|Home | About | Journals | Submit | Contact Us | Français|
While a variety of methods exist,1 target identification continues to pose a fundamental challenge in elucidating a natural product’s mode of action.2 The development of solutions to this problem is essential to advancing drugs into the discovery pipeline. The complex structural diversity that characterizes natural products complicates the adoption of routine protocols for these studies.3 A systematic approach to target elucidation would both facilitate the introduction of drugs into clinical trials, and challenge leads early on should undesired targets be identified.4
The process of determining the therapeutic mode of action of a natural product typically begins by developing probes bearing an affinity and/or fluorescent tag.5 The aim in this exercise is to construct a probe with activity comparable to its parent natural product. Often a loss in activity can be circumvented by use of bio-orthogonal approaches6 (Huisgen cycloaddition reactions or Click chemistry,7 Dielsalder cycloaddition,8 or olefin metathesis9), wherein the reporter is added bio-orthogonally after the natural product binds to its target. Though sufficient for cellular studies, the lack of a covalent link between the labeled natural product and its target commonly thwarts enrichment strategies (e.g., immunoprecipitation).
Crosslinking or photocrosslinking,10 a common solution to this problem, is frequently plagued by additional complications. First, the preparation of a dual-tag results in an enlarged probe,11 which often exhibits only a fraction of the activity of its parent natural product.2b Second, crosslinking methods are generally low yielding and unselective, increasing the complexity of downstream mass spectrometric (MS) analyses.12 Finally, and perhaps most critically, the crosslinked target still bears a covalently linked natural product that is subject to xenobiotic metabolism (e.g., cytochrome P450 oxidation).13 Such metabolic processing leads to the formation of multiple products, each of which is detected by MS, obscuring subsequent protein identification and binding-site mapping studies. In the present study, the interaction of a natural product with its target(s) is addressed via an elimination or displacement reaction, wherein the natural product is removed from the reaction milieu, acting as a leaving group.14
Given the use of acyl phenols in protein reactive dyes,15 we examined the feasibility of this motif as a transfer agent for the immunoaffinity fluorescent (IAF) tag (Fig. 1a). Though the attachment of a phenol-containing linker would enable this method to be applied to natural products in a general sense, the marinopyrroles, a family of phenolic 1,3′-bipyrroles including 6 and 7, were selected for this study.16 We chose these materials based on: (1) their availability from culture (strain CNQ-418) at mg/L scale, (2) their unknown mode of action, (3) their activity in tumor cell lines (GI50 values for 6 and 7 were 10 μM in HCT-116 cells and 620 and 500 nM for 6 and 7, respectively, in A549 breast cancer cells) and (4) their bis-phenol structure, previously shown to undergo clean acylation to 8 with acetic anhydride (Scheme 1) without a loss in activity (GI50 value of 0.8 μM for 8 in HCT-116 cells, ~12 fold increase over 6 or 7).16
Based on this evidence, mono-labeled 13 (45% yield) and bis-labeled 14 (traces) were prepared and analyzed using a combination of MS and NMR analysis. HMBC analysis confirmed the position of the tag in mono-labeled 14 (see Scheme 1). The low yield of the bis-labeled 14 was due to its instability, rapidly converting to 13 in aqueous media with a half-life t1/2 = 40 min in PBS pH 7.2 at 32 °C. Mono-labeled 13 was much less prone to hydrolysis with a half-life of t1/2 = ~2 d in PBS pH 7.2 at 32 °C.
Probes 13 and 14 mimicked the activity of marinopyrrole A (6). Probe 14, with a GI50 value of 1 μM, was ~10 fold more active than the parent natural products 6 and 7. Using two color confocal microscopy, we compared the native red fluorescence from marinopyrrole A (6) at 692 nm (Figs. 2a–2b) to the blue fluorescence from the IAF tag in 13 (Fig. 2d) and 14 (Fig. 2c). The presence of a similar pattern of red and blue staining confirmed that probes 13 and 14 shared the same subcelluar localization in HCT-116 cells, and therefore provided a sufficient mimic of 6. The co-localization of natural product 6 and dye-labeled probes 13 and 14 also showed that cells were not stained from the hydrolytic release of 9. Importantly, control 12 was readily washed from cells under each study (insets, Fig. 2a–2d).
The putative protein target of the marinopyrrole probe was initially established using an immunoprecipitation (IP) protocol.2c HCT-116 cells (~108 cells) were incubated with 10 μM 13 for 12 h. The cells were collected, lysed and screened for proteins that were labeled with the IAF tag. Two fluorescent bands at 40–45 kDa were apparent in the crude lysate (Fig. 3a). After IP, three bands were obtained (Fig. 3b). Mass spectral analysis (Fig. 3d) indicated that the two lower bands were actin, while the upper band arose from actinin. The fact that actinin was observed during IP experiments in the lysate, while not fluorescent, is expected given its tight complexation with actin.
The selective binding of the probe to actin was validated with purified proteins. Samples of rabbit muscle actin (Worthington Biochemicals) were treated with 13 or 14 to yield comparable fluorescently labeled bands (lanes L4-L7, Fig. 3e). Even with a 50-fold increase in concentration (lanes L1, L2, L3, L9 and L10 in Fig. 3e), the non-specific labels 10 and 11 failed to stain actin. The reactivity observed in Fig. 3e was indeed due to a specific binding event as pretreatment of actin with 6 effectively blocked the uptake of dye from 14 (lane L8, Fig. 3e).
Finally, mass spectral methods were applied to determine the site of dye transfer to actin. Protein from the reaction in lane L6 (Fig. 3e) was trypsin digested, purified by HPLC based on the absorbance of the IAF dye at 350 nm, and subjected to LTQ-MS/S with a NanoMate direct infusion system. Evaluation of the resulting spectra using the proteomics search tool, InSpecT, to find peptides that were modified by the IAF tag (M+286 Da), resulted in the identification of the modified peptide V98APEEHPTLLTEAPLNPKANR118. Because InSpecT annotated the modification to be anywhere from K115 to R118, we manually annotated the data to determine the site of localization. The manual annotation of the low-resolution data suggested that the modification could be found at K115. To further improve the localization of the probe, the same sample was reanalyzed by LTQ-FTICR-MS, giving rise to a mass of 861.7839 Da within 6.6 ppm of the calculated mass of the modified peptide. Furthermore, all the MS2 b and y fragment ions observed by LTQ-FTICR-MS matched to within 0.02 Da enabling us to localize the site of modification to K115. The unmodified peptide V98APEEHPTLLTEAPLNPK115 was also identified, which indicates that this residue was only partially modified.
Lysine 115 occupies a distinct site on actin (star, Fig. 4a) that does not overlap the ATP, gelsolin and profilin binding sites (see Supporting Information), nor does it overlap with the pocket targeted by other known natural products such as latrunculin A (dot, Fig. 4a), a marine natural product that binds to the nucleotide binding cleft of actin. Structural modeling identified a channel that exists adjacent to the amine terminus of K115, suggesting a putative marinopyrrole-binding site (Fig. 4b–c).
Transfer of the IAF dye from the natural product to actin was further supported by secondary analyses. First, microcalorimetry analyses verified the binding between marinopyrrole A (6) and rabbit muscle actin with a Kd of 0.12±0.2 μM. The strength of this complex compares favorably to a reported value of 0.2 μM for latrunculin A (albeit at a different binding site, see Fig. 4a).17 Second, using a fluorescence assay (Cytoskeleton Inc.),18 6 was shown to inhibit the polymerization of a pyrene-labeled G-actin with an IC50 value of 39.5±6.2 μM (see Supporting Information). Lastly, actin fibers in live HCT-116 cells treated with probe 13 were blue fluorescent (Fig. 2f) from the binding of 13, as evident by co-staining with FITC-phalloidin (Fig. 2e).
This investigation provides definitive support for the use of a dye transfer protocol via an acyl phenol motif for mode of action studies. By transferring the dye from the natural product to its protein target, we were able to streamline cellular, target elucidation, and binding site determination studies into a single workflow (see Fig. 1). Because bioactive natural products bind to proteins in selective binding pockets, the process of natural product mediated ligation may also be used to direct protein modifications site-specifically. Studies are now underway to evaluate the scope of this reaction centered on the electronic nature of the phenol and its ability to transfer its payload.
We thank Qishan Lin (CFG at University of Albany) for preliminary protein ID analyses. Support was provided by the NCI under grant R37 CA44848 (to WF), the V-foundation (to PCD) and NIH R01 GM086283 (to PCD).
Supporting Information Available: Synthetic methods, copies of NMR spectra, protocols for cell imaging studies, IP studies and actin assays, detailed methods for the mass spectral studies, copies of original images in Figs. 2–3 as well as additional references are available free of charge via the Internet at http://pubs.acs.org.