Application of transfer learning to predict drug-induced human in vivo gene expression changes using rat in vitro and in vivo data

The liver is the primary site for the metabolism and detoxification of many compounds, including pharmaceuticals. Consequently, it is also the primary location for many adverse reactions. As the liver is not readily accessible for sampling in humans; rodent or cell line models are often used to evaluate potential toxic effects of a novel compound or candidate drug. However, relating the results of animal and in vitro studies to relevant clinical outcomes for the human in vivo situation still proves challenging. In this study, we incorporate principles of transfer learning within a deep artificial neural network allowing us to leverage the relative abundance of rat in vitro and in vivo exposure data from the Open TG-GATEs data set to train a model to predict the expected pattern of human in vivo gene expression following an exposure given measured human in vitro gene expression. We show that domain adaptation has been successfully achieved, with the rat and human in vitro data no longer being separable in the common latent space generated by the network. The network produces physiologically plausible predictions of human in vivo gene expression pattern following an exposure to a previously unseen compound. Moreover, we show the integration of the human in vitro data in the training of the domain adaptation network significantly improves the temporal accuracy of the predicted rat in vivo gene expression pattern following an exposure to a previously unseen compound. In this way, we demonstrate the improvements in prediction accuracy that can be achieved by combining data from distinct domains.

Blood Milieu in Acute Myocardial Infarction Reprograms Human Macrophages for Trauma Repair

Acute myocardial infarction (AMI) is accompanied by a systemic trauma response that impacts the whole body, including blood. This study addresses whether macrophages, key players in trauma repair, sense and respond to these changes. For this, healthy human monocyte-derived macrophages are exposed to 20% human AMI (n = 50) or control (n = 20) serum and analyzed by transcriptional and multiparameter functional screening followed by network-guided data interpretation and drug repurposing. Results are validated in an independent cohort at functional level (n = 47 AMI, n = 25 control) and in a public dataset. AMI serum exposure results in an overt AMI signature, enriched in debris cleaning, mitosis, and immune pathways. Moreover, gene networks associated with AMI and with poor clinical prognosis in AMI are identified. Network-guided drug screening on the latter unveils prostaglandin E2 (PGE2) signaling as target for clinical intervention in detrimental macrophage imprinting during AMI trauma healing. The results demonstrate pronounced context-induced macrophage reprogramming by the AMI systemic environment, to a degree decisive for patient prognosis. This offers new opportunities for targeted intervention and optimized cardiovascular disease risk management.

Identification of a PRDM1-regulated T cell network to regulate T cell driving plaque inflammation in human and mouse atherosclerosis

 

T cells have a prominent role in the pathogenesis of atherosclerosis, although their exact function remains elusive. Here, we pursued a network-driven approach to identify T cell-associated gene programs driving the transition from low- to high-risk human plaques.

In this study 43 human carotid arterial plaques were collected and stratified based on absence (low-risk) or presence (high-risk) of intraplaque haemorrhage (IPH). Lesion RNA was subjected to microarray gene expression analysis and analysed by Weighted Gene Co-expression Network Analysis (WGCNA). We identified a co-expressed gene cluster displaying a strong T cell signalling signature in high- versus low-risk plaque, which was tightly connected to subnetworks of angiogenesis and interferon-signalling. WGCNA-based Bayesian network inference, cell-type deconvolution and single-cell gene expression revealed that this T cell-associated gene program was likely linked to effector-memory cytotoxic CD8+ T cells, underpinning the central role of T cells in plaque destabilization. Gene regulatory analysis identified cytotoxic T cell-related transcription factors, like PRDM1, regulating this plaque T cell gene program. Moreover, we demonstrated in LDL receptor knockout mice with T cell-specific Prdm1 deficiency, that lack of Prdm1 in T cells resulted in larger, more advanced plaques.

In conclusion, our study reveals a PRDM1-regulated T cell footprint in high- versus low-risk human atherosclerotic lesions and murine atherosclerotic plaque development, thereby identifying this network as a potential target for intervention in adverse T cell responses.

Funding Acknowledgement

Type of funding sources: Public grant(s) – EU funding. Main funding source(s): The European Research Area Network on Cardiovascular Diseases (ERA-CVD and Dutch Heart Foundation)

Identification of CD8+ T cell PRDM1 in high-risk human plaques and its regulatory role in murine lesion development

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Fritz Thyssen Stiftung

T cells have a prominent role in the pathogenesis of atherosclerosis, although their function in atherosclerotic plaques is only partly understood. In this study, we utilize the advantages of high-throughput techniques and data analytic strategies to compare the inherent biological changes of T cells during plaque transition from a stable, non-haemorrhaged (low-risk) to a rupture-prone, haemorrhaged (high-risk) phenotype.

We classified 43 human carotid arterial lesions into high- and low-risk plaques based on the presence/absence of intraplaque hemorrhages. RNA from these lesions was isolated and microarray gene expression data was obtained and analyzed by Weighted Gene Co-expression Network Analysis. A strong T cell signalling signature was identified in high- versus low-risk plaques, influencing angiogenesis and interferon-related processes. Bayesian network inference, cell type deconvolution and single-cell RNA sequencing analysis revealed that the T cell-associated gene program was linked to effector-memory cytotoxic, CD8+ T cells. This gene program appeared driven by CD8+ T cell-related transcription factors, including RUNX3, IRF7 and most importantly PRDM1. To validate these findings, we demonstrated in a murine model that T cell PRDM1 plays a key role in plaque formation, as atherosclerotic mice with a T cell specific Prdm1 deficiency developed larger and more advanced atherosclerotic plaques compared to control mice.

In conclusion, our study unveils a clear PRDM1-regulated effector-memory cytotoxic CD8+ T cell footprint in plaque development and the shift from low- to high-risk plaques, thereby revealing CD8+ T cells and PRMD1 as potential targets for intervention in adverse T cell responses in human atherosclerotic lesions.

Integrative multiomics analysis of human atherosclerosis reveals a serum response factor-driven network associated with intraplaque hemorrhage

BACKGROUND

While single-omics analyses on human atherosclerotic plaque have been very useful to map stage- or disease-related differences in expression, they only partly capture the array of changes in this tissue and suffer from scale-intrinsic limitations. In order to better identify processes associated with intraplaque hemorrhage and plaque instability, we therefore combined multiple omics into an integrated model.

METHODS

In this study, we compared protein and gene makeup of low- versus high-risk atherosclerotic lesion segments from carotid endarterectomy patients, as judged from the absence or presence of intraplaque hemorrhage, respectively. Transcriptomic, proteomic, and peptidomic data of this plaque cohort were aggregated and analyzed by DIABLO, an integrative multivariate classification and feature selection method.

RESULTS

We identified a protein-gene associated multiomics model able to segregate stable, nonhemorrhaged from vulnerable, hemorrhaged lesions at high predictive performance (AUC >0.95). The dominant component of this model correlated with αSMA- PDGFRα+ fibroblast-like cell content (p = 2.4E-05) and Arg1+ macrophage content (p = 2.2E-04) and was driven by serum response factor (SRF), possibly in a megakaryoblastic leukemia-1/2 (MKL1/2) dependent manner. Gene set overrepresentation analysis on the selected key features of this model pointed to a clear cardiovascular disease signature, with overrepresentation of extracellular matrix synthesis and organization, focal adhesion, and cholesterol metabolism terms, suggestive of the model's relevance for the plaque vulnerability. Finally, we were able to corroborate the predictive power of the selected features in several independent mRNA and proteomic plaque cohorts.

CONCLUSIONS

In conclusion, our integrative omics study has identified an intraplaque hemorrhage-associated cardiovascular signature that provides excellent stratification of low- from high-risk carotid artery plaques in several independent cohorts. Further study revealed suppression of an SRF-regulated disease network, controlling lesion stability, in vulnerable plaque, which can serve as a scaffold for the design of targeted intervention in plaque destabilization.

Use of deep learning methods to translate drug-induced gene expression changes from rat to human primary hepatocytes

In clinical trials, animal and cell line models are often used to evaluate the potential toxic effects of a novel compound or candidate drug before progressing to human trials. However, relating the results of animal and in vitro model exposures to relevant clinical outcomes in the human in vivo system still proves challenging, relying on often putative orthologs. In recent years, multiple studies have demonstrated that the repeated dose rodent bioassay, the current gold standard in the field, lacks sufficient sensitivity and specificity in predicting toxic effects of pharmaceuticals in humans. In this study, we evaluate the potential of deep learning techniques to translate the pattern of gene expression measured following an exposure in rodents to humans, circumventing the current reliance on orthologs, and also from in vitro to in vivo experimental designs. Of the applied deep learning architectures applied in this study the convolutional neural network (CNN) and a deep artificial neural network with bottleneck architecture significantly outperform classical machine learning techniques in predicting the time series of gene expression in primary human hepatocytes given a measured time series of gene expression from primary rat hepatocytes following exposure in vitro to a previously unseen compound across multiple toxicologically relevant gene sets. With a reduction in average mean absolute error across 76 genes that have been shown to be predictive for identifying carcinogenicity from 0.0172 for a random regression forest to 0.0166 for the CNN model (p < 0.05). These deep learning architecture also perform well when applied to predict time series of in vivo gene expression given measured time series of in vitro gene expression for rats.

Heterogeneous Domain Adaptation for IHC Classification of Breast Cancer Subtypes

Increasingly, multiple parallel omics datasets are collected from biological samples. Integrating these datasets for classification is an open area of research. Additionally, whilst multiple datasets may be available for the training samples, future samples may only be measured by a single technology requiring methods which do not rely on the presence of all datasets for sample prediction. This enables us to directly compare the protein and the gene profiles. New samples with just one set of measurements (e.g., just protein) can then be mapped to this latent common space where classification is performed. Using this approach, we achieved an improvement of up to 12 percent in accuracy when classifying samples based on their protein measurements compared with baseline methods which were trained on the protein data alone. We illustrate that the additional inclusion of the gene expression or protein expression in the training process enabled the separation between the classes to become clearer.

Fluctuations of pol I and fibrillarin contents of the nucleoli

Nucleoli are formed on the basis of ribosomal DNA (rDNA) clusters called Nucleolus Organizer Regions (NORs). Each NOR contains multiple genes coding for RNAs of the ribosomal particles. The prominent components of the nucleolar ultrastructure, fibrillar centers (FC) and dense fibrillar components (DFC), together compose FC/DFC units. These units are centers of rDNA transcription by RNA polymerase I (pol I), as well as the early processing events, in which an essential role belongs to fibrillarin. Each FC/DFC unit probably corresponds to a single transcriptionally active gene. In this work, we transfected human-derived cells with GFP-RPA43 (subunit of pol I) and RFP-fibrillarin. Following changes of the fluorescent signals in individual FC/DFC units, we found two kinds of kinetics: 1) the rapid fluctuations with periods of 2-3 min, when the pol I and fibrillarin signals oscillated in anti-phase manner, and the intensities of pol I in the neighboring FC/DFC units did not correlate. 2) fluctuations with periods of 10 to 60 min, in which pol I and fibrillarin signals measured in the same unit did not correlate, but pol I signals in the units belonging to different nucleoli were synchronized. Our data indicate that a complex pulsing activity of transcription as well as early processing is common for ribosomal genes.

Separation of replication and transcription domains in nucleoli

In mammalian cells, active ribosomal genes produce the 18S, 5.8S and 28S RNAs of ribosomal particles. Transcription levels of these genes are very high throughout interphase, and the cell needs a special strategy to avoid collision of the DNA polymerase and RNA polymerase machineries. To investigate this problem, we measured the correlation of various replication and transcription signals in the nucleoli of HeLa, HT-1080 and NIH 3T3 cells using a specially devised software for analysis of confocal images. Additionally, to follow the relationship between nucleolar replication and transcription in living cells, we produced a stable cell line expressing GFP-RPA43 (subunit of RNA polymerase I, pol I) and RFP-PCNA (the sliding clamp protein) based on human fibrosarcoma HT-1080 cells. We found that replication and transcription signals are more efficiently separated in nucleoli than in the nucleoplasm. In the course of S phase, separation of PCNA and pol I signals gradually increased. During the same period, separation of pol I and incorporated Cy5-dUTP signals decreased. Analysis of single molecule localization microscopy (SMLM) images indicated that transcriptionally active FC/DFC units (i.e. fibrillar centers with adjacent dense fibrillar components) did not incorporate DNA nucleotides. Taken together, our data show that replication of the ribosomal genes is spatially separated from their transcription, and FC/DFC units may provide a structural basis for that separation.

[Evidence-based treatment of combined rotator cuff and SLAP lesions]

PURPOSE

In the face of improved radiological and arthroscopic techniques the diagnosis and treatment of SLAP lesions has recently gained much interest. Originally described as an (isolated) injury of the overhead athlete, it was only recently that the association of SLAP and rotator cuff defects was described in up to 40 % of cases. This study addresses the question of the evidence-based treatment of such frequent, combined lesions.

METHODS

Based on a systematic review of the online databases PubMed, EMBASE, CINAHL and Cochrane Library we identified clinical studies on the treatment of combined SLAP and rotator cuff lesions. Study quality was assessed using levels of evidence and a modified Jadad score. Clinical outcome was assessed through scores and range of motion assessments.

RESULTS

We included 7 studies of 374 patients with a mean age of 53 ± 11 years followed for 35 ± 13 months. Combined lesions have a significant negative effect on isolated rotator cuff or SLAP repair. Patients older than 45 years of age had a significantly better clinical result after biceps tenotomy than SLAP repair with concomitant rotator cuff repair. Biceps tenotomy plus rotator cuff repair showed significantly better range of motion for flexion and rotation than SLAP plus rotator cuff repair.

CONCLUSION

The frequent combination of SLAP and rotator cuff injury should be considered during assessment and informed consent of shoulder patients. While young patients and isolated SLAP lesions show excellent clinical results after elective repair, combined lesions should be treated with biceps tenotomy and/or debridement plus rotator cuff repair in patients older than 45 years.

Positioning of the NOR-bearing chromosomes in relation to nucleoli in daughter cells after mitosis

It is known that chromosomes occupy non-random positions in the cell nucleus. However, it is not clear to what extent their nuclear positions, together with their neighborhood, are conserved in daughter cells. To address specific aspects of this problem, we used the model of the chromosomes carrying ribosomal genes that are organized in clusters termed Nucleolus Organizer Regions (NORs). We compared the association of chosen NOR-bearing chromosomes (NOR-chromosomes) with nucleoli, as well as the numbers of nucleoli, in the pairs of daughter cells, and established how frequently the daughter cells had equal numbers of the homologs of certain NOR-chromosomes associated with individual nucleoli. The daughter cells typically had different numbers of nucleoli. At the same time, using immuno-FISH with probes for chromosomes 14 and 15 in HeLa cells, we found that the cell pairs with identical combinations appeared significantly more frequently than predicted by the random model. Thus, although the total number of chromosomes associated with nucleoli is variable, our data indicate that the position of the NOR-bearing chromosomes in relation to nucleoli is partly conserved through mitosis.

NORs and their transcription competence during the cell cycle

In human cells ribosomal genes are organized as clusters, NORs, situated on the short arms of acrocentric chromosomes. It was found that essential components of the RNA polymerase I transcription machinery, including UBF, can be detected on some NORs, termed "competent" NORs, during mitosis. The competent NORs are believed to be transcriptionally active during interphase. However, since individual NORs were not observed in the cell nucleus, their interphase status remains unclear. To address this problem, we detected the competent NORs by two commonly used methods, UBF immunofluorescence and silver staining, and combined them with FISH for visualization of rDNA and/or specific chromosomes. We found that the numbers of competent NORs on specific chromosomes were largely conserved in the subsequent cell cycles, with certain NOR-bearing homologues displaying a very stable pattern of competence. Importantly, those and only those NORs that were loaded with UBF incorporated bromo-uridine in metaphase after stimulation with roscovitine and in telophase, suggesting that competent and only competent NORs contain ribosomal genes transcriptionally active during interphase. Applying premature chromosome condensation with calyculin A, we visualized individual NORs in interphase cells, and found the same pattern of competence as observed in the mitotic chromosomes.

Dynamics of replication foci in early S phase as visualized by cross-correlation function

To monitor gradual changes in the replication foci distribution during early S phase, different segments of newly synthesized DNA were visualized by immunocytochemical mapping of two consecutively incorporated deoxythymidine analogs in pulse-chase-pulse experiments in HeLa cells. The resulting dual-labeled fluorescence images were evaluated using cross-correlation function (CCF) analysis. General changes of CCF shape due to image deterioration caused by blur, noise, and lateral sampling (pixel size) were also discussed. Using CCF analysis of model images simulating either random initiation of new replication foci, or the firing of new foci in close proximity to completed ones, we were able to ascribe the changes in the early S replication foci distribution to the latter mechanism. In contrast to the data published previously, we monitored the dynamics of all replication foci for up to 3 h. In addition, we showed that the replication foci dynamics is well described by random walk model, so that the average de-localization of individual foci is proportional to square root of the applied chase.