Marc Thilo Figge

The rapid rise of artificial intelligence (AI) in both scientific research and everyday life over the last decade has renewed interest in how AI-driven methods can support fungal research, diagnostics, and clinical microbiology. At the same time, advances in imaging technologies and increasing access to large collections of microscopy and medical imaging data have created new opportunities for automated image analysis. In this review, we examine the recent development of AI-based image analysis methods for microscopy in general, with a particular focus on fungal imaging applications. We discuss how deep learning approaches, especially Convolutional Neural Networks (CNNs) and U-nets, have transformed image classification and object detection tasks by enabling automated recognition of fungal cells, spores, hyphae, and infection-related structures with high accuracy. Finally, we explore emerging developments in the field, particularly the introduction of vision transformers and hybrid transformer-based architectures.

Antimicrobial resistance and tolerance pose escalating global health threats, necessitating reproducible and accessible tools for antimicrobial susceptibility testing (AST). While disk diffusion assays (DDAs) and Epsilometer tests (Etests) are widely used, there are limited open-source tools to analyze them. We present J-AST, a free, open-source, web-based platform for analyzing both DDAs and Etests. It provides automated and interactive annotation of regions of interest and metadata management, and quantifies microbial resistance and tolerance. J-AST outputs correlate strongly with those of existing tools, and equivalent DDA and Etest results correlate strongly with each other. The automated MIC detection achieved >90% agreement with manual readouts. J-AST is deployable both as desktop software and cloud service, unifies automated analysis with interactive review, and advances both fundamental research and clinical AST workflows.

Ulcerative colitis (UC) is a chronic relapsing inflammatory bowel disease (IBD) with a complex and incompletely understood etiology. 1 It is widely assumed that environmental factors trigger an exa...

Antimicrobial resistance and tolerance pose escalating global health threats, necessitating reproducible and accessible tools for antimicrobial susceptibility testing (AST). While disk diffusion assays (DDAs) and Epsilometer tests (Etests) are widely used, there are limited open-source tools to analyze them. We present J-AST , a free, open-source, web-based platform for analyzing both DDAs and Etests. It provides automated and interactive annotation of regions of interest and metadata management, and quantifies microbial resistance and tolerance. J-AST outputs correlate strongly with those of existing tools, and equivalent DDA and Etest results correlate strongly with each other. The automated MIC detection achieved >90% agreement with manual readouts. J-AST is deployable both as desktop software and cloud service, unifies automated analysis with interactive review, and advances both fundamental research and clinical AST workflows.

Background Whether the lung microbiome represents a stable microbial colonization or a transient ecosystem shaped by continuous microbial turnover and controlled by host immunity remains unresolved. The murine lung microbiome largely consists of species from the former Lactobacillus genus with Ligilactobacillus murinus as a dominant species, bacterial genera such as Streptococcus , Staphylococcus , Mammaliicoccus , Enterococcus and other less frequently detected bacteria. Here, we directly addressed the question of persistence and host interaction of a dominant murine lung commensal in vivo and focused on the host immune response towards lung commensal bacteria. Results We developed a transformation strategy for stable genomic integration of a green fluorescent protein (GFP)-encoding gene to track the fate of a lung bacterium. Following intranasal administration of GFP-labeled L. murinus in mice, bacteria were readily detected in the lungs at early time points but declined rapidly and became undetectable after 72 hours, as determined by quantification of viable bacteria and qPCR. Flow cytometry and fluorescence imaging revealed efficient uptake of GFP-labeled bacteria by lung phagocytes. These findings indicate that even dominant members of the murine pulmonary microbiota normally detected at low abundances are transiently present in the lungs without causing infection. We further analyzed the effects of moderate and high bacterial concentrations. While moderate bacterial loads were efficiently controlled without clinical effects, high concentrations induced severe lethargy, indicating a threshold-dependent host response. Finally, we demonstrated that pulmonary commensals such as L. murinus , Staphylococcus xylosus , and Mammaliicoccus sciuri, as well as conidia of the opportunistic lung pathogen Aspergillus fumigatus , are phagocytosed at comparable rates in macrophage assays. Conclusions Our data demonstrate that even lung-adapted bacterial species fail to establish stable colonization and are instead subject to rapid immune-mediated elimination contributing to the maintenance of a low microbial burden in the lungs. While this homeostatic balance supports health, elevated bacterial loads trigger immune activation and, at high levels, lead to health deterioration. Together, these results support a model of a highly dynamic and transient lung microbiome, maintained by continual microbial immigration rather than long-term colonization. Accounting for the lung microbiome dynamics is essential for understanding host-microbiota interactions and respiratory health.

The regulation of human hematopoietic stem cell (HSC) function within its native bone marrow microenvironment remains poorly understood due to the limitations of existing humanized models. Here, we present a non-conditioned human HSC xenotransplantation platform that serves as a physiologically relevant in vivo surrogate to study these complex cellular interactions. Using this system, we uncover a dynamic, cross-species communication between human HSCs and the murine niche, revealing a profound cellular and molecular plasticity of the bone marrow microenvironment in response to humanization. Upon engraftment, platelet-derived growth factor receptor alpha positive (Pdgfra ⁺ ) mesenchymal stromal cells (MSCs) undergo significant numerical expansion and a major transcriptional shift, transitioning from a mixed adipo- and osteo-primed state toward a leptin receptor positive (Lepr ⁺ ) predominantly adipo-primed, hematopoiesis-supporting signature. Functional validation demonstrates that this niche plasticity is a critical determinant of stem cell engraftment: the targeted depletion of Lepr ⁺ MSCs or the genetic deletion of stem cell factor results in the rapid mobilization and lack of engraftment of human HSCs, respectively. Our findings establish niche plasticity as a primary regulator of human HSC function and demonstrate the utility of this cross-species platform as a modular genetic toolbox for precise in vivo engineering of the bone marrow microenvironment. Ultimately, this adaptable system provides a powerful framework for elucidating human stem cell biology and advancing bone marrow transplantation therapies.

Invasive candidiasis is a fungal infection characterized by a high mortality rate. Carcinoembryonic antigen-related cell adhesion molecule (CEACAM) family receptors play a crucial role in regulating innate responses of both leukocytes and epithelia. Human CEACAM3, CEACAM5 and CEACAM6 receptors recognize Candida albicans and are expressed in transgenic CEABAC10 mice. In a murine C. albicans infection model, CEABAC10 mice exhibited a shortened survival period attributed to an early cytokine storm, an exacerbated acute phase response, and heightened systemic inflammation compared to their wild-type littermates. The livers and kidneys of CEABAC10 mice displayed intensified purulent necrotizing inflammation, accompanied by increased infiltration of neutrophils and macrophages. Our in vivo and in vitro data indicated that the expression of CEACAM6 on monocytes of CEABAC10 mice caused the elevated cytokine levels and the subsequent exacerbation of the acute phase response upon C. albicans infection, resulting in decreased survival.

The ability to detect host cell factors during Staphylococcus aureus infection in vitro by immunofluorescence microscopy is severely hampered by staphylococcal protein A (SpA), a cell wall-anchored protein that binds the fragment crystallizable (Fc) region of immunoglobulins. This interaction generates strong nonspecific fluorescent signals on the bacterial surface, complicating data interpretation and limiting the accuracy of quantitative image analysis. Several measures have been put forward to overcome this obstacle, most importantly the pre-incubation with an anti-SpA antibody (αSpA) and the use of human serum (HS) as blocking agent and antibody diluent. To highlight this feature to general fluorescence microscopy users, we here systematically evaluated these two strategies. Using S. aureus coated on coverslips and S. aureus -infected A549 cells, we highlight the efficiencies of both approaches to markedly reduce nonspecific fluorescence, with HS treatment yielding the most profound suppression. Notably, HS, containing high levels of human immunoglobulins, offered a robust, cost-effective and broadly applicable solution for minimizing SpA-driven artifacts, thereby improving immunofluorescence microscopy in S. aureus infection models in vitro .

Background: The Gram-positive human pathogen Streptococcus pneumoniae adapts its metabolism to the environment during colonization and host invasion. Extracellular vesicles (EVs) are produced by S. pneumoniae in the process of infection but the exact interplay between metabolic adaptation and vesicle formation remains elusive. This study investigates the role of environmental cues in modulating pneumococcal EVs biogenesis and function. Results: Here, we demonstrate that exposure to normal human serum induced rearrangement of the pneumococcal cell wall and considerably increased Sp-EVs production. Temperature and pH were critical factors for Sp-EVs formation: 37 °C supported optimal EV production, while bacterial exposure to either basic or acidic environments slowed down pneumococcal EV biogenesis and led to a heterogeneous subpopulation profile. Proteomic analysis revealed that Sp-EVs are enriched in carbon metabolism-related proteins, specifically those associated with glycolysis (e.g. Eno, GapA, GapN, GpmA, PfkA, PykF, and Tpi). Moderate glucose availability enhanced Sp-EVs production and intracellular ATP level, underlying a relation between metabolic status and EV biogenesis. Functionally, Sp-EVs promoted biofilm formation in both S. pneumoniae and Streptococcus pyogenes . Sp-EVs isolated under glucose-rich conditions enhanced S. pneumoniae biofilms, whereas Sp-EVs from glucose-poor conditions strongly stimulated S. pyogenes biofilm formation. Conclusions: These findings underscore the role of host and environmental cues in shaping pneumococcal EV production, composition, and function, highlighting their potential involvement in metabolic adaptation and interspecies interactions.

The human-pathogenic fungus Aspergillus fumigatus can cause life-threatening infections, particularly in immunosuppressed or neutropenic patients with hematological malignancies. The human antifungal immune response primarily relies on neutrophils with contribution from macrophages, while the roles of unconventional lymphocytes such as Natural Killer (NK) cells and mucosal-associated invariant T (MAIT) cells remain less understood. Therefore, using an advanced microfluidic “invasive aspergillosis-on-chip” (IAC) model, we investigated the coordinated antifungal response of various immune cell types in combination with macrophages within a human tissue-like 3D environment. The IAC model, initially incorporating monocyte-derived macrophages, was enhanced by perfusing either primary neutrophils, NK cells, or MAIT cells for overnight infection experiments. In this model, hyphal growth of A. fumigatus was prevented only when both macrophages and neutrophils were present. However, uncontrolled growth was detected in neutropenic patient scenarios. We observed key interactions such as conidial shuttling and metaforosis between innate immune cells, accompanied by elevated cytokine levels of e.g. IL-1β and TNFα. Importantly, NK cells applied to the vascular compartment of the IAC model significantly reduced hyphal length, confirming their antifungal activity. Similarly, MAIT cells in sufficiently high quantities reduced hyphal growth and branching, although this effect depended on the recognition of riboflavin-derived antigens, as shown by the lack of response to the A. fumigatus ΔriboB mutant strain. Together, our findings demonstrate that the advanced IAC model is a powerful tool to dissect human antifungal immunity. The interactions observed between macrophages and other immune cells highlight the importance of multi-layered defense strategies against A. fumigatus.