The UPR inducer DPP23 inhibits the metastatic potential of MDA-MB-231 human breast cancer cells by targeting the Akt–IKK–NF-κB–MMP-9 axis

(E)-3-(3,5-dimethoxyphenyl)-1-(2-methoxyphenyl)prop-2-en-1-one (DPP23) is a synthetic polyphenol derivative that selectively induces apoptosis in cancer cells through the unfolded protein response pathway. In the present study, we evaluated the effect of DPP23 on tumour invasion and metastasis. Here, we show that DPP23 inhibited tumour necrosis factor alpha (TNFα)-induced motility, F-actin formation, and the invasive capability of MDA-MB-231 cells. DPP23 inhibited NF-κB-dependent MMP-9 expression at the transcriptional level. Akt is involved in the activation of IKK, an upstream regulator of NF-κB. DPP23 inhibited IKK and Akt, and knockdown of Akt2 significantly inhibited TNFα-induced IKK phosphorylation. We found that DPP23 bound to the catalytic domain of Akt2, as revealed by an in silico molecular docking analysis. These results suggest that DPP23 prevents TNFα-induced invasion of highly metastatic MDA-MB-231 breast cancer cells by inhibiting Akt–IKK–NF-κB axis-mediated MMP-9 gene expression. In addition, DPP23 attenuated experimental liver metastasis in a syngenic intrasplenic transplantation model using 4T1 mouse mammary carcinoma cells. Collectively, these results suggest that DPP23 could be used as a potential platform for the prevention of invasion and metastasis of early-stage breast cancer or as an adjuvant for chemo/radiotherapy.

. In the tumour microenvironment, TNFα plays a critical role in the migration and invasion of tumour cells 22 . An in vitro scratch wound-healing assay was performed to determine whether DPP23 affects the TNFα -induced migration of MDA-MB-231 cells. The MDA-MB-231 monolayer was scratched with a pipette tip, and the cells were then allowed to migrate in the presence and absence of TNFα and DPP23. As expected, TNFα enhanced the migration of MDA-MB-231 cells into the scratched area compared with the migration of the untreated cells (Fig. 1b). However, TNFα -induced cell migration was significantly reduced in the presence of DPP23 (P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test, n = 3; Fig. 1c). The actin cytoskeleton plays a key role in cell motility 23 . During cell movement, globular actin monomers (G-actin) are organized into filamentous bundles (F-actin). To determine whether DPP23 alters actin reorganization, the cells were treated with TNFα in the presence of DPP23. Polymerized F-actin was stained with rhodamine-labelled phalloidin, which selectively binds to F-actin. The cells treated with DPP23 exhibited a substantial reduction in TNFα -induced F-actin formation, mostly at the cell edges (Fig. 1d).

DPP23 inhibits TNFα-induced invasion of MDA-MB-231-derived spheroids.
The local invasive capability of tumour cells is considered to be an early step in the complex process of metastasis. A three-dimensional (3-D) spheroid culture system was used to determine the effect of DPP23 on tumour invasion. As shown in Fig. 2a, spheroids of MDA-MB-231 cells were not invasive under unstimulated conditions (Control). When the cells were treated with TNFα , they began to spread out of the spheroid into the surrounding matrix in a characteristic starburst pattern. However, these invasive protrusions were reduced in the presence of DPP23 (Fig. 2a). We calculated the invasive area using ImageJ software and showed that DPP23 significantly inhibited the invasive capability of MDA-MB-231 spheroids (all P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test) (Fig. 2b).
DPP23 downregulates TNFα-induced MMP-9 mRNA expression. MMPs are proteolytic enzymes that degrade ECM proteins and play an important role in tissue remodelling. The basement membrane is a fibrous ECM that is important for the structural support of epithelial cells and maintains tissue organization. Gelatinases, such as MMP-2 and -9, are members of the MMP family that are able to cleave type IV collagen, a major component of the basement membrane. Gelatinase expression is highly elevated in many malignant human tumours and is strongly correlated with cancer cell invasion and metastasis 24 . To examine whether DPP23 affects MMP-2 or -9 activity, we performed an in-gel gelatinase activity assay. Gelatin zymography showed that the gelatinase activity of MMP-9 was strongly increased upon TNFα stimulation, which was significantly suppressed in the presence of DPP23 (all P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test, n = 3; Fig. 3a). The invasive area was quantified using ImageJ software. Error bars represent the mean ± SD of three independent experiments. Statistical significant differences were determined by one-way ANOVA followed by Sidak's multiple comparisons test. **P < 0.001 (n = 3). In contrast, no effect on MMP-2 gelatinase activity was observed following TNFα or DPP23 treatment. TNFα upregulates the expression of the MMP-9 gene in many cancer cells [25][26][27][28] . We next investigated whether DPP23 alters TNFα -induced expression of the MMP-9 gene. MDA-MB-231 cells were treated with DPP23 prior to stimulation with TNFα . The reverse transcription-polymerase chain reaction (RT-PCR) analysis showed that DPP23 alone had no effect, but substantially reduced TNFα -induced MMP-9 mRNA expression (Fig. 3b). Real-time PCR analysis was performed to precisely quantify the effect of DPP23 on MMP-9 mRNA expression. As shown in Fig. 3c, the MMP-9 mRNA level was increased by 22.4 ± 2.50-fold in response to TNFα stimulation; however, in the presence of DPP23, the level of this mRNA was significantly decreased to 2.27 ± 0.473-fold compared with the vehicle control (P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test). To understand the possible mechanism by which DPP23 reduces MMP-9 expression, MDA-MB-231 cells were transfected with an MMP-9 promoter reporter, pMMP9-Luc(− 925/+ 13), and the effect of DPP23 on MMP-9 promoter activity was analysed. As shown in Fig. 3d, DPP23 significantly reduced TNFα -induced MMP-9 promoter activity (P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test). These results indicate that DPP23 has the ability to inhibit TNFα -induced MMP-9 expression at the transcriptional level. Conditioned media were collected, and gelatinolytic activity was measured using a 10% Zymogram gel (10% Tris-glycine/0.1% gelatin). Zymography band intensities were measured by ImageJ software and expressed as arbitrary units; background intensity level was set to "1". The data are plotted as the mean ± SD. P values were determined by one-way ANOVA followed by Sidak's multiple comparisons test (n = 3). (b) MDA-MB-231 cells were untreated or treated with 10 ng/mL TNFα in the presence or absence of 10 μ M DPP23 for 18 h. Total RNA was isolated, and RT-PCR was then performed. The GAPDH mRNA was used as an internal control. (c) MDA-MB-231 cells were treated as in (b), and the MMP-9 mRNA levels were measured by quantitative real-time PCR. The relative fold changes were normalized to the level of the GAPDH mRNA in the same sample. The data represent the mean ± SD of three independent experiments. **P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test (n = 3). ns, not significant versus vehicle control. (d) MDA-MB-231 cells were transfected with 0.2 μ g of pMMP9-Luc(− 925/+ 13) and 50 ng of pRL-null. After 48 h, the cells were treated with 10 ng/mL TNFα in the presence or absence of 10 μ M DPP23 for 8 h, and the luciferase activities were then determined. Firefly luciferase activity was normalized to Renilla luciferase activity. The data represent the mean ± SD of three independent experiments. **P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test (n = 3). ns, not significant versus vehicle control.
Scientific RepoRts | 6:34134 | DOI: 10.1038/srep34134 DPP23 suppresses TNFα-induced NF-κB activation. The MMP-9 gene is regulated by various transcription factors, including AP-1, NF-κ B, SP1, Egr-1, and Ets-1 27,[29][30][31] . Of these transcription factors, NF-κ B is activated by TNFα [31][32] . Mutation of the NF-κ B binding site within the MMP-9 gene promoter resulted in the suppression of TNFα -induced MMP-9 promoter activity (Fig. 4a), while transfection with empty vector did not affect TNFα -induced reporter activity. These data suggest that NF-κ B plays a critical role in TNFα -induced MMP-9 gene transcription. To assess the possibility that the DPP23-induced suppression of MMP-9 promoter activity is associated with the inhibition of NF-κ B, MDA-MB-231 cells were pretreated with DPP23 prior to the addition of TNFα , followed by the analysis of NF-κ B phosphorylation. The immunoblot analysis showed that DPP23 dose-dependently reduced TNFα -induced phosphorylation of Iκ B on Ser32 and p65 NF-κ B on both Ser536 and Ser438 (Fig. 4b), suggesting that DPP23 targets the Iκ B upstream kinase. We next tested whether the DPP23-induced inhibition of NF-κ B phosphorylation is functionally linked to the inhibition of its transcriptional activity. MDA-MB-231 cells were transfected with an NF-κ B cis-acting luciferase reporter containing five repeats of NF-κ B binding sites (5 × NFκ B-Luc). In this reporter system, luciferase activity represents NF-κ B-mediated transcriptional activity. TNFα increased luciferase reporter activity 7.4 ± 0.76-fold (Fig. 4c). However, this activity was significantly reduced in the presence of DPP23 compared with the activity in the TNFα -treated group After 48 h, the cells were treated with 10 ng/mL TNFα for 8 h, and the luciferase activities were then determined. Firefly luciferase activity was normalized to Renilla luciferase activity. The data represent the mean ± SD of three independent experiments performed in triplicate. **P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test (n = 9). (b) Serum-starved MDA-MB-231 cells were pretreated with DPP23 (10, 15, or 20 μ M) for 30 min before stimulation with 10 ng/mL TNFα . After 20 min, whole-cell lysates were prepared and immunoblotting was performed using phospho-specific antibodies against Iκ Bα (Ser32), p65 (Ser536), or p65 (Ser468). GAPDH was used as an internal control. (c) MDA-MB-231 cells were transfected with the 5 × NFκ B-Luc plasmid, along with 50 ng of pRL-null. At 48 h post-transfection, the cells were untreated or treated with 10 ng/mL TNFα in the presence or absence of DPP23 (10 or 20 μ M). The data represent the mean ± SD of three independent experiments performed in triplicate. **P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test (n = 9). (d) MDA-MB-231 cells were pretreated with 15 μ M DPP23 for 30 min before stimulation with 10 ng/mL TNFα . After 20 min, the cells were fixed and incubated with antibodies against α -tubulin or phospho-p65 (Ser536) for 2 h, followed by incubation with Alexa Fluor 488-conjugated (green signal) or Alexa Fluor 555-conjugated (red signal) secondary antibodies for 30 min. Nuclear DNA was stained with 1 μ g/mL Hoechst 33258 for 10 min (blue signal).
Scientific RepoRts | 6:34134 | DOI: 10.1038/srep34134 (P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test, n = 9). To further determine the effect of DPP23 on the inhibition of NF-κ B, p65 phosphorylation on Ser536 was analysed using immunofluorescence microscopy. Upon TNFα stimulation, fluorescent staining for phospho-p65 NF-κ B at Ser536 was remarkably increased in the nucleus. However, when the cells were pretreated with DPP23, this phosphorylation was substantially suppressed (Fig. 4d), supporting the assumption that DPP23 inhibits MMP-9 expression by inhibiting the NF-κ B-mediated pathway.

DPP23 binds to the catalytic domain of Akt2 in silico.
The phosphatidylinositol 3-kinase (PI3K)-Akt signalling pathway can activate NF-κ B by phosphorylating IKK [15][16][17][18][19] . Three Akt isoforms (Akt1, Akt2, and Akt3) are expressed in MDA-MB-231 cells 33 . As Akt2 is the predominant form amplified in breast cancer cells 34 and is involved in the invasion and metastasis of breast cancer cells 35 , it is tempting to speculate that DPP23 inhibits NFκ B by targeting Akt2. To test this possibility, we determined whether DPP23 could bind to Akt2 using in silico molecular docking simulations. A pan-Akt inhibitor, GSK690693, was used as a reference ligand 36 . The LigPlot analysis showed that twenty-two residues of Akt2 (Leu158, Gly159, Lys160, Phe163, Val166, Ala179, Lys181, Glu200, Leu204, Phe227, Met229, Glu230, Tyr231, Ala232, Glu236, Glu279, Met282, Thr292, Asp293, Phe294, Phe439, and Phe443) were involved in GSK690693 binding; four residues (Lys160, Glu230, Ala232, and Asp293) formed five hydrogen bonds (H-bonds), and eighteen residues formed hydrophobic interactions (Fig. 5a, left  panel). Similarly, ten residues (Lys160, Phe163, Val166, Lys181, Thr213, Met229, Glu279, Met282, Thr292, and Asp293) were located in close proximity to the DPP23 binding site (Fig. 5a, right panel). All of these residues except Thr213 were observed in both complexes, two residues (Lys181 and Asp293) formed two H-bonds, and ten residues formed hydrophobic interactions. PyMOL showed that five H-bonds were formed between Akt2 and GSK690693 (Fig. 5b, left panel), while two H-bonds were formed in the Akt2 and DPP23 complex (Fig. 5b, right  panel). Based on these data, we suggest that DPP23 may bind to the catalytic domain of Akt2. DPP23 inhibits TNFα-induced Akt activation. We next examined whether DPP23 inhibits Akt activity. For maximal activation, threonine phosphorylation in the catalytic domain (at Thr308 in Akt1 and Thr309 in Akt2, located in the activation loop) and serine phosphorylation (at Ser473 in Akt1 and Ser474 in the C-terminal hydrophobic motif of Akt) are required 37 . Figure 6a shows that TNFα increased the phosphorylation of Akt on Ser473/474 and of IKKα /β on Ser176/180 in serum-starved MDA-MB-231 cells, while treatment with DPP23 alone had no effect on their basal phosphorylation. However, when cells were pretreated with DPP23, TNFα -induced phosphorylation of Akt1/2 and IKKα /β was significantly reduced (all P < 0.0001 by one-way ANOVA followed by Sidak's multiple comparisons test, n = 3). Although Akt1 is important for tumour development, Akt2 is known to play a pivotal role in tumour invasion and metastasis [38][39] . To further investigate the possibility that inhibition of IKK phosphorylation in response to DPP23 treatment is due to the inhibition of Akt, we established MDA-MB-231 variant cell lines that stably expressed an Akt2 shRNA (MB231/shAkt2) or scrambled shRNA (MB231/shCT). Stable knockdown of Akt2 by shRNA was evaluated by immunoblotting (Fig. 6b). Silencing of endogenous Akt2 significantly attenuated the ability of TNFα to induce IKKα /β phosphorylation (all P < 0.05 by one-way ANOVA followed by Sidak's multiple comparisons test, n = 3). These findings suggest that DPP23 inhibits IKK activity by targeting Akt, resulting in the suppression of TNFα -induced NF-κ B phosphorylation.

Akt2 is involved in TNFα-induced MMP-9 transcription. RT-PCR analysis was performed on MB231/
shCT and MB231/shAkt2 cells to further determine whether the suppression of TNFα -induced MMP-9 expression by DPP23 is functionally linked to the inhibition of Akt2. As shown in Fig. 6c, the induction of MMP-9 mRNA expression by TNFα was substantially reduced in MB231/shAkt2 cells compared with MB231/shCT cells. Real-time PCR analysis demonstrated a 17.07 ± 1.96-fold increase in the level of the MMP-9 mRNA in the MB231/shCT cells, whereas a 6.30 ± 0.56-fold increase was observed in MB231/shAkt2 cells following TNFα stimulation compared with the basal levels (Fig. 6d). These data suggest that Akt plays an important role in the TNFα -induced regulation of MMP-9 transcription in MDA-MB-231 cells. Thus, DPP23 reduces TNFα -induced MMP-9 expression by inhibiting the Akt-IKK-NF-κ B axis in MDA-MB-231 cells.

DPP23 suppresses experimental syngenic liver metastasis in vivo.
We next investigated the effect of DPP23 on the anti-metastatic potential of cells using a mouse syngenic intrasplenic metastasis model. We used 4T1 mouse mammary carcinoma cells, which are known to be highly metastatic to the liver [40][41] . On day 1 after the intrasplenic injection of 4T1 cells, the mice were randomly divided into two groups, and phosphate-buffered saline (PBS; n = 4) or 10 mg/kg DPP23 (n = 6) was inoculated in the peritoneum daily for 7 days (Fig. 7a). All mice were euthanized on day 8 after implantation (Fig. 7b). Hematoxylin-Eosin (H&E) staining revealed the marked diffuse expansion of 4T1 cells around the red pulp region of the spleen in both vehicle-and DPP23-treated mice (Fig. 7c). Liver metastasis of 4T1 cells injected to spleen was observed based on markedly increased formation of tumour nodules in the liver of PBS-treated mice (Fig. 7d, middle panels). In contrast, smaller tumour nodules were observed in the liver of DPP23-treated mice (Fig. 7d, bottom panels). These data suggest that DPP23 has a potent inhibitory effect on the experimental liver metastasis of mouse mammary tumour cells.
In summary, this study shows that DPP23 inhibits the TNFα -induced invasion of highly metastatic MDA-MB-231 breast cancer cells by targeting the Akt-IKK-NF-κ B axis to inhibit the expression of the MMP-9 gene. We also found that DPP23 attenuates experimental liver metastasis in a syngenic intrasplenic transplantation model using 4T1 mouse mammary carcinoma cells. This finding suggests that DPP23 could be used to prevent the invasion and metastasis of early-stage breast cancer or as an adjuvant for chemo/radiotherapy. The pRL-null plasmid, which encodes Renilla luciferase, was also purchased from Promega.

Methods
Cell migration assay. The migration of MDA-MB-231 cells was determined using an in vitro scratch-wound healing assay, as previously described 42 . In brief, a scratch was made on a monolayer of cells with a micropipette tip, and the cells were untreated (control) or treated with 20 ng/mL TNFα in the presence or absence of 10 μ M DPP23. After 10 h, the cells were photographed with an EVOS ® FL Auto Cell Imaging System (Life Technologies, Carlsbad, CA, USA). Cell migratory ability was determined by counting the number of cells invading the scratch-wound area. P value was determined by one-way ANOVA followed by Sidak's multiple comparisons test.   with or without 10 μ M DPP23 for 30 min before stimulation with 10 ng/mL TNFα . After 20 min, whole-cell lysates were prepared and immunoblotting was performed using the phospho-specific antibodies against Akt (Ser473), p65 (Ser536), or p65 (Ser468). GAPDH was used as an internal control. Band intensities were analysed using a quantitative scanning densitometer. The data are plotted as the mean ± SD. P values were determined by one-way ANOVA followed by Sidak's multiple comparisons test (n = 3). (b) MB231/shCT and MB231/shAkt2 cells were serum-starved for 24 h and then pretreated with 10 μ M DPP23 for 30 min before stimulation with 10 ng/mL TNFα . After 20 min, the cells were lysed and immunoblotting was performed using antibodies against Akt2 and a phospho-specific antibody against IKKα /β (Ser176/180). GAPDH was used as an internal control. Band intensities were analysed using a quantitative scanning densitometer. The data are plotted as the mean ± SD. P values were determined by one-way ANOVA followed by Sidak's multiple comparisons test (n = 3). (c,d) MB231/shCT and MB231/shAkt2 cells were pretreated with 10 μ M DPP23 for 30 min before stimulation with 10 ng/mL TNFα . After 12 h, the cells were collected and the MMP-9 mRNA levels were measured by RT-PCR (c) and quantitative real-time PCR (d). The data are plotted as the mean ± SD. P values were determined by one-way ANOVA followed by Sidak's multiple comparisons test. **P < 0.0001 (n = 9).

3-D spheroid culture and invasion assay.
Scientific RepoRts | 6:34134 | DOI: 10.1038/srep34134 TNFα in the presence or absence of 10 μ M DPP23. Invasive protrusions into the ECM were visualized with an EVOS ® FL Auto Cell Imaging System (Life Technologies), and the invasive area was quantified using ImageJ software (NIH).

RT-PCR and quantitative real-time PCR.
Total RNA was extracted using the Isol-RNA lysis reagent (NucleoZOL; Clontech, Mountain View, CA, USA), and PCR was performed as previously described 43 . For quantitative real-time PCR, the TaqMan-iQ TM supermix kit (Bio-Rad, Hercules, CA, USA) was used with the Bio-Rad iCycler iQ TM thermal cycler, according to the manufacturer's instructions. The TaqMan TM fluorogenic probes and gene-specific PCR primers for MMP-9 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) are described elsewhere 43 . The relative fold changes were normalized to the level of the GAPDH mRNA in the same sample.
In-gel gelatinase assay. Gelatin zymography was performed using the Novex ® Zymogram gelatin gel system (Novex, San Diego, CA, USA) according to the manufacturer's instructions. Briefly, serum-starved MDA-MB-231 cells were treated with 10 ng/mL TNFα for 24 h in the presence or absence of DPP23. The conditioned medium was collected and subjected to polyacrylamide gel electrophoresis using 10% Novex ® pre-cast gels (10% Tris-glycine gel/0.1% gelatin) under nonreducing conditions. After electrophoresis, the proteolytic activity of the MMP enzymes on gelatin was measured as previously reported 31 . Zymography band intensities were expressed as arbitrary units, which set "1" as a background intensity.  antibodies. Primary antibodies against phospho-IKKα /β (Ser176/180), phospho-Iκ B (Ser32), phospho-p65 (Ser536), and phospho-p65 (Ser468) were obtained from Cell Signaling Technology (Beverly, MA, USA). The glyceraldehyde phosphate dehydrogenase (GAPDH) antibody was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). The blots were developed using an enhanced chemiluminescence detection system (GE Healthcare, Piscataway, NJ, USA). MMP-9 promoter reporter assay. The MMP-9 promoter reporters, wild-type pMMP9-Luc(− 925/+ 13), and disrupted NF-κ B binding site pMMP9-Luc(− 925/+ 13)mtNFκ B are described elsewhere 31 . MDA-MB-231 cells cultured on 12-well plates were transfected with 0.5 μ g of the MMP-9 promoter reporter constructs, along with 50 ng of the pRL-null plasmid encoding Renilla luciferase to monitor transfection efficiency. At 48 h post-transfection, the cells were treated with 10 ng/mL TNFα in the presence or absence of 10 μ M DPP23. After 8 h, the cells were collected, and the firefly luciferase activities were measured as previously described 31 . Luminescence was measured using a Centro LB960 dual luminometer (Berthold Technologies).

NF-κB-dependent
Immunofluorescence microscopy. MDA-MB-231 cells plated on coverslips were untreated or treated with 10 ng/mL TNFα in the presence or absence of 10 μ M DPP23 for 30 min, followed by fixation, permeabilization, and incubation with primary antibodies against α -tubulin and phospho-p65 (Ser536) for 2 h, as previously described 8 . Secondary antibodies conjugated to Alexa Fluor 488 (Invitrogen; green signal for α -tubulin) and Alexa Fluor 555 (Invitrogen; red signal for phospho-p65) were incubated with the cells for 30 min, followed by staining of the nuclear DNA (1 μ g/mL Hoechst 33258 (Sigma-Aldrich), blue signal) for 10 min. The stained cells were examined under an EVOS f1 fluorescence microscope (Advanced Microscopy Group; Bothell, WA, USA).

Molecular docking simulations.
Molecular docking experiments were performed on an Intel Core 2 Quad Q6600 (2.4 GHz) Linux PC with Sybyl 7.3 (Tripos, St. Louis, MO), as previously described 44 . 3D0E.pdb was used as the X-ray crystallographic structure of human Akt2 36 . The 3-D structure of DPP23 was determined using X-ray crystallography. All docking experiments were performed on an Intel Core 2 Quad Q6600 (2.4 GHz) Linux PC with Sybyl 7.3 (Tripos). The structure of the apo-protein of Akt2 (apo-3D0E) was prepared using the Sybyl program, and its solution structure was obtained by energy minimization using a previously reported detailed procedure 45 . The root-mean-square deviation between the crystal structure and the apo-protein was 0.1309 Å. The binding sites were determined based on the interpretation of the results analysed by the LigPlot program, as reported in the protein data bank 46 : Ala179, Lys181, Glu200, Met229, Ala232, Glu236, Asp293, and Phe294. A previously reported flexible docking method was used 45 . To confirm whether this docking process was correct, we docked 4-{2-(4-amino-1,2,5-oxadiazol-3-yl)-1-ethyl-7-[(3S)-piperidin-3-ylmethoxy]-1H-imidazo [4,5-c] pyridin-4-yl}-2-methylbut-3-yn-2-ol (GSK690693) as a reference ligand 36 . The radius for the active site was set to 1.6 Å. The docking process was iterated 30 times, and 30 complexes were generated ( Supplementary Fig. S1a). Of them, 23 complexes were bound to the same site of the crystal structure. Their binding energy ranged from − 21.94 to -10.42 kcal/mol. Similarly, DPP23 was docked into the apo-3D0E structure ( Supplementary Fig. S1b). Thirty complexes bound well, and their binding energy ranged from − 9.83 to -8. Syngenic experimental liver metastasis assay. Six-week-old female Balb/c mice were purchased from YoungBio (Seongnam, Gyeonggi-do, Republic of Korea). All animal experiments were conducted according to the standards and procedures approved by the Konkuk University Institutional Animal Care and Use Committee (no. KU15194). Balb/c mice were anaesthetized, and the spleen was exposed through a small incision in the left abdominal flank, as previously described 47 . The 4T1 mammary carcinoma cells (6 × 10 4 cells/50 μ L) were injected into the splenic lobe with a 30-gauge needle. The spleen was returned to the abdominal cavity, and the abdominal wall was closed using 6-0 nylon sutures. On day 1 post-implantation of the 4T1 cells, the mice were randomly divided into two groups. One group was intraperitoneally administered with PBS (n = 4), while the other group was intraperitoneally administered with 10 mg/kg DPP23 (n = 6) daily for 7 days. All mice were euthanized on day 8 after implantation, and the spleen and liver tissues were fixed with 10% phosphate buffered formalin and stained with haematoxylin and eosin (H&E). Statistical analysis. The data are presented as the mean ± S.D. Statistical analysis was analyzed by one-way analysis of variance (ANOVA) followed by Sidak's multiple comparisons test using GraphPad Prism version 7.0 software (GraphPad Software Inc., La Jolla, CA, USA). P values less than 0.05 were considered statistically significant.