with 21 μM Anti254_1 for different times (A) and 150 μM Anti197_1 for 6 h (B). (C) Fifty micromolar Necrostatin-1 (Nec-1), 50 μM inactive necrostatin-1 (Nec-1-inact), and 100 μM carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone (z-VAD-fmk; z-VAD) show little inhibition on the activity of Anti254_1. (D) One hundred micromolar z-VAD attenuates the effect of Anti197_2. Cell viability was measured after 24 h of incubation. For treatment in the presence of Nec-1, Nec-1-inact, or z-VAD, cells were treated with Nec-1, Nec-1-inact, or z-VAD for 3–4 h before addition of Anti254_1 or Anti197_2, and the reported viability is relative to treatment with only Nec-1, Nec-1-inact, or z-VAD, respectively. The viability of NUGC3 cells after treatment with Nec-1, Nec-1-inact, or z-VAD alone was 70 ± 4%, 91 ± 5%, and 92 ± 3%, respectively, relative to the DMSO control.

We conclude that there is significant p53-independent cytotoxicity of the peptides in our cell lines that are on 2D plates and the presence of mutant p53 enhances the toxicity somewhat in some cases.

Discussion

Denatured p53 Contains Several Aggregation-Prone Sequences.

Many proteins will form amorphous aggregates when their denatured state in urea is rapidly diluted into lower concentrations of denaturant, and p53 is no exception. Its aggregate formed from its denatured state in 3 M urea contained sequences that were highly protected against controlled proteolysis. They contained a site identified only by limited proteolysis (S5) and sites that were predicted to be prone to form fibrillar aggregates, among which were S7 and S10 in addition to those regions predicted by ZipperDB (40) (Table 2). Mutation of several of those sites to make them less aggregation-prone inhibited greatly the overall rate of aggregation in 3 M urea (Table 2). One of those sequences, residues 251–257 (ILTIITL) in β-strand S9, is proposed to be the key aggregation-prone sequence (32). Indeed, mutating I254 to R or D decreased the rate of aggregation in 3 M urea by an estimated greater than 10⁴-fold (extrapolated from Fig. S3A). Very large decreases were found for mutations in S5 and S10, and significant decreases were found elsewhere (Table 2). There is clearly not a unique aggregation sequence in the denatured state of p53, but a cooperative network of such sites.

Sequences involved in aggregation of p53 from its native state.

Using limited proteolysis in vitro, we found four main sequences protected in the aggregate core formed on incubation of the native state at 37 °C (Table 1): 249–267 (S9), 182–213 (H1S5S6), 268–282 (S10H2), and 102–120 (S1L1), which are sequences that overlap with those sequences found to be aggregation-prone in the denatured state. Mutating those sequences to be less aggregation-prone did not inhibit p53 aggregation in vitro; instead, they destabilized p53 and made it aggregate faster by increasing the rate of unfolding, just as did the I254D and I254R mutations.

The slow initial steps in aggregation from the native state are the sequential unfolding of two molecules of p53, followed by rapid polymerization to form ThT-binding oligomers, which then slowly rearrange to large aggregates (27, 28). The denatured states of all of the mutants aggregated far more rapidly in the absence of urea (Fig. S3A) than did the native state. Mutation of I254D greatly speeds up the rate of unfolding. Further, the rate of aggregation of denatured p53 is 500- to 1,000-fold faster in water than in 3 M urea, so that even alternative, less aggregation-prone sequences can aggregate sufficiently fast. Accordingly, although mutating I254 to D254 virtually eliminated aggregation of p53 in 3 M urea (Fig. 4), it actually speeded up aggregation, as did R254, in the absence of urea at 37 °C and neutral pH, which are conditions close to physiological (Fig. 2), and produced an aggregate of similar morphology to WT p53 (Fig. S2).

Scheme for multisite aggregation.

The kinetic and analytical results may be summed up in Fig. 15 in a mechanism that may be a basis for other proteins with multiple aggregation-prone sites. One molecule of p53 unfolds, partially or fully, to expose its aggregation-prone sequences (25, 27, 28, 50). Another molecule is induced to unfold to give the elongation-competent state. Aggregation then occurs by elongation and cross-linking, either by homologous pairing of sequences or by combination of homologous and heterologous pairing. If one of the aggregation sites is eliminated by mutation, then there are sufficient remaining sites for aggregation to proceed. We conclude that elimination of a single aggregation-prone site will not prevent aggregation of p53 in vitro because of the presence of other aggregation-prone sites.

Fig. 15.

Schematic mechanism for aggregation of p53. (A) Aggregation process. (B) Aggregation could not be blocked by only targeting one aggregation-driving region.

Implications for inhibiting aggregation in vivo.

Because eliminating the aggregation propensity of the major aggregation-prone site surrounding Ile254 or any of the other aggregation-prone sites did not inhibit aggregation of p53 in vitro, we wondered how a single peptide, ReACp53, that caps the buried Ile254 region in the native state could be effective in rescuing mutant p53 in cancer cells (40). Transthyretin, for example, has two aggregation-prone sites and requires two peptides, one to cap each and one to inhibit aggregation (51). To investigate the possibility of inhibiting aggregation in cancer cells, we designed peptides targeting three different sequences that drive aggregation of p53. They inhibited in vitro the self-aggregation of isolated peptides from p53 that contained the aggregation-prone sequences, but not the aggregation of full-length p53. Nevertheless, the peptides preferentially killed cancer cells carrying destabilized structural mutants of p53. The peptide Anti254_3 that targeted the S9 (251–257) region inhibited the self-aggregation of the isolated S9 peptide in vitro and in killing cancer cells to the same extent but at a fourfold lower concentration than the reported peptide ReACp53 (40). Peptides targeting the S9 region acted synergistically in cytotoxicity with peptides targeting other amyloidogenic regions. Both Anti254_1 itself and the synergistic combinations could selectively kill cancer cells at concentrations not affecting normal cells. At first sight, it seems a compelling story that peptides and peptide combinations that cap amyloidogenic sequences are anticancer drug leads, and the approach of targeting different aggregation-driving regions of p53 represents a promising anticancer strategy. However, we could not verify that p53 was the target. In vitro, the peptides did not inhibit the aggregation of full-length p53, consistent with our observations that knocking out amyloidogenic sequences by mutagenesis does not inhibit aggregation. We could effectively knock down the expression of the p53 gene using siRNA in cell lines but still could not eliminate, for example, the selective cytotoxicity of Anti254_1. At high concentrations, Anti254_1 killed cancer cells harboring a stable contact mutant of p53 or even cancer cells lacking p53.

There are many peptides that have anticancer effects. In particular, amphiphilic peptides, especially those amphiphilic peptides containing long cationic tails, selectively bind to cancer cells and cause cell death by cytoplasmic membrane disruption, whereas others trigger apoptosis via death receptor or mitochondrial pathways (52, 53). It should be noted, however, that a version of ReAcp53 with a shuffled sequence is not cytotoxic (40). We did a preliminary analysis of the mechanism of cell death induced by the peptides. The results were inconsistent with p53-induced apoptosis and consistent with known cytotoxic pathways of anticancer peptides. Anti254_1 killed cancer cells with a half-time of about 1 h, which is far faster than is usually found in p53-induced, transcription-dependent apoptosis (Fig. S8 A–C), and it induces a rapid decrease of mitochondrial membrane potential within 1 h (Fig. S8D). Mitochondrial p53 translocation can also induce fast transcription-independent apoptosis. However, we did not find translocation of p53 to mitochondria after a 30-min treatment of 21 μM Anti254_1 (Fig. S7). This fast cell death is not via activating caspase 3/7, but is likely to occur via necrosis, although leading to cell nucleus condensation (Fig. 14 and Fig. S7). Similarly, necrostatin-1 and pan-caspase inhibitor show only partial inhibition on toxicity of ReACp53 (40).

Fig. S8.

Time courses of peptide-induced rapid changes in viability and mitochondrial potential of cancer cells. NUGC3 and NUGC4 cancer cells were treated with 15 μM Anti254_1 (A) and 21 μM Anti254_1 (B) for the indicated times at 37 °C and with 10 μM Anti254_1 and 50 μM Anti197_2 (C). (D) Mitochondrial potential change was induced by 21 μM Anti254_1.

In contrast to our finding of multiple aggregation sites and that mutating I254 to R or D speeds up aggregation, Xu et al. (32) report only the finding of the I254-based sequence on proteolysis of harvested aggregates of p53, and find that mutation of I254R prevents aggregation, abrogates oncogenic gain of function, and inhibits binding of mutant p53 with p63/p73. However, others report p53I254R does bind to p63/p73, has gain of function, and promotes invasion (33).

How do the peptides act?

There is little doubt that ReACp53 (40), and now other peptides that target further amyloidogenic sequences within oncogenic mutant p53, are indeed cytotoxic toward cancer cells. The question now is how they act. There is not a simple explanation for several reasons, including that they may work simultaneously by multiple routes and that there are pleiotropic effects on mutating p53, including both loss and gain of function. An example of multiple routes is a small alkylator that covalently binds to mutant p53 in some cases, stabilizes it, and rescues its activity in some cancer cell lines, but kills other cancer cell lines via a reactive oxygen species (54). Similarly, several small molecules appear to reactivate p53 but, instead, cause cell death by other routes (3). There is major p53-independent activity in several cell lines we tested, but that finding does not preclude possible major p53-dependent activity in other cell lines or in tumors. Do the peptides prevent loss of function by directly inhibiting aggregation? If p53 aggregates in cells by the same process as in vitro, then the peptides are unlikely to function by inhibiting its aggregation from the native state. However, there may be as yet unknown effects of chaperones in vivo, perhaps by binding to partly denatured states. Do they affect gain of function? Mutant p53 may be oncogenic because of gain of function by interacting with other proteins or DNA. For example, peptides may function partly by screening the interaction between p53 and the transactivation inhibition domain of alpha isomers of p63/p73 (55), Indeed, a preliminary experiment knocking down p63 showed cell death was dependent on p63 (Fig. S9). Peptides might affect the oncogenic interplay between Nrf2 and mutant p53 (56). There is evidence for activity by more classical mechanisms of disruption of cell membranes and other processes, especially at high concentrations. We must emphasize that our results in cancer cell lines parallel the results found in a study by Soragni et al. (40), and that experiments in conventional 2D cultures as used by us may overestimate nonspecific cytotoxicity compared with 3D cultures and solid tumors. However, consistent with our results in cancer cells, our in vitro data do point to non-p53 entities as being important targets of the antiaggregation peptides, as well as possibilities of their interactions with p53.

Fig. S9.

Effects of p63 knockdown by siRNA on cytotoxicity of peptides. After knockdown for 48 h, NUGC3 cancer cells were washed with PBS and then incubated with compounds for 48 h.

Methods

Prediction of Aggregation Sites.

Aggregation sites were predicted using the programs Tango (41), Waltz (42), Aggrescan (43), FoldAmyloid (44), and Amylpred2 as well as by a consensus method combining the first three and NetCSSP, amyloid mutants, Pafig, amyloidogenic pattern, SecStr, average packing density, β-strand contiguity, and hexapeptide conformational energy (45). Sequences predicted to be aggregation-prone by at least six programs were chosen to be the aggregation sites. A threshold of 90% and pH 7.0 were used when predicted by Waltz. For prediction by FoldAmyloid, the expected number of contacts within 8 Å was used as a scale; the averaging frame and reliable frame are both 5, and the cutoff value was 21.4.

Kinetics of Aggregation.

Aggregation kinetics of native p53 variants were monitored by light scattering and ThT fluorescence using a Horiba FluoroMax-3 spectrophotometer as described previously (25). For aggregation of denatured p53 variants, proteins were denatured in 4 M urea on ice overnight before measurements. Experiments were generally performed with a protein concentration of 10 μM in 25 mM potassium or sodium phosphate (pH 7.2), 150 mM NaCl, 1 mM Tris(2-carboxyethyl)phosphine (TCEP), 5% (vol/vol) DMSO, 20 μM ThT, and 3 M urea, and were monitored as for the native proteins. To test the effect of antiaggregation peptides on aggregation of S9 peptide (Acetyl-RPILTIITLE-amide), S5–6 peptide, and S10 peptide, S9, S5–6, or S10 peptide in 6 M GdmCl was diluted into buffer finally containing 25 mM potassium phosphate (pH 7.2), 150 mM NaCl, 1 mM TCEP, 22.7 μM or 20 μM ThT, and certain concentrations of GdmCl and antiaggregation peptides as indicated. Antiaggregation peptide stocks were prepared with sterilized Milli-Q water. To test the initial aggregation rate of denatured p53 mutants in different concentrations of urea, 3 μM p53 mutant was diluted into the above buffer with various concentrations of urea at 37 °C.

Limited Proteolysis.

WTC, WTCI254D, WTFL, and WTFLG245S were incubated at 37 °C usually for 2 h (also at 1 and 3 h for WTC and overnight for WTCI254D) in 25 mM potassium phosphate (pH 7.2), 150 mM NaCl, 1 mM TCEP, and 5% (vol/vol) DMSO. Aggregates were harvested by centrifugation at 4 °C at 15,682 × g for 30 min and washed once with proteolysis buffer [20 mM Tris⋅HCl (pH 7.4), 1 mM TCEP] to remove residual phosphate buffer and soluble p53. Limited proteolysis of either the aggregates or soluble native states of WTC, WTCI254D, WTFL, and WTFLG245S with trypsin was carried out in 20 mM Tris⋅HCl (pH 7.4) and 1 mM TCEP at 20 °C. The enzyme/substrate (E/S) ratio for WTC and WTCI254D was 1:50 (wt/wt), and the enzyme/substrate ratio for WTFL and WTFLG245S was 1:100 (wt/wt). At a specified time, the remaining aggregate was separated from supernatant by centrifugation at 15,682 × g for 30 min. The reaction was quenched with acetic or trifluoroacetic acid. The remaining aggregate was dissolved in 70% (vol/vol) acetonitrile/3% (vol/vol) trifluoroacetic acid. Mass determination was performed using a MALDI-TOF mass spectrometer (Voyager-DE Pro; Applied Biosystems). Proteolysis by proteinase K of native WTC and aggregate of WTC was carried out using an E/S ratio of 1:22.5 (wt/wt) at 20 °C. The reaction was quenched by heating at 90 °C for 5 min. Mass and sequence determination of the resulted peptides was performed using both MALDI-TOF and Ultraflex III MALDI-TOF/TOF (Bruker Daltonics) MS.

Cell Lines and Culture Conditions.

NUGC3 (p53-Y220C^+/+), NUGC4 (WT p53^+/+), and MKN1 (p53-V143A^+/+) cells were obtained from the Japan Health Science Research Resources Bank, and they were maintained in RPMI medium. SKBR3 was purchased from the American Type Culture Collection, and MCF7 (WT p53^+/+), SW480, and WI38 fibroblast cells were maintained in DMEM. All of the media were supplemented with 10% FBS and 1% penicillin/streptomycin (10,000 U/mL penicillin, 10,000 μg/mL streptomycin). The FBS was heat-inactivated. Other cell lines were cultured in RPMI 1640 GlutaMAX medium with the same concentration of serum and antibiotics. All cell cultures were maintained at 37 °C and in 5% CO₂ in a humidified incubator.

Cell Viability Assay.

Cells (7,500 cells per well) were seeded in 96-well plates and cultured to about 60% confluence on the second day. Then, old medium was replaced by new medium with peptides or DMSO control. When test peptides were combined, peptides were added to the cells simultaneously. After 24-h treatment, except if indicated otherwise, cell viability was assessed by measuring the intracellular levels of ATP using a Cell Titer-Glo Luminescent Cell Viability Assay Kit (Promega) according to the manufacturer’s instructions.

Immunofluorescence.

Cells were treated with peptides or DMSO control for the indicated time and were then washed with PBS and fixed with 4% (vol/vol) paraformaldehyde for 10 min at room temperature. After being rinsed with PBS, cells were permeabilized with 0.5% (wt/vol) Triton X-100 in PBS for 5 min and blocked with 2% (wt/vol) BSA or 5% goat serum. The primary antibodies were incubated overnight at 4 °C, and secondary antibody goat anti-mouse Dylight488 was diluted to 1:1,000 and incubated for 1 h. The following primary antibodies were used: anti-p53 antibody Pab 1620 (Abcam) and anti-p53 antibody Pab 240 (Santa Cruz Biotechnology). Hoechst 33342 (Cell Signaling) or DAPI and MitoTracker Red (Lonza) were used to stain the nucleus and mitochondria of cells, respectively. Images were acquired using a Leica TCS SP8 confocal microscope.

Western Blots.

Cell lysates were prepared with radioimmunoprecipitation assay buffer (Sigma) containing a protease inhibitor mixture (Roche) after treatment of peptides or DMSO control. The lysates were run on SDS/PAGE and transferred onto polyvinylidene fluoride membranes. Membranes were blocked for 1 h with 5% (wt/vol) milk in Tris-buffered saline containing 0.1% Tween-20 (TBST) at room temperature before immunoblotting. The membranes were incubated with primary antibodies at 4 °C overnight. After three rinses with TBST, the membranes were stained with horseradish peroxidase-conjugated anti-mouse or anti-rabbit IgG (Dako) and imaged with ChemiDoc XRS (Bio-Rad). The following primary antibodies were used: p53 DO-7 (Dako), p21 (Millipore), MDM2 (Abcam), Bax (Cell Signaling), and β-actin (Abcam).

p53 Knockdown by siRNA.

Cancer cells were transfected with the siRNA directed against human p53 (Cell Signaling Technology), p63 (OriGene), or nontargeting negative control siRNA (Qiagen) using Lipofectamine RNAiMAX (Life Technologies) according to the manufacturer’s protocol. The final concentration of the siRNAs was 10 nM.

SI Methods

Caspase3/7 Activation Assay.

Caspase 3/7 activity was measured using the ApoTox-Glo Triplex Assay (Promega) according to the manufacturer’s instructions. Cells were cultured at 20,000 cells per well. Luminescence readings were made using a PHERAstar plate reader (BMG Labtech).

Measurement of Mitochondrial Membrane Potential Change.

Changes in mitochondrial membrane potential were measured using a TMRE-Mitochondrial Membrane Potential Assay Kit (Abcam) in 96-well plate following the manufacturer’s instructions.

RT-PCR.

NUGC-3 cells were treated with peptides or control for 4.5 h. Total RNA was extracted and purified using an RNeasy Mini Kit from Qiagen according to the manufacturer’s protocol. The cDNA was synthesized with an iScript cDNA Synthesis Kit (Bio-Rad). Real-time PCR was performed using a Rotor-Gene SYBR Green Kit (Qiagen) on a Rotor-Gene 6000 (Corbett Life Science) real-time PCR cycler. Each sample was measured in triplicate.

Transmission Electron Microscopy.

WTC or WTCI254D (3 μM) was incubated at 37 °C in 25 mM Tris⋅HCl (pH 7.2), 150 mM NaCl, 1 mM TCEP, and 5% DMSO overnight. Samples (5 μL) were adsorbed onto freshly glow-discharged formvar film 400 mesh copper grids, rinsed with deionized water, and stained with 1% uranyl acetate. Images were taken using an FEI Tecnai Spirit transmission electron microscope at 80 kV.