The significance of this finding lies in the peculiar epidemiology of sleeping sickness. Two subspecies of Trypanosoma brucei infect humans: Trypanosoma brucei gambiense, which accounts for more than 95 percent of all reported cases and is found primarily in West and Central Africa, and Trypanosoma brucei rhodesiense, which circulates in East and Southern Africa and also infects livestock and wild animals. Although gambiense sleeping sickness dominates the global case count, elimination of the disease as a public health problem requires confronting both parasites. A drug that works only against one subspecies would leave an entire region of the continent without a simple cure, undermining the ambitious elimination targets set by the World Health Organization and its partners.

Acoziborole, developed through a partnership led by the Drugs for Neglected Diseases initiative, represents a genuine breakthrough in chemotherapy for neglected diseases. Unlike earlier treatments for gambiense sleeping sickness, which required intravenous infusions of toxic compounds such as melarsoprol, or lengthy courses of oral drugs like fexinidazole, acoziborole is designed to be given as a single oral dose. The drug has already completed clinical development as a one-dose cure for Trypanosoma brucei gambiense infection, a milestone that could transform treatment programs by allowing community-level administration without the need for hospitalization or complex follow-up. But a single-dose strategy only delivers full elimination if it can be deployed against both subspecies of the parasite.

To address this gap, a team of researchers from the Swiss Tropical and Public Health Institute and the Drugs for Neglected Diseases initiative conducted an in-depth preclinical evaluation of acoziborole against bloodstream forms of all three Trypanosoma brucei subspecies, with particular attention to the rhodesiense parasite. The study, authored by Monica Cal, Sonja Keller, Romina Rocchetti, Stéphanie Braillard, Jean-Robert Ioset, Pascal Mäser and Marcel Kaiser, was designed to strengthen the drug’s preclinical dossier by systematically characterizing its potency, speed of action and killing profile across diverse parasite strains, including clinical isolates that had never been exposed to the compound in the laboratory.

The results were encouraging on every measure. Acoziborole displayed consistent and high activity against all tested strains of Trypanosoma brucei rhodesiense, with half-maximal inhibitory concentrations, or IC50 values, clustered around 600 nanomolar. This means that extremely low concentrations of the drug are sufficient to halve the growth of the parasite in vitro, a hallmark of a potent antitrypanosomal agent. Crucially, the activity was uniform across different strains, including a multidrug-resistant field isolate, suggesting that resistance mechanisms circulating in natural parasite populations are unlikely to compromise the drug’s effectiveness. For a disease whose control has repeatedly been set back by drug toxicity and emerging resistance, this consistency is a valuable asset.

Speed of kill is another critical parameter for any drug intended to cure a rapidly progressive infection, and here acoziborole also outperformed its closest oral competitor. The researchers found that acoziborole acts faster than fexinidazole, the ten-day oral treatment that has recently been introduced for both forms of sleeping sickness. More striking still, acoziborole exhibited irreversible cidality against Trypanosoma brucei rhodesiense at concentrations of 4.3 micromolar or higher, equivalent to 1.6 micrograms per milliliter, after just 15 hours of incubation. Irreversible cidality means that even after the drug is removed from the culture medium, the parasites cannot recover and resume growth. A compound that permanently disables the parasite within hours of exposure offers a substantial margin of safety in clinical practice, where drug levels in the blood fluctuate over time.

One nuance emerged from the side-by-side comparison of the two human-infective subspecies. At very short exposure times of eight hours or less, Trypanosoma brucei gambiense appeared to be more sensitive to acoziborole than its rhodesiense counterpart. In a conventional drug, such a difference in early kill kinetics might raise concerns about whether the same dosing regimen would suffice for both forms of the disease. However, the researchers argue that this discrepancy is clinically irrelevant, and the reasoning rests on a remarkable pharmacological property of the compound itself.

Acoziborole possesses an extremely long elimination half-life in the human body, meaning it persists in the bloodstream at active concentrations for weeks or even months after a single oral dose. This slow clearance is precisely what makes a one-dose cure possible: rather than requiring repeated administrations to maintain therapeutic levels, the drug lingers and continues to act on any parasites it encounters. Because patients effectively live with sustained drug exposure for extended periods, the modest difference in how quickly the two subspecies succumb during the first hours of contact becomes immaterial. Over the long residence time of acoziborole in the body, both parasites face the same relentless chemical assault, and both are expected to be eliminated.

Taken together, the laboratory findings support a confident conclusion: acoziborole is expected to be as effective against rhodesiense sleeping sickness, often abbreviated r-HAT, as it is against gambiense sleeping sickness, or g-HAT, in patient populations. This is more than a technical footnote. Rhodesiense sleeping sickness follows a more acute course than the gambiense form, with a shorter incubation period and faster progression to neurological disease, and it has historically been treated with the toxic arsenical melarsoprol, which itself kills a significant fraction of patients. A safe, single-dose oral alternative for this eastern form of the disease would eliminate one of the last justifications for using a nineteenth-century poison in modern medicine.

The broader context makes the timing of this work especially compelling. The incidence of sleeping sickness has fallen to a historic low, thanks to decades of sustained surveillance, vector control and treatment efforts, and the World Health Organization has recognized the disease as one of the very few infections targeted for elimination as a public health problem. Yet the endgame of any elimination campaign is the hardest phase: cases become rare, scattered and hard to find, and health systems in endemic regions struggle to justify maintaining specialized diagnostic and treatment capacity. A single-dose oral drug that works against both parasite subspecies could be administered presumptively or with minimal testing at the community level, closing the last gaps in coverage. With its clinical success against gambiense infection now backed by preclinical evidence of comparable activity against rhodesiense parasites, acoziborole stands as the most promising tool yet for delivering the final blow against a disease that has plagued tropical Africa for generations.

Subject of Research: Preclinical efficacy profiling of the antitrypanosomal drug acoziborole against Trypanosoma brucei rhodesiense and other African trypanosomes

Article Title: Winning the endgame against sleeping sickness: Efficacy profiling of acoziborole against African trypanosomes with the focus on Trypanosoma brucei rhodesiense

Article References: Cal, M., Keller, S., Rocchetti, R., Braillard, S., Ioset, J.-R., Mäser, P., & Kaiser, M. (2026). Winning the endgame against sleeping sickness: Efficacy profiling of acoziborole against African trypanosomes with the focus on Trypanosoma brucei rhodesiense. PLOS Neglected Tropical Diseases, 20(10), e0014762. https://doi.org/10.1371/journal.pntd.0014762

Image Credits: AI Generated

DOI: 10.1371/journal.pntd.0014762

Keywords: sleeping sickness, acoziborole, African trypanosomiasis, Trypanosoma brucei rhodesiense, Trypanosoma brucei gambiense, tsetse fly, neglected tropical diseases, drug resistance, fexinidazole, elimination, preclinical testing, antiparasitic drugs

The research, led by Nathalia Ferreira and Frauke Alves at the Max Planck Institute for Multidisciplinary Sciences in Göttingen, together with colleagues at Miltenyi Biotec and the University Medical Center Göttingen, focused on pancreatic ductal adenocarcinoma, or PDAC, one of the deadliest human malignancies. The team grew tumor organoids, miniature three-dimensional structures derived directly from patients, and co-cultured them with CAR T cells engineered to target CD318, a tumor-associated antigen frequently expressed on pancreatic cancer cells. Rather than relying on a single destructive endpoint measurement, the scientists filmed the interaction over time using time-lapse bright-field microscopy and let the algorithm do the quantitative heavy lifting.

Technically, the challenge is considerable. Organoids are irregular, translucent, three-dimensional objects embedded in a gel matrix, and distinguishing them from surrounding debris, T cells, and matrix components is far from trivial for image analysis software. OrganoIDNet, a deep learning-based framework originally developed for automated segmentation of organoid morphology, was extended in this study to perform label-free segmentation of individual organoids in bright-field images. Because the method requires no fluorescent dyes or genetic labeling of the tumor cells, it avoids the phototoxicity, spectral overlap, and perturbation of cell behavior that labeling can introduce, which is a significant advantage when the goal is to observe undisturbed immune-tumor interactions.

The co-culture system itself was built on a Matrigel-based sandwich arrangement, in which organoids are embedded within the extracellular-matrix-like gel and overlaid with effector cells. This configuration keeps the three-dimensional architecture of the tumor model intact while allowing CAR T cells to migrate toward and engage their targets. Using this setup, the researchers monitored the cultures across multiple effector-to-target ratios, meaning they varied the number of CAR T cells added per organoid to model different dosing intensities. The OrganoIDNet pipeline then continuously quantified two key parameters: the number of organoids present and their total area, providing a longitudinal readout of tumor killing that conventional assays, which typically destroy the sample at a single time point, cannot deliver.

The results were striking. CD318-directed CAR T cells, designated CAR-318, induced robust, antigen-dependent cytotoxicity against the PDAC organoids, visible as progressive reductions in both organoid number and organoid size over the course of the observation period. Crucially, the killing was antigen-specific: control conditions, including non-transduced T cells, did not produce the same destructive effect. This distinction matters because one of the persistent worries in CAR T development is off-tumor toxicity, and an assay that can distinguish antigen-driven killing from nonspecific activity in a physiologically relevant three-dimensional model offers a valuable preclinical filter.

Beyond the imaging data, the team corroborated the functional readouts with immunological measurements. At the endpoint of the co-cultures, the CAR T cells showed increased expression of T cell activation markers, consistent with genuine antigen engagement, and changes in the expression of TIM-3, a receptor often associated with T cell exhaustion or activation states. These flow cytometry-based findings dovetailed with the morphological evidence from the live imaging, painting a coherent picture: the engineered cells were not merely drifting near the organoids but were being activated by their target antigen and executing tumor cell elimination.

Spatial analysis added another dimension to the study. By examining the positions of T cells relative to organoids in the time-lapse images, the researchers documented close T cell-organoid interactions and evidence of early tumor cell elimination. This time-resolved, spatially aware information complements endpoint assays such as flow cytometry or viability staining, which reveal the final state of a culture but say little about the kinetics of the attack. Knowing when, not just whether, CAR T cells engage and destroy their targets could help developers compare candidate constructs, optimize effector-to-target dosing, and identify conditions under which tumor cells escape immune surveillance.

The implications for personalized medicine are among the most compelling aspects of the work. Patient-derived organoids retain many of the molecular and histological characteristics of the tumors from which they were grown, which makes them far more faithful stand-ins for a given patient’s cancer than conventional two-dimensional cell lines. An automated imaging assay that can quantify how a patient’s own tumor organoids respond to a given CAR construct could, in principle, inform treatment selection or guide the design of patient-specific engineered cells. The authors suggest that the platform has potential utility as a preclinical tool for CAR T cell development and future personalized immunotherapy studies, and the pancreatic cancer context is particularly significant given how few effective options exist for that disease.

The study also reflects a broader trend in preclinical oncology: the marriage of artificial intelligence with live-cell imaging to extract quantitative, longitudinal data from complex biological systems. Manual analysis of time-lapse microscopy is prohibitively labor-intensive and subject to observer bias, which has historically limited the throughput and reproducibility of such experiments. By delegating segmentation and quantification to a neural network, the Göttingen team transformed what would otherwise be hours of expert annotation into an automated pipeline capable of tracking dozens of organoids across multiple conditions and time points. The ethical and regulatory groundwork for the study was carefully laid as well: the patient material was obtained through the Molecular Pancreas Program at the University Medical Center Göttingen with appropriate approvals, and healthy donor blood for T cell isolation was collected with informed consent under a local ethics committee approval.

There are, of course, limits to what any organoid platform can model. Organoids lack the full vasculature, stromal complexity, and immune suppressive microenvironment of an intact tumor, and bright-field imaging, while label-free, offers lower molecular specificity than fluorescence-based approaches. The authors also note that the published version of their work was shared early as an accepted manuscript subject to further edits. Nevertheless, the combination of patient-derived three-dimensional tumor models, engineered immune cells, and AI-driven longitudinal image analysis represents a meaningful step toward preclinical assays that capture the dynamics of immunotherapy rather than just its endpoints. For a disease as lethal and treatment-resistant as pancreatic cancer, tools that reveal exactly how and when engineered immune cells dismantle a patient’s tumor in the laboratory could accelerate the translation of CAR T technology from the bench to the bedside, and the OrganoIDNet-based assay described in this study offers a template for how that acceleration might be achieved across a range of solid tumor types.

Subject of Research: Real-time AI-based imaging of CAR T cell cytotoxicity against pancreatic cancer patient-derived organoids

Article Title: Real-time monitoring of CAR T cell dynamics in tumor patient-derived organoids using the OrganoIDNet algorithm

Article References: Ferreira, N., Dourlens, C., Scodellaro, R., Stroebel, P., Schäfer, D., Hardt, O., & Alves, F. (2026). Real-time monitoring of CAR T cell dynamics in tumor patient-derived organoids using the OrganoIDNet algorithm. Journal of Translational Medicine, 24(1), Article 1169. https://doi.org/10.1186/s12967-026-08929-x

Image Credits: AI Generated

DOI: 10.1186/s12967-026-08929-x

Keywords: CAR T cells, patient-derived organoids, pancreatic cancer, PDAC, OrganoIDNet, deep learning, CD318, immunotherapy, live-cell imaging, preclinical testing, personalized medicine, T cell activation

The research, led by Franziska Köhler and Sonja Kallendrusch of Leipzig University and the HMU Health and Medical University, together with colleagues in Erfurt and Potsdam, compared two complementary patient-derived culture systems side by side. The first, patient-derived neurospheres (PDNS), are compact, free-floating spheroids grown from dissociated tumor cells under serum-free conditions. They offer a controlled, tumor-cell-intrinsic system in which variables can be manipulated with relative ease. The second, patient-derived tissue slice cultures (PDTC), preserve the tumor’s native architecture: thin, 350-micrometer slices of fresh surgical tissue are cultured at an air–liquid interface, retaining the extracellular matrix, vascular structures, immune cells, and stromal compartments that neurospheres inevitably discard. The team systematically exposed both models to two AAV serotypes, AAV2 and AAV6, each carrying a green fluorescent protein (GFP) reporter gene, and then tracked transduction with live confocal microscopy, quantitative PCR, flow cytometry, and immunofluorescence.

The technical rigor of the workflow is central to the study’s credibility. Thirteen surgical specimens from 25 collected samples ultimately entered the analysis, twelve of them IDH-wildtype glioblastomas under the WHO 2021 classification. Tissue not required for diagnosis was obtained with informed consent and ethical approval. For neurospheres, tumor tissue was mechanically and enzymatically dissociated with Accutase, filtered, and seeded at 100,000 cells per well; after 10 to 15 days, spheroids of 100 to 150 micrometers in diameter were ready for experiments. Slice cultures were punched into 2 to 3 millimeter discs and maintained on membrane inserts. After testing several media formulations, the researchers selected NeuroCult-XF, a serum-free proliferation medium, because it preserved tissue architecture and histopathological features better than serum-containing alternatives or other defined formulations. Blinded pathological review confirmed that fibrillary architecture and even characteristic gemistocytic features—enlarged cell bodies with eccentric nuclei—were retained in both models.

The cellular composition analysis revealed stark differences between the two systems, exactly as the team had hypothesized. Slice cultures retained expression of immune markers such as CD68, the vascular marker CD31, and the hypoxia marker HIF1α in organotypic patterns closely mirroring the original tumors. Neurospheres, by contrast, showed virtually no CD31, scarce and disorganized CD68-positive cells, and only weak HIF1α restricted to larger spheroids. More subtly, sphere-forming conditions selectively enriched stem-like populations: the mean SOX2-positive cell fraction rose from 0.114 in native tissue to 0.328 in PDNS, and Nestin positivity climbed from 0.132 to 0.393, while the mesenchymal marker CD44 fell. This finding matters because it confirms that neurosphere culture is not a neutral act of preservation but an active selection process that reshapes the cellular landscape before any vector is ever introduced.

When the vectors were applied, the serotype differences were unambiguous. In neurospheres, AAV6 drove significantly greater GFP expression than AAV2 across a dose range of 10^6 to 10^9 particles, with significant signal appearing as early as day 2 at higher doses, while AAV2 reached significance only at the highest dose after 5 days. Quantitative PCR corroborated the fluorescence data: five days after exposure to 10^8 particles, PLAT-normalized vector genome abundance averaged 146.7 copies per cell for AAV2 but 1,074.0 for AAV6 in EGF-containing medium. Notably, epidermal growth factor supplementation itself proved to be a critical experimental variable. Because AAV6 is known to use EGFR as a co-receptor for cellular entry, the team deliberately compared EGF-containing and EGF-free conditions. In neurospheres, EGF presence modified both the magnitude and temporal stability of AAV6-mediated GFP expression, and marker-defined analysis showed that EGF deprivation broadened GFP positivity across GFAP-, CD44-, Nestin-, and SOX2-positive populations, whereas under EGF-containing conditions only the Nestin-positive fraction showed significant GFP signal.

Exploratory flow cytometry from two independent experiments added an intriguing layer. GFP positivity was markedly enriched among SOX2-positive cells (40.8 percent versus 7.2 percent in SOX2-negative cells) and among cells triple-positive for CD133, Nestin, and SOX2 (41.0 percent versus 7.5 percent). The authors are careful to stress that these figures, based on limited sample numbers, are descriptive and cannot distinguish preferential vector uptake from differences in population abundance, survival, or promoter activity. Still, the pattern suggests that the most stem-like, aggressive cellular states within a glioblastoma may be among those most efficiently engaged by AAV6—a hypothesis with obvious therapeutic implications if it can be confirmed with larger datasets and mechanistic studies.

The most sobering results came from the tissue slice cultures. Overall GFP readouts were comparable to, or in some cases higher than, those in neurospheres, but interpatient variability was dramatically greater. Some patient specimens transduced robustly; others showed little or no detectable GFP despite identical conditions. Mean fluorescence intensities of 4,989 for AAV2, 4,232 for AAV6, and 6,448 for AAV6 without EGF in slice cultures contrasted with far lower values in matched neurospheres, but the spread between patients was wide enough that no single-patient result could be generalized. The authors argue that this variability should not be dismissed as experimental noise. Instead, it may represent a biologically faithful reflection of the heterogeneity that any real-world gene therapy for glioblastoma would confront. A vector that performs uniformly in a homogeneous spheroid panel may still fail in a substantial fraction of patients, and only multi-patient tissue-based validation can expose that vulnerability before clinical translation.

Spatial analysis added another dimension. In neurospheres, GFP-positive cells clustered predominantly at the spheroid periphery, consistent with diffusion-limited vector transport through densely packed three-dimensional tissue, although the authors caution that GFP fluorescence reflects the downstream consequence of delivery and expression rather than physical particle penetration itself. Slice cultures showed a more homogeneous distribution of fluorescence across the tissue, an effect not associated with preferential vascular localization and therefore likely driven by preserved microarchitecture rather than vascular entry routes. Equally notable was the observation of occasional GFP-positive cells in two samples of adjacent non-neoplastic brain tissue, a preliminary but cautionary signal that off-target transduction of tumor-adjacent cells can occur under experimental conditions, consistent with the broad central nervous system tropism of naturally occurring AAV serotypes.

The study’s self-imposed limitations are candidly acknowledged. Vector dose escalation could not be performed in slice cultures because of limited tissue availability, preventing fully matched comparisons between models. Flow cytometric findings rested on only two experiments. Vector genome quantification cannot distinguish entry, trafficking, persistence, or processing, and GFP expression additionally depends on transcription and translation. Slice cultures themselves, while architecturally faithful, remain ex vivo systems that cannot reproduce systemic immunity, vascular dynamics, or long-term tumor evolution. The significant increase in SOX2 expression observed in neurospheres after AAV6 treatment without EGF likewise supports an association rather than a demonstrated causal effect of vector exposure on tumor-cell state.

What the study ultimately delivers is a framework rather than a verdict. The authors explicitly do not claim that both models are mandatory for every preclinical AAV program, nor that either predicts in vivo performance. Instead, they advocate a question-driven approach: neurospheres, with their lower variability and experimental tractability, are well suited to initial serotype screening and dissection of tumor-cell-intrinsic effects; candidate vectors that pass that filter can then be challenged against panels of slice cultures from multiple patients, where preserved architecture, stromal and vascular compartments, and honest interpatient heterogeneity reveal how performance holds up in tissue that more closely resembles what a surgeon would leave behind in a resection cavity. In a disease where nearly every promising preclinical result has eventually collided with clinical reality, the insistence that model selection itself is a determinant of apparent vector efficacy may prove to be one of the more consequential methodological messages glioblastoma gene therapy has received in years.

Subject of Research: Preclinical evaluation of AAV-mediated gene delivery in patient-derived glioblastoma cultures

Article Title: Patient-derived tissue cultures complement neurospheres for preclinical evaluation of AAV-mediated gene delivery in glioblastoma

Article References: Köhler, F., Hess, K., Koloske, C., Gaunitz, F., Rosahl, S. K., Gerlach, R., & Kallendrusch, S. (2026). Patient-derived tissue cultures complement neurospheres for preclinical evaluation of AAV-mediated gene delivery in glioblastoma. Journal of Neuro-Oncology, 179(3), Article 93. https://doi.org/10.1007/s11060-026-05807-w

Image Credits: AI Generated

DOI: 10.1007/s11060-026-05807-w

Keywords: glioblastoma, AAV vectors, gene therapy, patient-derived models, neurosphere cultures, tissue slice cultures, tumor microenvironment, viral vectors, EGF, transduction, preclinical testing, gene delivery