Mapping human tissues with highly multiplexed RNA in situ hybridization.

In situ transcriptomic techniques promise a holistic view of tissue organization and cell-cell interactions. There has been a surge of multiplexed RNA in situ mapping techniques but their application to human tissues has been limited due to their large size, general lower tissue quality and high autofluorescence. Here we report DART-FISH, a padlock probe-based technology capable of profiling hundreds to thousands of genes in centimeter-sized human tissue sections. We introduce an omni-cell type cytoplasmic stain that substantially improves the segmentation of cell bodies. Our enzyme-free isothermal decoding procedure allows us to image 121 genes in large sections from the human neocortex in <10 h. We successfully recapitulated the cytoarchitecture of 20 neuronal and non-neuronal subclasses. We further performed in situ mapping of 300 genes on a diseased human kidney, profiled >20 healthy and pathological cell states, and identified diseased niches enriched in transcriptionally altered epithelial cells and myofibroblasts.

Gene panel selection for targeted spatial transcriptomics.

Targeted spatial transcriptomics hold particular promise in analyzing complex tissues. Most such methods, however, measure only a limited panel of transcripts, which need to be selected in advance to inform on the cell types or processes being studied. A limitation of existing gene selection methods is their reliance on scRNA-seq data, ignoring platform effects between technologies. Here we describe gpsFISH, a computational method performing gene selection through optimizing detection of known cell types. By modeling and adjusting for platform effects, gpsFISH outperforms other methods. Furthermore, gpsFISH can incorporate cell type hierarchies and custom gene preferences to accommodate diverse design requirements.

Mapping Human Tissues with Highly Multiplexed RNA in situ Hybridization.

In situ transcriptomic techniques promise a holistic view of tissue organization and cell-cell interactions. Recently there has been a surge of multiplexed RNA in situ techniques but their application to human tissues and clinical biopsies has been limited due to their large size, general lower tissue quality and high background autofluorescence. Here we report DART-FISH, a versatile padlock probe-based technology capable of profiling hundreds to thousands of genes in centimeter-sized human tissue sections at cellular resolution. We introduced an omni-cell type cytoplasmic stain, dubbed RiboSoma that substantially improves the segmentation of cell bodies. We developed a computational decoding-by-deconvolution workflow to extract gene spots even in the presence of optical crowding. Our enzyme-free isothermal decoding procedure allowed us to image 121 genes in a large section from the human neocortex in less than 10 hours, where we successfully recapitulated the cytoarchitecture of 20 neuronal and non-neuronal subclasses. Additionally, we demonstrated the detection of transcripts as short as 461 nucleotides, including neuropeptides and discovered new cortical layer markers. We further performed in situ mapping of 300 genes on a diseased human kidney, profiled >20 healthy and pathological cell states, and identified diseased niches enriched in transcriptionally altered epithelial cells and myofibroblasts.

Gene panel selection for targeted spatial transcriptomics.

Targeted spatial transcriptomics hold particular promise in analysis of complex tissues. Most such methods, however, measure only a limited panel of transcripts, which need to be selected in advance to inform on the cell types or processes being studied. A limitation of existing gene selection methods is that they rely on scRNA-seq data, ignoring platform effects between technologies. Here we describe gpsFISH, a computational method to perform gene selection through optimizing detection of known cell types. By modeling and adjusting for platform effects, gpsFISH outperforms other methods. Furthermore, gpsFISH can incorporate cell type hierarchies and custom gene preferences to accommodate diverse design requirements.

Tailoring Collagen to Engineer the Cellular Microenvironment.

Collagen is the most abundant protein in the extracellular matrix (ECM), and it can direct the behavior of the neighboring cells. By customizing properties of collagen, it is possible to control the cells that interact with it. Utilizing a bottom-up strategy, modular gene fragments are assembled and recombinantly processed to create collagen-mimetic variants that modulate proteolytic degradation, cell adhesion, and mechanical characteristics. The removal of the native MMP cleavage site results in MMP-1 resistant collagen. By introducing additional MMP-susceptible sequences, the degradation characteristics of collagen molecules are modified. Additional non-native functionality is also introduced into the collagen, including the IKVAV sequence, which has been implicated in neurite outgrowth. This mutation, which disrupts the Gly-X-Y tripeptide repeat of collagen, does not prevent the formation of triple-helical collagen. Non-native cysteines and the integrin binding sequence GFOGER are combined in the collagen, and encapsulation of normal human lung fibroblasts within collagen hydrogels are tested. Cells remain spherical, when encapsulated within hydrogels of collagen variants in which the native integrin binding sites are removed, but cell adhesion is restored with the introduction of non-native GFOGER binding sequences. This modular collagen system allows for the combination of multiple functionalities, and it enables the production of biomimetic scaffolds with customizable characteristics to modulate cellular microenvironments.

Recombinant collagen scaffolds as substrates for human neural stem/progenitor cells.

Adhesion to the microenvironment profoundly affects stem cell functions, including proliferation and differentiation, and understanding the interaction of stem cells with the microenvironment is important for controlling their behavior. In this study, we investigated the effects of the integrin binding epitopes GFOGER and IKVAV (natively present in collagen I and laminin, respectively) on human neural stem/progenitor cells (hNSPCs). To test the specificity of these epitopes, GFOGER or IKVAV were placed within the context of recombinant triple-helical collagen III engineered to be devoid of native integrin binding sites. HNSPCs adhered to collagen that presented GFOGER as the sole integrin-binding site, but not to IKVAV-containing collagen. For the GFOGER-containing collagens, antibodies against the ß1 integrin subunit prevented cellular adhesion, antibodies against the α1 subunit reduced cell adhesion, and antibodies against α2 or α3 subunits had no significant effect. These results indicate that hNSPCs primarily interact with GFOGER through the α1ß1 integrin heterodimer. These GFOGER-presenting collagen variants also supported differentiation of hNSPCs into neurons and astrocytes. Our findings show, for the first time, that hNSPCs can bind to the GFOGER sequence, and they provide motivation to develop hydrogels formed from recombinant collagen variants as a cell delivery scaffold. © 2018 Wiley Periodicals, Inc. J Biomed Mater Res Part A: 106A: 1363-1372, 2018.

Combination scaffolds of salmon fibrin, hyaluronic acid, and laminin for human neural stem cell and vascular tissue engineering.

UNLABELLED: Human neural stem/progenitor cells (hNSPCs) are good candidates for treating central nervous system (CNS) trauma since they secrete beneficial trophic factors and differentiate into mature CNS cells; however, many cells die after transplantation. This cell death can be ameliorated by inclusion of a biomaterial scaffold, making identification of optimal scaffolds for hNSPCs a critical research focus. We investigated the properties of fibrin-based scaffolds and their effects on hNSPCs and found that fibrin generated from salmon fibrinogen and thrombin stimulates greater hNSPC proliferation than mammalian fibrin. Fibrin scaffolds degrade over the course of a few days in vivo, so we sought to develop a novel scaffold that would retain the beneficial properties of fibrin but degrade more slowly to provide longer support for hNSPCs. We found combination scaffolds of salmon fibrin with interpenetrating networks (IPNs) of hyaluronic acid (HA) with and without laminin polymerize more effectively than fibrin alone and generate compliant hydrogels matching the physical properties of brain tissue. Furthermore, combination scaffolds support hNSPC proliferation and differentiation while significantly attenuating the cell-mediated degradation seen with fibrin alone. HNSPCs express two fibrinogen-binding integrins, αVß1 and α5ß1, and several laminin binding integrins (α7ß1, α6ß1, α3ß1) that can mediate interaction with the scaffold. Lastly, to test the ability of scaffolds to support vascularization, we analyzed human cord blood-derived endothelial cells alone and in co-culture with hNSPCs and found enhanced vessel formation and complexity in co-cultures within combination scaffolds. Overall, combination scaffolds of fibrin, HA, and laminin are excellent biomaterials for hNSPCs. STATEMENT OF SIGNIFICANCE: Interest has increased recently in the development of biomaterials as neural stem cell transplantation scaffolds to treat central nervous system (CNS) injury since scaffolds improve survival and integration of transplanted cells. We report here on a novel combination scaffold composed of fibrin, hyaluronic acid, and laminin to support human neural stem/progenitor cell (hNSPC) function. This combined biomaterial scaffold has appropriate physical properties for hNSPCs and the CNS, supports hNSPC proliferation and differentiation, and attenuates rapid cell-mediated scaffold degradation. The hNSPCs and scaffold components synergistically encourage new vessel formation from human endothelial cells. This work marks the first report of a combination scaffold supporting human neural and vascular cells to encourage vasculogenesis, and sets a benchmark for biomaterials to treat CNS injury.

Tuning cellular response by modular design of bioactive domains in collagen.

Collagen's ability to direct cellular behavior suggests that redesigning it at the molecular level could enable manipulation of cells residing in an engineered microenvironment. However, the fabrication of full-length collagen mimics of specified sequence de novo has been elusive, and applications still rely on material from native tissues. Using a bottom-up strategy, we synthesized modular genes and expressed recombinant human collagen variants in Saccharomyces cerevisiae. The resulting biopolymers contained prescribed cell-interaction sites that can direct and tune cellular responses, with retention of the important triple-helical self-assembled structure. Removal of the native integrin-binding sites GROGER, GAOGER, GLOGEN, GLKGEN, and GMOGER in human collagen III yielded collagen that did not support adhesion of mammalian cells. Introduction of GFOGER sequences to this scaffold at specified locations and densities resulted in varying degrees of cellular attachment. The recruitment of focal adhesion complexes on the different collagens ranged from a 96% reduction to a 56% increase over native collagen I. Adhesion to the GFOGER-containing variants was entirely dependent and partially dependent on the ß1 and α2 subunits of integrin, respectively, with cell adhesion on average reduced by 86% with anti-ß1 and 38% with anti-α2 integrin antibody incubation. Results support the importance of local context in collagen-cell interactions. The investigation demonstrates the flexibility of this approach to introduce targeted changes throughout the collagen polymer for producing fully-prescribed variants with tailored properties.

Expanding functionality of recombinant human collagen through engineered non-native cysteines.

Collagen is the most abundant protein in extracellular matrices and is commonly used as a tissue engineering scaffold. However, collagen and other biopolymers from native sources can exhibit limitations when tuning mechanical and biological properties. Cysteines do not naturally occur within the triple-helical region of any native collagen. We utilized a novel modular synthesis strategy to fabricate variants of recombinant human collagen that contained 2, 4, or 8 non-native cysteines at precisely defined locations within each biopolymer. This bottom-up approach introduced capabilities using sulfhydryl chemistry to form hydrogels and immobilize bioactive factors. Collagen variants retained their triple-helical structure and supported cellular adhesion. Hydrogels were characterized using rheology, and the storage moduli were comparable to fibrillar collagen gels at similar concentrations. Furthermore, the introduced cysteines functioned as anchoring sites, with TGF-ß1-conjugated collagens promoting myofibroblast differentiation. This approach demonstrates the feasibility to produce custom-designed collagens with chemical functionality not available from native sources.

Modification of biomaterials with a self-protein inhibits the macrophage response.

A biomaterial inhibits the host immune response by displaying an endo-genously expressed immunomodulatory molecule, CD200. Immobilization of CD200 onto biomaterial surfaces effectively suppresses macrophage activation and reduces inflammatory response to subcutaneously implanted materials.

Multiscale modeling of metabolism and macromolecular synthesis in E. coli and its application to the evolution of codon usage.

Biological systems are inherently hierarchal and multiscale in time and space. A major challenge of systems biology is to describe biological systems as a computational model, which can be used to derive novel hypothesis and drive experiments leading to new knowledge. The constraint-based reconstruction and analysis approach has been successfully applied to metabolism and to the macromolecular synthesis machinery assembly. Here, we present the first integrated stoichiometric multiscale model of metabolism and macromolecular synthesis for Escherichia coli K12 MG1655, which describes the sequence-specific synthesis and function of almost 2000 gene products at molecular detail. We added linear constraints, which couple enzyme synthesis and catalysis reactions. Comparison with experimental data showed improvement of growth phenotype prediction with the multiscale model over E. coli's metabolic model alone. Many of the genes covered by this integrated model are well conserved across enterobacters and other, less related bacteria. We addressed the question of whether the bias in synonymous codon usage could affect the growth phenotype and environmental niches that an organism can occupy. We created two classes of in silico strains, one with more biased codon usage and one with more equilibrated codon usage than the wildtype. The reduced growth phenotype in biased strains was caused by tRNA supply shortage, indicating that expansion of tRNA gene content or tRNA codon recognition allow E. coli to respond to changes in codon usage bias. Our analysis suggests that in order to maximize growth and to adapt to new environmental niches, codon usage and tRNA content must co-evolve. These results provide further evidence for the mutation-selection-drift balance theory of codon usage bias. This integrated multiscale reconstruction successfully demonstrates that the constraint-based modeling approach is well suited to whole-cell modeling endeavors.

Quantitative prediction of cellular metabolism with constraint-based models: the COBRA Toolbox v2.0.

Over the past decade, a growing community of researchers has emerged around the use of constraint-based reconstruction and analysis (COBRA) methods to simulate, analyze and predict a variety of metabolic phenotypes using genome-scale models. The COBRA Toolbox, a MATLAB package for implementing COBRA methods, was presented earlier. Here we present a substantial update of this in silico toolbox. Version 2.0 of the COBRA Toolbox expands the scope of computations by including in silico analysis methods developed since its original release. New functions include (i) network gap filling, (ii) (13)C analysis, (iii) metabolic engineering, (iv) omics-guided analysis and (v) visualization. As with the first version, the COBRA Toolbox reads and writes systems biology markup language-formatted models. In version 2.0, we improved performance, usability and the level of documentation. A suite of test scripts can now be used to learn the core functionality of the toolbox and validate results. This toolbox lowers the barrier of entry to use powerful COBRA methods.