Cuproptosis, first described as a distinct cell death modality in 2022, differs fundamentally from better-known death pathways such as apoptosis. Rather than dismantling the cell through caspase enzymes or membrane rupture, cuproptosis arises when copper ions bind directly to lipid-acylated components of the mitochondrial tricarboxylic acid cycle. This aberrant binding causes the affected proteins to aggregate, destabilizes iron-sulfur cluster proteins, and triggers a proteotoxic stress response that ultimately kills the cell. Because copper homeostasis is tightly regulated in healthy tissue, the discovery of cuproptosis opened an entirely new therapeutic frontier: if tumors can be pushed over the copper-death threshold, they might be eliminated by a mechanism they have never needed to resist during evolution.

The research team, led by Shuaiyuan Zhang, Xiaoyong Liu, Xu Jiang and corresponding author Zan Jiao, with affiliations spanning Sun Yat-Sen University, Shenzhen Longgang Otolaryngology Hospital and Southern Medical University, began by asking a simple question at the transcriptomic level. Do oral tumors from patients with a history of betel nut exposure look different, in terms of oxidative stress and copper-death biology, from oral tumors that arise without that exposure? The answer was emphatically yes. Betel nut-associated oral squamous cell carcinoma samples displayed enhanced oxidative stress signatures, reduced cuproptosis scores, and markedly elevated expression of a gene called MT2A, metallothionein 2A. Strikingly, high MT2A expression correlated with poorer overall survival and poorer progression-free survival, marking it as a potential prognostic indicator in its own right.

MT2A encodes a small, cysteine-rich metal-binding protein that can sequester heavy metal ions, including copper, and buffer them away from sensitive cellular targets. In the logic of cuproptosis, a protein that mops up free copper is exactly the kind of bodyguard a cancer cell would want. The researchers next turned to controlled laboratory experiments with oral cancer cell lines to test whether arecoline itself could induce this protective state. When cells were treated with arecoline, intracellular levels of reactive oxygen species, or ROS, rose. That surge in oxidative stress activated NRF2, nuclear factor erythroid 2-related factor 2, the master transcriptional regulator of the cellular antioxidant response. Once activated, NRF2 upregulated MT2A expression, and the cells became measurably more resistant to copper toxicity while their susceptibility to cuproptosis dropped.

This chain of events, from ROS to NRF2 to MT2A, constitutes what the authors define as a signaling axis, a causal pathway in which each link is necessary for the next. To prove causality rather than mere correlation, the team intervened at multiple points. When they scavenged ROS with pharmacological antioxidants, or when they genetically silenced NRF2 with interfering RNA, the arecoline-induced rise in MT2A was blunted. Crucially, blocking the axis restored the cells’ vulnerability to cuproptosis and weakened several hallmarks of malignancy: cell viability fell, colonies grew less efficiently in clonogenic assays, and the cells’ capacity to migrate and invade through surrounding matrix, the behaviors that make oral cancer so locally destructive, was attenuated in vitro.

To establish that NRF2 acts directly on the MT2A gene rather than through intermediaries, the researchers employed CUT&Tag profiling, a technique that maps where a transcription factor binds across the genome with base-level precision in living chromatin. The data showed direct occupancy of NRF2 on the MT2A promoter, the regulatory region that controls the gene’s transcription. This finding anchors the axis at the molecular level: arecoline-generated ROS liberate NRF2, NRF2 docks on the MT2A promoter, and MT2A protein accumulates to buffer copper. The team then performed the complementary rescue experiment. When NRF2 was knocked down, forcing MT2A back into the cells partially restored their pro-survival and anti-cuproptotic behavior, demonstrating that MT2A is a functional downstream effector, not merely a passive biomarker of NRF2 activity.

Cancer biology, however, is notoriously unforgiving of results that hold only in a culture dish. The team therefore extended their findings into xenograft models, in which human oral cancer cells are implanted into immunocompromised mice and allowed to form tumors. Knocking down either NRF2 or MT2A significantly inhibited tumor growth in these animals. Conversely, when the researchers enforced MT2A expression in tumors whose NRF2 had been depleted, the growth-suppressive effect was reversed, confirming the hierarchy of the axis in a living organism. In a final and particularly elegant test, the researchers used tetrathiomolybdate, or TTM, a clinical copper chelator originally developed for Wilson’s disease, to lower available copper in tumors lacking MT2A. Chelation rescued the tumor-suppressive effect of MT2A knockdown, meaning that removing the copper buffer mattered only when copper was present to be buffered, a result that further cements MT2A’s role in protecting oral cancer cells from cuproptosis in vivo.

Taken together, the study sketches a coherent and clinically provocative picture. Arecoline exposure floods oral epithelial cells with oxidative stress. The stress response, mediated by NRF2, inadvertently arms those cells with copper-binding metallothioneins. If and when malignant transformation occurs, the same machinery renders the resulting tumors resistant to copper-induced death and more aggressive in behavior. In this framing, the very alkaloid that helps initiate the cancer also appears to educate the tumor in how to survive one of the body’s most recently discovered lethal mechanisms. The paradox is that the ROS signal, usually cast as a tumor-promoting villain in cancer biology, here acts as the upstream trigger for a protective transcriptional program, illustrating how context determines whether oxidative stress harms or heals a cancer cell.

The therapeutic implications are twofold. First, the ROS-NRF2-MT2A axis offers a set of druggable nodes: disrupting NRF2 activation, silencing MT2A, or manipulating copper availability could each, in principle, resensitize arecoline-associated tumors to cuproptosis. Copper chelators and copper ionophores are already being explored in clinical trials for other malignancies, and the present findings suggest that betel nut-related oral cancers, which carry high MT2A expression and low cuproptosis scores, might represent a patient population particularly likely to benefit from copper-directed strategies. Second, MT2A expression could serve as a biomarker to stratify patients, identifying those whose tumors depend on this protective axis and who might therefore respond to combination approaches that pair conventional therapies with cuproptosis-inducing agents.

As with any preclinical study, important caveats remain. The work relies on cell lines, transcriptomic correlations and xenografts rather than human intervention trials, and the precise contribution of the axis relative to other arecoline-driven pathways, such as those involving epithelial-mesenchymal transition and inflammatory signaling, will require further dissection. Nevertheless, by connecting a specific environmental carcinogen to a specific cell death resistance mechanism, the study adds a compelling chapter to the fast-growing literature on copper and cancer. For the hundreds of millions of betel chewers worldwide, and for the clinicians treating the oral cancers that follow, the message is that the enemy’s chemistry may also reveal its weakness, and that the copper-handling machinery of the tumor cell is a vulnerability worth attacking.

Subject of Research: Arecoline-driven suppression of cuproptosis in oral squamous cell carcinoma via the ROS-NRF2-MT2A axis

Article Title: Arecoline suppresses cuproptosis and drives oral squamous cell carcinoma progression via a ROS-NRF2-MT2A axis

Article References: Zhang, S., Liu, X., Jiang, X., & Jiao, Z. (2026). Arecoline suppresses cuproptosis and drives oral squamous cell carcinoma progression via a ROS-NRF2-MT2A axis. Cell Death Discovery. https://doi.org/10.1038/s41420-026-03387-2

Image Credits: AI Generated

DOI: 10.1038/s41420-026-03387-2

Keywords: arecoline, betel nut, oral squamous cell carcinoma, cuproptosis, ROS, NRF2, MT2A, copper, cell death, oxidative stress, cancer progression, tetrathiomolybdate

The platform at the heart of the work is Artaban, a quadruped robot with twelve degrees of freedom, each leg driven by three Maxon motors. It carries four camera modules, two standard RGB and two depth-sensing RGB-D units, plus a time-of-flight camera in its chest, an inertial measurement unit, motor encoders, temperature sensors, and a built-in speaker. A locomotion controller based on Nonlinear Model Predictive Control translates high-level velocity commands into optimized foot trajectories. Onboard computing comes from an Intel NUC 11 with a Core i7 processor and an NVIDIA RTX 2060 GPU running Ubuntu 20.04 and ROS Noetic.

What makes SMaRTAban distinctive is how it stitches together the pieces of a complete voice-command pipeline. Spoken input passes through voice activity detection adapted from the Google Speech Recognition library, then to OpenAI’s hosted Whisper service, which automatically detects the language and transcribes the audio. The transcript is handed to a ReAct agent built on the LangGraph framework, where GPT-4o reasons step by step and selects from seven custom tools: moving the robot, stopping it, sitting down, standing up, interpreting camera images through GPT-4o vision, synthesizing speech, and waiting. Everything is displayed in a React-based graphical interface so the operator can see every tool call, every response, and every error in real time.

Safety receives unusually careful treatment for a language-driven system. Because large language models are probabilistic by nature, the researchers deliberately avoid relying on the model alone to keep the robot safe. Each tool has a typed schema whose arguments are validated before execution, and invalid parameters are rejected and returned to the model as errors without any physical action. More importantly, nearly every tool runs inside a subprocess guarded by a custom “interruptible” decorator. When the operator presses the interrupt button in the GUI, the main process kills that subprocess and independently issues a zero-velocity command to the robot — no LLM consultation required. Shutting down the application triggers the same stop, and a physical emergency-stop button cuts motor power as a final backstop. The researchers are candid that this layered approach still lacks a controller-level watchdog or velocity timeout, so it reduces risk without providing a formal safety guarantee.

The evaluation is unusually thorough for this field. A pilot voice study collected 120 recordings of four phrases — two short and two long, in both English and Slovak — spoken ten times each by three volunteers. English transcription was highly accurate, with word error rates of just 0.04 to 0.08. Slovak proved much harder at the acoustic level: the short phrase “Choď dopredu” (go forward) reached a word error rate of 0.97, occasionally being misidentified as a different Slavic language. Yet here is the striking part — even when the transcription looked mangled, meaning often survived. The short Slovak phrase still scored 0.87 on BERTScore, a semantic similarity measure, and the system correctly interpreted the intended command in 87 percent of recordings. The longer Slovak phrase reached a BERTScore of 0.96 and 90 percent understanding despite a word error rate of 0.25.

The core of the paper is a controlled ablation spanning 4,800 accepted tool-calling trials: two models, GPT-4o and GPT-4o-mini, across six agent and prompting configurations, two languages, ten command scenarios, and twenty repetitions each. The scenarios covered translation and turning movements, posture changes, a vision query, and one demanding multi-step command requiring the robot to sit, wait, stand, move forward, turn right, stop, and then describe what it sees. The results tell a clear story about what actually makes language-driven robots reliable.

Simply giving the model access to tools was not enough. Without a task-specific system prompt, GPT-4o completed the full expected action sequence in only 29.75 percent of trials. Adding an explicit system prompt describing the action contract — move, then wait, then stop — lifted that to 52.50 percent. But the decisive jump came from few-shot prompting: three worked examples demonstrating the movement pattern. With iterative ReAct execution, the system prompt, and those demonstrations combined, GPT-4o achieved a tool-call F1 of 0.9977 and 96.5 percent exact-sequence success. GPT-4o-mini followed the same trend, reaching 82 percent exact-sequence success at roughly one-twentieth of the model cost, though it failed to produce a single exact sequence for the complex multi-step command.

That gap between individual correctness and complete sequences is one of the paper’s most important lessons. On the complex scenario, GPT-4o-mini earned a respectable F1 of 0.8764 because most of its individual tool calls were right — but the ordering was wrong, typically invoking the camera before the required final stop. GPT-4o completed 26 of 40 complex sequences, including all 20 Slovak repetitions. The authors argue, convincingly, that robotics evaluations of language models must report task-level sequence success, not just per-call accuracy, because a command with a missing or misplaced stop action can be physically dangerous even when it scores well on superficial metrics.

Latency measurements complete the practical picture. On the deployed robot’s own computer, the first tool callback arrived a median of 2.74 seconds after the nominal end of a recorded utterance, rising to 4.10 seconds at the 95th percentile. By contrast, the local ROS stop-service round trip took a median of just 4.28 milliseconds. In other words, cloud transcription and model inference dominate the response path, while local robot communication is nearly instantaneous. This supports a sensible division of labor: the LLM handles interactive, high-level interpretation, while time-critical stabilization and locomotion stay with deterministic local controllers. The study also verified conversation history retention, hazard detection — the system correctly flagged a carpet cutter with an exposed blade as dangerous — graceful error recovery, and reliable mid-execution interruption.

The authors are careful about the limits of their claims. The functional study used scripted commands in dry-run mode, so physical motion quality was not quantified; the voice study involved only three speakers and four phrases; Slovak difficulty at the transcription stage remains an upstream failure mode when meaning is genuinely lost; and dependence on remote APIs introduces network latency and availability concerns. Still, SMaRTAban stands out as the first system in its field to report an evaluated combination of bilingual voice control in English and Slovak, on-demand GPT-4o vision, operator-controlled interruption of running actions, and rigorous quantitative analysis of the whole command path. Future work points toward native multimodal audio input, locally hosted open-source models to remove cloud dependency, the Model Context Protocol for portable tool definitions, integration with SLAM for autonomous navigation, and formal human-robot interaction studies with non-expert users. For now, the image of a quadruped robot that understands “Choď dopredu” as readily as “Go forward” — and stops the instant you tell it to — marks a meaningful step from Čapek’s century-old fiction toward an everyday robotic companion.

Subject of Research: A voice-controlled large language model agent enabling bilingual natural-language control of a quadruped mobile robot with integrated vision and safety mechanisms.

Article Title: SMaRTAban: a voice-controlled LLM agent for quadruped mobile robotics with integrated vision

Article References: Zelenay, E., Kocúr, M., Lukáč, M., Duchoň, F., & Marko, R. (2026). SMaRTAban: a voice-controlled LLM agent for quadruped mobile robotics with integrated vision. International Journal of Intelligent Robotics and Applications. https://doi.org/10.1007/s41315-026-00593-0

Image Credits: AI Generated

DOI: 10.1007/s41315-026-00593-0

Keywords: large language models, quadruped robotics, voice control, Whisper, GPT-4o, ReAct agent, human-robot interaction, natural language interface, LangGraph, robot safety, computer vision, ROS