Y.Z. I want to subset a specific cell type (cluster) and examine subtypes in this cell type. Qi, H., Liu, B., Wang, X. Commun. Nature 584, 437442 (2020). A recent question here gets into that particular problem a bit. Efficient recall of Omicron-reactive B cell memory after a third dose of SARS-CoV-2 mRNA vaccine. How to perform subclustering and DE analysis on a subset of an integrated object, Supervised clustering on a subset of integrated object (best practices?). 197, 10171022 (2016). Cell 185, 15881601.e14 (2022). ISSN 1529-2916 (online) f, Contour plots display FcRL4 expression in tonsillar and blood Bm cells gated as non-PB, non-GC (GC B cells identified as CD38+Ki-67+), IgD B cells and in tonsillar S+ Bm cells. Article ## [103] stringi_1.7.12 highr_0.10 desc_1.4.2 9c), indicating that S+ Bm cell subsets had comparable BCR repertoires, although the depth of our analysis was restricted by low cell numbers. | WhichCells(object = object, max.cells.per.ident = 500) | WhichCells(object = object, downsample = 500) | All plotting functions will return a ggplot2 plot by default, allowing easy customization with ggplot2. Now we can run a single integrated analysis on all cells! Is short-circuiting logical operators mandated? Standard edgeR workflow was used to create a linear model for the count data and to conduct statistical tests for differential segment usage between Bm cell subsets. 2b). Default is INF. No VH or VL chain segments were significantly differentially used between S+ Bm cell subsets. ## [124] gridExtra_2.3 parallelly_1.34.0 codetools_0.2-18 Altogether, these observations indicated that antigen reexposure by SARS-CoV-2 vaccination of SARS-CoV-2-recovered and SARS-CoV-2-vaccinated individuals stimulated S+ CD21CD27+ and CD21CD27 Bm cells. 6ac). Conversely, the frequency of S+ CD21CD27 Bm cells rose quickly and remained stable over 150days post-vaccination, accounting for about 20% of S+ Bm cells (Fig. Genewise statistics were conducted using empirical Bayes quasi-likelihood F-tests. If they had a confirmed SARS-CoV-2 infection and/or SARS-CoV-2 nucleocapsid-specific antibodies, they were considered SARS-CoV-2-recovered. Graphical representations were generated with BioRender.com. Med. I am also stuck on this issue too. The alternative would be to subset() the population of interest and run the complete preprocessing including integration only on those cells again. J. Exp. Thank you! d. Should ScaleData be run on the subset prior to PCA even though the subset comes from an integrated object prepped from SCT? In addition, since I am not integrating the subset, is it recommended to use the "scale.data" slot in the SCT assay for DE analysis or continue using the "data" slot in the SCT assay for this subset? SCT_not_integrated <- FindClusters(SCT_not_integrated) In d, severities were compared between the same timepoint using a Kruskal-Wallis test with a Dunns multiple comparison correction, with adjusted P values shown. My assumption was that it would start with 1 and if it does evaluate to "false" it would go on to 2 and than to 3, and if none matches the statement after == is "false" and if one of them matches, it is "true". SARS-CoV-2-nave healthy controls (n=11) were sampled before their SARS-CoV-2 mRNA vaccination, at week 2 post-second dose, month 6 post-second dose and at week 2 post-third dose. The flow cytometry and scRNA-seq subcohort characteristics are presented in Supplementary Tables 1 and 2, respectively. Generally, you'll want use different parameters for each sample. The pro of this approach is that it is fast and easy. Why are these constructs using pre and post-increment undefined behavior? ## [85] ragg_1.2.5 goftest_1.2-3 knitr_1.42 contributed reagents and interpreted data. Cell 177, 524540 (2019). d, Sorting strategy for S+ and S Bm cells, gated on CD19+ non-plasmablasts (non-PB, PB identified as CD38++CD27+) that were IgD and/or CD27+ and decoy, and for nave B cells, gated on CD19+ non-PB that were IgD+CD27 and S decoy. Lines connect paired samples. Analysis of SARS-CoV-2-specific GC Bcl-6+Ki-67+ B cells detected a trend towards elevated frequencies of S+ and N+ GC cells in recovered compared with vaccinated subjects (Extended Data Fig. Sci. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Connect and share knowledge within a single location that is structured and easy to search. Policy. What you could have written would have been something like: Which gives the same result as my earlier subset() call. RDocumentation. '||', where the operator is quoted. By clicking Post Your Answer, you agree to our terms of service, privacy policy and cookie policy. After determining the cell type identities of the scRNA-seq clusters, we often would like to perform a differential expression (DE) analysis between conditions within particular cell types. Functions reduce_dimension(), order_cells() and graph_test() were executed with default parameters. I did see batch effects here (cells from different batches did not share clusters). I just do not want to do manual subsetting on 10 genes, then manually getting @data matrix from each subset, and recreating seurat object afterwards. S+ Bm cells continued to show lower but still significantly increased proliferation at month 6, and only returned to background levels at month 12 post-infection (Fig. i, SHM counts are provided for nave B cells (n=1,607), blood (n=170) and tonsillar SWT+ Bm cells (n=1,128). Sci. I have a Seurat object that I have run through doubletFinder. 43, e47 (2015). Immunol. Can be used to downsample the data to a certain BCR variable gene segment usage was additionally quantified using the R package scRepertoire (v.1.3.5) (ref. Can the game be left in an invalid state if all state-based actions are replaced? 12, 6703 (2021). A multiple hypothesis correction procedure was applied to obtain adjusted P values. I have been subsetting a cluster from a Seurat object to find subclusters. 7, 83848410 (2021). Does it look right? Many thanks in advance. Using this subsetted data, I tried 4 different approaches: Approach 1: Default reintegration > Re-cluster (following, Approach 2: SCT reintegration > Re-cluster (following, Approach 3: No re-integration > Re-scale > Re-cluster (following, Approach 4: No re-integration > SC transform > Re-cluster (following. sessionInfo()## R version 4.2.0 (2022-04-22) Borcherding, N., Bormann, N. L. & Kraus, G. scRepertoire: an R-based toolkit for single-cell immune receptor analysis. Pape, K. A. et al. On what basis are pardoning decisions made by presidents or governors when exercising their pardoning power? Eight patients were vaccinated against SARS-CoV-2 (analyzed on average at day 144 after last vaccination), whereas the other eight patients were considered SARS-CoV-2-recovered based on a history of SARS-CoV-2 infection or positive anti-nucleocapsid (N) serum antibody measurement, with six of them additionally vaccinated against SARS-CoV-2 (assessed on average at day 118 post-last vaccination) (Extended Data Fig. batch effect correction), and to perform comparative scRNA-seq analysis of across experimental conditions. Sokal, A. et al. ), Deutsche Forschungsgemeinschaft (WA 1597/6-1 and WA 1597/7-1 to K.W. c, Frequency (median interquartile range) of S+ (left) and N+ (right) GC B cells within total B cells are given in tonsils of SARS-CoV-2-vaccinated and in recovered individuals. Google Scholar. And evaluation order? | object@data | GetAssayData(object = object) | 5c). Proc. 6f). Seurat continues to use t-distributed stochastic neighbor embedding (t-SNE) as a powerful tool to visualize and explore these datasets. We associated this with an incident during sample preparation in one of our experiments and decided to exclude most cells of this dataset from the analysis. All samples were analyzed by flow cytometry, and paired week 2, month 6 post-second dose and week 2 post-third dose samples from three patients were additionally assessed by scRNA-seq. 7, eabf5314 (2022). ; #323530-177975 to S.A.; #323530-191220 to C.C. Similar to issue #1547, (I ask because in the new integration vignette, it explicitly mentions not to run ScaleData after running the IntegrateData function)? Many, many thanks for the great package and continued support! As an aside, your middle two samples with a majority portion of cells with %mitochondrial reads > 10% are rather worrying, as they may largely be dead/dying. Making statements based on opinion; back them up with references or personal experience. b, Distribution of S+ Bm cell subsets in persistent and newly detected clones is shown at indicated timepoints. 7g). This work was funded by the Swiss National Science Foundation (#4078P0-198431 to O.B. 8b,c). Otherwise, will return an object consissting only of these cells, Parameter to subset on. The FCRL4hiENTPD1hiTNFRSF13Bhi cluster (cluster 6) probably represented the FcRL4+ B cell subset, and contained very few SWT+ Bm cells (Fig. 1e). But as you can see, %in% is far more useful and less verbose in such circumstances. Cao, J. et al. 59). the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in rev2023.4.21.43403. The DotPlot() function with the split.by parameter can be useful for viewing conserved cell type markers across conditions, showing both the expression level and the percentage of cells in a cluster expressing any given gene. But I am not sure which assay should be used for FindVariableFeatures of the subset cells, RNA, SCT, or Integrated? Notice that many of the top genes that show up here are the same as the ones we plotted earlier as core interferon response genes. 1g and Extended Data Fig. c, Heat map shows selected, significantly differentially expressed genes in indicated S+ Bm cell subsets. X-axis shows log-fold change and y-axis the adjusted P values (p<0.05 was considered significant). "~/Downloads/pbmc3k/filtered_gene_bc_matrices/hg19/", # Get cell and feature names, and total numbers, # Set identity classes to an existing column in meta data, # Subset Seurat object based on identity class, also see ?SubsetData, # Subset on the expression level of a gene/feature, # Subset on a value in the object meta data, # Downsample the number of cells per identity class, # View metadata data frame, stored in object@meta.data, # Retrieve specific values from the metadata, # Retrieve or set data in an expression matrix ('counts', 'data', and 'scale.data'), # Get cell embeddings and feature loadings, # FetchData can pull anything from expression matrices, cell embeddings, or metadata, # Dimensional reduction plot for PCA or tSNE, # Dimensional reduction plot, with cells colored by a quantitative feature, # Scatter plot across single cells, replaces GenePlot, # Scatter plot across individual features, repleaces CellPlot, # New things to try! Robbiani, D. F. et al. I have increased the resolution on FindClusters to analyze the integrated object and get my cluster of interested subclustered enough for DEG analysis but would simply like a new UMAP plot to visualize expression within that group of clusters. as.Seurat: Convert objects to 'Seurat' objects; as.SingleCellExperiment: Convert objects to SingleCellExperiment objects; as.sparse: Cast to Sparse; AugmentPlot: Augments ggplot2-based plot with a PNG image. Not the answer you're looking for? ## [70] labeling_0.4.2 rlang_1.0.6 reshape2_1.4.4 Gene set variation analysis with the package gsva (v1.42.0) was used to estimate gene set enrichments for more than two groups61. One limitation of our study is that we performed the clonal analysis after vaccination recall, because the numbers of S+ Bm cells during acute SARS-CoV-2 infection were too low for our sequencing approach. PubMed S+ CD21CD27+ activated Bm cells peaked in the first days post-vaccination, followed by a rapid decline over the subsequent 100days (Fig. ## [121] R6_2.5.1 promises_1.2.0.1 KernSmooth_2.23-20 But I'm also curious how others approach this! Weiss, G. E. et al. a, Scatter plot comparing binding scores (LIBRA-Score) was determined from scRNA-seq for SWT and RBD binding, with every dot representing a cell. ## [9] LC_ADDRESS=C LC_TELEPHONE=C But how do I subset a data before clustering? Samples in cf were compared using KruskalWallis test with Dunns multiple comparison, showing adjusted P values. 4c). If split.by is not NULL, the ncol is ignored so you can not arrange the grid. I would also like to know the recommended way of doing this. data.table vs dplyr: can one do something well the other can't or does poorly? (I assume if I just need to delete the 3 lines of code I just mentioned above and change I hope it is useful. The pro of this approach is that I use this method to solve the problem in the previous approach and now i have the genes that are primary markers for the cell sub types. Adv. The antibodies used are listed in Supplementary Tables 5 and 7. Compare: For your example, I believe the following should work: See the examples in ?subset for more. Goel, R. R. et al. 65). Cell 185, 18751887.e8 (2022). Zumaquero, E. et al. VH/VL were clustered hierarchically, with colors indicating frequencies. Sci. Of these, 35 received SARS-CoV-2 mRNA vaccination between month 6 and month 12, and 3 subjects between acute infection and month 6. e, Representative CD69 histograms in S+ Bm cells of patient CoV-T2 (left) and percentages of CD69+ S+ Bm cells (right) in blood and tonsils. Finally, we use a t-SNE to visualize our clusters in a two-dimensional space. Cells were sorted on a FACS Aria III 4L sorter using the FACS Diva software. When comparing dataset quality, we noticed a markedly lower median gene detection and unique molecular identifier count per cell in one of our datasets of the SARS-CoV-2 Infection Cohort. #2812 (comment). 6, 748 (2019). Another cohort (Extended Data Fig. 6, eabh0891 (2021). 4d). All study participants provided written informed consent. B cell clonality analysis was performed mainly with the changeo-10x pipeline from the Immcantation suite65 using the singularity image provided by Immcantation developers. You can read more on the concept here in Martin's paper. | WhichCells(object = object, ident = "ident.keep") | WhichCells(object = object, idents = "ident.keep") | I did integration with SCTransform. In this study, we demonstrated that individual clones of SARS-CoV-2-specific Bm cells harbored the capacity to follow phenotypically and functionally different trajectories after antigen reexposure, becoming CD21CD27+, CD21CD27 or CD21+CD27+/ Bm cells. It only takes a minute to sign up. Can the game be left in an invalid state if all state-based actions are replaced? On the basis of our data, we suggest a linearplastic model where the antigen stimulation and GC maturation of SARS-CoV-2-specific B cells resulted in the gradual adoption of a CD21+Ki-67lo resting Bm cell state at months 612 post-infection. Since the data I am analyzing comes from different diets as well as different batches, will batch-correction make me unable to determine differences in gene expression of cells from different diets? 5d,e). T-bet+ B cells are induced by human viral infections and dominate the HIV gp140 response. J. Subsetting the before integrating data to interested cells and then do the whole integration, followed by PCA, umap, findneighbors and findclusters seemed reasonale to me. 5a,b and Extended Data Fig. 4ac). The sequencing data have been deposited at Zenodo at https://doi.org/10.5281/zenodo.7064118. Cyster, J. G. & Allen, C. D. C. B cell responses: cell interaction dynamics and decisions. 124, 10171030 (1966). g, Heat map represents V heavy (VH) gene usage, in RBD+ and RBD Bm cells in scRNA-seq dataset from months 6 and 12. between condition A cluster 1 vs. condition B cluster 1 cells). After subsetting clusters of interest (subsetting by ident) I have a Seurat object with RNA, SCT and integrated assay, and dimensional reduction (pca, tsne, umap) coming from the original Seurat object. We performed scRNA-seq combined with feature barcoding, which allowed us to assess surface phenotype and to perform BCR-seq in sorted S+ Bm cells and S B cells from paired blood and tonsil samples of four patients (two SARS-CoV-2-recovered and two SARS-CoV-2-vaccinated). All tests were performed two-sided. New blog post from our CEO Prashanth: Community is the future of AI, Improving the copy in the close modal and post notices - 2023 edition, Manually define clusters in Seurat and determine marker genes, Trim Seurat object to contain expression info only for selected genes, Seurat VlnPlot presenting expression of multiple genes in a single cluster. Multi-Assay Features With Seurat, you can easily switch between different assays at the single cell level (such as ADT counts from CITE-seq, or integrated/batch-corrected data). Numbers indicate percentages of parent population. I think of this as if you FACS sorted cells and were to analyze them independently of the other cell types). Sci. f, Violin plots of IgG1+ (left) and IgG3+ percentages (right) are shown in each S+ Bm cell subset from the same samples as in e. g, Pie charts represent percentages of S+ Bm cells among all cells in scRNA-seq dataset, separated by Bm cell subsets. ; and #310030-200669 and #310030-212240 to O.B. Is it valid to set features.to.integrate to all the genes in the original Seurat object if I want run subclustering on the subset using its integrated assay? ## [130] mnormt_2.1.1 sctransform_0.3.5 multcomp_1.4-22 Why does Acts not mention the deaths of Peter and Paul? limma powers differential expression analyses for RNA-sequencing and microarray studies. Nat. and reading this issue I only got more confused. In e, two-sided Wilcoxon rank sum test was used and P values corrected by Bonferroni correction. Cells with LIBRA scores >0 for the respective antigens were defined as antigen-specific, and in the SARS-CoV-2 infection, cohort cells were considered S+ if any of the antigens used for baiting (SWT, Sbeta, Sdelta, RBD) were defined as specific. SubsetData( 1a). Gene expression data and TotalSeq surface proteome data were integrated separately. WNN clustering of all sequenced Bm cells identified ten clusters that, on the basis of the expression of cell surface markers and Ig isotype, were merged into five subsets annotated as CD21CD27+CD71+ activated Bm cells, CD21CD27FcRL5+ Bm cells, CD21+CD27 resting Bm cells, CD21+CD27+ resting Bm cells and unswitched CD21+ Bm cells (Fig. d, Clonality of S+ Bm cells was analyzed preVac and postVac in scRNA-seq dataset. 24, 389396 (2017). We found that SARS-CoV-2 infection and vaccination induced long-lived and stable antigen-specific Bm cells in the circulation that continued to mature up to 1year post-infection, as evidenced by their elevated proliferation rate at month 6, high SHM counts and improved breadth of SARS-CoV-2 antigen recognition. Seurat provides many prebuilt themes that can be added to ggplot2 plots for quick customization, | Theme | Function | Finally, we use a t-SNE to visualize our clusters in a two-dimensional space. filtration). Primary Handling Editor: Ioana Visan in collaboration with the Nature Immunology team. We thank the patients for their participation in our study, S. Hasler for assistance with patient recruitment, L. Brgi and R. Masek for help with sample processing, the Departments of Otorhinolaryngology and Anesthesiology, the Transplantation Immunology Laboratory of University Hospital Zurich, E. Baechli, A. Rudiger, M. Stssi-Helbling and L. Huber for help with patient recruitment, the Functional Genomics Center Zurich and Genomics Facility Basel for help with sample preparation and next-generation sequencing, and S. Chevrier, D. Pinschewer, L. Ceglarek, D. Caspar and the members of the Boyman and Moor Laboratories for helpful discussions. I wanted to base an analysis on data that that was matching one of a few criteria, e.g. to your account. Choose a subset of cells, and then split by samples and then re-run the integration steps (select integration features, find anchors and integrate data). Bm cells are colored by cluster (f, left), tissue origin (f, right) or SWT binding (g). Colors indicate frequency within RBD+ and RBD Bm cells. ## [31] xfun_0.37 dplyr_1.1.0 crayon_1.5.2 Making statements based on opinion; back them up with references or personal experience. ## [11] ifnb.SeuratData_3.1.0 hcabm40k.SeuratData_3.0.0 By using uniform manifold approximation and projection (UMAP) we visualized S+ Bm cells from the flow cytometry dataset obtained in nonvaccinated post-infection samples and performed a PhenoGraph clustering (Extended Data Fig. Nat. Med. Frequencies in g were compared using two-proportions z-test with Bonferronis multiple testing correction. I want to know: This is because the RNA slot is a true representative of biological variation, when someone tries to reproduce your findings they won't perform a negative binomial regression on their PCR. b, Cohort overview of SARS-CoV-2 Tonsil Cohort. Content Discovery initiative April 13 update: Related questions using a Review our technical responses for the 2023 Developer Survey, R: subsetting data frame by both certain column names (as a variable) and field values. e, Heat map shows enrichment scores of selected gene sets that are significantly different between CD27lo/hiCD21+ resting and CD21CD27FcRL5+ S+ Bm cell subsets in a pseudobulk analysis (n=5 individuals). Asking for help, clarification, or responding to other answers. We observed a strong increase in the frequency of S+ and RBD+ Bm cells in SARS-CoV-2-infected individuals at months 6 (median 0.14% and 0.033%, respectively) and 12 post-infection (median 0.068% and 0.02%) compared with acute infection (median 0.016% and 0.0023%) (Fig. and JavaScript. ), Digitalization Initiative of the Zurich Higher Education Institutions Rapid-Action Call #2021.1_RAC_ID_34 (to C.C. M.E.R. | rownames(x = object@data) | rownames(x = object) | Identification of resident memory CD8+ T cells with functional specificity for SARS-CoV-2 in unexposed oropharyngeal lymphoid tissue. f, Violin plots of percentages of Ki-67+ S+ Bm cells are shown at indicated timepoints. Any argument that can be retreived b, Distribution of S+ Bm cell subsets is provided at month 6 preVac, month 12 nonVac and month 12 postVac. Are || and ! Ritchie, M. E. et al. (default), then this list will be computed based on the next three Lung-resident memory B cells established after pulmonary influenza infection display distinct transcriptional and phenotypic profiles. Differential gene expression identified higher expression of CR2, CD44, CCR6 and CD69 in tonsillar SWT+ Bm cells compared with blood SWT+ Bm cells, whereas the activation-related genes FGR and CD52 were higher in blood SWT+ Bm cells compared with their tonsillar counterparts (Extended Data Fig. 5a and Extended Data Fig. GOPB, Gene Ontology Biological Process. # S3 method for Assay As cell identity is only available after intergration and clustering? Shared transcriptional profiles of atypical B cells suggest common drivers of expansion and function in malaria, HIV, and autoimmunity. ## [43] future.apply_1.10.0 BiocGenerics_0.44.0 abind_1.4-5 BCR diversity was slightly reduced in S+ CD21CD27FcRL5+ compared with S+ CD21+ resting Bm cells (Extended Data Fig. The joint analysis of two or more single-cell datasets poses unique challenges. Comprehensive analyses of B-cell compartments across the human body reveal novel subsets and a gut-resident memory phenotype. Samples in f were compared using two-proportions z-test. ## [133] parallel_4.2.0 grid_4.2.0 tidyr_1.3.0 ), A vector of cell names to use as a subset. ), Forschungskredit Candoc grant from UZH (FK-20-022; to S.A.), Young Talents in Clinical Research program of the SAMW and G. & J. Bangerter-Rhyner Foundation (YTCR 08/20; to M.E.R. Whereas S+ Bm cells were predominantly resting CD21+ Bm cells at month 6, vaccination strongly induced the appearance of S+ CD21CD27+ and CD21CD27 Bm cells in blood (Fig. Analysis of V heavy and light chain frequencies identified several chains enriched in RBD+ Bm cells compared with RBD Bm cells described to encode RBD-binding antibodies, including IGHV3-30, IGHV3-53, IGHV3-66, IGKV1-9 and IGKV1-33 (refs. Assa Yeroslaviz 1.8k. Unswitched CD21+ Bm cells were IgM+, whereas the other Bm cell subsets expressed mainly IgG, with IgG1 being the dominant subclass (Extended Data Fig. We found indication of increased BCR and IFN- signaling in S+ CD21CD27 Bm cells, in accord with the increased expression of T-bet and the T-bet target genes ZEB2 and ITGAX30. Academic theme for Collectively, these data identify a durable, IgG1-dominated S+ Bm cell response forming upon SARS-CoV-2 infection. This is in line with previous reports that SARS-CoV-2 infection and mRNA vaccination led to lasting Bm cell maturation through an ongoing GC reaction26,44,45,46. 4e). All the best, The SWT+ Bm cells in the IgG+CD27hiCD45RBhi cluster (cluster 5) were mainly from blood, in the IgG+CD21hi cluster (cluster 2) predominantly tonsillar, while the IgG+CD27lo cluster (cluster 4) contained SWT+ Bm cells from both compartments. Jordan. # One of these Assay objects is called the "default assay", meaning it's used for all analyses and visualization. Wang, Z. et al. Policy. Pseudotime-based trajectory analysis using Monocle 3 in our scRNA-seq dataset (Extended Data Fig. Invest. :) Thank you. English version of Russian proverb "The hedgehogs got pricked, cried, but continued to eat the cactus", Effect of a "bad grade" in grad school applications. At the transcriptional level, S+ Bm cells at month 6 post-infection upregulated genes associated with B cell activation and recent GC emigration35, such as NKFBIA, JUND, MAP3K8, CXCR4 and CD83, compared with S+ Bm cells at month 12 (Extended Data Fig. So, my here is my workflow: Mean diversity index (line) and confidence intervals (transparent shadings) are shown. k, Venn diagram shows clonal overlap of SWT+ and SWT Bm cells in tonsils and blood from scRNA-seq dataset. Find centralized, trusted content and collaborate around the technologies you use most. In the scRNA-seq dataset, CD21+CD27+ resting Bm cells were the main S+ Bm cell subset at months 6 and 12 post-infection in nonvaccinated individuals, whereas CD21CD27+CD71+ activated and CD21CD27FcRL5+ Bm cells became predominant post-vaccination at month 12 post-infection (Fig. ## Platform: x86_64-pc-linux-gnu (64-bit) 15, 149159 (2015). I can figure out what it is by doing the following: We did not assume normal distribution for the flow cytometry data and used nonparametric tests such as KruskalWallis to test for differences between continuous variables in more than two groups, and P values were adjusted for multiple testing using Dunns method.