Abstract
Glucocorticoids are the cornerstone of immunosuppressive and anti-inflammatory therapy in humans, yet the mechanisms of glucocorticoid immunoregulation and toxicity remain unclear. The response to glucocorticoids is highly cell type-dependent, so translating results from different experimental systems into a better understanding of glucocorticoid effects in humans would benefit from rapid access to high-quality data on the response to glucocorticoids by different cell types. We introduce GCgx, a web application that allows investigators to quickly visualize changes in transcript abundance in response to glucocorticoids in a variety of cells and species. The tool is designed to grow by the addition of datasets based on input from the user community. GCgx is implemented in R and HTML and packaged as a Docker image. The tool and its source code are publicly available.
Seven decades after their introduction to clinical practice, glucocorticoids remain the cornerstone of immunosuppressive and anti-inflammatory therapy in humans (Cain & Cidlowski 2017). They are routinely employed in most fields of clinical medicine and are the drug class of choice when rapid and potent control of an overactive immune system is necessary, as exemplified most recently by their use in hospitalized patients with severe COVID-19 (RECOVERY Collaborative Group 2020). A growing body of experimental evidence indicates that the response to glucocorticoids is highly cell type-dependent (Franco et al. 2019). An important consequence of this is that caution must be exerted before extrapolating experimental findings made in one species or cell type to the expected effects in other systems. Cell lines and animal models have been valuable tools for advancing our understanding of the biochemistry and molecular biology of glucocorticoids and the glucocorticoid receptor. The important task of translating results obtained in such systems to a better understanding of glucocorticoid effects in humans would benefit from rapid access by scientists in the field to data on the response to glucocorticoids by individual cell types. With this goal in mind, we have created GCgx, a web tool that allows investigators to quickly visualize changes in transcript abundance in response to glucocorticoids in a variety of cells and to compare the response across cell types.
GCgx will offer users the option of selecting from a list of relevant datasets for visualization. The tool is designed to grow by the addition of datasets, based on input and prioritization by the user community. In the initial version of GCgx, the default dataset will be an extensive study of total RNA sequencing (RNA-seq) in nine human primary cell types which include B cells, CD4+ T cells, endothelial cells, fibroblasts, monocytes, myoblasts, neutrophils, osteoblasts, and preadipocytes. Each cell type was obtained from four unrelated healthy donors, treated in vitro with the synthetic glucocorticoid methylprednisolone or a vehicle control, and sampled serially at 2 and 6 h for RNA-seq (Franco et al. 2019). The examples below are based on this default dataset, but the same principles apply to other datasets available for display.
GCgx allows users to search for glucocorticoid-responsive genes (upregulated or downregulated in response to glucocorticoids) with user-defined thresholds of log fold-change and adjusted P -value for differential expression (Fig. 1A). Importantly, the tool allows for cell type-based searches that are positive or negative. In the default dataset, for example, a user can search for genes that show increased transcript abundance 6 h after glucocorticoid treatment in hematopoietic cells, but not in non-hematopoietic cells, with an adjusted P -value threshold of <0.05 (Fig. 1A, left). In this example, a table with the 23 genes that match the search criteria is displayed (Fig. 1A, right). The results can be downloaded in table format or can be used as input for the visualization tools in GCgx.
Visualization of changes in transcript abundance in response to glucocorticoids with GCgx. (A) A search tool allows users to identify glucocorticoid-responsive genes. User-defined parameters include the direction of change (upregulation or downregulation), the threshold to use for log2 fold-change and adjusted P -value for differential expression, and the specific cell types. In the example provided, these parameters were set to search for glucocorticoid-responsive genes that are upregulated at 6 h in hematopoietic cells, but not in non-hematopoietic cells, with an adjusted P -value lower than 0.05 and log2 fold-change greater than 0. A table displays all the genes that meet the search criteria. Search results can be saved or can be used as input for the GCgx visualization tools. A heatmap tool allows users to visualize the expression of sets of genes across multiple cell types, either at baseline (B) or in response to glucocorticoids (C). In the example provided, the results of the search for genes whose expression is induced by glucocorticoids in hematopoietic cells only (A) were used as input for heatmap visualization. A dot plot tool allows users to visualize the expression of individual genes and to assess the level of variation across biological replicates. Users can specify the RNA-seq read count normalization method (FPKM (fragments per kilobase of exon per million reads), TPM (transcripts per million reads), or median-of-ratios), the specific cell types, and whether to display the data by time point (D) or by cell type (E). In the example provided, one of the genes identified as glucocorticoid-induced in hematopoietic cells only, DYNLT1, is shown. In (D), the visualization parameters were set to display the expression of this gene across the nine cell types at baseline, with FPKM as the normalization method. In (E), the visualization parameters were set to display the transcript-level response of DYNLT1 to glucocorticoid treatment at 2 and 6 h, with FPKM as the normalization method, in three hematopoietic cells (CD4+ T cells, monocytes, and neutrophils) and three non-hematopoietic cells (endothelial cells, fibroblasts, and osteoblasts).
Citation: Journal of Molecular Endocrinology 68, 2; 10.1530/JME-21-0107
Visualization of changes in transcript abundance in response to glucocorticoids with GCgx. (A) A search tool allows users to identify glucocorticoid-responsive genes. User-defined parameters include the direction of change (upregulation or downregulation), the threshold to use for log2 fold-change and adjusted P -value for differential expression, and the specific cell types. In the example provided, these parameters were set to search for glucocorticoid-responsive genes that are upregulated at 6 h in hematopoietic cells, but not in non-hematopoietic cells, with an adjusted P -value lower than 0.05 and log2 fold-change greater than 0. A table displays all the genes that meet the search criteria. Search results can be saved or can be used as input for the GCgx visualization tools. A heatmap tool allows users to visualize the expression of sets of genes across multiple cell types, either at baseline (B) or in response to glucocorticoids (C). In the example provided, the results of the search for genes whose expression is induced by glucocorticoids in hematopoietic cells only (A) were used as input for heatmap visualization. A dot plot tool allows users to visualize the expression of individual genes and to assess the level of variation across biological replicates. Users can specify the RNA-seq read count normalization method (FPKM (fragments per kilobase of exon per million reads), TPM (transcripts per million reads), or median-of-ratios), the specific cell types, and whether to display the data by time point (D) or by cell type (E). In the example provided, one of the genes identified as glucocorticoid-induced in hematopoietic cells only, DYNLT1, is shown. In (D), the visualization parameters were set to display the expression of this gene across the nine cell types at baseline, with FPKM as the normalization method. In (E), the visualization parameters were set to display the transcript-level response of DYNLT1 to glucocorticoid treatment at 2 and 6 h, with FPKM as the normalization method, in three hematopoietic cells (CD4+ T cells, monocytes, and neutrophils) and three non-hematopoietic cells (endothelial cells, fibroblasts, and osteoblasts).
Citation: Journal of Molecular Endocrinology 68, 2; 10.1530/JME-21-0107
Visualization of changes in transcript abundance in response to glucocorticoids with GCgx. (A) A search tool allows users to identify glucocorticoid-responsive genes. User-defined parameters include the direction of change (upregulation or downregulation), the threshold to use for log2 fold-change and adjusted P -value for differential expression, and the specific cell types. In the example provided, these parameters were set to search for glucocorticoid-responsive genes that are upregulated at 6 h in hematopoietic cells, but not in non-hematopoietic cells, with an adjusted P -value lower than 0.05 and log2 fold-change greater than 0. A table displays all the genes that meet the search criteria. Search results can be saved or can be used as input for the GCgx visualization tools. A heatmap tool allows users to visualize the expression of sets of genes across multiple cell types, either at baseline (B) or in response to glucocorticoids (C). In the example provided, the results of the search for genes whose expression is induced by glucocorticoids in hematopoietic cells only (A) were used as input for heatmap visualization. A dot plot tool allows users to visualize the expression of individual genes and to assess the level of variation across biological replicates. Users can specify the RNA-seq read count normalization method (FPKM (fragments per kilobase of exon per million reads), TPM (transcripts per million reads), or median-of-ratios), the specific cell types, and whether to display the data by time point (D) or by cell type (E). In the example provided, one of the genes identified as glucocorticoid-induced in hematopoietic cells only, DYNLT1, is shown. In (D), the visualization parameters were set to display the expression of this gene across the nine cell types at baseline, with FPKM as the normalization method. In (E), the visualization parameters were set to display the transcript-level response of DYNLT1 to glucocorticoid treatment at 2 and 6 h, with FPKM as the normalization method, in three hematopoietic cells (CD4+ T cells, monocytes, and neutrophils) and three non-hematopoietic cells (endothelial cells, fibroblasts, and osteoblasts).
Citation: Journal of Molecular Endocrinology 68, 2; 10.1530/JME-21-0107
For simultaneous visualization of the response to glucocorticoids across multiple genes and cell types, GCgx can generate heatmaps. Lists of genes can be input manually, or as csv files (one gene symbol per row) which can be generated by the user or obtained directly from the output of the glucocorticoid-responsive gene search function of GCgx. Users can select the cell types to be displayed and can visualize the expression of their genes of choice at baseline or in response to glucocorticoids. In the example provided, the baseline expression across the 9 cell types of the 23 genes found by GCgx to be induced by glucocorticoid treatment in hematopoietic but not in non-hematopoietic cells, is shown in Fig. 1B. It is clear from this plot that a majority of the 23 genes are expressed at baseline in all cell types. A GCgx heatmap of the glucocorticoid response at 6 h for this set of 23 genes is shown in Fig. 1C. As expected, increased transcript abundance for this set of genes after glucocorticoid treatment is only evident in hematopoietic cells. It is also visually clear from this heatmap that the magnitude of the response varies by gene and cell type, highlighting the cell type dependence of the response to glucocorticoids.
For visualization of the transcript abundance of a single gene of interest before and after glucocorticoid treatment, including the level of variation across biological replicates, GCgx can generate dot plots. To facilitate comparison with other datasets, users can choose among three common RNA-seq normalization methods, that is FPKM, TPM, and the median-of-ratios method implemented in the R package DESeq2. Transcript abundance dot plots can be displayed by time point (Fig. 1D) or, depending on the dataset, by cell type (Fig. 1E). In the example shown, the expression of the gene DYNLT1 at baseline (Fig. 1D) or in response to glucocorticoid treatment is shown. DYNLT1 encodes one of the light chain polypeptides of the cytoplasmic dynein motor that performs retrograde transport of vesicles and organelles along microtubules. This and other non-catalytic components are thought to function by linking dynein to different forms of cargo and to adapter proteins that regulate dynein function. This gene has not been reported as a glucocorticoid target, as indicated by a PubMed search with the Boolean expression (DYNLT1 OR TCTEL1 OR TCTEX1) AND (glucocorticoid OR corticosteroid). This gene also provides a good example of the cell-type dependence of the glucocorticoid response, as it is expressed at the baseline in the nine cell types studied (Fig. 1B and D), yet induced by glucocorticoids only in hematopoietic cells, with a stronger induction in B cells, CD4+ T cells, and neutrophils than in monocytes (Fig. 1C and E). The level of variation of the expression values for this gene among the four biological replicates in this dataset is low (Fig. 1D and E). This example highlights the ability of GCgx to facilitate the discovery of new glucocorticoid-responsive genes in humans and to offer insight into the patterns of biological variation in the response.
The computational infrastructure behind GCgx is summarized in Fig. 2. The program is written in the R web application framework shiny (Chang et al. 2020), with additional custom html code, and it is packaged as a Docker image. It is hosted in the Amazon Web Services Elastic Container Service on an externally managed Elastic Compute Cloud (EC2) cluster. GCgx is built and deployed via the NIH NIAID Monarch Platform, which provides software tools, application programming interfaces, and automation frameworks for managing the application lifecycle. This computational infrastructure is designed to maximize security and longevity for GCgx and to simplify the incorporation over time of additional datasets that are relevant to the glucocorticoid research community. GCgx is publicly available free of charge and without the need for registration, at https://gcgx.niaid.nih.gov or through the NIAID Bioinformatics Portal, at https://bioinformatics.niaid.nih.gov/applications. The GCgx source code is hosted at the NIAID GitHub Enterprise Server and is also publicly available.
Topology diagram of GCgx. The source code is maintained in GitHub Enterprise and it is built and deployed via the NIAID Monarch platform. Users can freely access GCgx through the internet, and the application container runs behind an ALB that distributes incoming web requests. The application is hosted in AWS, on an externally managed EC2 cluster. The computational infrastructure behind GCgx is designed to maximize security, apply standardized and automated methods for quality assurance, maintenance and update, and facilitate the incorporation over time of additional datasets for query and visualization. ALB, Application Load Balancer; AWS, Amazon Web Services; ECS, Elastic Container Service; VPC, Virtual Private Cloud.
Citation: Journal of Molecular Endocrinology 68, 2; 10.1530/JME-21-0107
Topology diagram of GCgx. The source code is maintained in GitHub Enterprise and it is built and deployed via the NIAID Monarch platform. Users can freely access GCgx through the internet, and the application container runs behind an ALB that distributes incoming web requests. The application is hosted in AWS, on an externally managed EC2 cluster. The computational infrastructure behind GCgx is designed to maximize security, apply standardized and automated methods for quality assurance, maintenance and update, and facilitate the incorporation over time of additional datasets for query and visualization. ALB, Application Load Balancer; AWS, Amazon Web Services; ECS, Elastic Container Service; VPC, Virtual Private Cloud.
Citation: Journal of Molecular Endocrinology 68, 2; 10.1530/JME-21-0107
Topology diagram of GCgx. The source code is maintained in GitHub Enterprise and it is built and deployed via the NIAID Monarch platform. Users can freely access GCgx through the internet, and the application container runs behind an ALB that distributes incoming web requests. The application is hosted in AWS, on an externally managed EC2 cluster. The computational infrastructure behind GCgx is designed to maximize security, apply standardized and automated methods for quality assurance, maintenance and update, and facilitate the incorporation over time of additional datasets for query and visualization. ALB, Application Load Balancer; AWS, Amazon Web Services; ECS, Elastic Container Service; VPC, Virtual Private Cloud.
Citation: Journal of Molecular Endocrinology 68, 2; 10.1530/JME-21-0107
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the work reported.
Funding
This work was supported by the Intramural Research Programs at the National Institute of Arthritis and Musculoskeletal and Skin Diseases and the National Institute of Allergy and Infectious Diseases, NIH.
Author contribution statement
Qilin Cao: conceptualization, software, visualization, and writing – original draft. Yamil Boo Irizarry: software, writing – review and editing. Svetlana Yazhuk: software, resources, and writing – review and editing. Thai Tran: data curation, validation, and writing – review and editing. Manasi Gadkari: investigation, validation, and writing – review and editing. Luis M Franco: conceptualization, validation, visualization, and writing – original draft.
Acknowledgements
The authors would like to thank Michael A Dolan of the NIAID Bioinformatics and Computational Biosciences Branch and Christopher Campanale of the NIAID Operations and Engineering Branch for their work on the implementation and deployment of GCgx.
References
Cain DW & Cidlowski JA 2017 Immune regulation by glucocorticoids. Nature Reviews: Immunology 17 233–247. (https://doi.org/10.1038/nri.2017.1)
Chang W, Cheng J, Allaire JJ, Sievert C, Schloerke B, Xie Y, Allen J, McPherson J, Dipert A & Borge B 2020 shiny: Web Application Framework for R. R Package version 1.6.0. (available at: https://cran.r-project.org/package=shiny)
Franco LM, Gadkari M, Howe KN, Sun J, Kardava L, Kumar P, Kumari S, Hu Z, Fraser IDC & Moir S et al.2019 Immune regulation by glucocorticoids can be linked to cell type-dependent transcriptional responses. Journal of Experimental Medicine 216 384–406. (https://doi.org/10.1084/jem.20180595)
RECOVERY Collaborative Group, Horby P, Lim WS, Emberson JR, Mafham M, Bell JL, Linsell L, Staplin N, Brightling C & Ustianowski A et al.2021 Dexamethasone in hospitalized patients with Covid-19. New England Journal of Medicine 384 693–704. (https://doi.org/10.1056/NEJMoa2021436)
