Marcel Deckert

Cytoskeletal plasticity is a defining feature of cancer progression, enabling tumor cells to adapt their morphology, mechanics, and migratory behavior during invasion and metastasis. Although actin filaments, microtubules, and intermediate filaments are known to cooperate in these processes, the molecular mechanisms coordinating their dynamics remain incompletely understood, particularly the role of post-translational modifications (PTMs). Here, we developed CytoFRET2, a multiparametric cytometry-based FRET platform that enables real-time and simultaneous monitoring of the dynamics of actin filaments, microtubules, and vimentin in living suspension cells. The system combines fluorescently tagged cytoskeletal reporters with spectral flow cytometry, allowing simultaneous high-content analysis of multiple cytoskeletal networks while overcoming autofluorescence and fluorescence interference from small molecules. Using spectral CytoFRET2, we screened a small library of epigenetic compounds targeting regulators of acetylation and methylation pathways. The screen revealed that inhibition of lysine deacetylases (KDACs) and sirtuins promoted stabilization of both microtubules and vimentin filaments, without impacting actin filament organization. In contrast, inhibition of lysine acetyltransferases (KATs), particularly with garcinol and anacardic acid, induced rapid vimentin disassembly. Mechanistically, the study reveals acetylation as a key post-translational modification regulating the dynamics of microtubules and vimentin filaments, with KAT inhibitors emerging as potent modulators of vimentin organization. Together, the findings establish spectral CytoFRET2 as a versatile platform for systematic investigation of cytoskeletal regulatory networks and therapeutic vulnerabilities in cancer.

Extracellular matrix (ECM) stiffening is a biophysical hallmark of solid tumors. Cutaneous melanoma is an aggressive malignancy characterized by high heterogeneity and phenotypic plasticity in which melanoma cells switch from a proliferative and differentiated phenotype to an invasive, dedifferentiated and therapy-resistant state. However, the impact of ECM stiffness on the diverse cellular phenotypes of melanoma remains poorly defined. Here, we show that melanoma cell subpopulations differ in their responses to mechanical signals. Compared to melanocytic/transitory cells, dedifferentiated cells exhibited heightened sensitivity to stiff collagen matrices, characterized by increased cell spreading, focal adhesion maturation, YAP nuclear translocation and contractility. ECM stiffening enhanced proliferation, migration and invasion in dedifferentiated cells, whereas highly proliferative and poorly migratory melanocytic/transitory cells were less affected by collagen stiffness. Importantly, a soft ECM sensitized dedifferentiated cells, but not melanocytic/transitory cells, to BRAF/MEK inhibition. Mechanistically, the mechanosensitivity of dedifferentiated cells relies on collagen receptors DDR1 and DDR2, which control cytoskeleton reorganization and YAP mechanosignaling. Genetic or pharmacological inhibition of DDR, actomyosin contractility, or YAP suppressed stiffness-induced proliferation and migration, reduced traction forces, and restored sensitivity to targeted therapy in dedifferentiated cells. Conversely, ectopic DDR1/DDR2 expression confers mechanosensitive properties to melanocytic cells. Our results thus reveal that phenotypic plasticity endows dedifferentiated melanoma cells with increased addiction to mechanical cues and implicate DDR1/2-YAP-dependent signaling in this aggressive behavior.

Melanoma progression and resistance to targeted therapies remain major clinical challenges, driven in part by the remarkable phenotypic plasticity of melanoma cells. Identifying molecular mechanisms that couple tumor survival to adaptative drug responses is therefore essential. Here, we identify PSMD14, a proteasome-associated deubiquitinase, as an essential regulator of melanoma plasticity, growth, survival, and therapeutic resistance with strong prognostic significance in metastatic disease. An unbiased siRNA screen targeting the human deubiquitinase family revealed PSMD14 (proteasome 26S subunit, non-ATPase 14) as a top regulator of melanoma cell proliferation. Integrative analyses of DepMap, TCGA, and patient-derived datasets revealed that PSMD14 is frequently upregulated in melanoma, enriched in metastatic lesions, and significantly associated with poor patient outcome. Functional and pharmacological studies demonstrated that genetic depletion or inhibition of PSMD14 suppresses proliferation, clonogenic and long-term growth, and viability of melanoma cells across BRAF-, NRAS-, and NF1-driven genotypes, while inducing DNA damage and apoptosis. Consistently, PSMD14 inhibition markedly reduced tumor growth in Nras and Braf syngeneic mouse models. Mechanistically, we uncover a non-proteolytic role for PSMD14 as an epigenetic regulator of chromatin state. Proteomic and biochemical analyses identified histone H2A as a direct interactor and substrate of PSMD14. PSMD14 deubiquitinates H2A at lysine 119 independently of the proteasome, antagonizing the Polycomb E3 ligase RING1B. Loss of PSMD14 allows the increment of H2AK119 ubiquitination, transcriptional repression of pro-survival genes, including MCL1 and BCL2, and apoptotic cell death, effects rescued by RING1B depletion. Importantly, we demonstrate that the PSMD14-H2A axis governs melanoma adaptation to MAPK pathway inhibition. PSMD14 expression and H2AK119 ubiquitination dynamically correlate with therapeutic response, drug-tolerant persistence, and acquired resistance. Targeting PSMD14 genetically or pharmacologically enhances the efficacy of BRAF and MEK inhibitors, suppresses the emergence of drug-tolerant persister cells, and prevents tumor relapse in vivo . Together, this findings establish PSMD14 as a chromatin-rewiring enzyme that links proteostasis to epigenetic control of melanoma plasticity and therapy resistance, highlighting PSMD14 as a promising biomarker and therapeutic target in aggressive and drug-resistant melanoma.

Extracellular matrix (ECM) stiffening is a biophysical hallmark of solid tumors. Cutaneous melanoma is an aggressive malignancy characterized by high heterogeneity and phenotypic plasticity in which melanoma cells switch from a proliferative and differentiated phenotype to an invasive, dedifferentiated and therapy-resistant state. The impact of ECM stiffness on the diverse cellular phenotypes of melanoma remains poorly defined. Here, we show that melanoma cell subpopulations differ in their responses to mechanical signals. Compared to melanocytic cells, dedifferentiated cells exhibited increased proliferation, migration, and resistance to MAP kinase-targeted therapy when seeded on stiff collagen. In contrast, a soft ECM impaired their proliferation and migration, and sensitized them to targeted therapy. The mechanosensitivity of dedifferentiated cells relies on collagen receptors DDR1 and DDR2, which control cytoskeleton reorganization and YAP mechanosignaling. Our results thus reveal that phenotypic plasticity endows dedifferentiated melanoma cells with increased addiction to ECM stiffness and implicate DDR1/2-YAP-dependent signaling in this aggressive behavior.

Post-translational modification by ubiquitin is crucial for protein turnover. Deubiquitinases (DUBs) remove ubiquitin chains from target proteins to prevent their degradation by the proteasome, thus acting as gatekeepers of protein homeostasis alongside the ubiquitin-proteasome system (UPS). Tumor cells exhibit remarkable plasticity, enabling them to adapt to anticancer treatments and the conditions of the tumor microenvironment, including mechanical cues from the extracellular matrix (ECM). However, the role of DUBs in mechanotransduction remains unexplored. To identify DUBs involved in cancer cell mechanosignaling, we used melanoma cells grown on collagen matrices with varying stiffnesses and an activity-based ubiquitin probe to profile DUB activities. Our approach, combined with quantitative proteomics, revealed that ubiquitin-specific protease 9X (USP9X) is sensitive to ECM stiffness through discoidin domain receptors (DDR)/actomyosin signaling pathway. In silico analysis further indicated that the mechanosensor YAP is part of the USP9X interactome, and USP9X expression correlates with the YAP transcriptional signature in melanoma. We hypothesized that mechanical signals regulate YAP levels through USP9X DUB activity. Consistently, low collagen stiffness reduced YAP expression, and siRNA-mediated depletion or pharmacological inhibition of USP9X decreased YAP protein expression in tumor cells. Conversely, knockdown of the ubiquitin E3 ligase βTrCP increased YAP protein levels. Affinity purification of polyubiquitinated proteins using Tandem Ubiquitin Binding Entities (TUBEs) showed that combined USP9X and proteasome inhibition increased YAP poly-ubiquitination, revealing that USP9X deubiquitinates YAP to prevent its proteasomal degradation. Targeting USP9X impaired stiffness-mediated responses, including YAP nuclear translocation and transcriptional activity, cell migration and invasion, and drug resistance. An experimental metastasis assay showed that stable knockdown of USP9X impaired melanoma cell lung colonization. Finally, targeting USP9X in a syngeneic BRAF-mutant melanoma model counteracted targeted therapy-induced ECM remodeling, enhanced treatment efficacy, and delayed tumor relapse. Our findings reveal a novel role of USP9X in cancer cell mechanobiology and drug resistance through stiffness-dependent stabilization of the oncoprotein YAP, proposing USP9X as a targetable "mechano-DUB" in cancer.

Extracellular matrix (ECM) stiffening, resulting from increased collagen deposition and cross-linking, is a key biophysical factor of the tumor microenvironment. Cutaneous melanoma is a deadly metastatic cancer. Its aggressiveness stems from high intratumoral heterogeneity, resulting from the plasticity of melanoma cells, which transit from a melanocytic state to dedifferentiated therapy-resistant and invasive phenotypes, characterized by mesenchymal and/or neural crest stem cell-like features. Phenotypic plasticity is regulated by stroma-derived soluble factors, but the functional impact of ECM stiffening on melanoma cell phenotypes remains ill defined. Here, we found that melanoma cell subpopulations display difference in mechanical responsiveness. Compared to melanocytic cells, mesenchymal dedifferentiated cells showed increased proliferation, migration and resistance to MAP kinase-targeted therapy when seeded on stiff collagen. By contrast, a soft ECM impaired their proliferation and migration and sensitized them to targeted therapy. In addition, extracellular mechanical signals are required to sustain melanoma cell identity and dedifferentiation features. Further analyses indicated that the mechanosensitivity nature of dedifferentiated cells relies on the expression and activation of collagen receptors DDR1 and DDR2 that control actomyosin cytoskeleton reorganization and YAP mechanotransduction pathway. Inhibiting both DDR in dedifferentiated melanoma cells abrogated their mechano-induced behavior and drug-resistant phenotype, while forcing their expression in melanocytic cells induced mechanical responsiveness and a less differentiated phenotype. Our results reveal that phenotypic reprogramming endows dedifferentiated melanoma cells with increased sensitivity and addiction to ECM stiffness. We propose that mechano-addiction mediated by DDR collagen receptors may represent a novel vulnerability for aggressive dedifferentiated cancer cells that can be exploited for therapeutic benefits.

Background: Large language models (LLMs) have raised both interest and concern in the academic community. They offer the potential for automating literature search and synthesis for systematic reviews but raise concerns regarding their reliability, as the tendency to generate unsupported (hallucinated) content persist. Objective: The aim of the study is to assess the performance of LLMs such as ChatGPT and Bard (subsequently rebranded Gemini) to produce references in the context of scientific writing. Methods: The performance of ChatGPT and Bard in replicating the results of human-conducted systematic reviews was assessed. Using systematic reviews pertaining to shoulder rotator cuff pathology, these LLMs were tested by providing the same inclusion criteria and comparing the results with original systematic review references, serving as gold standards. The study used 3 key performance metrics: recall, precision, and F1-score, alongside the hallucination rate. Papers were considered “hallucinated” if any 2 of the following information were wrong: title, first author, or year of publication. Results: In total, 11 systematic reviews across 4 fields yielded 33 prompts to LLMs (3 LLMs×11 reviews), with 471 references analyzed. Precision rates for GPT-3.5, GPT-4, and Bard were 9.4% (13/139), 13.4% (16/119), and 0% (0/104) respectively (P<.001). Recall rates were 11.9% (13/109) for GPT-3.5 and 13.7% (15/109) for GPT-4, with Bard failing to retrieve any relevant papers (P<.001). Hallucination rates stood at 39.6% (55/139) for GPT-3.5, 28.6% (34/119) for GPT-4, and 91.4% (95/104) for Bard (P<.001). Further analysis of nonhallucinated papers retrieved by GPT models revealed significant differences in identifying various criteria, such as randomized studies, participant criteria, and intervention criteria. The study also noted the geographical and open-access biases in the papers retrieved by the LLMs. Conclusions: Given their current performance, it is not recommended for LLMs to be deployed as the primary or exclusive tool for conducting systematic reviews. Any references generated by such models warrant thorough validation by researchers. The high occurrence of hallucinations in LLMs highlights the necessity for refining their training and functionality before confidently using them for rigorous academic purposes.

Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: Large language models (LLMs) have raised both interest and concern in the academic community. They offer the potential for automating literature search and synthesis for systematic reviews but raise concerns regarding their reliability, as the tendency to generate unsupported (hallucinated) content persist. Objective: The aim of the study is to assess the performance of LLMs such as ChatGPT and Bard (subsequently rebranded Gemini) to produce references in the context of scientific writing. Methods: The performance of ChatGPT and Bard in replicating the results of human-conducted systematic reviews was assessed. Using systematic reviews pertaining to shoulder rotator cuff pathology, these LLMs were tested by providing the same inclusion criteria and comparing the results with original systematic review references, serving as gold standards. The study used 3 key performance metrics: recall, precision, and F₁-score, alongside the hallucination rate. Papers were considered “hallucinated” if any 2 of the following information were wrong: title, first author, or year of publication. Results: In total, 11 systematic reviews across 4 fields yielded 33 prompts to LLMs (3 LLMs×11 reviews), with 471 references analyzed. Precision rates for GPT-3.5, GPT-4, and Bard were 9.4% (13/139), 13.4% (16/119), and 0% (0/104) respectively (P<.001). Recall rates were 11.9% (13/109) for GPT-3.5 and 13.7% (15/109) for GPT-4, with Bard failing to retrieve any relevant papers (P<.001). Hallucination rates stood at 39.6% (55/139) for GPT-3.5, 28.6% (34/119) for GPT-4, and 91.4% (95/104) for Bard (P<.001). Further analysis of nonhallucinated papers retrieved by GPT models revealed significant differences in identifying various criteria, such as randomized studies, participant criteria, and intervention criteria. The study also noted the geographical and open-access biases in the papers retrieved by the LLMs. Conclusions: Given their current performance, it is not recommended for LLMs to be deployed as the primary or exclusive tool for conducting systematic reviews. Any references generated by such models warrant thorough validation by researchers. The high occurrence of hallucinations in LLMs highlights the necessity for refining their training and functionality before confidently using them for rigorous academic purposes.

Supplementary Figure 1. b-AP15 and its derivative VLX1570 inhibit USP14 activity in melanoma cells. Supplementary Figure 2. USP14 inhibition irreversibly induces growth inhibition and cytotoxicity in melanoma cells. Supplementary Figure 3. USP14 overexpression reduces the cytotoxic effect of b-AP15 on melanoma cell. Supplementary Figure 4. siRNA-mediated depletion of USP14 reduces melanoma cell growth and focus formation. Supplementary Figure 5. USP14 targeting induces caspase-independent and ROS-dependent cell death. Supplementary Figure 6. USP14 inhibition induces a rapid accumulation of polyubiquitinated proteins and HSP70 in melanoma cells