A Quantitative Chemical Proteomics Approach to Profile the Specific Cellular Targets of Andrographolide, a Promising Anticancer Agent that Suppresses Tumor Metastasis

seeds. The interactions of complex p50-Andro conformations were analyzed using Pymol.

tain post-translational modifications. (17)(18)(19) These studies shed light on how quantitative proteomics can improve the specificity of the target protein identification. Nevertheless, due to the inherent limitation of SILAC, such an approach takes a long time for complete incorporation of isotopic amino acids. Furthermore, it is also extremely difficult to apply the SILAC approach to tissue and body fluid samples, which are of particular relevance to biomedical research.

<Suggested location for figure 1>
Here we introduce an isobaric tag for relative and absolute quantitation (iTRAQ)-based Clickable ABPP (ICABPP) approach for unbiased, specific and comprehensive identification of target proteins in live cells. iTRAQ is a stable-isotope labeling approach for multiplexed quantitative proteome profiling. (20) An overview of the technique is illustrated in Fig. 1a. In this assay, cells are first incubated with a clickable probe or with DMSO, which serves as a negative control. After the probe permeates the cell and covalently binds to its dedicated in situ targets, the washed cells are lysed, clicked with biotin-N 3 tag and enriched through avidin pull-down in parallel. The beads are washed thoroughly and the bond proteins are directly digested on beads with trypsin.
The resulting peptides are labeled with respective iTRAQ reagents, and pooled together for further identification and quantification by LC-MS/MS. This technique enabled us to discriminate specific protein targets from non-specific and endogenously biotinylated proteins. Biological replicates of probe-or DMSO-treated samples are included to overcome experimental variations.
As shown in Fig. 1b, non-specific binding proteins' iTRAQ reporters have equal or similar intensities, whereas specific target proteins enriched by the probe show highly differential intensities compared to DMSO-treated control samples (as illustrated by the significantly higher reporter intensities of 116 and 117 vs. 113 and 114 shown in Fig. 1b). The multiplexing nature of iTRAQ-based chemical proteomics method allows replicated enrichments to be compared within by guest on May 8, 2020 https://www.mcponline.org Downloaded from a single LC-MS/MS analysis, hence increasing the accuracy of identifying specific targets and minimizing experimental errors.
In this context, ICABPP approach was applied to identify protein targets of andrographolide (Andro, Fig. 2), a natural product with known anti-inflamation and anti-cancer effects, (21)(22)(23)(24)(25) in live cancer cells. A spectrum of 75 potential Andro targets was identified with high confidence, which suggested that Andro may exert anti-cancer effects by acting on multiple targets to interfere with several cellular signaling pathways. Two targets, NF-B and β-actin were validated by in vitro binding assay and direct binding site mapping. Furthermore, our data revealed a novel mechanism of Andro in suppressing tumor metastasis.

Inhibition of Cancer Cell Proliferation:
HCT116, HeLa and HepG2 Cell Lines: 20,000 cells was seeded into 96 well plate and allowed to attach for 24 hrs. Cells were then treated with Andro at various concentrations for 48hrs. At the end of the treatment, media was removed and the wells were washed once with PBS. The cells were then stained with 0.5% crystal violet in 20% methanol for 10 min. Excess crystal violet was washed off with PBS and the wells were allowed to dry. Solubilization was done using 1% SDS for 30 min and the absorbance measured at 550nm. MV4-11 Cell Lines: Cells were treated with increasing concentration of Andro for 48 hrs. Following the intended treatment, the cells were centrifuged and resuspended in 2 mL medium. A volume of cell suspension containing 1x10 5 cells was centrifuged into a pellet. After dislodging the pellet, the samples were added with 10 μL of 50 μg/mL propidium iodide (PI) and subjected to flow cytometric analysis using the Beckman Counter (EPICS-Xl MCL).

In situ Fluorescence Labeling Experiments:
HCT116 cells were grown to 80-90% confluence in 6-well plates. After the media was removed, cells were washed twice with PBS. P1 or P2 (100 μM) in 2 ml medium with a final DMSO concentration of 1% was added and cells were incubated for 4 hrs at 37 °C and 5% CO 2 . Equal vol-by guest on May 8, 2020 ume of DMSO was used as a negative control. For concentration optimization experiment ( Fig.   S3 in Supporting Information), increasing concentrations (20-200 μM) of P2 were used to culture the cells for 4 hrs. Subsequently, the medium was removed and cells were washed with PBS and detached with trypsin. The cell pellet was resuspended in PBS, washed, followed by sonication in 150 µl of PBS to lyse cells. The resultant cell lysate was cleared by centrifuging at 13,000 rpm for 30 min. Protein concentrations of the cell lysates were determined using the Bradford assay.
The samples were incubated at room temperature for 2 hrs. Next, clicked proteins were precipitated by acetone and air dried. 100 μL 1× SDS loading buffer was added to dissolve the sample and 50 μL of sample was separated by SDS gel electrophoresis on 10% polyacrylamide gel. After SDS-PAGE, gels were visualized using a Typhoon 9410 laser scanner (GE, health care) and images were analyzed by TotalLab software.

Cells Labeling:
In the subsequent ICABPP study, two biological duplicate of P2 treated and two DMSO treated samples were pulled down and digested in parallel. The two DMSO control samples were labeled with iTRAQ reagent and quantified by iTRAQ ratios. Briefly, HCT116 cells were grown to 80-90% confluence in T175 flasks. Spent medium was then aspirated and the cells washed twice with PBS. P2 (100 μM) in 20 ml medium with a final DMSO concentration of 1% was added to the cells in the flasks and incubated for 4 hrs in the CO 2 incubator.. Culture medium containing 1% DMSO was used as negative control. Subsequently, P2-and DMSO-containing media were removed, and then the cells were washed with PBS and detached with trypsin. The cell pellet was resuspended in PBS, washed and lysed by sonication in PBS. The cell lysates were clarified by centrifugation at 13,000 rpm for 30 min followed by Bradford protein assay.

On-beads Digestion:
The beads were washed a total of 9 times; thrice with 1% SDS, followed by 3 times with 6M urea and thrice with PBS. The extensively washed beads were resuspended in 25mM ammonium bicarbonate (NH 4 HCO 3 ) and 2 µL tris-(2-carboxyethyl) phosphine (TCEP, 100 mM stock solution) added. The beads were placed in a 65°C heat block for 60 min. Next, 1µL methyl methanethiosulfonate (MMTS, 200mM stock solution) was added and the samples left in the dark and allowed to react for 15 min at room temperature. Following reduction and alkylation, trypsin (12.5 ng/ µL, Promega) was added and incubated at 37°C overnight. The digested peptides were separated from the beads using a filter-spin column (GE, healthcare). These digested peptides by guest on May 8, 2020 could be stored at -20 °C for several months pending iTRAQ labeling and mass spectrometry analysis.
iTRAQ Labeling of the Digested Pull-Down Samples: iTRAQ labeling was performed out using iTRAQ Reagent kit (AB SCIEX, Foster City, CA, USA) based on the vender's instruction manual with minor modifications. The two biological replicates of the negative control pull-down samples were labeled with iTRAQ reagent 113 and 114, respectively. Similarly, two biological replicate of digested Andro pull-down samples were labeled with reagent 116 and 117, respectively. Briefly, the on-beads digested peptides were dried and reconstituted with equal volume of dissolving buffer (0.5M TEAB). The peptides were then labeled with the respective iTRAQ reagents and incubated at room temperature for 2 hrs before all the samples were pooled together. The iTRAQ workflow is shown in Fig. 1 in the main text.

Proteins Identification and Quantification:
Nano LC−ESI-MS: by guest on May 8, 2020 The detailed methods for LC-MS/MS was described previously. (26) Briefly, separation of the iTRAQ labeled peptides was carried out on an Eksigent nanoLC Ultra and ChiPLC-nanoflex

ProteinPilot Analysis:
The detailed method of ProteinPilot analysis was described previously. 2 Briefly, the protein identification and iTRAQ quantification were performed with ProteinPilot™ 4.5 (AB SCIEX, Foster City, CA) which uses the Paragon™ algorithm to perform database searches. The database used includes the International Protein Index (IPI) v3.87 human protein sequences (total 91 468 entries). The search parameters used were as follows: Cysteine alkylation of MMTS; Trypsin Digestion; TripleTOF 5600; Biological modifications. Redundancy was eliminated by the grouping of identified proteins using the ProGroup algorithm in the software. A decoy database search strategy was used to determine the false discovery rate (FDR) for peptide identification. A corresponding randomized database was generated using the Proteomics System Performance Evaluation Pipeline (PSPEP) feature in the ProteinPilot™ Software 4.5. In this study, a strict total score cut-off >1.3 was adopted as the qualification criterion, which corresponded to a peptide confidence level of 95%. The identification and quantification results were then exported into Microsoft Excel for manual data analysis. by guest on May 8, 2020 Data Analysis: To determine the cut-off threshold for the fold change of proteins identified from the iTRAQ study to be considered as significantly regulated, two equal amounts of six-protein mixtures (Applied Biosystems) were trypsin-digested and labeled with the iTRAQ reagents. 3 The standard deviation (S.D) of all the ratios of the labeled peptides was computed to be 0.15. Thus by using a 1 + 2 S.D formula the fold-change cut-off thresholds were set as 1.3 for up-regulated proteins and reciprocally 0.77 for down-regulated proteins. This strategy was adopted for our quantitative study. This cut-off was used to eliminate protein targets where the two biological replicate samples showed significant change (ratio >1.3 or <0.77). Basing on this strategy, 208 proteins were considered to be the statistically reliable hits (Fig. 3d), and the distribution of the enrichment ratios of these proteins were further presented as the colored heat map as illustrated in Fig. 3b. The 4 set ratios of Andro pull-down vs. DMSO pull-down were presented as colored heatmap, using the MultiExperiment Viewer (MeV). (29,30) Proteins with enrichment ratio closed to 1 are denoted in blue and likely to exhibit non-specific binding. In contrast, proteins labeled in red showed enrichment ratio above 2 or close to 3, suggesting that they are likely the specific binding targets.
To reduce the likelihood of selecting the false positive drug targets, we chose a stringent ratio equivalent to 2 as the cut-off to identify specific protein targets for subsequent experiments.
Meanwhile, proteins identified based on a single peptide are considered unreliable and were removed. Using these criteria, 75 proteins were identified and selected (Fig. 3d). The full list of the 75 potential targets is shown in Table S1 in supporting information.

Pathway Analysis of Andro Targets:
The specific Andro targets identified using the ICABPP approach were analyzed using the Ingenuity Pathway Analysis software (Ingenuity® Systems, Redwood city, CA, USA). A spreadsheet containing the list of Andro targets was uploaded into IPA. The software mapped each of the proteins to the repository of information in the Ingenuity Pathways Knowledge base. Molecular networks and canonical pathways regulated by these drugs targets were obtained using IPA core analysis.

Validation of Drug Target using Western Blot:
Andro probe affinity pull-down sample was separated by 1D-SDS PAGE together with DMSO pull-down sample. After SDS-PAGE, the proteins were transferred onto PVDF membranes (Bio-Rad). The blots were blocked with 5% (w/v) BSA in PBS with 0.1% Tween 20 (PBS-T) for 4hrs at room temperature. The membranes were incubated with rabbit anti-NF-κB p50 (1:1500), from Santa Cruz Biotechnology, Inc. as well as mouse anti--actin (1:4000) from BD Transduction Laboratories. HRP-conjugated from Pierce Biotechnology, or HRP-conjugated anti-mouse IgG (1:5000) from GE Healthcare were used as secondary antibodies and incubated for 2 hrs at room temperature. The membrane was washed 3 times in PBS-T between each antibody incubation step. Subsequent visualization was performed using ECL substrate (Pierce Biotechnology). The eluted p50 was dialyzed overnight against 1xPBS pH 7.3 to remove imidazole. This protein was subsequently further purified using gel filtration column Superdex 200 (Amersham).

In vitro Labeling of Human Recombinant NF-κB p50:
Recombinant NF-κB p50 was reconstituted with PBS as 1 mg/ml. 1 μL protein solution was diluted with 42 μL PBS and incubated with 1 μL P2 at a final concentration of 0, 10, 20, 40, 80, 160 μM for 4 hrs, followed by Click reaction, SDS-PAGE and fluorescence scanning. For heat denatured sample, 1 μL protein solution was diluted with 40 μL PBS and 2 μL of 25% SDS and the sample was heated at 96 °C for 10 min. Then the heated sample was cooled to room temperature and reacted with 80 μM P2 at. Competition assay was carried out with the pre-treatment of 10× excess free Andro for 4 hrs, and then labeled with P2. 1mM DTT or 5mM BME cotreatment together with Andro were also included in our experiment. Mutated C62A p50 was labeled with 80 μM to test this critical amino acid in Andro reaction. Cys62 are shown in Fig. 5d (top). The MS/MS spectra of the Andro labeled NF-κB peptide containing Cys62 are shown in Fig. 5d (bottom). C A* (red) represents the Andro modified Cys. Peptide ions containing modified Cys are indicated A* in red.

Molecular Docking:
Autodock is an automated procedure for predicting optical conformations and orientations for the ligand, protein or DNA with the target proteins at the binding site.  Excess stain was washed off and the inserts were allowed to dry. Solubilization was done using 1% SDS for 2 hrs and the absorbance were measured at 550nm. age was also synthesized. This probe would have better in vitro and in vivo stability, as it would be less prone to breakage or hydrolysis. (Figure 2, Synthetic scheme shown in Scheme S3 in Supporting Information). Because the introduction of amino group partially reverted the configuration of C14 during the synthesis, P2 was a mixture of two isomers with the ration of 1:0.8. We did not separate these two isomers as we found that the anti-cancer activity was still retained (vide infra). To investigate whether the structural modifications would influence the anti-cancer potency of the probe, we conducted in vitro growth inhibitory assay on HCT116 using P1 and P2.
Our data confirmed that P1 and P2 still possess antiproliferative activity ( Figure S2b in Supporting Information). Therefore, both P1 and P2 meet the essential criteria for target protein identification.

In situ proteome profiling:
To visualize the native cellular targets of Andro using fluorescence gel profiling ( Figure S1 in Supporting Information), we treated live HCT116 cells with P1 or P2. The probe-labeled proteomes were reacted with Cy3 azide via click chemistry before being resolved on SDS-PAGE for fluorescence detection. As shown in Figure 3a, in sharp contrast to the DMSO control, P2 labeled proteome yielded high fluorescence intensity bands, signifying that there were proteins interacted with P2. As expected, P1 labeled proteome showed rather weak intensity bands, possibly due to the instability of the ester linker. These observations supported our hypothesis and were consistent with a recent review, which postulated that the elimination of the β-hydroxy group of Andro might occur during the Andro alkylation.(15) Based on our results, the stronger labeling intensity of P2 in comparison to P1 suggests that the ester bond of P1 might be broken or hydrolyzed in the reaction, while the amide bond linkage in P2 is more stable.(15) Therefore, P2 was chosen for subsequent analyses. We next performed ICABPP using P2 to identify specific cellular protein targets of Andro (Figure 1a), which yielded 4 sets of Andro probe vs. control pull-down ratios (116/113; 117/113; 116/114; 117/114). A total of 291 proteins was successfully identified and quantified in our experiment (Figure 3d). Outliers were identified using p-value >0.05 and 114/113 ratio >1.3 or <0.77. This resulted in 208 proteins being considered to be statistically reliable hits (Figure 3d).
The distribution of the enrichment ratios of these proteins were further presented as a coloured heat map in Figure 3b. To reduce false positive targets, we chose a highly stringent ratio of 2 as the cut-off to differentiate specific (red) from non-specific (blue) binding targets. Meanwhile, proteins identified based on a single peptide are considered unreliable and were removed. Consequently, 75 proteins were regarded as the specific targets of Andro based on the above criteria ( Figure 3d). The complete list of the 75 potential targets was shown in Table S1 in Supporting Information. NF-κB p50, a known Andro target, (21,33,34) was found to be enriched 2.8 fold in Andro pull-down when compared to DMSO control, validating our approach. The subsequent pathway analysis suggested that Andro may exert its anti-cancer effects through multiple targets and pathways: more than 30 targets were involved in cancer cell death pathways; 15 hits were involved in cell migration and metastasis (Figure 3c and Figure 4); 20 hits were related to inflammation and 10 to protein synthesis pathway (the respective pathway analyses are shown in Figure S5 and Figure S6  Previous studies showed that several small molecules can bind to actin by forming a covalent bond with Cys via similar Michael addition reactions. (47,51,52) To validate the interaction between actin and Andro, -actin in G-buffer and F-buffer were incubated with Andro, respectively. by guest on May 8, 2020 The results demonstrated that Andro selectively bound to polymerized F-actin (Figure 6a). The binding site of Andro on actin was subsequently confirmed to be Cys272 by MS/MS, which was reported as a highly reactive cysteine due to its full solvent accessibility (Figure 6c) (53)(54)(55) Finally we verified the anti-metastatic potential of Andro using cell migration and invasion assays. HCT 116 cells were treated with RA and Andro at a non-cytotoxic concentration of 5µM for 30 hrs. The results showed that Andro can effectively suppress the migration and invasion of HCT116 cells (Figure 7). These results further supported our novel finding on the role of Andro in inhibiting tumor metastasis, thus broadening its therapeutic applications as an anti-cancer agent.
As actin plays an important role in cytokinesis, we further examined Andro's effect on cell cycle using flow cytometry. Our data showed that HCT116 cells were arrested at G2/M phase upon Andro treatment in a time-dependent manner (Figure 6b). We also analyzed the cell cycle using other cell lines including HeLa and HepG2. Significant cell cycle arrest was also observed at G2/M phase upon Andro treatment ( Figure S8 in Supporting Information).

Conclusion:
In conclusion, our results demonstrated that the ICABPP method, which combines clickable ABPP and iTRAQ, is a powerful approach to identify specific drug targets in live cells. In this study, a spectrum of specific targets of Andro was identified using this new method. In particular, we have identified the novel anti-metastasis potential of Andro through targets and pathway analysis, which have been validated through subsequent migration and invasion assays. To the best of our knowledge, this is the first report that systematically combines the iTRAQ-based by guest on May 8, 2020 quantitative proteomics with the clickable activity-based probe to profile specific drug targets in live cells.
In practice, ICABPP method can be easily optimized and used in all kinds of affinity chromatography or ABPP-based target identification. The multiplexing property of the iTRAQ enables the precise and accurate quantitation of up to 8 samples simultaneously, allowing the inclusion of the biological replicates or drug competition of pull-down samples. Moreover, it can be used to study drug targets using tissue or body fluid samples in clinical research. We anticipate this ICABPP approach to be widely applied in drug development and optimization to refine the ther-