OpenFold: retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization.

AlphaFold2 revolutionized structural biology with the ability to predict protein structures with exceptionally high accuracy. Its implementation, however, lacks the code and data required to train new models. These are necessary to (1) tackle new tasks, like protein-ligand complex structure prediction, (2) investigate the process by which the model learns and (3) assess the model's capacity to generalize to unseen regions of fold space. Here we report OpenFold, a fast, memory efficient and trainable implementation of AlphaFold2. We train OpenFold from scratch, matching the accuracy of AlphaFold2. Having established parity, we find that OpenFold is remarkably robust at generalizing even when the size and diversity of its training set is deliberately limited, including near-complete elisions of classes of secondary structure elements. By analyzing intermediate structures produced during training, we also gain insights into the hierarchical manner in which OpenFold learns to fold. In sum, our studies demonstrate the power and utility of OpenFold, which we believe will prove to be a crucial resource for the protein modeling community.

SCIMAP: A Python Toolkit for Integrated Spatial Analysis of Multiplexed Imaging Data.

Multiplexed imaging data are revolutionizing our understanding of the composition and organization of tissues and tumors. A critical aspect of such tissue profiling is quantifying the spatial relationship relationships among cells at different scales from the interaction of neighboring cells to recurrent communities of cells of multiple types. This often involves statistical analysis of 10^7 or more cells in which up to 100 biomolecules (commonly proteins) have been measured. While software tools currently cater to the analysis of spatial transcriptomics data, there remains a need for toolkits explicitly tailored to the complexities of multiplexed imaging data including the need to seamlessly integrate image visualization with data analysis and exploration. We introduce SCIMAP, a Python package specifically crafted to address these challenges. With SCIMAP, users can efficiently preprocess, analyze, and visualize large datasets, facilitating the exploration of spatial relationships and their statistical significance. SCIMAP's modular design enables the integration of new algorithms, enhancing its capabilities for spatial analysis.

SpatialCells: automated profiling of tumor microenvironments with spatially resolved multiplexed single-cell data.

Cancer is a complex cellular ecosystem where malignant cells coexist and interact with immune, stromal and other cells within the tumor microenvironment (TME). Recent technological advancements in spatially resolved multiplexed imaging at single-cell resolution have led to the generation of large-scale and high-dimensional datasets from biological specimens. This underscores the necessity for automated methodologies that can effectively characterize molecular, cellular and spatial properties of TMEs for various malignancies. This study introduces SpatialCells, an open-source software package designed for region-based exploratory analysis and comprehensive characterization of TMEs using multiplexed single-cell data. The source code and tutorials are available at https://semenovlab.github.io/SpatialCells. SpatialCells efficiently streamlines the automated extraction of features from multiplexed single-cell data and can process samples containing millions of cells. Thus, SpatialCells facilitates subsequent association analyses and machine learning predictions, making it an essential tool in advancing our understanding of tumor growth, invasion and metastasis.

A general algorithm for consensus 3D cell segmentation from 2D segmented stacks.

Cell segmentation is the fundamental task. Only by segmenting, can we define the quantitative spatial unit for collecting measurements to draw biological conclusions. Deep learning has revolutionized 2D cell segmentation, enabling generalized solutions across cell types and imaging modalities. This has been driven by the ease of scaling up image acquisition, annotation and computation. However 3D cell segmentation, which requires dense annotation of 2D slices still poses significant challenges. Labelling every cell in every 2D slice is prohibitive. Moreover it is ambiguous, necessitating cross-referencing with other orthoviews. Lastly, there is limited ability to unambiguously record and visualize 1000s of annotated cells. Here we develop a theory and toolbox, u-Segment3D for 2D-to-3D segmentation, compatible with any 2D segmentation method. Given optimal 2D segmentations, u-Segment3D generates the optimal 3D segmentation without data training, as demonstrated on 11 real life datasets, >70,000 cells, spanning single cells, cell aggregates and tissue.

A CRISPRi/a screening platform to study cellular nutrient transport in diverse microenvironments.

Blocking the import of nutrients essential for cancer cell proliferation represents a therapeutic opportunity, but it is unclear which transporters to target. Here we report a CRISPR interference/activation screening platform to systematically interrogate the contribution of nutrient transporters to support cancer cell proliferation in environments ranging from standard culture media to tumours. We applied this platform to identify the transporters of amino acids in leukaemia cells and found that amino acid transport involves high bidirectional flux dependent on the microenvironment composition. While investigating the role of transporters in cystine starved cells, we uncovered a role for serotonin uptake in preventing ferroptosis. Finally, we identified transporters essential for cell proliferation in subcutaneous tumours and found that levels of glucose and amino acids can restrain proliferation in that environment. This study establishes a framework for systematically identifying critical cellular nutrient transporters, characterizing their function and exploring how the tumour microenvironment impacts cancer metabolism.

Chemotherapy induces myeloid-driven spatial T-cell exhaustion in ovarian cancer.

To uncover the intricate, chemotherapy-induced spatiotemporal remodeling of the tumor microenvironment, we conducted integrative spatial and molecular characterization of 97 high-grade serous ovarian cancer (HGSC) samples collected before and after chemotherapy. Using single-cell and spatial analyses, we identify increasingly versatile immune cell states, which form spatiotemporally dynamic microcommunities at the tumor-stroma interface. We demonstrate that chemotherapy triggers spatial redistribution and exhaustion of CD8+ T cells due to prolonged antigen presentation by macrophages, both within interconnected myeloid networks termed "Myelonets" and at the tumor stroma interface. Single-cell and spatial transcriptomics identifies prominent TIGIT-NECTIN2 ligand-receptor interactions induced by chemotherapy. Using a functional patient-derived immuno-oncology platform, we show that CD8+T-cell activity can be boosted by combining immune checkpoint blockade with chemotherapy. Our discovery of chemotherapy-induced myeloid-driven spatial T-cell exhaustion paves the way for novel immunotherapeutic strategies to unleash CD8+ T-cell-mediated anti-tumor immunity in HGSC.

Genomic and Immunophenotypic Landscape of Acquired Resistance to PD-(L)1 Blockade in Non-Small-Cell Lung Cancer.

PURPOSE: Although immune checkpoint inhibitors (ICI) have extended survival in patients with non-small-cell lung cancer (NSCLC), acquired resistance (AR) to ICI frequently develops after an initial benefit. However, the mechanisms of AR to ICI in NSCLC are largely unknown. METHODS: Comprehensive tumor genomic profiling, machine learning-based assessment of tumor-infiltrating lymphocytes, multiplexed immunofluorescence, and/or HLA-I immunohistochemistry (IHC) were performed on matched pre- and post-ICI tumor biopsies from patients with NSCLC treated with ICI at the Dana-Farber Cancer Institute who developed AR to ICI. Two additional cohorts of patients with intervening chemotherapy or targeted therapies between biopsies were included as controls. RESULTS: We performed comprehensive genomic profiling and immunophenotypic characterization on samples from 82 patients with NSCLC and matched pre- and post-ICI biopsies and compared findings with a control cohort of patients with non-ICI intervening therapies between biopsies (chemotherapy, N = 32; targeted therapies, N = 89; both, N = 17). Putative resistance mutations were identified in 27.8% of immunotherapy-treated cases and included acquired loss-of-function mutations in STK11, B2M, APC, MTOR, KEAP1, and JAK1/2; these acquired alterations were not observed in the control groups. Immunophenotyping of matched pre- and post-ICI samples demonstrated significant decreases in intratumoral lymphocytes, CD3e+ and CD8a+ T cells, and PD-L1-PD1 engagement, as well as increased distance between tumor cells and CD8+PD-1+ T cells. There was a significant decrease in HLA class I expression in the immunotherapy cohort at the time of AR compared with the chemotherapy (P = .005) and the targeted therapy (P = .01) cohorts. CONCLUSION: These findings highlight the genomic and immunophenotypic heterogeneity of ICI resistance in NSCLC, which will need to be considered when developing novel therapeutic strategies aimed at overcoming resistance.

Cell state dependent effects of Bmal1 on melanoma immunity and tumorigenicity.

The circadian clock regulator Bmal1 modulates tumorigenesis, but its reported effects are inconsistent. Here, we show that Bmal1 has a context-dependent role in mouse melanoma tumor growth. Loss of Bmal1 in YUMM2.1 or B16-F10 melanoma cells eliminates clock function and diminishes hypoxic gene expression and tumorigenesis, which could be rescued by ectopic expression of HIF1α in YUMM2.1 cells. By contrast, over-expressed wild-type or a transcriptionally inactive mutant Bmal1 non-canonically sequester myosin heavy chain 9 (Myh9) to increase MRTF-SRF activity and AP-1 transcriptional signature, and shift YUMM2.1 cells from a Sox10high to a Sox9high immune resistant, mesenchymal cell state that is found in human melanomas. Our work describes a link between Bmal1, Myh9, mouse melanoma cell plasticity, and tumor immunity. This connection may underlie cancer therapeutic resistance and underpin the link between the circadian clock, MRTF-SRF and the cytoskeleton.

Phase II Study of Eribulin plus Pembrolizumab in Metastatic Soft-tissue Sarcomas: Clinical Outcomes and Biological Correlates.

PURPOSE: Eribulin modulates the tumor-immune microenvironment via cGAS-STING signaling in preclinical models. This non-randomized phase II trial evaluated the combination of eribulin and pembrolizumab in patients with soft-tissue sarcomas (STS). PATIENTS AND METHODS: Patients enrolled in one of three cohorts: leiomyosarcoma (LMS), liposarcomas (LPS), or other STS that may benefit from PD-1 inhibitors, including undifferentiated pleomorphic sarcoma (UPS). Eribulin was administered at 1.4 mg/m2 i.v. (days 1 and 8) with fixed-dose pembrolizumab 200 mg i.v. (day 1) of each 21-day cycle, until progression, unacceptable toxicity, or completion of 2 years of treatment. The primary endpoint was the 12-week progression-free survival rate (PFS-12) in each cohort. Secondary endpoints included the objective response rate, median PFS, safety profile, and overall survival (OS). Pretreatment and on-treatment blood specimens were evaluated in patients who achieved durable disease control (DDC) or progression within 12 weeks [early progression (EP)]. Multiplexed immunofluorescence was performed on archival LPS samples from patients with DDC or EP. RESULTS: Fifty-seven patients enrolled (LMS, n = 19; LPS, n = 20; UPS/Other, n = 18). The PFS-12 was 36.8% (90% confidence interval: 22.5-60.4) for LMS, 69.6% (54.5-89.0) for LPS, and 52.6% (36.8-75.3) for UPS/Other cohorts. All 3 patients in the UPS/Other cohort with angiosarcoma achieved RECIST responses. Toxicity was manageable. Higher IFNα and IL4 serum levels were associated with clinical benefit. Immune aggregates expressing PD-1 and PD-L1 were observed in a patient that completed 2 years of treatment. CONCLUSIONS: The combination of eribulin and pembrolizumab demonstrated promising activity in LPS and angiosarcoma.

Illuminating function of the understudied druggable kinome.

The human kinome, with more than 500 proteins, is crucial for cell signaling and disease. Yet, about one-third of kinases lack in-depth study. The Data and Resource Generating Center for Understudied Kinases has developed multiple resources to address this challenge including creation of a heavy amino acid peptide library for parallel reaction monitoring and quantitation of protein kinase expression, use of understudied kinases tagged with a miniTurbo-biotin ligase to determine interaction networks by proximity-dependent protein biotinylation, NanoBRET probe development for screening chemical tool target specificity in live cells, characterization of small molecule chemical tools inhibiting understudied kinases, and computational tools for defining kinome architecture. These resources are available through the Dark Kinase Knowledgebase, supporting further research into these understudied protein kinases.

Loss of Pip4k2c confers liver-metastatic organotropism through insulin-dependent PI3K-AKT pathway activation.

Liver metastasis (LM) confers poor survival and therapy resistance across cancer types, but the mechanisms of liver-metastatic organotropism remain unknown. Here, through in vivo CRISPR-Cas9 screens, we found that Pip4k2c loss conferred LM but had no impact on lung metastasis or primary tumor growth. Pip4k2c-deficient cells were hypersensitized to insulin-mediated PI3K/AKT signaling and exploited the insulin-rich liver milieu for organ-specific metastasis. We observed concordant changes in PIP4K2C expression and distinct metabolic changes in 3,511 patient melanomas, including primary tumors, LMs and lung metastases. We found that systemic PI3K inhibition exacerbated LM burden in mice injected with Pip4k2c-deficient cancer cells through host-mediated increase in hepatic insulin levels; however, this circuit could be broken by concurrent administration of an SGLT2 inhibitor or feeding of a ketogenic diet. Thus, this work demonstrates a rare example of metastatic organotropism through co-optation of physiological metabolic cues and proposes therapeutic avenues to counteract these mechanisms.

Qualification of a multiplexed tissue imaging assay and detection of novel patterns of HER2 heterogeneity in breast cancer.

Emerging data suggests that HER2 intratumoral heterogeneity (ITH) is associated with therapy resistance, highlighting the need for new strategies to assess HER2 ITH. A promising approach is leveraging multiplexed tissue analysis techniques such as cyclic immunofluorescence (CyCIF), which enable visualization and quantification of 10-60 antigens at single-cell resolution from individual tissue sections. In this study, we qualified a breast cancer-specific antibody panel, including HER2, ER, and PR, for multiplexed tissue imaging. We then compared the performance of these antibodies against established clinical standards using pixel-, cell- and tissue-level analyses, utilizing 866 tissue cores (representing 294 patients). To ensure reliability, the CyCIF antibodies were qualified against HER2 immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH) data from the same samples. Our findings demonstrate the successful qualification of a breast cancer antibody panel for CyCIF, showing high concordance with established clinical antibodies. Subsequently, we employed the qualified antibodies, along with antibodies for CD45, CD68, PD-L1, p53, Ki67, pRB, and AR, to characterize 567 HER2+ invasive breast cancer samples from 189 patients. Through single-cell analysis, we identified four distinct cell clusters within HER2+ breast cancer exhibiting heterogeneous HER2 expression. Furthermore, these clusters displayed variations in ER, PR, p53, AR, and PD-L1 expression. To quantify the extent of heterogeneity, we calculated heterogeneity scores based on the diversity among these clusters. Our analysis revealed expression patterns that are relevant to breast cancer biology, with correlations to HER2 ITH and potential relevance to clinical outcomes.

Systematic characterization of antibody-drug conjugate targets in central nervous system tumors.

BACKGROUND: Antibody-drug conjugates (ADCs) enhance the specificity of cytotoxic drugs by directing them to cells expressing target antigens. Multiple ADCs are FDA-approved for solid and hematologic malignancies, including those expressing HER2, TROP2, and NECTIN4. Recently, an ADC targeting HER2 (Trastuzumab-Deruxtecan) increased survival and reduced growth of brain metastases in treatment-refractory metastatic breast cancer, even in tumors with low HER2 expression. Thus, low-level expression of ADC targets may be sufficient for treatment responsiveness. However, ADC target expression is poorly characterized in many central nervous system (CNS) tumors. METHODS: We analyzed publicly available RNA-sequencing and proteomic data from the children's brain tumor network (N = 188 tumors) and gene-expression-omnibus RNA-expression datasets (N = 356) to evaluate expression of 14 potential ADC targets that are FDA-approved or under investigation in solid cancers. We also used immunohistochemistry to measure the levels of HER2, HER3, NECTIN4, TROP2, CLDN6, CLDN18.2, and CD276/B7-H3 protein in glioblastoma, oligodendroglioma, meningioma, ependymoma, pilocytic astrocytoma, medulloblastoma, atypical teratoid/rhabdoid tumor (AT/RT), adamantinomatous craniopharyngioma (ACP), papillary craniopharyngioma (PCP), and primary CNS lymphoma (N = 575). RESULTS: Pan-CNS analysis showed subtype-specific expression of ADC target proteins. Most tumors expressed HER3, B7-H3, and NECTIN4. Ependymomas strongly expressed HER2, while meningiomas showed weak-moderate HER2 expression. ACP and PCP strongly expressed B7-H3, with TROP2 expression in whorled ACP epithelium. AT/RT strongly expressed CLDN6. Glioblastoma showed little subtype-specific marker expression, suggesting a need for further target development. CONCLUSIONS: CNS tumors exhibit subtype-specific expression of ADC targets including several FDA-approved for other indications. Clinical trials of ADCs in CNS tumors may therefore be warranted.

Residency Octree: A Hybrid Approach for Scalable Web-Based Multi-Volume Rendering.

We present a hybrid multi-volume rendering approach based on a novel Residency Octree that combines the advantages of out-of-core volume rendering using page tables with those of standard octrees. Octree approaches work by performing hierarchical tree traversal. However, in octree volume rendering, tree traversal and the selection of data resolution are intrinsically coupled. This makes fine-grained empty-space skipping costly. Page tables, on the other hand, allow access to any cached brick from any resolution. However, they do not offer a clear and efficient strategy for substituting missing high-resolution data with lower-resolution data. We enable flexible mixed-resolution out-of-core multi-volume rendering by decoupling the cache residency of multi-resolution data from a resolution-independent spatial subdivision determined by the tree. Instead of one-to-one node-to-brick correspondences, each residency octree node is mapped to a set of bricks from different resolution levels. This makes it possible to efficiently and adaptively choose and mix resolutions, adapt sampling rates, and compensate for cache misses. At the same time, residency octrees support fine-grained empty-space skipping, independent of the data subdivision used for caching. Finally, to facilitate collaboration and outreach, and to eliminate local data storage, our implementation is a web-based, pure client-side renderer using WebGPU and WebAssembly. Our method is faster than prior approaches and efficient for many data channels with a flexible and adaptive choice of data resolution.

Quality Control for Single Cell Analysis of High-plex Tissue Profiles using CyLinter.

Tumors are complex assemblies of cellular and acellular structures patterned on spatial scales from microns to centimeters. Study of these assemblies has advanced dramatically with the introduction of high-plex spatial profiling. Image-based profiling methods reveal the intensities and spatial distributions of 20-100 proteins at subcellular resolution in 103-107 cells per specimen. Despite extensive work on methods for extracting single-cell data from these images, all tissue images contain artefacts such as folds, debris, antibody aggregates, optical aberrations and image processing errors that arise from imperfections in specimen preparation, data acquisition, image assembly, and feature extraction. We show that these artefacts dramatically impact single-cell data analysis, obscuring meaningful biological interpretation. We describe an interactive quality control software tool, CyLinter, that identifies and removes data associated with imaging artefacts. CyLinter greatly improves single-cell analysis, especially for archival specimens sectioned many years prior to data collection, such as those from clinical trials.

Multiplexed 3D Analysis of Immune States and Niches in Human Tissue.

Tissue homeostasis and the emergence of disease are controlled by changes in the proportions of resident and recruited cells, their organization into cellular neighbourhoods, and their interactions with acellular tissue components. Highly multiplexed tissue profiling (spatial omics) 1 makes it possible to study this microenvironment in situ , usually in 4-5 micron thick sections (the standard histopathology format) 2 . Microscopy-based tissue profiling is commonly performed at a resolution sufficient to determine cell types but not to detect subtle morphological features associated with cytoskeletal reorganisation, juxtracrine signalling, or membrane trafficking 3 . Here we describe a high-resolution 3D imaging approach able to characterize a wide variety of organelles and structures at sub-micron scale while simultaneously quantifying millimetre-scale spatial features. This approach combines cyclic immunofluorescence (CyCIF) imaging 4 of over 50 markers with confocal microscopy of archival human tissue thick enough (30-40 microns) to fully encompass two or more layers of intact cells. 3D imaging of entire cell volumes substantially improves the accuracy of cell phenotyping and allows cell proximity to be scored using plasma membrane apposition, not just nuclear position. In pre-invasive melanoma in situ 5 , precise phenotyping shows that adjacent melanocytic cells are plastic in state and participate in tightly localised niches of interferon signalling near sites of initial invasion into the underlying dermis. In this and metastatic melanoma, mature and precursor T cells engage in an unexpectedly diverse array of juxtracrine and membrane-membrane interactions as well as looser "neighbourhood" associations 6 whose morphologies reveal functional states. These data provide new insight into the transitions occurring during early tumour formation and immunoediting and demonstrate the potential for phenotyping of tissues at a level of detail previously restricted to cultured cells and organoids.

Development and validation of time-to-event models to predict metastatic recurrence of localized cutaneous melanoma.

BACKGROUND: The recent expansion of immunotherapy for stage IIB/IIC melanoma highlights a growing clinical need to identify patients at high risk of metastatic recurrence and, therefore, most likely to benefit from this therapeutic modality. OBJECTIVE: To develop time-to-event risk prediction models for melanoma metastatic recurrence. METHODS: Patients diagnosed with stage I/II primary cutaneous melanoma between 2000 and 2020 at Mass General Brigham and Dana-Farber Cancer Institute were included. Melanoma recurrence date and type were determined by chart review. Thirty clinicopathologic factors were extracted from electronic health records. Three types of time-to-event machine-learning models were evaluated internally and externally in the distant versus locoregional/nonrecurrence prediction. RESULTS: This study included 954 melanomas (155 distant, 163 locoregional, and 636 1:2 matched nonrecurrences). Distant recurrences were associated with worse survival compared to locoregional/nonrecurrences (HR: 6.21, P < .001) and to locoregional recurrences only (HR: 5.79, P < .001). The Gradient Boosting Survival model achieved the best performance (concordance index: 0.816; time-dependent AUC: 0.842; Brier score: 0.103) in the external validation. LIMITATIONS: Retrospective nature and cohort from one geography. CONCLUSIONS: These results suggest that time-to-event machine-learning models can reliably predict the metastatic recurrence from localized melanoma and help identify high-risk patients who are most likely to benefit from immunotherapy.

SpatialCells: Automated Profiling of Tumor Microenvironments with Spatially Resolved Multiplexed Single-Cell Data.

Background: Cancer is a complex cellular ecosystem where malignant cells coexist and interact with immune, stromal, and other cells within the tumor microenvironment. Recent technological advancements in spatially resolved multiplexed imaging at single-cell resolution have led to the generation of large-scale and high-dimensional datasets from biological specimens. This underscores the necessity for automated methodologies that can effectively characterize the molecular, cellular, and spatial properties of tumor microenvironments for various malignancies. Results: This study introduces SpatialCells, an open-source software package designed for region-based exploratory analysis and comprehensive characterization of tumor microenvironments using multiplexed single-cell data. Conclusions: SpatialCells efficiently streamlines the automated extraction of features from multiplexed single-cell data and can process samples containing millions of cells. Thus, SpatialCells facilitates subsequent association analyses and machine learning predictions, making it an essential tool in advancing our understanding of tumor growth, invasion, and metastasis.

Cell Spotter (CSPOT): A machine-learning approach to automated cell spotting and quantification of highly multiplexed tissue images.

Highly multiplexed tissue imaging and in situ spatial profiling aim to extract single-cell data from specimens containing closely packed cells of diverse morphology. This is challenging due to the difficulty of accurately assigning boundaries between cells (segmentation) and then generating per-cell staining intensities. Existing methods use gating to convert per-cell intensity data to positive and negative scores; this is a common approach in flow cytometry, but one that is problematic in imaging. In contrast, human experts identify cells in crowded environments using morphological, neighborhood, and intensity information. Here we describe a computational approach (Cell Spotter or CSPOT) that uses supervised machine learning in combination with classical segmentation to perform automated cell type calling. CSPOT is robust to artifacts that commonly afflict tissue imaging and can replace conventional gating. The end-to-end Python implementation of CSPOT can be integrated into cloud-based image processing pipelines to substantially improve the speed, accuracy, and reproducibility of single-cell spatial data.