Unsupervised pattern identification in spatial gene expression atlas reveals mouse brain regions beyond established ontology.

The rapid growth of large-scale spatial gene expression data demands efficient and reliable computational tools to extract major trends of gene expression in their native spatial context. Here, we used stability-driven unsupervised learning (i.e., staNMF) to identify principal patterns (PPs) of 3D gene expression profiles and understand spatial gene distribution and anatomical localization at the whole mouse brain level. Our subsequent spatial correlation analysis systematically compared the PPs to known anatomical regions and ontology from the Allen Mouse Brain Atlas using spatial neighborhoods. We demonstrate that our stable and spatially coherent PPs, whose linear combinations accurately approximate the spatial gene data, are highly correlated with combinations of expert-annotated brain regions. These PPs yield a brain ontology based purely on spatial gene expression. Our PP identification approach outperforms principal component analysis and typical clustering algorithms on the same task. Moreover, we show that the stable PPs reveal marked regional imbalance of brainwide genetic architecture, leading to region-specific marker genes and gene coexpression networks. Our findings highlight the advantages of stability-driven machine learning for plausible biological discovery from dense spatial gene expression data, streamlining tasks that are infeasible by conventional manual approaches.

Staphylococcus aureus proteases trigger eosinophil-mediated skin inflammation.

Staphylococcus aureus skin colonization and eosinophil infiltration are associated with many inflammatory skin disorders, including atopic dermatitis, bullous pemphigoid, Netherton's syndrome, and prurigo nodularis. However, whether there is a relationship between S. aureus and eosinophils and how this interaction influences skin inflammation is largely undefined. We show in a preclinical mouse model that S. aureus epicutaneous exposure induced eosinophil-recruiting chemokines and eosinophil infiltration into the skin. Remarkably, we found that eosinophils had a comparable contribution to the skin inflammation as T cells, in a manner dependent on eosinophil-derived IL-17A and IL-17F production. Importantly, IL-36R signaling induced CCL7-mediated eosinophil recruitment to the inflamed skin. Last, S. aureus proteases induced IL-36α expression in keratinocytes, which promoted infiltration of IL-17-producing eosinophils. Collectively, we uncovered a mechanism for S. aureus proteases to trigger eosinophil-mediated skin inflammation, which has implications in the pathogenesis of inflammatory skin diseases.

Frontotemporal lobar degeneration targets brain regions linked to expression of recently evolved genes.

In frontotemporal lobar degeneration (FTLD), pathological protein aggregation in specific brain regions is associated with declines in human-specialized social-emotional and language functions. In most patients, disease protein aggregates contain either TDP-43 (FTLD-TDP) or tau (FTLD-tau). Here, we explored whether FTLD-associated regional degeneration patterns relate to regional gene expression of human accelerated regions (HARs), conserved sequences that have undergone positive selection during recent human evolution. To this end, we used structural neuroimaging from patients with FTLD and human brain regional transcriptomic data from controls to identify genes expressed in FTLD-targeted brain regions. We then integrated primate comparative genomic data to test our hypothesis that FTLD targets brain regions linked to expression levels of recently evolved genes. In addition, we asked whether genes whose expression correlates with FTLD atrophy are enriched for genes that undergo cryptic splicing when TDP-43 function is impaired. We found that FTLD-TDP and FTLD-tau subtypes target brain regions with overlapping and distinct gene expression correlates, highlighting many genes linked to neuromodulatory functions. FTLD atrophy-correlated genes were strongly enriched for HARs. Atrophy-correlated genes in FTLD-TDP showed greater overlap with TDP-43 cryptic splicing genes and genes with more numerous TDP-43 binding sites compared with atrophy-correlated genes in FTLD-tau. Cryptic splicing genes were enriched for HAR genes, and vice versa, but this effect was due to the confounding influence of gene length. Analyses performed at the individual-patient level revealed that the expression of HAR genes and cryptically spliced genes within putative regions of disease onset differed across FTLD-TDP subtypes. Overall, our findings suggest that FTLD targets brain regions that have undergone recent evolutionary specialization and provide intriguing potential leads regarding the transcriptomic basis for selective vulnerability in distinct FTLD molecular-anatomical subtypes.

Network Connectivity Alterations across the MAPT Mutation Clinical Spectrum.

OBJECTIVE: Microtubule-associated protein tau (MAPT) mutations cause frontotemporal lobar degeneration, and novel biomarkers are urgently needed for early disease detection. We used task-free functional magnetic resonance imaging (fMRI) mapping, a promising biomarker, to analyze network connectivity in symptomatic and presymptomatic MAPT mutation carriers. METHODS: We compared cross-sectional fMRI data between 17 symptomatic and 39 presymptomatic carriers and 81 controls with (1) seed-based analyses to examine connectivity within networks associated with the 4 most common MAPT-associated clinical syndromes (ie, salience, corticobasal syndrome, progressive supranuclear palsy syndrome, and default mode networks) and (2) whole-brain connectivity analyses. We applied K-means clustering to explore connectivity heterogeneity in presymptomatic carriers at baseline. Neuropsychological measures, plasma neurofilament light chain, and gray matter volume were compared at baseline and longitudinally between the presymptomatic subgroups defined by their baseline whole-brain connectivity profiles. RESULTS: Symptomatic and presymptomatic carriers had connectivity disruptions within MAPT-syndromic networks. Compared to controls, presymptomatic carriers showed regions of connectivity alterations with age. Two presymptomatic subgroups were identified by clustering analysis, exhibiting predominantly either whole-brain hypoconnectivity or hyperconnectivity at baseline. At baseline, these two presymptomatic subgroups did not differ in neuropsychological measures, although the hypoconnectivity subgroup had greater plasma neurofilament light chain levels than controls. Longitudinally, both subgroups showed visual memory decline (vs controls), yet the subgroup with baseline hypoconnectivity also had worsening verbal memory and neuropsychiatric symptoms, and extensive bilateral mesial temporal gray matter decline. INTERPRETATION: Network connectivity alterations arise as early as the presymptomatic phase. Future studies will determine whether presymptomatic carriers' baseline connectivity profiles predict symptomatic conversion. ANN NEUROL 2023;94:632-646.

Accurate Hellmann-Feynman forces from density functional calculations with augmented Gaussian basis sets.

The Hellmann-Feynman (HF) theorem provides a way to compute forces directly from the electron density, enabling efficient force calculations for large systems through machine learning (ML) models for the electron density. The main issue holding back the general acceptance of the HF approach for atom-centered basis sets is the well-known Pulay force which, if naively discarded, typically constitutes an error upward of 10 eV/Å in forces. In this work, we demonstrate that if a suitably augmented Gaussian basis set is used for density functional calculations, the Pulay force can be suppressed, and HF forces can be computed as accurately as analytical forces with state-of-the-art basis sets, allowing geometry optimization and molecular dynamics to be reliably performed with HF forces. Our results pave a clear path forward for the accurate and efficient simulation of large systems using ML densities and the HF theorem.

Predicting accurate ab initio DNA electron densities with equivariant neural networks.

One of the fundamental limitations of accurately modeling biomolecules like DNA is the inability to perform quantum chemistry calculations on large molecular structures. We present a machine learning model based on an equivariant Euclidean neural network framework to obtain accurate ab initio electron densities for arbitrary DNA structures that are much too large for conventional quantum methods. The model is trained on representative B-DNA basepair steps that capture both base pairing and base stacking interactions. The model produces accurate electron densities for arbitrary B-DNA structures with typical errors of less than 1%. Crucially, the error does not increase with system size, which suggests that the model can extrapolate to large DNA structures with negligible loss of accuracy. The model also generalizes reasonably to other DNA structural motifs such as the A- and Z-DNA forms, despite being trained on only B-DNA configurations. The model is used to calculate electron densities of several large-scale DNA structures, and we show that the computational scaling for this model is essentially linear. We also show that this machine learning electron density model can be used to calculate accurate electrostatic potentials for DNA. These electrostatic potentials produce more accurate results compared with classical force fields and do not show the usual deficiencies at short range.

RAS oncogene signal strength regulates matrisomal gene expression and tumorigenicity of mouse keratinocytes.

Environmental and molecular carcinogenesis are linked by the discovery that chemical carcinogen induced-mutations in the Hras or Kras genes drives tumor development in mouse skin. Importantly, enhanced expression or allele amplification of the mutant Ras gene contributes to selection of initiated cells, tumor persistence, and progression. To explore the consequences of Ras oncogene signal strength, primary keratinocytes were isolated and cultured from the LSL-HrasG12D and LSL-KrasG12D C57BL/6J mouse models and the mutant allele was activated by adeno-Cre recombinase. Keratinocytes expressing one (H) or two (HH) mutant alleles of HrasG12D, one KrasG12D allele (K), or one of each (HK) were studied. All combinations of activated Ras alleles stimulated proliferation and drove transformation marker expression, but only HH and HK formed tumors. HH, HK, and K sustained long-term keratinocyte growth in vitro, while H and WT could not. RNA-Seq yielded two distinct gene expression profiles; HH, HK, and K formed one cluster while H clustered with WT. Weak MAPK activation was seen in H keratinocytes but treatment with a BRAF inhibitor enhanced MAPK signaling and facilitated tumor formation. K keratinocytes became tumorigenic when they were isolated from mice where the LSL-KrasG12D allele was backcrossed from the C57BL/6 onto the FVB/N background. All tumorigenic keratinocytes but not the non-tumorigenic precursors shared a common remodeling of matrisomal gene expression that is associated with tumor formation. Thus, RAS oncogene signal strength determines cell-autonomous changes in initiated cells that are critical for their tumor-forming potential.

A dynamic gradient architecture generates brain activity states.

The human brain exhibits a diverse yet constrained range of activity states. While these states can be faithfully represented in a low-dimensional latent space, our understanding of the constitutive functional anatomy is still evolving. Here we applied dimensionality reduction to task-free and task fMRI data to address whether latent dimensions reflect intrinsic systems and if so, how these systems may interact to generate different activity states. We find that each dimension represents a dynamic activity gradient, including a primary unipolar sensory-association gradient underlying the global signal. The gradients appear stable across individuals and cognitive states, while recapitulating key functional connectivity properties including anticorrelation, modularity, and regional hubness. We then use dynamical systems modeling to show that gradients causally interact via state-specific coupling parameters to create distinct brain activity patterns. Together, these findings indicate that a set of dynamic, intrinsic spatial gradients interact to determine the repertoire of possible brain activity states.

RAS induced senescence of skin keratinocytes is mediated through Rho-associated protein kinase (ROCK).

Cellular senescence is a well-documented response to oncogene activation in many tissues. Multiple pathways are invoked to achieve senescence indicating its importance to counteract the transforming activities of oncogenic stimulation. We now report that the Rho-associated protein kinase (ROCK) signaling pathway is a critical regulator of oncogene-induced senescence in skin carcinogenesis. Transformation of mouse keratinocytes with oncogenic RAS upregulates ROCK activity and initiates a senescence response characterized by cell enlargement, growth inhibition, upregulation of senescence associated ß-galactosidase (SAßgal) expression, and release of multiple pro-inflammatory factors comprising the senescence-associated secretory phenotype (SASP). The addition of the ROCK inhibitor Y-27632 and others prevents these senescence responses and maintains proliferating confluent RAS transformed keratinocyte cultures indefinitely. Mechanistically, oncogenic RAS transformation is associated with upregulation of cell cycle inhibitors p15Ink4b , p16Ink4a , and p19Arf and downregulation of p-AKT, all of which are reversed by Y-27632. RNA-seq analysis of Y-27632 treated RAS-transformed keratinocytes indicated that the inhibitor reduced growth-inhibitory gene expression profiles and maintained expression of proliferative pathways. Y-27632 also reduced the expression of NF-κB effector genes and the expression of IκBζ downstream mediators. The senescence inhibition from Y-27632 was reversible, and upon its removal, senescence reoccurred in vitro with rapid upregulation of cell cycle inhibitors, SASP expression, and cell detachment. Y-27632 treated cultured RAS-keratinocytes formed tumors in the absence of the inhibitor when placed in skin orthografts suggesting that factors in the tumor microenvironment can overcome the drive to senescence imparted by overactive ROCK activity.

Epipyrone A, a Broad-Spectrum Antifungal Compound Produced by Epicoccum nigrum ICMP 19927.

We have isolated a filamentous fungus that actively secretes a pigmented exudate when growing on agar plates. The fungus was identified as being a strain of Epicoccum nigrum. The fungal exudate presented strong antifungal activity against both yeasts and filamentous fungi, and inhibited the germination of fungal spores. The chemical characterization of the exudate showed that the pigmented molecule presenting antifungal activity is the disalt of epipyrone A-a water-soluble polyene metabolite with a molecular mass of 612.29 and maximal UV-Vis absorbance at 428 nm. This antifungal compound showed excellent stability to different temperatures and neutral to alkaline pH.

First-Principles Atomic Force Microscopy Image Simulations with Density Embedding Theory.

We present an efficient first-principles method for simulating noncontact atomic force microscopy (nc-AFM) images using a "frozen density" embedding theory. Frozen density embedding theory enables one to efficiently compute the tip-sample interaction by considering a sample as a frozen external field. This method reduces the extensive computational load of first-principles AFM simulations by avoiding consideration of the entire tip-sample system and focusing on the tip alone. We demonstrate that our simulation with frozen density embedding theory accurately reproduces full density functional theory simulations of freestanding hydrocarbon molecules while the computational time is significantly reduced. Our method also captures the electronic effect of a Cu(111) substrate on the AFM image of pentacene and reproduces the experimental AFM image of Cu2N on a Cu(100) surface. This approach is applicable for theoretical imaging applications on large molecules, two-dimensional materials, and materials surfaces.

The Microwave Spectrum and Molecular Structure of (E)-1-Chloro-2-Fluoroethylene-HF: Revealing the Balance among Electrostatics, Sterics, and Resonance in Intermolecular Interactions.

The structure of the gas-phase bimolecular complex formed between (E)-1-chloro-2-fluoroethylene and hydrogen fluoride is determined via Fourier transform microwave spectroscopy from 6.9-21.6 GHz. Although the complex adopts a geometry very similar to that of previously studied dihalosubstituted ethylene-HF species, trends observed in the values of structural parameters such as bond lengths, bond angles, and deviations of the primary hydrogen-bonding interaction from linearity provide information regarding the balance among electrostatic, steric, and resonance effects in the structures of these complexes. Consideration of the ab initio interaction potential between (E)-1-chloro-2-fluoroethylene and hydrogen fluoride suggests that it is the strength of the intermolecular bond formed between the hydrogen atom of HF and the fluorine atom of the substituted ethylene that plays the significant role in determining the geometry. In addition to determining the complete nuclear quadrupole coupling constant tensor for the (E)-1-chloro-2-fluoroethylene-HF complex, the corresponding tensor for (E)-1-chloro-2-fluoroethylene itself was measured with greater precision than previously availabe, including the first reported determination of the single, nonzero off-diagonal element, χab.

Interaction range of P-dopants in Si[110] nanowires: determining the nondegenerate limit.

It is well known that the activation energy of dopants in semiconducting nanomaterials is higher than in bulk materials owing to dielectric mismatch and quantum confinement. This quenches the number of free charge carriers in nanomaterials. Though higher doping concentration can compensate for this effect, there is no clear criterion on what the doping concentration should be. Using P-doped Si[110] nanowires as the prototypical system, we address this issue by establishing a doping limit by first-principles electronic structure calculations. We examine how the doped nanowires respond to charging using an effective capacitance approach. As the nanowire gets thinner, the interaction range of the P dopants shortens and the doping concentration can increase concurrently. Hence, heavier doping can remain nondegenerate for thin nanowires.

Building an ab initio solvated DNA model using Euclidean neural networks.

Accurately modeling large biomolecules such as DNA from first principles is fundamentally challenging due to the steep computational scaling of ab initio quantum chemistry methods. This limitation becomes even more prominent when modeling biomolecules in solution due to the need to include large numbers of solvent molecules. We present a machine-learned electron density model based on a Euclidean neural network framework that includes a built-in understanding of equivariance to model explicitly solvated double-stranded DNA. By training the machine learning model using molecular fragments that sample the key DNA and solvent interactions, we show that the model predicts electron densities of arbitrary systems of solvated DNA accurately, resolves polarization effects that are neglected by classical force fields, and captures the physics of the DNA-solvent interaction at the ab initio level.

Antifungal Activity of Disalt of Epipyrone A from Epicoccum nigrum Likely via Disrupted Fatty Acid Elongation and Sphingolipid Biosynthesis.

Multidrug-resistant fungal pathogens and antifungal drug toxicity have challenged our current ability to fight fungal infections. Therefore, there is a strong global demand for novel antifungal molecules with the distinct mode of action and specificity to service the medical and agricultural sectors. Polyenes are a class of antifungal drugs with the broadest spectrum of activity among the current antifungal drugs. Epipyrone A, a water-soluble antifungal molecule with a unique, linear polyene structure, was isolated from the fungus Epiccocum nigrum. Since small changes in a compound structure can significantly alter its cell target and mode of action, we present here a study on the antifungal mode of action of the disalt of epipyrone A (DEA) using chemical-genetic profiling, fluorescence microscopy, and metabolomics. Our results suggest the disruption of sphingolipid/fatty acid biosynthesis to be the primary mode of action of DEA, followed by the intracellular accumulation of toxic phenolic compounds, in particular p-toluic acid (4-methylbenzoic acid). Although membrane ergosterol is known to be the main cell target for polyene antifungal drugs, we found little evidence to support that is the case for DEA. Sphingolipids, on the other hand, are known for their important roles in fungal cell physiology, and their biosynthesis has been recognized as a potential fungal-specific cell target for the development of new antifungal drugs.

Data-driven fine-grained region discovery in the mouse brain with transformers.

Technologies such as spatial transcriptomics offer unique opportunities to define the spatial organization of the mouse brain. We developed an unsupervised training scheme and novel transformer-based deep learning architecture to detect spatial domains across the whole mouse brain using spatial transcriptomics data. Our model learns local representations of molecular and cellular statistical patterns which can be clustered to identify spatial domains within the brain from coarse to fine-grained. Discovered domains are spatially regular, even with several hundreds of spatial clusters. They are also consistent with existing anatomical ontologies such as the Allen Mouse Brain Common Coordinate Framework version 3 (CCFv3) and can be visually interpreted at the cell type or transcript level. We demonstrate our method can be used to identify previously uncatalogued subregions, such as in the midbrain, where we uncover gradients of inhibitory neuron complexity and abundance. Notably, these subregions cannot be discovered using other methods. We apply our method to a separate multi-animal whole-brain spatial transcriptomic dataset and show that our method can also robustly integrate spatial domains across animals.

Metabolic Reprogramming of Tumor-Associated Macrophages Using Glutamine Antagonist JHU083 Drives Tumor Immunity in Myeloid-Rich Prostate and Bladder Cancers.

Glutamine metabolism in tumor microenvironments critically regulates antitumor immunity. Using the glutamine-antagonist prodrug JHU083, we report potent tumor growth inhibition in urologic tumors by JHU083-reprogrammed tumor-associated macrophages (TAMs) and tumor-infiltrating monocytes. We show JHU083-mediated glutamine antagonism in tumor microenvironments induced by TNF, proinflammatory, and mTORC1 signaling in intratumoral TAM clusters. JHU083-reprogrammed TAMs also exhibited increased tumor cell phagocytosis and diminished proangiogenic capacities. In vivo inhibition of TAM glutamine consumption resulted in increased glycolysis, a broken tricarboxylic acid (TCA) cycle, and purine metabolism disruption. Although the antitumor effect of glutamine antagonism on tumor-infiltrating T cells was moderate, JHU083 promoted a stem cell-like phenotype in CD8+ T cells and decreased the abundance of regulatory T cells. Finally, JHU083 caused a global shutdown in glutamine-utilizing metabolic pathways in tumor cells, leading to reduced HIF-1α, c-MYC phosphorylation, and induction of tumor cell apoptosis, all key antitumor features. Altogether, our findings demonstrate that targeting glutamine with JHU083 led to suppressed tumor growth as well as reprogramming of immunosuppressive TAMs within prostate and bladder tumors that promoted antitumor immune responses. JHU083 can offer an effective therapeutic benefit for tumor types that are enriched in immunosuppressive TAMs.

Machine Learning for Uncovering Biological Insights in Spatial Transcriptomics Data.

Development and homeostasis in multicellular systems both require exquisite control over spatial molecular pattern formation and maintenance. Advances in spatially-resolved and high-throughput molecular imaging methods such as multiplexed immunofluorescence and spatial transcriptomics (ST) provide exciting new opportunities to augment our fundamental understanding of these processes in health and disease. The large and complex datasets resulting from these techniques, particularly ST, have led to rapid development of innovative machine learning (ML) tools primarily based on deep learning techniques. These ML tools are now increasingly featured in integrated experimental and computational workflows to disentangle signals from noise in complex biological systems. However, it can be difficult to understand and balance the different implicit assumptions and methodologies of a rapidly expanding toolbox of analytical tools in ST. To address this, we summarize major ST analysis goals that ML can help address and current analysis trends. We also describe four major data science concepts and related heuristics that can help guide practitioners in their choices of the right tools for the right biological questions.

Dysfunctional Cortical Gradient Topography in Treatment-Resistant Major Depressive Disorder.

BACKGROUND: Treatment-resistant depression (TRD) refers to patients with major depressive disorder who do not remit after 2 or more antidepressant trials. TRD is common and highly debilitating, but its neurobiological basis remains poorly understood. Recent neuroimaging studies have revealed cortical connectivity gradients that dissociate primary sensorimotor areas from higher-order associative cortices. This fundamental topography determines cortical information flow and is affected by psychiatric disorders. We examined how TRD impacts gradient-based hierarchical cortical organization. METHODS: In this secondary study, we analyzed resting-state functional magnetic resonance imaging data from a mindfulness-based intervention enrolling 56 patients with TRD and 28 healthy control subjects. Using gradient extraction tools, baseline measures of cortical gradient dispersion within and between functional brain networks were derived, compared across groups, and associated with graph theoretical measures of network topology. In patients, correlation analyses were used to associate measures of cortical gradient dispersion with clinical measures of anxiety, depression, and mindfulness at baseline and following the intervention. RESULTS: Cortical gradient dispersion was reduced within major intrinsic brain networks in patients with TRD. Reduced cortical gradient dispersion correlated with increased network degree assessed through graph theory-based measures of network topology. Lower dispersion among default mode, control, and limbic network nodes related to baseline levels of trait anxiety, depression, and mindfulness. Patients' baseline limbic network dispersion predicted trait anxiety scores 24 weeks after the intervention. CONCLUSIONS: Our findings provide preliminary support for widespread alterations in cortical gradient architecture in TRD, implicating a significant role for transmodal and limbic networks in mediating depression, anxiety, and lower mindfulness in patients with TRD.

Increasing empathic concern relates to salience network hyperconnectivity in cognitively healthy older adults with elevated amyloid-ß burden.

Enhanced emotional empathy, the ability to share others' affective experiences, can be a feature of Alzheimer's disease (AD), but whether emotional empathy increases in the preclinical phase of the disease is unknown. We measured emotional empathy over time (range = 0 - 7.3 years, mean = 2.4 years) in 86 older adults during a period in which they were cognitively healthy, functionally normal, and free of dementia symptoms. For each participant, we computed longitudinal trajectories for empathic concern (i.e., an other-oriented form of emotional empathy that promotes prosocial actions) and emotional contagion (i.e., a self-focused form of emotional empathy often accompanied by feelings of distress) from informant ratings of participants' empathy on the Interpersonal Reactivity Index. Amyloid-ß (Aß) positron emission tomography (PET) scans were used to classify participants as either Aß positive (Aß+, n = 23) or negative (Aß-, n = 63) based on Aß-PET cortical binding. Participants also underwent structural and task-free functional magnetic resonance imaging approximately two years on average after their last empathy assessment, at which time most participants remained cognitively healthy. Results indicated that empathic concern, but not emotional contagion, increased more over time in Aß+ participants than in Aß- participants despite no initial group difference at the first measurement. Higher connectivity between certain salience network node-pairs (i.e., pregenual anterior cingulate cortex and periaqueductal gray) predicted longitudinal increases in empathic concern in the Aß+ group but not in the Aß- group. The Aß+ participants also had higher overall salience network connectivity than Aß- participants despite no differences in gray matter volume. These results suggest gains in empathic concern may be a very early feature of AD pathophysiology that relates to hyperconnectivity in the salience network, a system that supports emotion generation and interoception. A better understanding of emotional empathy trajectories in the early stages of AD pathophysiology will broaden the lens on preclinical AD changes and help clinicians to identify older adults who should be screened for AD biomarkers.