Unravelling the anti-cancer mechanisms elicited by non-covalent thioredoxin reductase inhibitors for triple negative breast cancer therapy.

1
Introduction
The idea of targeting redox homeostasis as an anti-cancer strategy is not new. However, despite extensive research, translating this concept into effective treatments remains challenging. Redox homeostasis refers to the balance between reactive oxygen species (ROS) and the cell's antioxidant defense mechanisms. ROS influence every cancer hallmark, including DNA damage and genetic alterations, cell death, metabolic rewiring, therapy resistance, escape from tumor immune microenvironment, and metastasis [1,2]. In cancer cells low to moderate ROS levels induce genetic changes that are essential for cancer initiation, proliferation, and progression, as well as the development of therapeutic resistance. This positions the thioredoxin (Trx) and the glutathione (GSH) antioxidant systems as pivotal regulators of redox homeostasis and oncogenesis.
In breast cancer, studies using murine mammary gland tumor models have shown that GSH is required for tumor initiation [3]. However, once tumors are established, inhibiting GSH alone is ineffective due to compensatory upregulation of the Trx pathway [3]. In contrast, targeting the Trx pathway has demonstrated therapeutic benefits in limiting tumor progression [3]. The Trx system depends on the ubiquitously expressed selenoproteins TXNRD1 (cytosolic) and TXNRD2 (mitochondrial) isoenzymes along with NAPDH to maintain reduced Trx levels (Trx1 in cytosol; Trx2 in mitochondria) and support essential housekeeping and antioxidant defenses [[4], [5], [6]]. Our findings further underscore the importance of the Trx-TXNRD axis, showing that TXNRDs’ expression is elevated in triple-negative breast cancer (TNBC) and correlates with poor patient outcomes. TNBC is a heterogeneous, highly proliferative breast cancer subtype characterized by a high risk of recurrence and metastasis, the absence of targeted therapies, and the poorest prognosis among breast cancer types. Given the urgent need for new treatment strategies, targeting the Trx-TXNRD pathway offers a promising therapeutic approach for TNBC.
TXNRD enzymes reduce the oxidized disulfide-form to regenerate the catalytic dithiol of Trx, their main endogenous substrate. This reduced Trx, in turn, supports the activity of multiple downstream targets, including peroxiredoxins and other protein substrates. TXNRDs can also reduce a variety of other thioredoxin-fold proteins, such as protein disulfide-isomerase (PDI), as well as small-molecule substrates like menadione and lipoic acid (reviewed in Ref. [7]). Reduced Trx interacts with different downstream factors that play central roles in apoptosis, survival and gene transcription, all processes especially relevant to cancer. These include nuclear factor κB (NFκB), p53, apoptosis signal-regulating kinase (ASK1) and nuclear factor erythroid 2-related factor 2 (NRF2), among others [6]. Critically, Trx also supports the function of ribonucleotide reductase (RNR), the enzyme responsible for the rate-limiting step in the conversion of ribonucleotides (NTPs) into deoxyribonucleotides (dNTPs), which are required for DNA replication and repair [8]. Trx was originally discovered by Peter Reichard and colleagues in 1963 in the context of RNR activity in E.coli, where it serves as its reducing agent [9]. Most of our understanding of RNR regulation by the Trx system comes from studies in E.coli and yeast, with relatively limited insights from mammalian systems. Of note, in addition to Trx, glutaredoxin (Grx) is also reported to support RNR [8]. RNR is a heterotetramer composed of two homodimeric subunits: RRM1, which contains the catalytic site, and RRM2, which harbors a diferric iron center essential for generating the tyrosyl radical required for RRM1 activity [8]. Notably, a recent study showed that depletion of Trx1 or TXNRD1 in lung cancer cells that harbor a defective GSH-Grx system was synthetically lethal when combined with CHK1 inhibition, owing to disrupted RNR function involving both RRM1 and RRM2 [10].
Despite its therapeutic potential, clinical progress in targeting TXNRDs has been hindered by the lack of selective and safe inhibitors. Traditional TXNRD inhibitors are covalent and/or irreversible, often interacting with off-target proteins and macromolecules, resulting in significant toxicity (reviewed in Ref. [11]). The use of these inhibitors also complicates the interpretation of biological outcomes. We recently discovered a new class of non-covalent TXNRD inhibitors (TXNRD(i)s) that bind to a newly identified allosteric regulatory site on the enzyme, termed the “doorstop pocket” [11,12]. These novel inhibitors show potent anti-cancer activity in TNBC cells by reducing cell viability, increasing cell death, impairing invasion, and suppressing both clonogenic and mammosphere formation, while sparing normal breast epithelial cells [13]. Importantly, these compounds demonstrated specific inhibition of TXNRD activity both in vitro and in vivo, providing strong evidence for a defined on-target molecular mechanism of action. In preclinical TNBC models, two of these inhibitors, 8VP101 and 9VP19, significantly suppressed tumor growth, offering compelling proof-of-concept and validating the therapeutic promise of this pharmacologic strategy [13]. The pleiotropic anti-cancer effects observed with these non-covalent TXNRD(i)s underscore the critical role of TXNRD enzymes in driving oncogenic processes. These findings prompted us to further investigate the cellular mechanisms underlying their therapeutic activity.

2
Results
2.1
The transcriptome regulated by non-covalent TXNRD(i)s’ in TNBC largely overlaps with silencing of TXNRD1
The first-in-class non-covalent TXNRD(i)s, 8VP101 and 9VP19, were designed to bind to an allosteric regulatory site within the enzyme disrupting the electron flow from NADPH to reduce the disulfide of Trx and other substrates [11,12]. As TXNRDs are selenoenzymes, the addition of exogenous selenium is expected to mitigate inhibition, potentially by increasing TXNRD abundance or activity [14]. Consistent with this notion, 8VP101, previously characterized in TNBC models, demonstrated diminished inhibitory effects in the presence of selenium (Supp. Fig. 1A). To investigate the global transcriptomic changes induced by TXNRD inhibition, RNA sequencing was performed on MDA-MB-231 TNBC cells treated with 10 μM of either 8VP101 or 9VP19 for 24 h, with vehicle-treated cells as controls. TXNRD1 is regarded as the primary therapeutic target and non-covalent TXNRD(i)s effectively suppress TXNRD1 activity in TNBC cells [13]. Concurrently, RNA sequencing was conducted in cells transfected with either siTXNRD1 or non-silencing control siNeg for 48 h (Supp. Fig. 1B). Each experimental condition was assessed in triplicate biological replicates (Supp. Fig. 1C–D), and hierarchical clustering of the resulting transcriptomes is presented in Supp. Fig. 1E. A Venn diagram (Fig. 1A) illustrates that the majority of transcripts altered by TXNRD1 silencing or pharmacologic inhibition via TXNRD(i)s are shared. However, the distinct pools of significantly altered transcripts (log2FC ≥ 0, p-value ≤0.05) depicted in the table shown in Fig. 1B reveal that the transcriptomic impact of TXNRD(i)s is approximately threefold greater than that of siTXNRD1. Comprehensive gene enrichment and pathway analyses were then applied to the significantly regulated transcriptomes.
Reactome gene enrichment analysis reveals that cell cycle control and DNA replication processes are prominently represented in both siTXNRD1 and TXNRD(i)-treated cells. These include M Phase, S Phase, G1/S Transition, Cell Cycle Checkpoints, and DNA Replication signatures (Fig. 1C, Supp. Fig. 2A). STRING analysis further underscores the prominence of cell cycle processes, indicated by green bubbles in Fig. 1D and Supp 2B, which are regulated by siTXNRD1, 8VP101 and 9VP19. Additionally, Cell Cycle processes are linked to Protein Folding (red circles) and Regulation of Cell Death (purple circles), both similarly influenced by siTXNRD1, 8VP101 (Fig. 1D) and 9VP19 (Supp. Fig. 2B). The enrichment of Cell Death processes by 8VP101 and 9VP19 aligns with our previous findings [13]. Likewise, Metascape analysis highlights the Mitotic Cell Cycle, DNA Metabolic Process and Regulation of Cell Cycle as the top downregulated processes (Supp. Fig. 3A), whereas Response to Oxidative Stress and Cellular Catabolic Process as the top upregulated processes (Supp. Fig. 3B). No distinct pathway is identified within the smaller, unique sets of transcripts regulated by either TXNRD(i)s or siTXNRD1 alone (Supp. Fig. 4). The top transcription factors implicated in mediating the transcriptional changes with siTXNRD1 or TXNRD(i)s include SP1, E2F1, ATF4, MYC, RELA and TP53 (Fig. 1E and Supp. Fig. 2C), aligning with the fact that the Trx-TXNRD axis supports transcription factor such as NFκB and p53. Of note, ATF4, a transcription factor of the Unfolded Protein Response (UPR), is consistent with Protein Folding process noted earlier and further substantiated by significant upregulation of UPR genes such as, PERK, CHOP and BIP (Supp. Fig. 5B). E2F1, specifically linked to cell cycle regulation, is investigated again in the subsequent sections. Given the similarities in the transcriptome regulated by either of TXNRD(i)s, 8VP101 or 9VP19, an overlapping signature is used and referred to as TXNRD(i) signature in the rest of the text.

2.2
TXNRD(i) treatment increases redox stress, but it is not critical to cytotoxicity
Oxidative Stress Response emerged as a prominent pathway in siTXNRD1-transfected cells and was similarly enriched in TNBC cells treated with TXNRD(i)s (Fig. 2A) consistent with inhibition of the antioxidant thioredoxin system, which mitigates ROS and peroxides via electron donation to peroxiredoxins. To assess intracellular redox stress levels, we utilized the fluorescent, cell-permeable probe H2DCFDA and analyzed cells via flow cytometry after treatment with 10 μM 8VP101 for 2, 6, 18, and 24 h. Following treatment, cells were incubated with 1 μM H2DCFDA for 15 min before FACS analysis, with the gating strategy outlined in Supp. Fig. 6A. This probe detects various ROS, including hydrogen peroxide, hydroxyl radicals, and peroxyl radicals [15]. We observed a time-dependent increase in intracellular redox stress in TNBC cell lines MDA-MB-231 and HCC1806, as well as the non-tumorigenic breast epithelial line MCF-10A upon 8VP101 treatment, with ROS levels plateauing at approximately a two-fold increase across all lines (Fig. 2B). We confirmed a similar redox stress level using a second probe, dihydroethidium (DHE, Supp. Fig. 6B). However, despite similar redox stress induction, cell viability varied significantly, with TNBC cells displaying greater sensitivity to inhibition by 8VP101 compared to MCF-10A (Fig. 2C).
To assess the role of ROS induction in the anti-cancer effects of TXNRD(i)s, we employed two ROS scavengers: N-acetyl cysteine (NAC) and α-tocopherol (α-T, vitamin E). NAC replenishes glutathione (GSH), the most abundant cellular ROS detoxifier [16], and effectively scavenges hydrogen peroxide-induced redox stress in TNBC cells (Supp. Fig. 6C). α-T, a naturally occurring lipophilic antioxidant, quenches ROS and interrupts peroxidation chain reactions by neutralizing peroxyl radicals [17]. Treatment with either NAC or α-T effectively restored intracellular redox homeostasis to baseline levels in TXNRD(i)-treated cells, with α-T showing greater efficacy (Fig. 3A and C). However, despite mitigating ROS, neither NAC nor α-T rescued cell viability in MDA-MB-231 or HCC1806 cells treated with low concentrations of 8VP101 (Fig. 3B and D) or a high concentration of 8VP101 (Supp. Fig. 6D). These findings indicate that while TXNRD(i) treatment elevates redox stress, this alone is not the primary driver of the anti-cancer effects, as ROS scavenging does not reverse the cytotoxicity. Lastly, we assessed basal redox stress levels in these cells. The mean fluorescence intensity (MFI) of DCF was normalized to an unstained control to account for differences in autofluorescence across cell lines. MDA-MB-231 cells exhibited higher basal redox stress compared to MCF-10A, consistent with the notion that cancer cells generally experience elevated oxidative stress (Supp. Fig. 6E). In contrast, HCC1806 cells displayed significantly lower basal redox stress, contrary to expectations. These data suggest that redox buffering capacity alone does not determine sensitivity to TXNRD(i)s.

2.3
TXNRD(i)s block proliferation, induce cell cycle arrest and trigger cell death by depleting dNTP pools and impairing RNR function
Given the prominent DNA Replication (Fig. 4A) and Cell Cycle signatures enriched in the transcriptomics data following TXNRD(i) treatments, we further investigated these pathways. We observe a marked reduction in DNA synthesis as evidenced by decreased incorporation of 5-ethynyl-2′-deoxyuridine (EdU), a thymidine nucleoside analog, into newly synthesized DNA in cells treated with increasing concentrations of 8VP101 (Fig. 4B) with incorporation levels dropping from 57 % to 24 %. A broad reduction in TNBC cell proliferation following 8VP101 treatment was further supported by gene expression data (Fig. 4C); both Ki67, a marker of proliferation, and E2F1, a transcription factor promoting cell cycle progression, were downregulated, while p21, a cell cycle inhibitor, was significantly upregulated consistent with changes observed at the protein level (Fig. 4E). Similarly, cell cycle analyses in MDA-MB-231 and HCC1806 cells (gating strategy in Supp. Fig. 6F), reveal a significant reduction in S phase and an accumulation in G1 phase (Fig. 4D). This effect is more profound at the IC90 concentration used here to achieve near-complete inhibition of TXNRD, although it should be noted that significant cell death is also observed at this concentration. This profile contrasts with the cell cycle effects induced by AF (Supp. Fig. 6G). While the immortalized MCF-10A cells exhibited a similar trend, the magnitude of cell cycle disruption was notably less pronounced (Fig. 4D, bottom panels).
The Trx/TXNRD system is known to regulate DNA synthesis through redox control of ribonucleotide reductase (RNR), as demonstrated in studies using E. coli and yeast; however, evidence in mammalian systems remains limited. RNR catalyzes the reduction of ribonucleotides to deoxyribonucleotides (dNTPs), which are essential precursors for DNA synthesis and repair [8]. We observed a significant downregulation of Purine and Pyrimidine Metabolism signatures following TXNRD(i) treatment (Fig. 5A). Notably, the expression of RNR subunits RRM1 and RRM2 is significantly elevated in breast tumors compared to normal tissue, with a pronounced increase observed in TNBC tumors (Supp. Fig. 7A–B). Supplementation with exogenous dNTPs, the products of RNR activity, rescued both cell viability (Fig. 5B) and the proliferation arrest induced by TXNRD(i)s in the two TNBC cell lines by restoring progression through the G1 and S phases of the cell cycle (Fig. 5C, third and fourth bars). Given the role of dNTPs in DNA repair, we assessed DNA integrity using γH2AX as a marker of DNA damage [18]. TXNRD(i) treatment significantly increased γH2AX levels, which were partially reduced by dNTP supplementation (Fig. 5D). As DNA damage can lead to cell death, we examined this outcome and found that TXNRD(i)s increased both apoptosis and overall cell death (Fig. 5E, first and third flow panel), consistent with our prior findings [13]. Importantly, exogenous dNTPs significantly mitigated cell death (Fig. 5E), indicating that impaired RNR function and subsequent dNTP depletion contributes to the cytotoxic effects of TXNRD(i)s.
Next, we evaluated the effect of 8VP101 in combination with standard-of-care therapies commonly used to treat TNBC, including the cytotoxic chemotherapies paclitaxel, doxorubicin, and cisplatin, as well as the poly ADP-ribose polymerase (PARP) inhibitor olaparib. Synergy scores were calculated using SynergyFinder, with scores ≥10 indicating synergy and scores between 1 and 10 considered additive [19]. While the three cytotoxic agents showed additive effects with synergy scores just below 10, olaparib demonstrated a synergy score >10, indicating a synergistic interaction with TXNRD(i) in suppressing TNBC cell growth with (Supp. Fig. 7C).
To assess whether TXNRD(i)s inhibit tumor growth in vivo by blocking proliferation, xenograft tumors were established in immunocompromised mice through the injection of human TNBC cells, MDA-MB-231 and HCC1806. Once tumors were established, mice were randomized to receive either 8VP101, 50 mg/kg, daily IP injection, or vehicle control, following the treatment scheme outlined in Fig. 6A. After three consecutive days of treatment, tumors were excised and analyzed for proliferation markers. We previously reported that 8VP101 significantly suppressed MDA-MB-231 tumor growth [13]. Similarly, we observed comparable suppression, and in some cases regression, of HCC1806 xenograft tumor growth (Fig. 6B, left panels), with minimal impact on mouse body weight (Fig. 6B, right panels). Additionally, despite the brief treatment duration, proliferation markers were reduced (Fig. 6C). Collectively, these findings support halted proliferation via RNR dysfunction as the primary mechanism of action for non-covalent TXNRD(i)s in TNBC.

2.4
TXNRD(i)s engage and inhibit both cytosolic and mitochondrial TXNRD enzymes for effective and broad anti-cancer effects
Given the pronounced anti-cancer effects of TXNRD(i)s and the three-fold larger transcriptomic response regulated by TXNRD(i)s compared to siTXNRD1 (Fig. 1B), we investigated whether TXNRD(i)s may have additional targets. The doorstop pocket of TXNRD1 is predicted to be conserved in other members of the thioredoxin reductase family, including the ubiquitously expressed mitochondrial TXNRD2 [11]. We employed the highly selective off-on fluorescent probe TRFS-green to measure TXNRD activity in live TNBC cells, as previously described [13,20]. To distinguish between cytosolic and mitochondrial TXNRD pools, we performed simultaneous labeling with MitoTracker-red, which marks active mitochondria [21]. Co-localization of TRFS-green and MitoTracker-red signals appearing as orange enables quantification of mitochondrial TXNRD activity attributed to TXNRD2, whereas TRFS-green signal outside mitochondria reflects cytosolic TXNRD1 activity (representative images shown in Fig. 7A, and general scheme in Supp. Fig. 8A). To validate this method, TNBC cells were transfected with siNeg, siTXNRD1, or siTXNRD2, and silencing efficiency was confirmed by RT-QPCR (Supp. Fig. 8E). Silencing either TXNRD1 or TXNRD2 partially reduced the total TRFS-green signal (Supp. Fig. 8B, top left panel). Total MitoTracker-red intensity was slightly increased in siTXNRD1 cells but otherwise unchanged (Supp. Fig. 8B, top right panel). However, when the TRFS-green signal was stratified by cytosolic or mitochondrial activity corresponding to TXNRD1 or TXNRD2, respectively, we observed a complete reduction of the TRFS-green signal to baseline levels upon silencing of each enzyme in its respective compartment, with minimal impact on the complementary pools (Supp. Fig. 8B, bottom panels). Using this dual-labeling method, we confirmed that AF inhibits both TXNRD1 and TXNRD2 (Supp. Fig. 8C), consistent with its classification as a pan-TXNRD inhibitor [22]. Applying the same approach to 8VP101 and 9VP19, we found that these TXNRD(i)s also effectively inhibit both cytosolic and mitochondrial pools of TXNRDs (Fig. 7A and Supp. Fig. 8D), confirming that they function as pan-TXNRD inhibitors.
This finding prompted us to investigate the individual and combined roles of TXNRD1 and TXNRD2 in TNBC cells. We used small interfering RNA (siRNA)-mediated silencing and direct cell counting with the BioTek BioSpa system to assess cell growth in two TNBC cell lines, MDA-MB-231 and HCC1806. Silencing TXNRD1 alone (Fig. 7B, blue lines) resulted in significant growth suppression, although the effect diminished at later time points in HCC1806 cells, consistent with previous reports that identify TXNRD1 as the main cancer-relevant target [6], particularly in TNBC [13]. In contrast, silencing TXNRD2 (Fig. 7B, red lines) led to a significant and sustained reduction in cell growth over time in both cell lines, indicating that TXNRD2 is also required for TNBC proliferation and represents a previously underappreciated therapeutic target. Additionally, combined silencing of TXNRD1 and TXNRD2 (Fig. 7B, purple lines) produced a more pronounced effect, resulting in regression of the growth curves suggestive of cytotoxic, rather than merely cytostatic, effects. Notably, MCF-10A non-tumorigenic breast epithelial cells were unaffected by either individual or combined silencing (Fig. 7B, last panel), suggesting a potential therapeutic window. The anti-proliferative effects of TXNRD(i)s in TNBC cells closely mimicked the effects of dual TXNRD1/2 silencing (Fig. 7C). Lastly, the cell cycle profiles following individual or dual TXNRD silencing closely mirrored those observed with TXNRD(i) treatment (Fig. 7D). Together, these results support the conclusion that the enhanced anti-cancer activity of TXNRD(i)s arises from pan-TXNRD inhibition, which emerges as a more effective therapeutic strategy for TNBC. Other solid tumors show elevated expression of TXNRD1, TXNRD2 or both, including lung cancer (Supp. Fig. 9A). We evaluated and found that several lung cancer cell models express both TXNRD1 and TXNRD2 (Supp. Fig. 9B). Furthermore, using two TXNRD(i)s, 8VP101 and 9VP19, the viability of lung cancer cells is inhibited in a concentration dependent manner with IC50 values ranging from 6 to 10 μM (Supp. Fig. 9C). In summary, non-covalent pan-TXNRD inhibitors that bind to the doorstop pocket of these enzymes exhibit significant and broad anti-cancer activity, driven by RNR dysfunction rather than the pro-oxidant mechanism typically associated with traditional inhibitors.

3
Discussion
Targeting TXNRD1, and to a lesser extent TXNRD2, for cancer therapy is well supported by multiple preclinical studies [6]. The highly reactive selenocysteine residue in the active site of TXNRD1 and TXNRD2 renders these enzymes susceptible to electrophilic attack from compounds such as AF and TRi-1. Covalent modification of TXNRD1 by these agents has been shown to convert TXNRD1 into a selenium compromised thioredoxin reductase-derived apoptotic proteins (secTRAPs) by conferring a gain-of-function NADPH oxidase pro-oxidant activity that promotes cell death [23]. Notably, this pro-oxidant cytotoxicity is often more pronounced with pharmacological inhibition than with siRNA-mediated knockdown of TXNRD1 [24,25]. Unlike traditional covalent TXNRD inhibitors, our first-in-class non-covalent TXNRD(i)s target a distinct allosteric site known as the doorstop pocket [11,12]. Although treatment with TXNRD(i)s increases intracellular ROS levels, our findings indicate that ROS are not the primary mediators of cytotoxicity in TNBC cells, as antioxidant fails to reverse the observed cell death. In contrast, AF-induced cell death in TNBC is ROS dependent [26]. Similar ROS dependent mechanisms of AF have been reported in other cancers [22] including high grade serious ovarian cancer [27], gastric cancer [28], mesothelioma [29], lung cancer [30,31], acute lymphoblastic leukemia [32], and B-cell lymphomas [33]. However, it is important to note that the mode of AF-induced cell death varies significantly across cancer types and is influenced by factors such as drug concentration, cancer cell type, and genetic background [34].
Treatment of TNBC cells with TXNRD(i)s activated programs involving several key transcription factors, including NFκB, p21, and p53. This is consistent with the established role of the Trx system in modulating transcription factor activity, as Trx influences the DNA-binding capacity and regulatory functions of NFκB, p21, and p53 [6]. Our data further suggest that TXNRD(i)s induce endoplasmic reticulum (ER) stress and activate the UPR, specifically through the PERK–ATF4–CHOP axis. This aligns with previous findings that cytosolic TXNRD1 is required for reducing non-native disulfide bonds in proteins entering the ER [35]. Under prolonged stress conditions, activation of this pathway promotes apoptotic signaling and upregulation of autophagy-related genes [36]. The induction of autophagy likely serves as a compensatory mechanism to degrade misfolded proteins and alleviate proteotoxic stress, which may explain the enrichment of catabolic processes identified by Metascape analysis. However, the precise contribution of UPR activation to TXNRD(i)-mediated cytotoxicity remains unclear and requires further investigation. These findings also suggest that, in addition to apoptosis, autophagy may represent an alternative mechanism of cell death induced by these inhibitors. Notably, AF has been shown to induce paraptosis in breast cancer cells through dual inhibition of TXNRD1 and the proteasome [34]. Paraptosis is a caspase-independent form of programmed cell death characterized by cytoplasmic vacuolation and swelling of the ER and mitochondria. It is typically triggered by sustained ER stress, particularly when the accumulation of misfolded proteins overwhelms the ER's folding capacity [37]. Given that ER stress emerged as a dominant signature in our transcriptomic analysis, it will be important to determine whether TXNRD(i)s also elicit paraptosis.
Contrary to the anticipated pro-oxidant mechanism involving ROS-mediated cytotoxicity, the anticancer activity of these TXNRD(i)s appears to be primarily driven by disruption of RNR function, as evidenced by a pronounced replication arrest phenotype. We found that exogenous supplementation with dNTPs, the direct substrates of RNR, fully rescued the cell cycle arrest and partially mitigated cell death. The induction of cell death following TXNRD(i) treatment is likely mediated, at least in part, by DNA damage, as indicated by the accumulation of the DNA damage marker γH2AX, which was partially reduced by dNTP supplementation. These findings suggest that impaired RNR activity leads to replication stress and genotoxicity, ultimately activating apoptotic cell death pathways. Supporting the role of RNR dysfunction, we observed additive effects when TXNRD(i)s were combined with standard-of-care DNA-damaging agents used in TNBC treatment. This suggests that co-treatment with TXNRD(i)s may enhance therapeutic efficacy and potentially allow for reduced chemotherapy dosing, thereby improving patient outcomes and quality of life. Notably, TXNRD(i)s synergized with PARP inhibitors, possibly by exacerbating DNA repair deficiencies. These findings further support the rationale for a combination strategy that concurrently targets both DNA damage and repair pathways in TNBC.
RNR requires dithiol electron donors to maintain its catalytic activity. In E. coli. RNR both Trx and Grx serve this function [8]. Grx is a small redox enzyme that facilitates thiol-disulfide exchange reactions in the presence of GSH, glutathione reductase (GSR) and NADPH [38]. In human lung cancer cells, high doses of AF induced oxidation of RRM1 leading to dNTP depletion and impaired replication fork elongation [39]. This phenotype was not rescued by co-treatment with the NAC, but was significantly mitigated by supplementation with exogenous nucleotides. Similarly, genetic depletion of TXNRD1 results in RNR oxidation and impaired function in lung cancer cells [39]. In mouse models, conditional deletion of TXNRD1 in T cells leads to defective RNR catalysis and impaired T cell expansion [40]. These findings highlight the critical role of the Trx-TXNRD axis in maintaining mammalian RNR function. Notably, in lung cancers, frequent deletion of the chromosome 8p12 locus, which contains the GSR gene [41], suggests that loss of the GSH-Grx axis may force these cancers to rely more heavily on the Trx–TXNRD pathway. Our data now extend these observations to TNBC, providing the first evidence that the Trx-TXNRD axis is similarly essential for sustaining RNR activity and cellular proliferation in TNBC cells. The mechanistic basis for this reliance in TNBC warrants further investigation. The GSH-GSR axis has been studied in breast cancer, but findings remain inconsistent. Some studies reported elevated GSH levels and increased expression of related enzymes, including GSR, in breast tumors compared to normal tissue, albeit with substantial interindividual variation [[42], [43], [44]]. Conversely, another study comparing GSR levels in breast tumors and matched adjacent normal tissues found that GSR was downregulated in as many cases as it was elevated [45]. Analysis of TCGA data estimated the prevalence of GSR alterations or deletions in breast cancer to be approximately 6 % [46]. The status of Grx in breast cancer, particularly in the TNBC subtype, is even less well defined. Why the GSH-Grx system cannot substitute for Trx in sustaining RNR activity in TNBC also remains unclear. Notably, the inability of NAC supplementation, which primarily supports GSH, to rescue the phenotype suggests non-redundant functions between Trx and GSH-Grx, and/or a dysfunctional GSH-Grx axis in these cells. None of the TXNRD(i)s used in this study inhibit recombinant GSR [13]. Interestingly, the TNBC cell line MDA-MB-468, which shows the highest GSR levels (see Fig. 2C in Ref. [13]), was among the least sensitive to TXNRD(i)s. This observation may indicate that a small subset of TNBCs with elevated GSR-GSH-Grx activity can partially compensate for Trx in supporting RNR, rendering them less responsive to TXNRD(i)s. However, these possibilities should be interpreted with caution and warrant further investigation.
To address how TXNRD(i)s affect RNR activity, one possibility is that they oxidize RRM1, similar to what has been observed in lung cancer models treated with AF or upon TXNRD1 depletion. [38]. Alternatively, transcriptional regulation of RNR subunits may play a role. Transcriptomic profiling of TNBC cells treated with TXNRD(i)s revealed significant downregulation of multiple E2F family transcription factors. Given that the promoters of both RRM1 and RRM2 harbor E2F-binding sites [47], suppression of E2F activity may contribute to reduced RNR gene expression. Further studies are needed to elucidate how these transcriptional and post-translational mechanisms converge to regulate RNR function downstream of the Trx-TXNRD axis in TNBC.
Although TXNRD(i)s exhibit broad activity against TNBC, whether specific TNBC subtypes display heightened sensitivity remains to be determined. Transcriptomic analysis revealed modulation of androgen receptor (AR) signaling, suggesting that the luminal androgen receptor (LAR) subtype may derive added benefit from TXNRD(i) treatment. The LAR subtype is characterized by elevated AR expression and has been associated with a poor pathological complete response compared to other TNBC subtypes in patients receiving neoadjuvant chemotherapy [48]. Whether TXNRD(i)s exert enhanced anticancer effects in LAR TNBC warrants further investigation. Given the essential role of RNR in DNA synthesis and repair, it has long been considered a compelling target for cancer therapy [47]. Current clinically approved RNR inhibitors include nucleoside analogs such as gemcitabine and clofarabine, which target the RRM1 subunit, and the radical scavenger hydroxyurea, which inhibits RRM2. However, these agents are limited by significant drawbacks: nucleoside analogs often exhibit dose-limiting toxicity and a narrow therapeutic window, while radical scavengers lack both target specificity and sustained efficacy [47]. Our novel TXNRD(i)s may provide an alternative strategy to disrupt RNR activity, particularly in tumors such as TNBC that exhibit elevated RNR subunit expression and depend on the Trx-TXNRD system.
TXNRD1 is commonly presumed to be the primary therapeutic target in cancer, including TNBC. However, our data suggest that TXNRD2 may be equally relevant, as supported by both genetic and pharmacological approaches. We found that TXNRD(i)s effectively inhibit both cytosolic (TXNRD1) and mitochondrial (TXNRD2) pools of TXNRD, consistent with the prediction that the allosteric doorstop pocket is also present in TXNRD2. Notably, siRNA-mediated knockdown of TXNRD2 induced a robust G1 cell cycle arrest, closely resembling the phenotype observed following TXNRD(i) treatment. This raises the intriguing possibility that TXNRD2 plays a direct role in regulating cell cycle progression. Further studies are warranted to elucidate the precise function of TXNRD2 and to evaluate its potential as a therapeutic target in TNBC, as it may represent an underappreciated vulnerability. In addition to their anticancer activity, our TXNRD(i)s serve as valuable pharmacological tools for advancing our understanding of TXNRD biology. Both TXNRD1 and TXNRD2 are embryonically essential, and shRNA-based depletion of TXNRDs in cancer cells often triggers compensatory upregulation of alternative antioxidant pathways [48]. Our approach offers a unique pharmacological opportunity to dissect TXNRD function and uncover new tumor biology related to TXNRD inhibition in TNBC.
In conclusion, this study demonstrates that our novel first-in-class non-covalent TXNRD(i)s exhibit potent anticancer activity in TNBC cells through mechanisms distinct from those of traditional TXNRD inhibitors. Our TXNRD(i)s do not rely on pro-oxidant or ROS-dependent mechanisms to exert their effects. Instead, our data support a model in which their primary mechanism of action involves disruption of RNR function, leading to reduced proliferation and induction of DNA damage-mediated cell death. This mechanistic insight has important implications: (i) it can inform the design of next-generation inhibitors with improved anticancer efficacy and safety profiles, (ii) it offers the potential to develop therapeutic biomarkers for identifying TNBC patients most likely to benefit from TXNRD(i) treatment, and (iii) provides a rationale for investigating drug combination strategies against TNBC.

4
Materials and methods
4.1
Reagents
The TXNRD inhibitors, 8VP101 and 9VP19, were synthesized as previously reported [12]. The purity of inhibitors was determined to be >95 % using HPLC analyses performed on a Shimadzu HPLC system with a NUCLEODUR 100-5 CN-RP column (pore size: 110 Å; particle size: 5 μm; dimensions: 4.6 × 100 mm). Auranofin (AF, A6733) was purchased from Millipore Sigma. Propidium iodide (PI, P1304MP) and PureLinkTM RNase A (12091-021) were purchased from Invitrogen. 2,7-dichloroluorescein diacetate (H2DCFDA, D6883) and N-acetyl cysteine (NAC, A7250) were purchased from Sigma Aldrich. 7-Aminoactinomycin D (7-AAD, A1310) and dNTPs (10297018) were purchased from ThermoFisher Scientific. Dihydroethidium (DHE; cat# HY-D0079) and α-tocopherol (HY-16686) were purchased from MedChemExpress. Silencer® Select Negative Control #2 siRNA (4390846), TXNRD1 Silencer Select Validated siRNA (4427038) and TXNRD2 Silencer Select Validated siRNA (s20782) were purchased from ThermoFisher Scientific. Lipofectamine RNAiMAX Reagent (13778-075) and Opti-MEM (11058-021) were purchased from Fisher Scientific. TRFS-green (HY-115640) was purchased from MedChemExpress. MitoTracker-Red CMXROS (9082P) was purchased from Cell Signaling.

4.2
Cell lines, culture conditions and treatments
MDA-MB-231 cells were obtained from Dr. Clodia Osipo (Loyola University Chicago) and were maintained in RPMI 1640 medium (Gibco, 11875-093) supplemented with 5 % fetal bovine serum (FBS), 2 mmol/L l-glutamine, 1 % antibiotics penicillin-streptomycin, and 1 % non-essential amino acids. HCC1806 cells were obtained from ATCC and maintained in RPMI 1640 medium with HEPES and sodium pyruvate (ATCC, 30–2001) supplemented with 10 % FBS, 2 mmol/L l-glutamine, 1 % antibiotics penicillin-streptomycin, and 1 % non-essential amino acids. MCF-10A cells were obtained from ATCC and maintained in DMEM-F12 medium (Gibco, 11330032) supplemented with 10 % fetal bovine serum, 20 ng/mL epithelial growth factor, 0.5 mg/mL hydrocortisone, 100 ng/mL cholera toxin, and 10 μg/mL insulin. All cells were maintained at 37 °C in a humidified atmosphere with 5 % CO2. H1437, H1299, and A549 cells were obtained from Dr. Maurizio Bocchetta (Loyola University Chicago) and were maintained in RPMI 1640 medium supplemented with 10 % FBS. All cell lines were tested for mycoplasma using the LookOut PCR detection kit (Sigma) and validated by STR through ATCC. Inhibitors were dissolved in dimethyl sulfoxide (DMSO) and the final culture volume of inhibitors or vehicle was less than or equal to 0.1 %. siRNA transfections were performed according to the manufacturer's instructions.

4.3
RNA sequencing and bioinformatic analysis
Total RNA was extracted using TRIzol according to the manufacturer's instructions. RNA quantity and quality were assessed using a NanoDrop spectrophotometer. RNA-seq library preparation, sequencing, and initial quality control were performed by Novogene. Libraries were sequenced using the Illumina NovaSeq platform employing a paired-end 150bp sequencing strategy. Downstream analysis was conducted using the DESeq2 R package (1.20.0) for differential expression analysis. Gene set enrichment analysis (GSEA) was conducted using the GSEA desktop application (version 4.3.3, Broad Institute). The full gene expression dataset, quantified as FPKM (Fragments Per Kilobase of transcript per Million mapped reads), was used as input. Analysis was performed using the standard (non-pre-ranked) mode with 1000 phenotype permutations to assess statistical significance. Functional enrichment analysis was performed using Metascape (https://www.metascape.org/) to identify biological pathways and regulatory networks associated with transcriptional changes following TXNRD inhibition. Genes submitted for analysis are specified in figure legend. Default parameters were used: a minimum gene overlap of 3, a p-value cutoff of 0.01, and an enrichment factor threshold of 1.5. Results were visualized using Metascape's clustering and pathway annotation tools.

4.4
Cell viability
Cells were seeded in 24-well plates at densities optimized for assay duration: 20,000 cells/well for MDA-MB-231 and MCF-10A, 40,000 cells/well for HCC1806 in 72-h assays, and 75,000 cells/well for all cell lines in 6-h and 24-h assays. After overnight attachment, cells were treated with the indicated TXNRD(i)s for the designated times. Cells were stained with 1 % crystal violet in methanol and water (1:4), solubilized in 1 % sodium dodecyl sulfate (SDS) and the absorbance was measured at 570 nm using the BioTek Synergy H1 Plate Reader.

4.5
Flow cytometry
Detection of redox stress was measured using H2DCFDA. H2DCFDA was used as previously described in the literature [37]. MDA-MB-231, HCC1806, and MCF-10A cells were seeded in 6-well plates at a density of 150,000 cells/well. Cells were harvested by trypsinization and loaded with 1 μM H2DCFDA in serum free media for 15 min at 37 °C. Cells were then washed twice with ice-cold PBS, resuspended in ice-cold PBS with 3 μg/mL of 7-AAD and analyzed immediately using NovoCyte Quanteon Flow Cytometer by Agilent Technologies. Mean fluorescence intensity (MFI) was used as a measure of ROS and data were analyzed with Flowjo v11. A second redox probe, dihydroethidium (DHE) was also used. Cells were stained with 2.5 μM DHE for 30 min before analysis by flow cytometry as described above using the same gating strategy.
For 5-ethynyl-2′-deoxyuridine (EdU) incorporation, cells were seeded in 6-well plates at densities optimized for assay duration: 100,000 cells/well for MDA-MB-231 and MCF-10A, and 150,000 cells/well for HCC1806. Cells were stained according to the manufacturer's instructions using an EdU Staining Proliferation Kit iFluor 488 (ab219801) from Abcam. Proliferation was immediately analyzed using NovoCyte Quanteon Flow Cytometer by Agilent Technologies. A minimum of 30,000 events were recorded for each sample.
For cell cycle analysis, cells were seeded in 6-well plates at densities optimized for assay duration: 100,000 cells/well for MDA-MB-231 and MCF-10A, and 150,000 cells/well of HCC1806. Cells were collected using trypsin and washed twice with ice-cold PBS. Cells were resuspended in 0.5 mL of ice-cold PBS and 1 mL of 100 % ethanol was added in dropwise while vortexing, followed by incubation for 1 h at 4 °C. Cells were then resuspended in 100 μg/mL RNase A and 50 μg/mL PI for 1 h at room temperature, then immediately analyzed using NovoCyte Quanteon Flow Cytometer by Agilent Technologies. A minimum of 30,000 events were recorded for each sample.
Cell death was measured using AlexaFluorTM 488 Annexin V/Dead Cell Apoptosis Kit (V13241) from ThermoFisher Scientific according to the manufacturer's instructions. Cells were seeded in 6-well plates at densities optimized for assay duration: 100,000 cells/well for MDA-MB-231 and MCF-10A, and 150,000 cells/well of HCC1806. Apoptosis was immediately analyzed using NovoCyte Quanteon Flow Cytometer by Agilent Technologies. A minimum of 30,000 events were recorded for each sample.

4.6
Microscopy in live cells
Cells were seeded at high confluency in a 24-well plate. The next day, cells were pretreated with the inhibitors for 2 h. Cells were then incubated with 10 μM TRFS-green at 37 °C for 4 h followed by 100 nM MitoTracker-red CMXRos (CellSignaling Technology, cat# 9082) for 30 min and imaged on the Nikon Ti2E inverted microscope at 20x using the GFP/FITC filter cube with excitation of 480 nm for TRFS-green and TRITC/CY3 filter cube with excitation of 540 nm for MitoTracker-red. Utilizing the NIS-Element application, regions of interest (ROIs) were drawn around the perimeter of each cell in each combined image's field. TRFS-green and MitoTracker-red intensity were thresholded and used as reference binaries for intersection (TXNRD2 activity pool) and subtraction (TXNRD1 activity pool) binaries. Average intensity per cell per field was calculated from the total SumGreen intensity divided by the total cell per field for TRFS-green, intersection and subtraction binaries.

4.7
Western blot
Whole cell extracts were prepared using the M-PER reagent (ThermoFisher Scientific). Proteins were separated by SDS-PAGE (Invitrogen), transferred to nitrocellulose membranes using an iBlot 2 instrument (Invitrogen), blocked for 1 h in TBS/T buffer containing 5 % non-fat dry milk and incubated with the indicated primary antibody (diluted in 2.5 % milk in TBS/T) overnight. The next day, membranes were incubated with secondary antibody for 1 h and the signal was visualized on the iBright CL1000 Imaging System (Invitrogen) using the Pierce Supersignal West Pico chemiluminescent substrate (ThermoFisher Scientific). The antibody for β-actin (A5441) was purchased from Sigma Aldrich. The antibodies for γ-H2AX (2577S), TXNRD1 (15140S), and TXNRD2 (12029S) were purchased from Cell Signaling. The antibody for p21 (10355-1-AP) was purchased from ProteinTech.

4.8
RT-Quantitative PCR (QPCR)
Total RNA was isolated using TRIzol according to the manufacturer's instructions. RNA (0.5 μg) was reverse transcribed in a total volume of 10 μL using 200U of M-MLV reverse transcriptase, 100 ng random hexamer, 0.5 mM deoxy-NTP and 10 mM DTT. The resulting cDNA was mixed with SYBR Green Master mix, forward and reverse primers and the amplifications were performed using a QuantStudio3 instrument (ThermoFisher Scientific) according to manufacturer's instructions. Fold change was calculated using the ΔΔCt method with β-actin serving as the internal control. All QPCR primers used were validated and the sequences are included in Supplementary Table 1.

4.9
In vivo studies
Mouse experiments were performed at the Loyola University Chicago animal facility and conducted in accordance with institutional procedures and guidelines after prior approval from the Institutional Animal Care and Use Committee (IACUC). Female athymic nude mice (nu/nu), aged 5-weeks-old, were purchased from Envigo and allowed to acclimate for one week. One or two million MDA-MB-231 or HCC1806 cells, respectively in 100 μL PBS were bilaterally injected orthotopically into the thoracic mammary glands. Tumor formation was monitored by palpation and once tumors were detected (∼50 mm3 in size), mice were randomized into either vehicle control (10 % DMSO, 10 % tween-80, and 80 % PBS for 100 μL volume) or treatment groups of 8VP101, 50 mg/kg. Mice received daily treatments via intraperitoneal (IP) injection. Tumor sizes were measured daily with an electronic caliper. The tumor volume was calculated as length/2 × width2 × π. Tumor were excised and processed for gene expression.

4.10
Statistical analysis
Experiments were done in biological triplicate unless otherwise stated. Data are presented as mean ± sem from at least three independent determinations. Statistical analysis consisted of 1-way ANOVA followed by Tukey posttest, unpaired t-test, and Kruskal-Wallis test followed by Dunns posttest, as appropriate.

CRediT authorship contribution statement
Abigail Rullo: Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft. Brenna Flowers: Data curation, Investigation, Methodology. Keacha Chang: Investigation, Methodology. An Zhang: Investigation, Methodology. Valentina Z. Petukhova: Investigation, Methodology. Luke Harding: Investigation. Sammy Y. Aboagye: Methodology. Maurizio Bocchetta: Data curation. Wei Qiu: Data curation. David L. Williams: Data curation, Writing – review & editing. Francesco Angelucci: Data curation, Writing – review & editing. Pavel A Petukhov: Formal analysis, Resources, Supervision, Writing – review & editing. Irida Kastrati: Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Resources, Software, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.