Recifin A, Initial Example of the Tyr-Lock Peptide Structural Family, Is a Selective Allosteric Inhibitor of Tyrosyl-DNA Phosphodiesterase I

Abstract

Tyrosyl-DNA phosphodiesterase 1 (TDP1) is a molecular target for the sensitization of cancer cells to the FDA-approved topoisomerase inhibitors topotecan and irinotecan. High-throughput screening of natural product extract and fraction libraries for inhibitors of TDP1 activity resulted in the discovery of a new class of knotted cyclic peptides from the marine sponge Axinella sp. Bioassay-guided fractionation of the source extract resulted in the isolation of the active component which was determined to be an unprecedented 42-residue cysteine-rich peptide named recifin A. The native NMR structure revealed a novel fold comprising a four strand antiparallel β-sheet and two helical turns stabilized by a complex disulfide bond network that creates an embedded ring around one of the strands. The resulting structure, which we have termed the Tyr-lock peptide family, is stabilized by a tyrosine residue locked into three-dimensional space. Recifin A inhibited the cleavage of phosphodiester bonds by TDP1 in a FRET assay with an IC₅₀ of 190 nM. Enzyme kinetics studies revealed that recifin A can specifically modulate the enzymatic activity of full-length TDP1 while not affecting the activity of a truncated catalytic domain of TDP1 lacking the N-terminal regulatory domain (Δ1–147), suggesting an allosteric binding site for recifin A on the regulatory domain of TDP1. Recifin A represents both the first of a unique structural class of knotted disulfide-rich peptides and defines a previously unseen mechanism of TDP1 inhibition that could be productively exploited for potential anticancer applications.

Graphical Abstract

INTRODUCTION

Relaxation of supercoiled DNA by topoisomerases is necessary for the normal cell functions of DNA transcription, replication, recombination, and repair.¹ Topoisomerase I (TOP1) mediates both DNA strand break and religation by forming a transient, covalent 3′-;-phosphotyrosyl bond with the DNA substrate (Figure 1).² This TOP1-DNA cleavage complex is the target of chemotherapeutic TOP1 inhibitors such as the natural product camptothecin. Irinotecan, an analogue of camptothecin, is a widely used anticancer agent that stabilizes the TOP1-DNA cleavage complex, causing irreversible double-strand DNA breaks, eventually leading to the death of replicating cancer cells.^3,4 Tyrosyl-DNA phosphodiesterase 1 (TDP1) is an enzyme that, upon recognizing stalled TOP1-DNA cleavage complexes, catalyzes the cleavage of the 3′-phosphotyrosyl bond between DNA and TOP1 (Figure 1).⁵ TDP1 is composed of an as-yet nonstructurally characterized N-terminal regulatory domain whose function has been reported to be modulated by both phosphorylation and SUMOylation and a C-terminal catalytic domain that utilizes two histidine residues to effect phosphodiester cleavage at Tyr723 of TOP1 (Figure 1).⁶ After removal of the 3′ adduct, polynucleotide kinase phosphatase prepares the degraded DNA strands for further repair by DNA polymerase β and DNA ligase III.⁷ The clearance of TOP1-DNA complexes (Figure 1) results in escape from TOP1 inhibitor-induced cell death. This activity has led to the consideration of TDP1 as a molecular target for the sensitization of replicating cancer cells to camptothecin and related chemotherapeutic agents.^6,8,9

Figure 1.

TDP1 processing of 3′-TOP1-DNA adducts and inhibition by recifin A. TOP1 catalyzes single-strand DNA breaks via a transitory, covalent phosphotyrosine linkage involving tyrosine 723 (pTyr723). The cleavage complex is stabilized by the natural product camptothecin, which inhibits DNA religation, trapping TOP1 on the DNA strand, ultimately leading to double-strand breaks and cell death. TDP1 removes the 3′-TOP1-pTyr-DNA adducts via a nucleophilic attack on the phosphodiester bond by histidine 263 (HIS263) and subsequent hydrolysis by histidine 493 (HIS493). After removal of the 3′ adduct, the DNA strand is further enzymatically repaired and religated. Inhibitors of TDP1 catalytic activity, such as recifin A (depicted here by its electrostatic surface potential model), have the potential to sensitize cancer cells to TOP1 poisons.

Marine organisms are known to be a diverse source of bioactive peptides and proteins including lectins, defensins, conotoxins, cyclic peptides, and depsipeptides.^10,11 Previously reported cysteine-rich peptides (CRPs) isolated from marine sponges include asteropine A,¹² the asteropsins,^13–15 the neopetrosiamides,^16,17 and the barrettides.¹⁸ Within the Porifera, the genus Axinella itself has yielded many bioactive molecules, including cyclic peptides, lipopolysaccharides, lectins, alkaloids, terpenes, lipids, and polyether macrolides.^19–27 We recently conducted a high-throughput screen to identify natural product inhibitors of TDP1 activity.²⁸ During this effort, inhibitory activity was identified in the aqueous extract of the marine sponge Axinella sp. Here we report the first CRP isolated from the genus Axinella, which we have named recifin A after the Cape Recife Nature Reserve in South Africa from which the source sponge was harvested. The 42-residue peptide was sequenced, the disulfide connectivity was elucidated, and the unique three-dimensional structure of the peptide was solved using solution state homonuclear NMR spectroscopy. The isolated peptide recifin A is shown to specifically modulate the enzymatic activity of full-length TDP1 but not an enzymatically active N-terminal truncated variant (Δ147TDP1) lacking the regulatory domain, suggesting an allosteric recifin A binding site within the regulatory domain of TDP1.

RESULTS AND DISCUSSION

Isolation and Primary Structure Determination.

The aqueous extract of the marine sponge Axinella sp. was identified as active in a previously published high throughput screen for TDP1 inhibitory activity.²⁸ The aqueous extract was first subjected to vacuum-assisted, wide-pore C₄ chromatography and eluted using a stepwise methanol gradient. Fractions were tested for TDP1 inhibitory activity and subjected to liquid chromatography-mass spectrometry (LC-MS) analysis. Peptidic molecular charge envelopes were observed in LC-MS analysis of the active fractions, and the proteinaceous composition of active components was confirmed by SDS-PAGE. A family of four main Axinella peptides was isolated from the initial chromatographic step with observed average masses of 4683.87, 4785.89, 4915.95, and 5674.47 Da (Figure S1).

A combination of reversed-phase-high performance liquid chromatography (RP-HPLC) and bioassay-guided fractionation²⁸ was used to isolate the most abundant and most active peptide, recifin A (MW 4915.95 Da), to homogeneity. The yield of purified recifin A from the crude aqueous extract was approximately 0.1% w/w. Recifin A was found to inhibit full-length recombinant human TDP1 enzymatic activity in a concentration-dependent manner with an IC₅₀ of 2.4 μM in a biochemical assay for cleavage of a 5′-radiolabeled oligonucleotide DNA substrate containing a 3′-phosphotyrosyl residue (Figure 2).²⁹ Importantly, recifin A retained the ability to inhibit TDP1 processing of the radiolabeled oligonucleotide within a whole-cell extract assay context, indicating the specificity and stability of the molecule. This is significant as it shows that recifin A could exert its inhibitory activity against native TDP1 in the presence of other cellular macromolecules and against an enzyme whose regulatory domain had potentially been post-translationally modified. The other main Axinella-derived peptides showed weaker TDP1 inhibitory activity, indicating they are likely additional members of the same structural class of peptides and will be characterized more fully in subsequent research studies (Figure S2).

Figure 2.

Recifin A inhibition of full-length human TDP1 enzymatic activity. Serial dilutions of purified recifin A were combined with a synthetic 5′[³²P]-labeled, 3′-phosphotyrosine capped oligonucleotide DNA substrate and incubated with either full-length recombinant human TDP1 (rhTDP1) or human TDP1 complemented DT40 knockout whole cell extracts (hTDP1 WCE). Reactions were subjected to polyacrylamide gel electrophoresis and phosphorimaging. Activity was calculated as percent of noninhibited substrate cleavage reaction control.

The primary amino acid sequence of recifin A was determined using a combination of tandem mass spectrometry (MS/MS) and automated Edman degradation. The monoisotopic mass of the intact recifin A peptide was observed to be 4912.9661 Da. Upon disulfide reduction and alkylation with 4-vinylpyridine (4-VP, 105.06 Da), a mass increase of 636.38 Da was observed, which indicated the reduction of three disulfide bonds and conversion of six cysteine residues to S-pyridylethyl cysteine (Figure S3). Neither the native peptide nor the 4-VP alkylated peptide was amenable to N-terminal amino acid sequencing by Edman degradation, which suggested a blocked N-terminus. A trypsin digest of the 4-VP alkylated peptide was performed which generated three fragments A, B, and C with molecular weights of 1205.48, 2136.94, and 2242.95 Da, respectively (Figure S4).

Sequencing of the tryptic fragments by LC/MS/MS and confirmation by Edman degradation (Figure S4A–C)³⁰ indicated the presence of a pyroglutamic acid residue (pGlu) on the N-terminus of fragment A explaining the lack of success with N-terminal Edman degradation of recifin A. This was confirmed by selective cleavage of the pGlu with Pfu pyroglutamate aminopeptidase. Upon successful enzymatic removal of the N-terminal pGlu from the reduced, alkylated recifin A, 35 contiguous amino acids of the N-terminally truncated peptide were able to be sequenced by Edman degradation. In addition to trypsin digestion, the alkylated peptide was subjected to digestion with chymotrypsin, glutamic acid C-terminal (Glu-C), and proline endopeptidases. The resultant fragments were sequenced by CID MS/MS only (Figure S5) and confirmed the full sequence of recifin A. The theoretical mass of the proposed amino acid sequence of recifin A was 4918.9994 Da, which differed from the observed mass by 6.0333 Da, confirming the presence of three disulfide bonds (2.1 ppm mass error).

Disulfide Bond Mapping.

The disulfide bonding pattern of recifin A was determined using a partial reduction and alkylation approach as previously reported.³¹ A combination of 80 μM recifin A and 50 mM tris(2-carboxyethyl)phosphine (TCEP) yielded one (two intact cystines, 2-SS), and two disulfide bond (one intact cystine, 1-SS) reduction events and the fully reduced species (zero intact cystines, 0-SS) as shown in Figure 3A.

Figure 3.

Disulfide mapping of recifin A. (A) RP-HPLC analysis of partially reduced and alkylated recifin A. The mixture of native recifin A (3 intact disulfides/3-SS), partially reduced and alkylated isoforms (2 intact disulfides/2-SS, 1 intact disulfide/1-SS), and completely reduced and alkylated recifin A (0-SS) was desalted and separated by RP-HPLC prior to further analysis. (B) MS/MS sequencing results for the 2-SS recifin A isoform trypsin fragments established the Cys IV–VI disulfide linkage. (C) Proposed recifin A disulfide bonding pattern: Cys I–III, Cys II–V, and Cys IV–VI. Abbreviations: pGlu, pyroglutamic acid; IAA, iodoacetamide alkylated cysteine; NEM, N-ethylmaleimide alkylated cysteine.

The main 2-SS, N-ethylmaleimide (NEM) alkylated peptide isoform (Figure 3A) was fully reduced, alkylated, and digested with trypsin. The resultant trypsin fragments were sequenced to map the positions of the alkylation events (Figure 3B and Figure S6). NEM-alkylated cysteine residues were found to be at positions 22 and 42, which mapped a projected cystine linkage at Cys IV–VI. The 2-SS, NEM-alkylated peptide was digested with chymotrypsin to map the remaining, intact disulfide linkages by LC-MS. MassHunter (Agilent Technologies, Inc.) software was used to construct a database of the three possible disulfide-linked sequence permutations (Cys I–II, Cys III–V, Cys IV–VI; Cys I–III, Cys II–V, Cys IV–VI; and Cys I–V, Cys II–III, Cys IV–VI) and to match the observed chymotrypsin fragment masses to a set of theoretical digest fragment masses. A limitation of 5 ppm mass error was applied to the fragment matching process. Only fragments that linked Cys I–III and Cys II–V were observed (Supporting Table 1). Taken together, the data indicated the disulfide bond connectivity of recifin A to be Cys I–III, Cys II–V, and Cys IV–VI (Figure 3C).

The molecular weight, number of cysteine residues, and the stability of recifin A are similar to those reported for members of the inhibitory cystine knot (ICK) family, comprising protease inhibitors, toxins, and antimicrobial peptides. However, the ICK family is characterized by the intertwined or “knotted” Cys I–IV, Cys II–V, Cys III–VI disulfide bond arrangement.³² The recifin A disulfide bond framework is Cys I–III, Cys II–V, and Cys IV–VI, so while recifin A is a CRP, the peptide is not a member of the ICK family. Indeed, the primary amino acid sequence of recifin A is not homologous to any sequence within the nonredundant GenBank translated protein database (BLASTp search). While recifin A does display some component similarities to CRPs isolated from the marine sponge genus Asteropus,^12–15 there were no identified amino acid sequence alignments of recifin A to Asteropus-derived CRPs (or ICK peptides) within the KNOTTIN database (Figure S7).³³

Solution NMR Spectroscopy Structure Determination.

Given the lack of sequence homology to known proteins and unexpected disulfide array when compared to other CRPs, we subjected recifin A to solution NMR spectroscopy in order to characterize its three-dimensional structure. The one-dimensional ¹H NMR spectrum showed excellent signal dispersion across the entire spectral region indicating a well-structured peptide (Figure S8A). Homonuclear ¹H TOCSY and NOESY data were used for sequential assignments as described previously (Figure S9).³⁴ This process proved a significant challenge due to a number of unusual chemical shifts and features in the NMR data for recifin A. Chemical shift anomalies included the Gly16 HN proton at 5.51 ppm, upfield of several Hα protons. The Hβ resonances of Tyr40 and Pro35 were observed at 1.15 and −0.33 ppm, respectively, the latter being the most upfield resonance in the spectrum. Finally, the Hα of Cys11 was essentially overlapping one of the Hβ resonance at 2.68 ppm. Resonances observed at 4.94 and 5.58 were, after identification of TOCSY peaks to their respective Hβ protons, assigned as the hydroxyl protons of Ser27 and Ser29, and a resonance at 7.96 was assigned as the phenolic proton of Tyr6 because of a lack of TOCSY peaks but strong NOESY connections to Tyr6 Hε. These protons are all not expected to be visible in the spectra due to fast exchange with the solvent but in the recifin A structure must clearly be involved in strong hydrogen bonds and protected from the solvent. Finally, four individual aromatic ¹H signals were identified for Tyr6 (Hδ1, Hδ2, Hε1, Hε2) revealing that it is positioned in a tightly packed environment where ring-flips are sufficiently slowed down to prevent averaging into the typically observed single Hδ* and Hε* resonances. Line broadening, suggesting dynamics, was also observed around residues 21–25, with the HN proton of Arg25 broadened beyond detection. In addition to the homonuclear data, a ¹H–¹³C HSQC data set was recorded at natural abundance, which was essential for confirming all proton assignments and provided ¹³C chemical shift information for dihedral restraints.

Initial analysis of secondary Hα chemical shifts suggested secondary structural features in the form of short β-strands and α-helices/turns, as indicated by significant positive and negative shifts, respectively (Figure S8B).³⁵ The three-dimensional solution structure of recifin A was calculated using torsion angle dynamics in Cyana followed by refinement in a watershell using crystallography and NMR system (CNS).³⁶ A total of 425 distance restraints, including 403 distance restraints derived from NOEs, 22 hydrogen bond restraints, and 75 dihedral angle restraints (ϕ, ψ, χ) were included in the calculations (Table 1). A family of 20 structures were chosen to represent the solution structure of recifin A based on energies, stereochemical quality, and consistency with the experimental data (Table 1). As seen from the superposition of the ensemble, the structure is well-defined except a loop region comprising residues 21–25, consistent with the observed line broadening (Figure 4). The structure is dominated by a central, antiparallel β-sheet comprising four strands involving residues 4–6, 14–16, 27–29, and 40–41 and two short 3₁₀ helical turns involving residues 21–23 and 36–38. The elements of secondary structure are stabilized by the three disulfide bonds, with the Cys5-Cys21 and Cys22-Cys42 disulfides bracing the 21–23 turn to strands 1 and 4, respectively, and the Cys11-Cys39 cross-bracing two loops. Intriguingly, the disulfides form an embedded ring together with their backbone segments, through which the third strand (27–29) is threaded. This arrangement gives rise to a previously not observed fold and represents a new type of cysteine-rich peptide knot. Although this is somewhat reminiscent of the inhibitory cystine knot, where two of the disulfide bonds form a ring structure through, which the third disulfide bond is threaded forming the knot,^37,38 it bears perhaps even more resemblance to the lasso peptides, in which the peptide backbone is threaded through a ring formed by an N-terminus to side-chain carboxyl lactam bond (Figure 5) and locked in place by two aromatic residues.³⁹

Table 1.

Statistical Analysis of the 20 Best Structural Models of Recifin A Based on MolProbity Scores

distance restraints
intraresidue (|i – j| = 0)	126
sequential (|i – j| = 1)	115
medium range (|i – j| ≤ 5)	50
long range (|i – j| > 5)	112
hydrogen bonds	22
total	425
dihedral angle restraints
Φ	25
φ	25
χ	25
total	75
structure statistics
energies (kcal/mol, mean ± SD)
overall	−1422.3 ± 36.5
bonds	16.9 ± 1.2
angles	49.8 ± 4.4
improper	19.3 ± 2.4
dihedral	181.4 ± 1.6
van de Waals	−221.2 ± 3.9
electrostatic	−1469.1 ± 35.2
NOE (experimental)	0.03 ± 0.01
constrained dihedrals (experimental)	0.6 ± 0.3
atomic RMSD (Å)
mean global backbone (1–42)a	0.93 ± 0.31
mean global heavy (1–42)a	1.57 ± 0.26
mean global backbone (3–17, 27–42)	0.44 ± 0.08
mean global heavy (3–17, 27–42)	1.17 ± 0.15
MolProbity statistics
Clash score, all atomsb	10.02 ± 1.9
poor rotamers	0 ± 0
favored rotamers	95.8 ± 1.6
Ramachandran outliers (%)	0 ± 0
Ramachandran favored (%)	94.9 ± 3.3
MolProbityc score	1.8 ± 0.2
MolProbity percentile	82.9 ± 8.3
violations
distance constraints (>0.5 Å)	0
dihedral-angle constraints (>5°)	0

^aPairwise RMSD from 20 refined structures over amino acids 1–42.

^bNumber of steric overlaps (>0.4 Å)/1000 atoms.

^c100% is the best among structures of comparable resolution. 0% is the worst.

Figure 4.

NMR solution structure of recifin A. (A) The 20 best structures based on MolProbity scores superposed over residues 3–18 and 26–42, emphasizing the well-ordered core. (B) Ribbon structure highlighting the four antiparallel β-strands (I–IV). (C) Ribbon structure depicted in panel B rotated 90°, illustrating the threading of the third β-strand through the ring formed by the three disulfide bonds and β-strands I and IV.

Figure 5.

Structural comparison of representative “knotted” peptide families including (A, D) the new β-strand threaded Tyr-lock peptide recifin A (this study), (B, E) the lasso peptide microcin J25 (PDB code 1Q71),⁴⁰ and (C, F) the cyclic ICK peptide kalata B1 (PDB code 1NB1).⁴¹ Recifin A is stabilized by three disulfide bonds, Cys I–III, Cys II–V, and Cys IV–VI, forming a ring together with two of the β-strands, which is penetrated by a third β-strand. Recifin A structure is further stabilized by a central Tyr6 residue (cyan) locking the structure in place, reminiscent of microcin J25. Microcin J25 lacks disulfide bonds, but a threaded structure is formed by a cyclization via an amide-bond between the N-terminal amino group and the side chain carboxyl group of Glu8 (royal blue), which creates a circle that wraps around the C-terminal part of the sequence. The threaded structure is locked in place by two aromatic residues, Phe19 and Tyr20, (cyan), making it sterically impossible for the structure to unravel. Kalata B1, a prototypical plant cyclotide, contains an ICK motif and a head-to-tail backbone cyclization. The ICK is formed by three disulfide bonds (Cys I–IV, Cys II–V, Cys III–VI), two of which together with the backbone form a ring that the third disulfide bond is threaded through.

However, the embedded ring in recifin A (Figure 5A) is bigger than both lasso-peptides (e.g., microcin J25)⁴⁰ and prototypic ICK peptides (e.g., kalata B1) (Figure 5B and Figure 5C, respectively).⁴¹ The unusual fold of recifin A is further stabilized by Tyr6, which is deeply buried in the middle of the peptide (Figure 6) and locked in place by a number of residues, most notably Cys11, Tyr14, Ser29, and Leu32. It is because of this tight packing that Tyr6 does not undergo the usual fast “ring flips” typically observed for aromatic residues, which results in single resonance lines and sets of NOEs for the geminal Hδ* and Hε* protons. Instead, recifin A has extensive individual NOEs from surrounding residues to each of the four aromatic (Hδ1/2 and Hε1/2) protons locking Tyr6 in a specific conformation. In addition, a series of NOEs from the phenolic proton of Tyr6 to other surrounding residues can be observed, further highlighting the structurally stabilizing role of Tyr6 as these types of NOEs are rarely seen in a NOESY spectrum. The buried Tyr6 phenol group serves both as hydrogen bond donor, to the backbone carbonyl of Glu31, and as hydrogen bond acceptor for the HN proton of Gln33, while the hydroxyl groups of Ser27 and Ser29 serve as hydrogen bond donors to the carbonyls of Asp8 and Glu31, respectively. Ring current effects from aromatic residues are responsible for the unusual chemical shifts with Tyr6 packing against the Hα of Cys11, while the positioning of the side chains of Tyr14, Tyr28, and Trp37 is consistent with ring current effects on the HN of Gly16 and the Hβ resonances of Tyr40 and Pro35, respectively. This unprecedented structural arrangement and integral locked position of Tyr6 in the recifin A structure led us to the term structural class Tyr-lock peptides of which recifin A is the first.

Figure 6.

Stabilizing function of the Tyr6 residue in the overall, Tyr-lock structure of recifin A: backbone (blue), hydrogens (white), Tyr6 (pink), Cys11 (orange), Tyr14 (cyan), Ser29 (red), and Leu32 (green). Side chains of residues that pack around Tyr6 are shown with black lines indicating confirmed inter-residual NOEs.

Although there is currently no information available as to which region of recifin A interacts with the regulatory domain of TDP1, there are potential points on recifin A where residues commonly involved in protein-protein interactions (i.e., Arg9, Phe10, Arg25, and Trp37) are clustered.^42,43 The identification of key binding residues will require the capability to produce this peptide either synthetically or recombinantly, facilitating thorough structure-activity relationship analysis.

Biological Activity and Mechanism of Inhibition.

To characterize the effect of recifin A on TDP1 catalytic activity, a steady-state enzyme kinetics study was undertaken. The initial TDP1 screening assay was conducted at a single time-point involving product capture and optimized for use with crude natural product extracts which can result in inflated effective concentrations; it was therefore necessary to employ a previously described fluorescence resonance energy transfer (FRET) assay²⁸ to more accurately measure the kinetic parameters of the recifin A-TDP1 interaction. Recifin A inhibitory activity was confirmed in the FRET assay format as shown in Figure 7. Recifin A inhibited full-length TDP1 enzymatic activity in a concentration-dependent manner with an apparent IC₅₀ of 190 nM. We also evaluated the ability of recifin A to inhibit the enzymatic activity of a N-terminal truncated form of TDP1 (Δ147TDP1) in which the regulatory domain had been removed⁴⁴ and found only a minimal effect (approximately 20% maximal inhibition) at the highest concentration (1500 nM) assessed. Initial kinetic evaluation of the effect of recifin A on full-length TDP1 activity revealed that submicromolar concentrations of recifin A increased the K_m for the substrate, broadly defined as an inhibitory characteristic. In addition, a modest increase of the observed V_max value was also detected. This second observation is most often associated with allosteric enzymatic activators (Figure 8A).^45,46

Figure 7.

Biological activity and specificity of recifin A. Recifin A inhibited full-length TDP1 but not N-terminally truncated TDP1 (Δ147TDP1), enzymatic activity in a concentration-dependent manner with an IC₅₀ of 0.19 μM.

Figure 8.

Characterization of the effect of recifin A on full-length and truncated (Δ147TDP1) TDP1 kinetic parameters. (A) Steady-state analysis of recifin A modulation of full-length TDP1. Recifin A increased both the K_m and V_max kinetic constants of the TDP1 FRET assay,²⁸ exhibiting characteristics of both an enzyme inhibitor and activator. (B) Effect of recifin A on Δ147TDP1 kinetic parameters. Addition of recifin A did not affect either the K_m or V_max kinetic constants of the Δ147TDP1 FRET assay. As the Δ147TDP1 enzyme retained the identical substrate binding and catalytic sites as full-length TDP1, this suggested allosteric modulation of TDP1, dependent on the N-terminal 147 amino acid residues.

Further analysis of the data revealed that the recifin A-dependent increase in the K_m for the substrate is significantly more pronounced (approximately 6-fold higher) than the modest effect on the observed V_max (approximately 1.6-fold higher, Figure S10), consistent with our initial discovery of this peptide as a TDP1 inhibitor. To further characterize the inhibitory effects of recifin A on TDP1, we used the FRET based assay to determine whether recifin A had any effect on the enzymatic activity of Δ147TDP1, lacking the regulatory domain of TDP1. While smaller, Δ147TDP1 retains the substrate binding cleft and dual histidine-lysine-aspartic acid (HKD) motifs responsible for phosphodiesterase catalysis.^47,48 As shown in Figure 8B, recifin A had overlapping 95% confidence intervals for both K_m and V_max with the untreated controls, indicating that it does not affect the enzymatic activity of truncated TDP1. This suggests that the binding site for recifin A on TDP1 is outside the active site region common to both the truncated and full-length forms of TDP1, consistent with our results suggesting that recifin A acts as an allosteric modulator of TDP1 enzymatic activity and is binding to the N-terminal TDP1 regulatory domain. Additionally, evaluation of extract of the marine sponge Axinella sp. that yielded recifin A, in an assay to identify inhibitors of the related enzyme tyrosyl-DNA phosphodiesterase II (TDP2),⁶ indicated lack of inhibition of TDP2 (data not shown). The lack of activity against this related phosphodiesterase suggests another level of specificity for recifin A against TDP1.

Mechanistically, the recifin A-TDP1 interaction is interesting in that modulators that increase the K_m of an enzyme for the substrate are most often characterized as competitive inhibitors. However, the fact that an enzymatically active but truncated form of the protein, with an identical active site, was unaffected by recifin A indicates that the peptide was not directly competing for substrate binding at the active site. Additionally, the observation that recifin A treatment increased the V_max of the enzyme is a general characteristic of an enzymatic activator, further highlighting the novelty of the recifin A-TDP1 interaction and reinforcing the evidence that recifin A does not compete with the phosphotyrosyl-DNA TDP1 substrate in contrast to recently discovered TDP1 inhibitors.⁴⁹ These attributes together in a single interaction are unusual but not without precedent when considering that recifin A is not a small molecule but a complex peptide. There are several classes of enzymes for which a protein-protein interaction is known to change substrate specificity, catalytic efficiency, or both.^50–53 That recifin A may bind TDP1 allosterically suggests that there may be more to understand about the allosteric regulation of cellular TDP1 activity and that more of the TDP1 protein may be both pharmacologically accessible and therapeutically relevant. It is worth noting that the importance and major topological features present in the first 147 amino acids (deleted from the truncated variant) have not been resolved in a published crystal structure. The few existing publications about this region suggest that it has several known and potential post-translational modification sites (12 predicted according to prosite.expasy.org), including phosphorylation of serine 81 and SUMOylation of lysine 111, which are important for the regulation of TDP1 intracellular activity.^54–57 As our data demonstrate, there are substantial enzymatic differences with regard to both K_m and V_max of the truncated and full-length TDP1 enzymes. These catalytic differences as well as the specific interaction between the full-length TDP1 and recifin A suggest that a more complete enzymatic and structural characterization of the effect of modifications and protein-protein interactions in the regulatory region of TDP1 is warranted.

In conclusion, we have isolated and characterized recifin A, the first member of a new family of cysteine-rich peptides which we have named Tyr-lock peptides, from the genus Axinella, and which exhibit the ability to selectively inhibit the enzyme TDP1. Recifin A was determined to have a novel, non-ICK peptide disulfide-bonding pattern. The unusual disulfide linkage gives recifin A an unprecedented three-dimensional structure, with a central four-stranded β-sheet motif sandwiched by two helical turns. The disulfides create an embedded ring that encloses one of the strands, defining a new type of cystine knot peptide that also locks Tyr6 in three-dimensional space. Furthermore, recifin A was shown to modulate the activity of full-length human TDP1 enzyme but not an N-terminal truncated form of TDP1 (Δ147TDP1) lacking the regulatory domain. Further kinetic analysis confirmed recifin A to be an allosteric modulator of TDP1 suggesting that the N-terminal regulatory domain of TDP1 could be of possible pharmacological relevance for therapeutic inhibition of TDP1 to enhance sensitivity to anticancer topoisomerase I inhibitors.

EXPERIMENTAL SECTION

General Procedures.

All purification solvents were of HPLC and spectrophotometry grade. Mass spectrometry solvents were LC-MS grade and purchased from either Thermo Fisher or Burdick & Jackson. Mass spectrometry measurements were performed using an Accurate-Mass Q-TOF Dual-ESI 6530B instrument with an online 1260 Infinity binary HPLC system (Agilent Technologies, Inc., Santa Clara, CA), calibrated daily and operated with continual, internal calibration using reference mass ions at 121 and 1221 m/z. For MS, chromatographic separations were performed using linear gradients from 0 to 60% acetonitrile (0.1% v/v formic acid modified) at 1.00 mL/min on a Poroshell 300SB-C18, 5 μm, 2.1 mm × 75 mm column (Agilent Technologies, Inc., Santa Clara, CA) maintained at 40 °C. Source parameters for dual-ESI(+) were the following: capillary 4000 V, fragmentor 150–175 V, skimmer 65 V. Nitrogen flow was 12 L/min at 350 °C. High-resolution measurements (minimum of 20 000 resolution at 1521 m/z) were acquired in the range 100–3200 m/z at a scan rate of 1 spectra/s, and for MS/MS it was 50–3200 m/z at a scan rate of 3 spectra/s for both MS and MS/MS. Collision induced dissociation was accomplished using nitrogen gas and ramped collision energies (CE) calculated using the equation

$$\text{CE}=\frac{4\left(\frac{m}{z}\right)}{100}$$

Extraction and Isolation.

The sponge Axinella sp., (voucher ID 0CDN7410, NSC C020686) was harvested at a depth of 40 m at the Thunderbolt reef, south-southwest of Cape Recife Nature Reserve, Port Elizabeth, South Africa. A voucher specimen for this collection is maintained at the Smithsonian Institution (Suitland, MD). Aqueous extracts of Axinella sp. were provided by the Natural Products Branch of the National Cancer Institute and were prepared as previously reported.⁵⁸ The dried extract was reconstituted in water at a concentration of 10 mg/mL and then subjected to vacuum-assisted chromatography using Bakerbond C4 wide-pore media (Mallinckrodt Baker, Inc., Phillipsburg, NJ). Compounds were eluted using a stepwise methanol gradient of five column volumes (CV) each of 100% water, 40% methanol, 60% methanol, and 100% methanol, and the resulting fractions were evaporated under vacuum and then lyophilized to dryness. A high-throughput biochemical assay for inhibition of TDP1 enzymatic activity was utilized to track fraction activity.²⁸ Active fractions were subjected to RP-HPLC at room temperature, first using a Dynamax 300 Å, 5 μm, C4 column (Rainin, Woburn, MA), eluted with a 0–60% methanol gradient over 20 CV and then purified to homogeneity using a Vydac Protein&Peptide, 300 Å, 5 μm, C18 column (Grace Davison Discovery Science, Deerfield, IL), eluted with either a 0–60% methanol, 20 CV gradient or a 5–40%, 20 CV acetonitrile gradient. Purified peptides were lyophilized and stored at −20 °C.

Amino Acid Sequencing and Disulfide Assignments.

Purified recifin A was dissolved in 0.25 M Tris HCl, 1 mM EDTA, 6 M guanidine HCl, reduced at room temperature with 2-mercaptoethanol, and alkylated with 4-vinylpyridine according to standard techniques.⁵⁹ The peptide was purified by RP-HPLC using a Vydac C18 column and eluted using an acetonitrile gradient as described above. Reduced and alkylated peptide was subjected to digestion with various proteases per manufacturer’s protocol. Roche Diagnostics, Indianapolis, IN: trypsin, chymotrypsin. Thermo Scientific, Rockford, IL: glu-c. Sigma-Aldrich, St. Louis, MO: proline specific endopeptidase. Clontech Takara Bio USA, Inc., Mountain View, CA: pfu-pyroglutamate aminopeptidase. Peptide fragments were sequenced by MS/MS CID or purified by RP-HPLC and sequenced by automated N-terminal Edman degradation on an Applied Biosystems 494 protein sequencer (Applied Biosystems, Foster City, CA) according to the manufacturer’s protocols. PEAKS software version 7.5 was used for de novo peptide sequencing (Bioinformatics Solutions, Inc., Waterloo, ON, Canada). Precursor mass error tolerances were set to 5 ppm, and fragment ion error tolerance was set to 0.1 Da.

Disulfide bonds were mapped using a partial reduction and sequential alkylation technique.³¹ A quantity of 1 nmol of recifin A (81 μM final concentration) was incubated in 0.1 M glycine HCl, pH 2.5, with 5, 10, 20, or 50 mM TCEP at 37 °C for 30 min. N-Ethylmaleimide, freshly prepared in acetonitrile, was added to the reaction to a final concentration of 250 mM and incubated at 37 °C for 15 min. Partially alkylated species were desalted and separated by RP-HPLC using a Vydac Protein & Peptide, 300 Å, 5 μm, C18 column at 40 °C, using a linear gradient of water with 0.05% (v/v) TFA to 50% acetonitrile with 0.05% (v/v) TFA. The partially reduced/alkylated species were either combined with 0.1 M Tris HCl, pH 8.0, 1 M urea and digested with chymotrypsin for 18 h at room temperature or fully reduced with 5 mM DTT, alkylated with 14 mM iodoacetamide, and digested with trypsin for 18 h at 37 °C. Fragments were sequenced by LC-MS/MS collision induced dissociation and PEAKS de novo sequencing software as described above. Intact disulfide-bridged peptides were analyzed by LC-MS and assigned using MassHunter qualitative analysis software with BioConfirm, version B.07.00 (Agilent Technologies, Inc., Santa Clara, CA). Input amino acid sequences of the disulfide isoforms were constructed with a fixed N-terminal pyroglutamic acid residue, and amino acid numbers 22 and 42 were fixed as N-ethylmaleimide alkylated cysteine residues.

NMR Spectroscopy and Structure Determination.

All spectra were acquired on a 600 MHz Bruker Avance III equipped with a cryogenically cooled probe (Bruker Biospin, Billerica, MA). An approximately 1 mg sample of recifin A was dissolved in 90% H₂O/10% D₂O at pH 4.85, and 1D ¹H and 2D ¹H–¹H TOCSY (mixing time 80 ms) and ¹H–¹H NOESY (mixing time 200 ms) experiments were acquired at 298 K. In addition, a series of ¹H–¹H TOCSY experiments were acquired over 24 h directly after adding lyophilized recifin A to 100% D₂O to investigate slow exchange of HN protons. This was followed by acquisition of ¹H–¹³C HSQC and ¹H–¹H NOESY (200 ms mixing time) experiments in 100% D₂O. TopSpin 3.5 (Bruker) was used to process the spectra, and the data were referenced to water at δ_H 4.76 ppm. Sequential assignments were completed using CCPNMR analysis 2.4.1 (CCPN, University of Cambridge, Cambridge, U.K.) and Xeasy.⁶⁰ Distance restraints were derived from ¹H–¹H NOESY experiments acquired in 90% H₂O/10% D₂O and 100% D₂O, and ϕ and ψ dihedral angle restraints were derived from chemical shifts from ¹H–¹H NOESY and ¹H–¹³C HSQC experiments analyzed by the online version of TALOS-N⁶¹ to derive ϕ and ψ dihedral angle restraints. χ¹ and χ² dihedral restraints for Cys residues were derived from DISH,⁶² and additional χ¹ dihedral restraints were derived from a combination of TALOS-N, patterns of NOE intensities, and preliminary structure calculations. Hydrogen bonds were introduced based on D₂O exchange data or in the case of hydroxyl groups based on exchange behavior in the H₂O sample and preliminary structure calculations. An initial 20 structures were calculated using the using automated assignments in Cyana.^63,64 After manual assessment of the output all remaining NOEs could be unambiguously assigned. Structural refinement was carried out in a watershell using CNS⁶⁵ where 50 structures were calculated and 20 representative structures selected based on MolProbity scores⁶⁶ and energies. Root mean square deviations (RMSDs) were calculated using MOLMOL,⁶⁷ and structural visualization was carried out using MOLMOL and PyMOL (the PyMOL Molecular Graphics System, version 1.7.4, Schrödinger, LLC). Recifin A structure has been deposited into the PDB⁶⁸ (rcsb.org, code 6XN9), and NMR data have been deposited into the Biological Magnetic Resonance Bank⁶⁹ (bmrb.wisc.edu, code 30767).

Biological Activity and Kinetics Assays.

TDP1 enzymatic activity inhibition assays using a radiolabeled oligonucleotide DNA substrate have been previously described.²⁹ Briefly, serial dilutions of recifin A were incubated with 1 nM 5′–³²P-labeled DNA oligos (P14Y:5′-[³²P]-GATCTAAAAGACTT(3′-pTyr)-3′), 30 pM recombinant human TDP1 or 2 μg/mL of hTDP1 WCE which were collected from TDP1 knockout (TDP1−/−) DT40 cells complemented with human TDP1. The reactions were carried out in a final volume of 10 μL in 1× LMP 1 reaction buffer (50 mM Tris HCl, pH 7.5, 80 mM KCl, 2 mM EDTA, 1 mM DTT, 40 μg/mL BSA, 0.01% Tween 20) at room temperature for 15 min and terminated by adding 10 μL of 2× stop buffer (99.5% formamide, 10 mM EDTA, 0.01% methylene blue, 0.01% bromophenol blue). A 20% DNA sequencing gel was used to load the samples and exposed to a PhosphorImager screen for further analysis by Typhoon FLA 9500 (GE Healthcare).

FRET-based TDP1 enzymatic activity inhibition assays have been described in detail previously.²⁸ Briefly, for Michaelis-Menten analysis, an eight-point FRET substrate concentration response was used (from 0.01–3 μM substrate) in the presence of 0, 0.2, 0.5, 1, and 2 μM recifin A. Quadruplicate reactions were set up in which a 1.25× concentration of either full-length TDP1 or Δ1–147TDP1 was diluted to 1× by the addition of a 6× solution of substrate and recifin A to reach a final concentration of 0.5 nM TDP1 (full length or truncated) and the indicated substrate and recifin A concentration in 1× phosphate buffered saline (PBS), pH 7.4, 80 mM potassium chloride, 1 mM TCEP, referred to as “1× TDP1 buffer”. After dilution these reactions were transferred to a black small volume 384-well plate (Greiner Bio-One, Monroe, NC). Fluorescence measurements (excitation 520 nM, emission 550 nm) were taken at 30 s intervals for 1 h using a i3x SpectraMax plate reader (Molecular Devices, Sunnyvale, CA). Reaction progression curves for each condition were examined for linearity over the time course, and the reaction rate for each condition was determined by linear regression using GraphPad Prism software (version 8.3.1, San Diego, CA). Reaction rates were replotted in terms of substrate concentration, and kinetic parameters for each recifin A treatment concentration were calculated by nonlinear regression (GraphPad Prism) according to the following equation:

$$v=\frac{V_{\max }[\mathrm{S}]}{K_{\text{m}}+[\mathrm{S}]}$$

For IC₅₀ determinations, a 12-point concentration-response curve was prepared over a recifin A concentration range of 0–15 μM. This was accomplished by diluting a 5× stock solution of recifin A and TDP1 FRET substrate into a stock solution of 1.25× TDP1 buffer containing 0.625 nM full-length TDP1 or Δ147TDP1, bringing the final concentration to 1× TDP1 buffer, 0.5 nM enzyme (or a no enzyme control), 1 μM FRET substrate, and 0–15 μM recifin A. Reactions were set up in triplicate using the same plates and plate reader described above for the kinetic measurements. Reaction wells were read at 0 (T₀) and 15 (T₁₅) min after initiation. The T₁₅ data were background corrected by subtracting T₀ fluorescence measurements. Corrected data were normalized to a control with no enzyme present (0% activity) and a vehicle control (100% activity). Recifin A concentrations were converted to log₁₀ values, normalized data were fitted to the following equation by nonlinear regression (least-squares fit with a variable slope), and an IC₅₀ value was calculated using GraphPad Prism software:

$$\%\text{normalized}\text{activity}=\frac{100}{1+{10}^{(\left(\log \mathrm{I}{\mathrm{C}}_{50}-[\text{recifin}]\right)\mathrm{Hill}\mathrm{slope})}}$$