Defining and benchmarking open problems in single-cell analysis.

With the growing number of single-cell analysis tools, benchmarks are increasingly important to guide analysis and method development. However, a lack of standardisation and extensibility in current benchmarks limits their usability, longevity, and relevance to the community. We present Open Problems, a living, extensible, community-guided benchmarking platform including 10 current single-cell tasks that we envision will raise standards for the selection, evaluation, and development of methods in single-cell analysis.

Efficient cytometry analysis with FlowSOM in Python boosts interoperability with other single-cell tools.

MOTIVATION: We describe a new Python implementation of FlowSOM, a clustering method for cytometry data. RESULTS: This implementation is faster than the original version in R, better adapted to work with single-cell omics data including integration with current single-cell data structures and includes all the original visualizations, such as the star and pie plot. AVAILABILITY AND IMPLEMENTATION: The FlowSOM Python implementation is freely available on GitHub: https://github.com/saeyslab/FlowSOM_Python.

SpatialData: an open and universal data framework for spatial omics.

Spatially resolved omics technologies are transforming our understanding of biological tissues. However, the handling of uni- and multimodal spatial omics datasets remains a challenge owing to large data volumes, heterogeneity of data types and the lack of flexible, spatially aware data structures. Here we introduce SpatialData, a framework that establishes a unified and extensible multiplatform file-format, lazy representation of larger-than-memory data, transformations and alignment to common coordinate systems. SpatialData facilitates spatial annotations and cross-modal aggregation and analysis, the utility of which is illustrated in the context of multiple vignettes, including integrative analysis on a multimodal Xenium and Visium breast cancer study.

Uncertainty-aware single-cell annotation with a hierarchical reject option.

MOTIVATION: Automatic cell type annotation methods assign cell type labels to new datasets by extracting relationships from a reference RNA-seq dataset. However, due to the limited resolution of gene expression features, there is always uncertainty present in the label assignment. To enhance the reliability and robustness of annotation, most machine learning methods address this uncertainty by providing a full reject option, i.e. when the predicted confidence score of a cell type label falls below a user-defined threshold, no label is assigned and no prediction is made. As a better alternative, some methods deploy hierarchical models and consider a so-called partial rejection by returning internal nodes of the hierarchy as label assignment. However, because a detailed experimental analysis of various rejection approaches is missing in the literature, there is currently no consensus on best practices. RESULTS: We evaluate three annotation approaches (i) full rejection, (ii) partial rejection, and (iii) no rejection for both flat and hierarchical probabilistic classifiers. Our findings indicate that hierarchical classifiers are superior when rejection is applied, with partial rejection being the preferred rejection approach, as it preserves a significant amount of label information. For optimal rejection implementation, the rejection threshold should be determined through careful examination of a method's rejection behavior. Without rejection, flat and hierarchical annotation perform equally well, as long as the cell type hierarchy accurately captures transcriptomic relationships. AVAILABILITY AND IMPLEMENTATION: Code is freely available at https://github.com/Latheuni/Hierarchical_reject and https://doi.org/10.5281/zenodo.10697468.

Myelodysplastic neoplasms dissected into indolent, leukaemic and unfavourable subtypes by computational clustering of haematopoietic stem and progenitor cells.

Myelodysplastic neoplasms (MDS) encompass haematological malignancies, which are characterised by dysplasia, ineffective haematopoiesis and the risk of progression towards acute myeloid leukaemia (AML). Myelodysplastic neoplasms are notorious for their heterogeneity: clinical outcomes range from a near-normal life expectancy to leukaemic transformation or premature death due to cytopenia. The Molecular International Prognostic Scoring System made progress in the dissection of MDS by clinical outcomes. To contribute to the risk stratification of MDS by immunophenotypic profiles, this study performed computational clustering of flow cytometry data of CD34+ cells in 67 MDS, 67 AML patients and 49 controls. Our data revealed heterogeneity also within the MDS-derived CD34+ compartment. In MDS, maintenance of lymphoid progenitors and megakaryocytic-erythroid progenitors predicted favourable outcomes, whereas expansion of granulocyte-monocyte progenitors increased the risk of leukaemic transformation. The proliferation of haematopoietic stem cells and common myeloid progenitors with downregulated CD44 expression, suggestive of impaired haematopoietic differentiation, characterised a distinct MDS subtype with a poor overall survival. This exploratory study demonstrates the prognostic value of known and previously unexplored CD34+ populations and suggests the feasibility of dissecting MDS into a more indolent, a leukaemic and another unfavourable subtype.

A TCF4-dependent gene regulatory network confers resistance to immunotherapy in melanoma.

To better understand intrinsic resistance to immune checkpoint blockade (ICB), we established a comprehensive view of the cellular architecture of the treatment-naive melanoma ecosystem and studied its evolution under ICB. Using single-cell, spatial multi-omics, we showed that the tumor microenvironment promotes the emergence of a complex melanoma transcriptomic landscape. Melanoma cells harboring a mesenchymal-like (MES) state, a population known to confer resistance to targeted therapy, were significantly enriched in early on-treatment biopsies from non-responders to ICB. TCF4 serves as the hub of this landscape by being a master regulator of the MES signature and a suppressor of the melanocytic and antigen presentation transcriptional programs. Targeting TCF4 genetically or pharmacologically, using a bromodomain inhibitor, increased immunogenicity and sensitivity of MES cells to ICB and targeted therapy. We thereby uncovered a TCF4-dependent regulatory network that orchestrates multiple transcriptional programs and contributes to resistance to both targeted therapy and ICB in melanoma.

An Introduction to Adversarially Robust Deep Learning.

The widespread success of deep learning in solving machine learning problems has fueled its adoption in many fields, from speech recognition to drug discovery and medical imaging. However, deep learning systems are extremely fragile: imperceptibly small modifications to their input data can cause the models to produce erroneous output. It is very easy to generate such adversarial perturbations even for state-of-the-art models, yet immunization against them has proven exceptionally challenging. Despite over a decade of research on this problem, our solutions are still far from satisfactory and many open problems remain. In this work, we survey some of the most important contributions in the field of adversarial robustness. We pay particular attention to the reasons why past attempts at improving robustness have been insufficient, and we identify several promising areas for future research.

A bispecific Clec9A-PD-L1 targeted type I interferon profoundly reshapes the tumor microenvironment towards an antitumor state.

Despite major improvements in immunotherapeutic strategies, the immunosuppressive tumor microenvironment remains a major obstacle for the induction of efficient antitumor responses. In this study, we show that local delivery of a bispecific Clec9A-PD-L1 targeted type I interferon (AcTaferon, AFN) overcomes this hurdle by reshaping the tumor immune landscape.Treatment with the bispecific AFN resulted in the presence of pro-immunogenic tumor-associated macrophages and neutrophils, increased motility and maturation profile of cDC1 and presence of inflammatory cDC2. Moreover, we report empowered diversity in the CD8+ T cell repertoire and induction of a shift from naive, dysfunctional CD8+ T cells towards effector, plastic cytotoxic T lymphocytes together with increased presence of NK and NKT cells as well as decreased regulatory T cell levels. These dynamic changes were associated with potent antitumor activity. Tumor clearance and immunological memory, therapeutic immunity on large established tumors and blunted tumor growth at distant sites were obtained upon co-administration of a non-curative dose of chemotherapy.Overall, this study illuminates further application of type I interferon as a safe and efficient way to reshape the suppressive tumor microenvironment and induce potent antitumor immunity; features which are of major importance in overcoming the development of metastases and tumor cell resistance to immune attack. The strategy described here has potential for application across to a broad range of cancer types.

A robust pipeline for high-content, high-throughput immunophenotyping reveals age- and genetics-dependent changes in blood leukocytes.

High-dimensional flow cytometry is the gold standard to study the human immune system in large cohorts. However, large sample sizes increase inter-experimental variation because of technical and experimental inaccuracies introduced by batch variability. Our high-throughput sample processing pipeline in combination with 28-color flow cytometry focuses on increased throughput (192 samples/experiment) and high reproducibility. We implemented quality control checkpoints to reduce technical and experimental variation. Finally, we integrated FlowSOM clustering to facilitate automated data analysis and demonstrate the reproducibility of our pipeline in a study with 3,357 samples. We reveal age-associated immune dynamics in 2,300 individuals, signified by decreasing T and B cell subsets with age. In addition, by combining genetic analyses, our approach revealed unique immune signatures associated with a single nucleotide polymorphism (SNP) that abrogates CD45 isoform splicing. In summary, we provide a versatile and reliable high-throughput, flow cytometry-based pipeline for immune discovery and exploration in large cohorts.

Functions of the RIP kinase family members in the skin.

The receptor interacting protein kinases (RIPK) are a family of serine/threonine kinases that are involved in the integration of various stress signals. In response to several extracellular and/or intracellular stimuli, RIP kinases engage signaling cascades leading to the activation of NF-κB and mitogen-activated protein kinases, cell death, inflammation, differentiation and Wnt signaling and can have kinase-dependent and kinase-independent functions. Although it was previously suggested that seven RIPKs are part of the RIPK family, phylogenetic analysis indicates that there are only five genuine RIPKs. RIPK1 and RIPK3 are mainly involved in controlling and executing necroptosis in keratinocytes, while RIPK4 controls proliferation and differentiation of keratinocytes and thereby can act as a tumor suppressor in skin. Therefore, in this review we summarize and discuss the functions of RIPKs in skin homeostasis as well as the signaling pathways involved.

Adjuvant dendritic cell-based immunotherapy after cytoreductive surgery and hyperthermic intraperitoneal chemotherapy in patients with malignant peritoneal mesothelioma: a phase II clinical trial.

BACKGROUND: Malignant peritoneal mesothelioma (MPM) is an aggressive malignancy with a poor prognosis. Cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) improves survival outcomes, but recurrence rates remain high. Dendritic cell-based immunotherapy (DCBI) showed promising results in patients with pleural mesothelioma. The primary aim of this trial was to determine feasibility of adjuvant DCBI after CRS-HIPEC. METHODS: This open-label, single-center, phase II clinical trial, performed in the Erasmus MC Cancer Institute Rotterdam, the Netherlands, included patients with epithelioid MPM. 4-6 weeks before CRS-HIPEC leukapheresis was performed. 8-10 weeks after surgery, DCBI was administered three times biweekly. Feasibility was defined as administration of at least three adjuvant vaccinations in 75% of patients. Comprehensive immune cell profiling was performed on peripheral blood samples prior to and during treatment. RESULTS: All patients who received CRS-HIPEC (n=16) were successfully treated with adjuvant DCBI. No severe toxicity related to DCBI was observed. Median progression-free survival (PFS) was 12 months (IQR 5-23) and median overall survival was not reached. DCBI was associated with increased proliferation of circulating natural killer cells and CD4+ T-helper (Th) cells. Co-stimulatory molecules, including ICOS, HLA-DR, and CD28 were upregulated predominantly on memory or proliferating Th-cells and minimally on CD8+ cytotoxic T-lymphocytes (CTLs) after treatment. However, an increase in CD8+ terminally differentiated effector memory (Temra) cells positively correlated with PFS, whereas co-expression of ICOS and Ki67 on CTLs trended towards a positive correlation. CONCLUSIONS: Adjuvant DCBI after CRS-HIPEC in patients with MPM was feasible and safe, and showed promising survival outcomes. DCBI had an immune modulatory effect on lymphoid cells and induced memory T-cell activation. Moreover, an increase of CD8+ Temra cells was more pronounced in patients with longer PFS. These data provide rationale for future combination treatment strategies. TRIAL REGISTRATION NUMBER: NTR7060; Dutch Trial Register (NTR).

Ferroptosis and pyroptosis signatures in critical COVID-19 patients.

Critical COVID-19 patients admitted to the intensive care unit (ICU) frequently suffer from severe multiple organ dysfunction with underlying widespread cell death. Ferroptosis and pyroptosis are two detrimental forms of regulated cell death that could constitute new therapeutic targets. We enrolled 120 critical COVID-19 patients in a two-center prospective cohort study to monitor systemic markers of ferroptosis, iron dyshomeostasis, pyroptosis, pneumocyte cell death and cell damage on the first three consecutive days after ICU admission. Plasma of 20 post-operative ICU patients (PO) and 39 healthy controls (HC) without organ failure served as controls. Subsets of COVID-19 patients displayed increases in individual biomarkers compared to controls. Unsupervised clustering was used to discern latent clusters of COVID-19 patients based on biomarker profiles. Pyroptosis-related interleukin-18 accompanied by high pneumocyte cell death was independently associated with higher odds at mechanical ventilation, while the subgroup with high interleuking-1 beta (but limited pneumocyte cell death) displayed reduced odds at mechanical ventilation and lower mortality hazard. Meanwhile, iron dyshomeostasis with a tendency towards higher ferroptosis marker malondialdehyde had no association with outcome, except for the small subset of patients with very high catalytic iron independently associated with reduced survival. Forty percent of patients did not have a clear signature of the cell death mechanisms studied in this cohort. Moreover, repeated moderate levels of soluble receptor of advanced glycation end products and growth differentiation factor 15 during the first three days after ICU admission are independently associated with adverse clinical outcome compared to sustained lower levels. Altogether, the data point towards distinct subgroups in this cohort of critical COVID-19 patients with different systemic signatures of pyroptosis, iron dyshomeostasis, ferroptosis or pneumocyte cell death markers that have different outcomes in ICU. The distinct groups may allow 'personalized' treatment allocation in critical COVID-19 based on systemic biomarker profiles.

Cell type-specific attenuation of brassinosteroid signaling precedes stomatal asymmetric cell division.

In Arabidopsis thaliana, brassinosteroid (BR) signaling and stomatal development are connected through the SHAGGY/GSK3-like kinase BR INSENSITIVE2 (BIN2). BIN2 is a key negative regulator of BR signaling but it plays a dual role in stomatal development. BIN2 promotes or restricts stomatal asymmetric cell division (ACD) depending on its subcellular localization, which is regulated by the stomatal lineage-specific scaffold protein POLAR. BRs inactivate BIN2, but how they govern stomatal development remains unclear. Mapping the single-cell transcriptome of stomatal lineages after triggering BR signaling with either exogenous BRs or the specific BIN2 inhibitor, bikinin, revealed that the two modes of BR signaling activation generate spatiotemporally distinct transcriptional responses. We established that BIN2 is always sensitive to the inhibitor but, when in a complex with POLAR and its closest homolog POLAR-LIKE1, it becomes protected from BR-mediated inactivation. Subsequently, BR signaling in ACD precursors is attenuated, while it remains active in epidermal cells devoid of scaffolds and undergoing differentiation. Our study demonstrates how scaffold proteins contribute to cellular signal specificity of hormonal responses in plants.

Machine Learning Pipeline for Predicting Bone Marrow Edema Along the Sacroiliac Joints on Magnetic Resonance Imaging.

OBJECTIVE: We aimed to develop and validate a fully automated machine learning (ML) algorithm that predicts bone marrow edema (BME) on a quadrant level in sacroiliac (SI) joint magnetic resonance imaging (MRI). METHODS: A computer vision workflow automatically locates the SI joints, segments regions of interest (ilium and sacrum), performs objective quadrant extraction, and predicts presence of BME, suggestive of inflammatory lesions, on a quadrant level in semicoronal slices of T1/T2-weighted MRI scans. Ground truth was determined by consensus among human readers. The inflammation classifier was trained using a ResNet18 backbone and five-fold cross-validated on scans of patients with spondyloarthritis (SpA) (n = 279), postpartum individuals (n = 71), and healthy subjects (n = 114). Independent SpA patient MRI scans (n = 243) served as test data set. Patient-level predictions were derived from aggregating quadrant-level predictions, ie, at least one positive quadrant. RESULTS: The algorithm automatically detects the SI joints with a precision of 98.4% and segments ilium/sacrum with an intersection over union of 85.6% and 67.9%, respectively. The inflammation classifier performed well in cross-validation: area under the curve (AUC) 94.5%, balanced accuracy (B-ACC) 80.5%, and F1 score 64.1%. In the test data set, AUC was 88.2%, B-ACC 72.1%, and F1 score 50.8%. On a patient level, the model achieved a B-ACC of 81.6% and 81.4% in the cross-validation and test data set, respectively. CONCLUSION: We propose a fully automated ML pipeline that enables objective and standardized evaluation of BME along the SI joints on MRI. This method has the potential to screen large numbers of patients with (suspected) SpA and is a step closer towards artificial intelligence-assisted diagnosis and follow-up.

Gut Inflammation in Axial Spondyloarthritis Patients is Characterized by a Marked Type 17 Skewed Mucosal Innate-like T Cell Signature.

OBJECTIVE: Patients with spondyloarthritis (SpA) often present with microscopic signs of gut inflammation, a risk factor for progressive disease. We investigated whether mucosal innate-like T cells are involved in dysregulated interleukin-23 (IL-23)/IL-17 responses in the gut-joint axis in SpA. METHODS: Ileal and colonic intraepithelial lymphocytes (IELs), lamina propria lymphocytes (LPLs), and paired peripheral blood mononuclear cells (PBMCs) were isolated from treatment-naive patients with nonradiographic axial SpA with (n = 11) and without (n = 14) microscopic gut inflammation and healthy controls (n = 15) undergoing ileocolonoscopy. The presence of gut inflammation was assessed histopathologically. Immunophenotyping of innate-like T cells and conventional T cells was performed using intracellular flow cytometry. Unsupervised clustering analysis was done by FlowSOM technology. Serum IL-17A levels were measured via Luminex. RESULTS: Microscopic gut inflammation in nonradiographic axial SpA was characterized by increased ileal intraepithelial γδ-hi T cells, a γδ-T cell subset with elevated γδ-T cell receptor expression. γδ-hi T cells were also increased in PBMCs of patients with nonradiographic axial SpA versus healthy controls and were strongly associated with Ankylosing Spondylitis Disease Activity Score. The abundance of mucosal-associated invariant T cells and invariant natural killer T cells was unaltered. Innate-like T cells in the inflamed gut showed increased RORγt, IL-17A, and IL-22 levels with loss of T-bet, a signature that was less pronounced in conventional T cells. Presence of gut inflammation was associated with higher serum IL-17A levels. In patients treated with tumor necrosis factor blockade, the proportion of γδ-hi cells and RORγt expression in blood was completely restored. CONCLUSION: Intestinal innate-like T cells display marked type 17 skewing in the inflamed gut mucosa of patients with nonradiographic axial SpA. γδ-hi T cells are linked to intestinal inflammation and disease activity in SpA.

QT-GWAS: A novel method for unveiling biosynthetic loci affecting qualitative metabolic traits.

Although the plant kingdom provides an enormous diversity of metabolites with potentially beneficial applications for humankind, a large fraction of these metabolites and their biosynthetic pathways remain unknown. Resolving metabolite structures and their biosynthetic pathways is key to gaining biological understanding and to allow metabolic engineering. In order to retrieve novel biosynthetic genes involved in specialized metabolism, we developed a novel untargeted method designated as qualitative trait GWAS (QT-GWAS) that subjects qualitative metabolic traits to a genome-wide association study, while the conventional metabolite GWAS (mGWAS) mainly considers the quantitative variation of metabolites. As a proof of the validity of QT-GWAS, 23 and 15 of the retrieved associations identified in Arabidopsis thaliana by QT-GWAS and mGWAS, respectively, were supported by previous research. Furthermore, seven gene-metabolite associations retrieved by QT-GWAS were confirmed in this study through reverse genetics combined with metabolomics and/or in vitro enzyme assays. As such, we established that CYTOCHROME P450 706A5 (CYP706A5) is involved in the biosynthesis of chroman derivatives, UDP-GLYCOSYLTRANSFERASE 76C3 (UGT76C3) is able to hexosylate guanine in vitro and in planta, and SULFOTRANSFERASE 202B1 (SULT202B1) catalyzes the sulfation of neolignans in vitro. Collectively, our study demonstrates that the untargeted QT-GWAS method can retrieve valid gene-metabolite associations at the level of enzyme-encoding genes, even new associations that cannot be found by the conventional mGWAS, providing a new approach for dissecting qualitative metabolic traits.

A redundant transcription factor network steers spatiotemporal Arabidopsis triterpene synthesis.

Plant specialized metabolites modulate developmental and ecological functions and comprise many therapeutic and other high-value compounds. However, the mechanisms determining their cell-specific expression remain unknown. Here we describe the transcriptional regulatory network that underlies cell-specific biosynthesis of triterpenes in Arabidopsis thaliana root tips. Expression of thalianol and marneral biosynthesis pathway genes depends on the phytohormone jasmonate and is limited to outer tissues. We show that this is promoted by the activity of redundant bHLH-type transcription factors from two distinct clades and coactivated by homeodomain factors. Conversely, the DOF-type transcription factor DAG1 and other regulators prevent expression of the triterpene pathway genes in inner tissues. We thus show how precise expression of triterpene biosynthesis genes is determined by a robust network of transactivators, coactivators and counteracting repressors.

Single-cell molecular profiling using ex vivo functional readouts fuels precision oncology in glioblastoma.

BACKGROUND: Functional profiling of freshly isolated glioblastoma (GBM) cells is being evaluated as a next-generation method for precision oncology. While promising, its success largely depends on the method to evaluate treatment activity which requires sufficient resolution and specificity. METHODS: Here, we describe the 'precision oncology by single-cell profiling using ex vivo readouts of functionality' (PROSPERO) assay to evaluate the intrinsic susceptibility of high-grade brain tumor cells to respond to therapy. Different from other assays, PROSPERO extends beyond life/death screening by rapidly evaluating acute molecular drug responses at single-cell resolution. RESULTS: The PROSPERO assay was developed by correlating short-term single-cell molecular signatures using mass cytometry by time-of-flight (CyTOF) to long-term cytotoxicity readouts in representative patient-derived glioblastoma cell cultures (n = 14) that were exposed to radiotherapy and the small-molecule p53/MDM2 inhibitor AMG232. The predictive model was subsequently projected to evaluate drug activity in freshly resected GBM samples from patients (n = 34). Here, PROSPERO revealed an overall limited capacity of tumor cells to respond to therapy, as reflected by the inability to induce key molecular markers upon ex vivo treatment exposure, while retaining proliferative capacity, insights that were validated in patient-derived xenograft (PDX) models. This approach also allowed the investigation of cellular plasticity, which in PDCLs highlighted therapy-induced proneural-to-mesenchymal (PMT) transitions, while in patients' samples this was more heterogeneous. CONCLUSION: PROSPERO provides a precise way to evaluate therapy efficacy by measuring molecular drug responses using specific biomarker changes in freshly resected brain tumor samples, in addition to providing key functional insights in cellular behavior, which may ultimately complement standard, clinical biomarker evaluations.

LXR signaling controls homeostatic dendritic cell maturation.

Dendritic cells (DCs) mature in an immunogenic or tolerogenic manner depending on the context in which an antigen is perceived, preserving the balance between immunity and tolerance. Whereas the pathways driving immunogenic maturation in response to infectious insults are well-characterized, the signals that drive tolerogenic maturation during homeostasis are still poorly understood. We found that the engulfment of apoptotic cells triggered homeostatic maturation of type 1 conventional DCs (cDC1s) within the spleen. This maturation process could be mimicked by engulfment of empty, nonadjuvanted lipid nanoparticles (LNPs), was marked by intracellular accumulation of cholesterol, and was highly specific to cDC1s. Engulfment of either apoptotic cells or cholesterol-rich LNPs led to the activation of the liver X receptor (LXR) pathway, which promotes the efflux of cellular cholesterol, and repressed genes associated with immunogenic maturation. In contrast, simultaneous engagement of TLR3 to mimic viral infection via administration of poly(I:C)-adjuvanted LNPs repressed the LXR pathway, thus delaying cellular cholesterol efflux and inducing genes that promote T cell-mediated immunity. These data demonstrate that conserved cellular cholesterol efflux pathways are differentially regulated in tolerogenic versus immunogenic cDC1s and suggest that administration of nonadjuvanted cholesterol-rich LNPs may be an approach for inducing tolerogenic DC maturation.

Meta-analysis of (single-cell method) benchmarks reveals the need for extensibility and interoperability.

Computational methods represent the lifeblood of modern molecular biology. Benchmarking is important for all methods, but with a focus here on computational methods, benchmarking is critical to dissect important steps of analysis pipelines, formally assess performance across common situations as well as edge cases, and ultimately guide users on what tools to use. Benchmarking can also be important for community building and advancing methods in a principled way. We conducted a meta-analysis of recent single-cell benchmarks to summarize the scope, extensibility, and neutrality, as well as technical features and whether best practices in open data and reproducible research were followed. The results highlight that while benchmarks often make code available and are in principle reproducible, they remain difficult to extend, for example, as new methods and new ways to assess methods emerge. In addition, embracing containerization and workflow systems would enhance reusability of intermediate benchmarking results, thus also driving wider adoption.