MTT cell viability assays were performed by plating cells at 7,500/well in 96-well plates, followed by treatment of the cells with increasing concentrations of TA (Chem-Impex International, Inc., Wood Dale, IL, United States) for 72 h at 37°C. After removal of the cell media, cells were incubated with 0.5 mg/ml MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) (Biomatik, Wilmington, DE, United States) in PBS for 3 h. The MTT solution was subsequently removed and the MTT formazan product was dissolved in 100 μl of DMSO. Absorbance (570 and 690 nm) of the MTT formazan product was measured in a plate reader.

012/LVM2/LR10 tumors were generated by injection of 1 × 10⁶ 012/LVM2/LR10 cells into the #4 mammary fat pads of NOD-SCID-gamma (NSG) mice from Jackson Laboratories (Bar Harbor, ME, United States), as described previously (Wang et al., 2019b). In the tumor study in Figures 1F,G, mice bearing HCI-012/LVM2/LR10 breast tumors were treated once daily with 375 mg/kg TA by oral gavage for 12 days followed by breast tumor removal. In the tumor study in Figure 4A, tumors were generated using the 012/LVM2/LR10 cancer cell model as above and the animals were treated with vehicle (peanut oil), 375 mg/kg TA in water, 10 mg/kg dMtcyDTDO in peanut oil, or 375 mg/kg TA + 10 mg/kg dMtcyDTDO once daily by oral gavage. Tumor dimensions were recorded daily at the time of treatment.

Protein synthesis assays were performed by measurement of the incorporation of ³H-Leucine (Perkin Elmer, Waltham, MA, United States) into protein, as detailed previously (Law et al., 2000a; Law et al., 2000b).

Thymidine incorporation assays were described in a previous publication (Corsino et al., 2009).

MDA-MB-468 cells were plated at 10⁵/well in 24 well plates and incubated overnight to permit attachment. Cells were gently washed with PBS and fed with serum-free and Arg and Lys-free DMEM (Thermo Fisher, Waltham, MA, United States, Cat. # 88364) containing the indicated concentrations of TA and incubated for 24 h. Cells were pulsed for 2 h with 0.5 μCi/well ³H-Arg (PerkinElmer, Waltham, MA, United States, Cat. # NET1123250UC) or 0.2 μCi/well ¹⁴C-Lys (PerkinElmer, Cat. # NEC280E050UC). The cells were washed twice with PBS and lysed in 300 μl/well 0.1% SDS in 0.5 N NaOH. Uptake of radiolabeled amino acids was quantified by counting 100 μl of the lysate from each well in a scintillation counter.

P300 Acetylation assays were performed by incubating the reaction buffer (50 mM Tris, pH 8.0, 10% glycerol, 0.1 mM EDTA, and 1 mM DTT), 0.5 μg p300 (ENZO Life Sciences, Farmingdale, NY, United States), 0.5 μg Histone H3 (NEB, Ipswich, MA, United States), and 60 μM Acetyl CoA (Sigma-Aldrich, St. Louis, MO, United States) in the presence or absence of TA, NacTA, or tcyDTDO for 2 h at 30°C. Reactions were terminated by the addition of 2X SDS sample buffer and boiling for 10 min, followed by immunoblot analysis employing a 45 min transfer to 0.2 μm nitrocellulose using CAPS buffer, pH 11.

The BT474, MDA-MB-231, MDA-MB-468, MCF10A, HMEC, B16, MALME-3M, DU145, SH-SY5Y, K 562, and MOLT4 cells lines were purchased from American Type Culture Collection (ATCC) (Manassas, VA USA). The KMS-11, KMS-18, JIM-3, and 4937 cell lines were kindly provided by Dr. Olga Guryanova, University of Florida Health Cancer Center and the WM793 cell line was generously provided by Dr. W. Douglas Cress, Moffitt Cancer Center. Derivation of the HCI-012/LVM2/LR10 cell line was previously described (Wang et al., 2019a; Wang et al., 2019b). MCF10A cell lines stably expressing MYC were generated as detailed in previous reports (Law et al., 2002; Wang et al., 2019a; Wang et al., 2019b), as was the generation of the 231/E-Cad and 231/E-Cad-GFP cells lines (Law et al., 2013). Cell lysates were prepared, as described previously (Law et al., 2002). The antibodies used for immunoblot analyses are listed in Supplementary Table S1.

Compounds were docked into the Arg binding pocket of the bacterial homolog of CAT (CATh). The crystal structure was downloaded from the RCSB protein databank (Accession # 6f34, (Jungnickel et al., 2018)) in .pdb format and all heteroatoms were removed. The molecular surface of the protein was created using the Chimera program and spheres representing potential docking sites were created using SPHGEN. Spheres that did not correspond to the Arg binding pocket were removed and the SHOWBOX program created a 3-dimensional box around the spheres with a 4 Å buffer on all sides. The GRID program was then used to generate files representing the electrostatic and van der Waals forces produced by the protein in the region of interest. The DOCK program was used to calculate GRID scores for the ligands of interest, consisting of electrostatic and van der Waals interactions, and ranked depending on predicted binding affinity. All programs were from UCSF and, except for Chimera, are included in the DOCK 6.7 suite (Allen et al., 2015). Images were generated using PyMol from Schrodinger.

Chemical structures of TA and the other compounds used or mentioned in the work described here are shown in Figure 1A. MTT cell viability assays showed that TA suppressed the viability of a variety of cancer cell lines including melanoma (B16, WM793, and MALME-3M), breast (MDA-MB-468 and BT474), prostate (DU145), and neuroblastoma (SH-SY5Y) lines (Figure 1B). DU145 was one of the more TA-sensitive lines and exhibited statistically significant TA-mediated viability suppression at the lowest TA concentration tested 0.25% (w/v) compared with the vehicle control (Supplementary Figure S1A). Analysis of TA effects on cell proliferation, as measured by ³H-thymidine incorporation, showed that TA inhibited DNA synthesis in the BT474 and MDA-MB-468 cells over the same concentration range that it reduces cell viability (Figure 1C). Studies of TA effects on the viability of leukemia, lymphoma, and multiple myeloma lines using the XTT assay showed concentration-dependent reductions in cancer cell viability (Figure 1D). The MYC oncogene is frequently overexpressed in human cancers (Rothberg 1987; Saksela 1990; Shiu et al., 1993). However, MYC overexpression renders cancer cells vulnerable to some therapeutics, resulting in collateral vulnerabilities amenable to synthetic lethal targeting approaches (Hsieh and Dang 2016; Lee et al., 2019). Therefore, we examined if MYC overexpression increased the sensitivity of the MCF10A human breast epithelial cell line to TA. MYC overexpression enhanced the TA suppression of cell viability across an array of TA concentrations suggesting that MYC overexpression may be useful as a marker to identify TA-sensitive tumors (Figure 1E).

We employed a previously described metastatic, Lapatinib-resistant, orthotopic xenograft model, 012/LVM2/LR10 (Wang et al., 2019a; Wang et al., 2019b), to examine if TA reduces the growth of breast tumors in experimental animals (Figure 1F, upper panel). Tumor growth studies revealed that 375 mg/kg TA suppressed the growth of 012/LVM2/LR10 tumors without evidence of treatment toxicity (Figure 1F, lower panel). At the 12-day treatment endpoint, tumors in the TA treatment group were approximately one-ninth the size of the tumors in the vehicle group (Figure 1G).

We next sought to examine the mechanisms through which TA impacts tumor growth. TA may exhibit anticancer activity through effects on Plasmin activity and via Plasmin-independent actions (O'Grady et al., 1981; Tanaka et al., 1982; Stonelake et al., 1997; Dunbar et al., 2000; Suojanen et al., 2009). Previous work indicated that Plasmin cleaves the extracellular domain of the transmembrane protein CUB Domain-Containing Protein 1 (CDCP1) to produce a truncated protein that exhibits differential protein-protein interactions and signaling functions compared with the full-length protein, suggesting that cleaved CDCP1 (cCDCP1) may contribute to tumor growth and progression (Casar et al., 2014; Law et al., 2016; Wright et al., 2016). However, the ability of TA to suppress CDCP1 cleavage has not been examined. TA reduced CDCP1 cleavage in MDA-MB-231 breast cancer cells (Figure 2A). Unlike Transforming Growth Factor-β (TGFβ) or the glucocorticoid dexamethasone, which block CDCP1 cleavage by upregulating Plasminogen Activator Inhibitor-1 (PAI-1) (Law et al., 2013; Law et al., 2016), TA did not increase PAI-1 levels (Figure 2A). cCDCP1 binds E-cadherin and interferes with its cell-cell adhesive functions, while the full-length protein does not (Law et al., 2016). Blocking CDCP1 cleavage via TA administration may be useful on restoring the anti-invasive functions of E-cadherin. Since full-length CDCP1 does not bind E-cadherin, we did not expect enforced E-cadherin expression to alter TA inhibition of CDCP1 cleavage positively or negatively. To test this, we examined if enforced expression of E-cadherin or an E-Cadherin-Green Fluorescent Protein (GFP) fusion protein influences TA blockade of CDCP1 cleavage (Figure 2B). TA (0.5% w/v) decreased CDCP1 cleavage in MDA-MB-231 cells irrespective of E-cadherin overexpression and did not significantly impact CDCP1 tyrosine phosphorylation on multiple sites previously implicated in regulating downstream signaling proteins including Src (He et al., 2010; Liu et al., 2011). Consistent with this, TA did not alter Src phosphorylation on activating (Y416) or inhibitory (Y527) sites (Figure 2B). FAK and CDCP1 exhibit reciprocal tyrosine phosphorylation patterns in which FAK is heavily phosphorylated in cells attached to substrata, while CDCP1 is heavily phosphorylated in cells growing in suspension (Wortmann et al., 2011). TA did not alter FAK phosphorylation on the Y397 site that reflects FAK activation (Thamilselvan et al., 2007). Both the full length, 135 kDa, and cleaved, 85 kDa, forms of CDCP1 exhibited tyrosine phosphorylation at multiple sites.

We also screened the effects of TA on several other signaling pathways important in tumor growth and progression including the oncogenic transcription factors STAT3 and MYC, the mitogen-activated serine/threonine kinases S6 protein kinase (S6K), ERK, and AKT, and the levels and phosphorylation status of proteins that control cell cycle progression including Cyclin D1, Cyclin B1, and cleavage of PARP and Caspase 3 as markers of apoptotic cell death. TA reduced STAT3 tyrosine phosphorylation on its activating site, Y705, decreased MYC protein expression, and suppressed S6K1 phosphorylation on the activating site, T389 (Figure 2C). TA reduction of cell viability in MTT assays closely paralleled suppression of proliferation in thymidine incorporation assays (see Figures 1B,C). Consistent with this observation, TA did not increase the apoptotic markers Cleaved Caspase 3 (CC3) or PARP cleavage. However, tcyDTDO increased CC3 to detectable levels and elevated PARP cleavage as expected. TA had no consistent effect on ERK or AKT phosphorylation, Cyclin D1 or Cyclin B1 levels, or PARP or Caspase 3 cleavage across the MDA-MB-468, BT474, and WM793 human cancer cell lines (Supplementary Figure S1B).

TA treatments were also carried out in the presence of 10X excesses of Lys or Arg to examine the ability of high levels of these amino acids to overcome TA effects. High Lys or Arg levels did not alter STAT3 phosphorylation, but increased S6K1 phosphorylation and MYC levels consistent with the sensitivity of S6K1 activation (Iiboshi et al., 1999; Hidayat et al., 2003) and MYC expression (Yue et al., 2017) to amino acid levels (Figure 2C). However, TA still reduced S6K1 phosphorylation and MYC levels in the presence of excess Lys or Arg. Treatment with the protein synthesis inhibitor cycloheximide (CHX) was included as a control to determine if inhibition of protein synthesis was sufficient to mimic any TA effects. CHX decreased STAT3 phosphorylation, but did not decrease MYC levels or S6K1 phosphorylation. Control protein synthesis experiments showed that 1.25 μM CHX inhibits protein synthesis by approximately 60% (Supplementary Figure S1C).

Our group identified a class of compounds with anticancer activity termed DDAs that interfere with disulfide bond formation, induce endoplasmic reticulum (ER) stress, and inhibit protein synthesis (Ferreira et al., 2015; Ferreira et al., 2017; Wang et al., 2019a; Wang et al., 2019b). Results posted in a preprint suggest that DDAs act by inhibiting a subset of Protein Disulfide Isomerases (Law et al., 2021). Therefore, we examined if DDAs induce overlapping responses with TA and therefore may be useful in combination regimens against cancer. The DDA tcyDTDO decreased STAT3 and S6K1 phosphorylation and MYC levels (Figure 2C). Combined tcyDTDO/TA treatment reduced overall S6K1 phosphorylation more than either treatment alone as evidenced by electrophoretic mobility shifts of both the 70 and 85 kDa forms of S6K1 (Figure 2C, expanded region). We next examined the effect of TA on the phosphorylation of STAT3 and S6K1 in the MCF10A human mammary epithelial cell line overexpressing MYC and the MDA-MB-468 breast cancer line. TA induced a concentration-dependent decrease in STAT3 phosphorylation on Y705 but had no effect on phosphorylation on S727 (Figure 2D). The AGC-family kinases AKT and S6K1 have similar Arg-rich consensus substrate recognition sequences. Immunoblot analyses using antibodies recognizing phospho-AKT substrates and the S6K1 substrate S6 showed a TA-induced decrease in phosphorylation at these sites. This was paralleled by reduced phosphorylation of the S6K1 site S411 that is also associated with kinase S6K1 activation (Han et al., 1995). Together, the results discussed thus far in Figure 2 indicate that TA inhibits a subset of mitogenic/oncogenic signaling pathways. Quantification studies demonstrated that 2% w/v TA decreases phosphorylation of S6K1 on T389 and STAT3 on Y705, which are sites required for their kinase and transcriptional activities, respectively (Figure 2E). TA reduced CDCP1 cleavage in a concentration-dependent and statistically significant manner (Figure 2E). TA also decreased MYC expression levels, but this trend did not reach statistical significance.

Since TA and protein synthesis inhibition by CHX both decreased STAT3 tyrosine phosphorylation (Figure 2C) and TA mimics Lys and Arg, we hypothesized that TA may inhibit protein synthesis. TA inhibited protein synthesis in a concentration-dependent manner, with 2% TA reducing protein synthesis by 60% (Figure 3A). DDAs also inhibit protein synthesis (Ferreira et al., 2017; Wang et al., 2019a; Wang et al., 2019b). Therefore, we examined if combining low concentrations of the DDA tcyDTDO and TA would suppress protein synthesis in an additive manner. Treatment with 1.25 μM tcyDTDO inhibited protein synthesis by 57% and addition of increasing concentrations of TA further reduced protein synthesis (Figure 3B). Similarly, combining concentrations of TA and tcyDTDO that have little effect on cancer cell viability alone resulted in larger reductions in viability (Figure 3C).

Suppression of histone acetylation by the p300/CBP enzymes is an emerging strategy for cancer therapy (Yang et al., 2013; Gu et al., 2016; Lasko et al., 2017; Wang et al., 2017; van Gils et al., 2021). We hypothesized that TA may alter Lys acetylation through competitive inhibition of Lys acetyltransferases and next examined if TA reduces acetylation of proteins on Lys. TA at 0.5 or 1% partially blocked protein Lys acetylation of a prominent ∼20 kDa band as detected with an acetyl-Lys antibody (Figure 3D, arrow). This band is likely acetylated histones, which are the most abundant acetylated cellular proteins in this molecular size range (Hansen et al., 2019). TcyDTDO also reduced acetylation of this band, and acetylation was strongly increased by treatment of the cells with the histone deacetylase inhibitor Vorinostat/SAHA. We next examined if TA could overcome the effects of SAHA. TA treatment reduced Lys acetylation, but TA did not override the increased acetylation induced by SAHA co-treatment (Figure 3E). In contrast, treatment with the DDA dMtcyDTDO (Law et al., 2021) decreased baseline Lys acetylation and blocked the SAHA-driven increase in acetylation. The p300 protein is a major cellular histone acetyltransferase (HAT). We next performed HAT enzyme assays to determine if TA or tcyDTDO directly inhibit p300 activity. TA (2%) did not affect p300 acetylation of Histone 3 in vitro (Figure 3F). TcyDTDO also did not directly inhibit p300 activity. TA is partly metabolized in the body by acetylation (0.5% of oral dose; http://www.accessdata.fda.gov/drugsatfda_docs/label/2020/022430s009lbl.pdf), and perhaps also in cultured cells. Therefore, we examined if the metabolite N-acetyl-TA (NacTA) alters p300 activity. NacTA partially inhibited p300 HAT activity and interestingly, combining NacTA and tcyDTDO strongly inhibited overall p300 activity (AcK), as well as acetylation of Histone H3 on Lys residues 9 and 27. Further studies are required to determine if TA, NacTA, or DDAs reduce cellular protein acetylation through effects on p300 or other HAT enzymes. However, the results in Figures 3D–F suggest that TA and/or its metabolites and DDAs may mediate their effects in part by altering protein acetylation.

The findings presented in Figures 1–3 indicated that TA suppressed the growth of tumors derived from 012/LVM2/LR10 cells in a 12-day study, and cell culture studies showed that TA may suppress cancer cell viability through multiple mechanisms. The results further suggest that TA biochemical mechanisms of anticancer action partially overlap with those of DDAs. Therefore, we examined the anticancer efficacy of 375 mg/kg TA, 10 mg/kg of the DDA dMtcyDTDO, or TA + dMtcyDTDO administered once daily by oral gavage in longer-term studies. Tumors were initiated by orthotopic injections of 10⁶ 012/LVM2/LR10 cells and treatment was initiated when tumors were detectable by palpation. By 20 days of treatment, tumors in vehicle-treated mice averaged over 200 mm³, while tumors in drug-treated animals were less than 50 mm³ (Figure 4A). Immunoblot analysis of tumor extracts from each group showed that TA reduced S6K1 phosphorylation on Thr³⁸⁹, which is required for its kinase activity (Dennis et al., 1998; McMahon et al., 2002), to a greater extent on average than the other treatments (Figure 4B). Quantification of the results in Figure 4B indicated that TA suppressed S6K1 phosphorylation in a statistically significant manner (Figure 4C). Rapamycin analogs (“rapalogs”) are in clinical use against human cancer and act in part by inhibiting S6K (Ferrari et al., 1993; Meng and Zheng 2015), and S6K inhibitors are under development as cancer therapeutics (Pearce et al., 2010; Nam et al., 2019), consistent with the well-established role of S6K activation in the molecular pathogenesis of multiple cancers.

We previously observed that treatment with the DDA tcyDTDO caused widespread apoptosis of cancer cells in primary tumors and metastases without inducing apoptosis in adjacent normal tissues, resulting in substantial regions of dead tumor tissue (Wang et al., 2019b). This indicates that calculating tumor volumes without taking into account viability of the cancer cells may underestimate the efficacy of anticancer regimens. Further, sampling portions of such heterogeneous tumors may introduce errors into downstream proteomic or transcriptomic analyses. Therefore, we examined the morphology of Hematoxylin and Eosin (H&E)-stained tumors from each of the four treatment groups. Two representative tumors from vehicle-treated mice (Figure 5A) exhibited predominantly viable tumor tissue (T), with small regions of dead cancer cells (N). Higher magnification images showed mitotic figures, demonstrating cancer cell proliferation in the vehicle-treated tumors (yellow arrows). Representative tumors from dMtcyDTDO-treated animals in contrast, exhibited predominantly dead tumor tissue (N), with small areas of apparently viable tumor cells (T) (Figure 5B). Closer inspection of the boundary between the dead and viable cancer cells showed tumor cells undergoing nuclear condensation and fragmentation (red arrows), consistent with apoptosis. Interestingly, tumors from TA-treated animals exhibited somewhat similar morphology to the tumors from dMtcyDTDO-treated animals with large regions of dead tumor tissue and tumor cells exhibiting nuclear condensation and fragmentation (red arrows, Figure 5C). Tumors from mice treated with TA + dMtcyDTDO (Figure 5D) exhibited similar morphology to the tumors from mice treated with TA or dMtcyDTDO with large regions of dead tumor tissue, tumor cells exhibiting nuclear condensation and fragmentation, and a lack of mitotic figures. A few previous studies have noted TA anticancer activity in xenograft models (Astedt and Trope 1980; Iwakawa and Tanaka 1982; Ogawa et al., 1982), and in a case report (Astedt et al., 1977), against several solid tumor types, including breast cancer. In these reports, TA anticancer activity was attributed to its antifibrinolytic actions. Our results demonstrate TA efficacy against breast tumors in animal models and identify several novel TA-induced biochemical responses that may contribute to TA anticancer activity.

As mentioned above, S6K1 activation is sensitive to the levels of amino acids, as is MYC expression and the initiation of protein synthesis. Given the similarity of TA to Lys and Arg, we hypothesized that TA may compete with Lys or Arg for uptake by Cationic Amino acid Transporter (CATs). If correct, this hypothesis may also explain how TA enters cells since TA contains charged carboxyl and amino groups and is thus unlikely to traverse the plasma membrane unaided. The structure of a bacterial homolog of the mammalian CAT-family proteins (CATh) has been obtained with Arg bound (Jungnickel et al., 2018). We deleted Arg from this structure to perform in silico molecular docking to determine if TA is predicted to bind to this pocket, and if so, how its predicted binding energy compares with that of Arg or Lys. Figure 6A, upper-left and upper-middle shows a model of how TA might bind to CATh as viewed in the plane of the plasma membrane, or looking down into the transporter pore. The DOCK 6.7 program estimated Arg to have the highest binding energy for the CATh pocket, followed by TA, and Lys. It must be noted that if CATh bound to Lys or TA had served as the original model rather than CATh bound to Arg the results may have differed. However, the results obtained suggest that TA might bind to CATh with an affinity that is similar in magnitude with that of Arg and Lys. We next examined if TA suppresses the uptake of radiolabeled Arg or Lys by breast cancer cells. TA reduced the absorption of ³H-Arg by MDA-MB-468 cells in a concentration-dependent manner (Figure 6B). Similarly, TA decreased ¹⁴C-Lys uptake by MDA-MB-468 cells in a concentration-dependent manner (Figure 6C). These results suggest that TA may reduce cancer cell viability in part by inhibiting import of Lys and Arg through competition for CATs or other amino acid transporters, and that TA might enter cancer cells through these same transporters.