Study approval

All experiments were performed in accordance with protocols approved by the University of North Carolina Institutional Animal Care and Use Committee (application number 14-001). Prior to sample collection, all human patients signed informed consent under Duke University IRB study protocol Pro00110250. Healthy donor control cells were purchased from Memorial Blood Center (St. Paul, MN) where all commercially available products are collected with IRB approval or exemption.

C57BL/6 (strain 0000664) and C57BL/6JDBA/2 F1 (B6D2, strain 100006) mice were purchased from The Jackson Laboratory. The generation of enhanced GFP-expressing C57BL/6 mice has been described previously26. Donor and recipient mice were age-matched males between 8 and 16 weeks. Animals were housed under specific pathogen-free (SPF) conditions on a 12-h dark/light cycle at 2122C and 3070% humidity. Where applicable in all mouse studies, animals were euthanized via CO2-compressed carbon dioxide gas in cylinders followed by physical cervical dislocation to ensure death.

Eight- to 16-week-old B6 mice were given 0.4g recombinant mouse IL-17E/IL-25 (R&D Systems) by i.p. injection for 4 days. On day 5, cells were isolated from the mesenteric lymph nodes and peritoneum by peritoneal lavage using RPMI-1640 supplemented with 10% FBS, 2mM L-glutamine, 12mM HEPES, 0.1mM non-essential amino acids, 1mM sodium pyruvate, 1% Pen/Strep, and 50M 2-mercaptoethanol (complete media). ILC2s were isolated by negative selection with a MACS column using the following antibodies (anti-CD8 [clone 53-6.7], anti-CD4 [RM 4.4], anti-CD3 [clone 145-2C11], anti-TCR [UC7-13DS], anti-TER119 [TER-119], anti-B220 [RA3-6B2], anti-CD11b [M1/70], anti-NK1.1 [PK136], eBioscience; anti-CD11c [N418], anti-CD19 [MB19-1], anti-Ly6G [1A8], and anti-CD49b [DX5], BioLegend) and Streptavidin Microbeads (Miltenyi 130-048-101). Cells were expanded in culture at 2.25x105 cells/mL in 24 well flat bottom TC-treated plates or flasks (Corning) as described below.

ILC2s were cultured at 2.25x105 cells/mL for 6 days in complete media (RPMI-1640 supplemented with 10% FBS, 2mM L-glutamine, 12mM HEPES, 0.1mM non-essential amino acids, 1mM sodium pyruvate, 1% Pen/Strep, and 50M 2-mercaptoethanol) and supplemented with 10ng/ml rIL-7 and rIL-33 (PeproTech), with the media changed every 2 days. ILC2 activation was evaluated using flow cytometry on day 6 by surface and intracellular cytokine staining with antibodies against Lineage (eBioscience) and ST2 (Thermo Scientific). For experiments in which cells were generated via cytokine-mediated skewing (pcILC2s), cells were cultured at 2.25x105 cells/mL for 48h in complete media supplemented with 10ng/ml rIL-7 and rIL-33 (PeproTech). On Days 2 and 4, the media was replaced with complete R10 containing 10ng/ml rIL-7, 10ng/ml rIL-12, 10ng/ml rIL-1b, 10ng/ml rIL-15, 10ng/ml rIL-2, and 5ng/ml rIL-18.

Non-mobilized healthy donor peripheral blood (PB) leukapheresis products were purchased from Memorial Blood Center (St. Paul, MN), and human ILC2 cells were captured via the RosetteSep human ILC2 enrichment kit (STEMCELL Technologies), per the manufacturers instructions.

ILC2s were cultured at 2.25x105 cells/mL for 6 days in complete media (RPMI-1640 supplemented with 10% FBS, 2mM L-glutamine, 12mM HEPES, 0.1mM non-essential amino acids, 1mM sodium pyruvate, 1% Pen/Strep, and 50M 2-mercaptoethanol) and supplemented with 10ng/ml rIL-7 and rIL-33 (PeproTech), with the media changed every 2 days. ILC2 activation was evaluated using flow cytometry on day 6 by surface and intracellular cytokine staining with antibodies against Lineage (eBioscience) and ST2 (Thermo Scientific). For experiments in which cells were generated via cytokine-mediated skewing (pcILC2s), cells were cultured at 2.25x105 cells/mL for 48h in complete media supplemented with 10ng/ml rIL-7 and rIL-33 (PeproTech). On Days 2 and 4, the media was replaced with complete R10 containing 10ng/ml rIL-7, 10ng/ml rIL-12, 10ng/ml rIL-1b, 10ng/ml rIL-15, 10ng/ml rIL-2, and 5ng/ml rIL-18.

The phenotype and function of murine ILC2s were evaluated by flow cytometry with antibodies as listed in Supplementary Data2. Prior to transplantation T cells were evaluated by surface staining of CD4 (GK1.5) and CD8 (clone 53-6.7). Sample acquisition was performed using a BD LSRII or BD LSRFortessa (BD Bioscience) or a MACS Quant (Miltenyi Biotec), and data were analyzed by FlowJo v9/10 (TreeStar, BD) and Prism v10 (GraphPad).

Total T cells were isolated using a Cedarlane T cell recovery column kit (Cedarlane Laboratories), followed by antibody depletion using PE-conjugated anti-mouse B220 (RA3-B62) and anti-mouse CD25 (3C7) antibodies (eBioscience) and magnetic bead selection using anti-PE beads (130-0480801, Miltenyi Biotec). TCD bone marrow was prepared as described previously27. The day prior to transplantation, recipient mice received 950cGy of total body irradiation. Recipients were intravenously injected with 4106T cells and 3x106 TCD BM cells. For ILC2 treatment groups, B6D2 recipients also received 3 4x106 ILC2s, respectively. Recipients were monitored three times a week and scored for clinical GVHD symptoms (designated clinical score) using a semiquantitative scoring system as previously described28,29; animals were coded for these evaluations. The sample size was chosen for the effect size needed based on our previous experience with sample sizes needed to demonstrate a significant difference in GVHD scoring between control and treated groups. For the scoring evaluation experiments, the inclusion of 912 recipients provided a power of 90% to detect a difference of 14 days in the median GVHD score of 5 with an error of <0.05 between control and treated groups. For all experiments, a control group received TCD bone marrow alone without additional T cells, which controlled for the presence of T cells in the marrow inoculum and potential infectious complications during aplasia.

Animals were euthanized with CO2 followed by cervical dislocation,and spleen, liver, lungs, mesenteric lymph nodes (mLN), and lamina propria (LP) were excised. LP lymphocytes were isolated using the Miltenyi LP dissociation kit (catalog 130-097-410) as per the manufacturers instructions. Livers and lungs were digested in a solution of 1mg/ml collagenase A (Roche) and 75 U DNase I (Sigma-Aldrich) in RPMI 1640 with 5% newborn calf serum. Digested tissues were treated with ACK lysis buffer to remove RBCs and were passed through 100m cell strainers. Leukocytes were collected at the interface of a 40%:80% Percoll (Sigma-Aldrich) gradient in RPMI 1640 with 5% NCS. The pelleted cells were washed in 1x DPBS with 2% FBS. Spleens and mLN were teased apart, treated with ACK lysis buffer, and washed in 1 x DPBS with 2% FBS.

Animals were sacrificed, and lymphocytes were isolated from the SI, mLN, and IP lavage as described above. Single-cell suspensions were stained with an e450 Lineage antibody cocktail (Invitrogen, 88-7772-72) and BD Horizon AlexaFluor 700 Fixable Viability Stain (BD 564997). Cells were sorted based on GFP expression on a BD FACSAria II (BD Bioscience), and GFP+ cells were collected into cR10 prior to downstream processing.

Peripheral blood samples were collected from 12 adults who underwent HSCT at the Duke Adult Bone Marrow Transplant Clinic in Durham, NC between January 2015 and April 2017. All patients signed informed consent under Duke University IRB study protocol Pro00110250. Of the patients who remained stable after allo-HSCT, two out of three received myeloablative conditioning regiments while one underwent non-myeloablative conditioning. Similarly, of the patients who remained went on to experience an episode of acute GVHD in the first four months following their allo-HSCT, two out of three received myeloablative conditioning regiments while one underwent non-myeloablative conditioning. The average patient age at the time of HSCT was 56 years, and all included patients had a transplant indication of myelodysplastic syndrome. Sex and/or gender were not considered in the study design as specimen selection was limited to a small pool of experimentally eligible samples. All patients received calcineurin inhibition and short-course methotrexate for GvHD prophylaxis. We requested samples from 36 patients after alloHSCT with a diagnosis of acute graft-versus-host disease and 36 that were stable. In addition, we requested a pre-HSCT sample and then one drawn as close as possible to the time of the aGVHD diagnosis, as well as 3 months and 1 year where available.

2.55x106 cells were pelleted prior to fixation with and fixed with a 1% formaldehyde using the ChIP-IT High Sensitivity Kit (Active Motif). After quenching and washing, pellets were frozen at 80C. Upon thawing, cells were sheared using a chilled dounce homogenizer using with the ChIP-IT High Sensitivity Kit (Active Motif), and lysed cells were sonicated with nanodroplets for 60s with the Covaris Le220 prior to clarified by centrifuging at full speed for 15min. Input DNA was prepared with RNase A and Proteinase K prior to clean up and concentration with the Zymo Chip DNA Clean and Concentrate kit. 1030ug of ILC chromatin were treated with an anti-H3K4me3 antibody (Cell Signaling Technologies, rabbit mAb 9751, clone C42D8)+blocker mix along with a protease inhibitor cocktail prior to overnight end-to-end rotation at 4C. 30L of Protein G Dynabeads were added to each 240L immunoprecipitation reaction, and reactions were incubated for 3h. Samples were passed through a ChIP Filtration Column and then washed 5x with Wash Buffer AM1. After the removal of the residual wash buffer, samples were eluted in pre-warmed elution buffer AM4. De-crosslinking was carried out for 2.5h with Proteinase K prior to DNA extraction with the ChIP DNA Clean & Concentrator (Zymo Research) protocol. Chromatin immunoprecipitation quantitative real-time PCR (ChIP-qPCR) was performed using the QuantStudio 6K from Applied Biosystems.

Cells of interest were prepared in suspension and nuclei were isolated by pelleting 25,000200,000 in a fixed-angle centrifuge. Cells were lysed with 10mM Tris-HCl (pH 7.4), 10mM NaCl, and 3mM MgCl2 prior to 30min of transposition with an in house Tn5 transposase (generated at UNC CICBDD with Addgene construct #60240 as adapted from Picelli et al., Genome Res, 2014)30 at 37C. Immediately following transposition, samples were purified with Zymo Conc & Clean (Genesee) and stored at 20C prior to library amplification. Fragments were amplified using 1 NEBnext PCR master mix (New England BioSciences) and custom Nextera PCR primers 1 and 2 (Illumina, Supplementary Data3). Full libraries were amplified for five cycles (Thermo Fisher), after which a test aliquot of each sample was taken to test 20 cycles to determine the additional number of cycles needed for the remaining 45L reaction (QuantStudio 6K, Applied Biosystems). After the additional cycles were complete (average of 515 additional cycles), libraries were purified using a Zymo Conc & Clean (Genesee) prior to a two-sided Ampure bead (Beckman Coulter) size selection to enrich nucleosome-free fragments. Fragment size and concentration were determined by TapeStation 2000 and Qubit 4 (Agilent). Following QAQC, samples were sequenced on the Illumina HiSeq4000 (75x paired-end HO).

Following FACS sorting, paired multiome Single Cell ATAC (scATAC) and Single Cell Gene Expression (scGEX) libraries were prepared using the 10x Chromium Single Cell Multiome ATAC+Gene Expression kit (CG000338) with the low cell input nuclei isolation protocol (CG000365 Rev C). Lysis buffer strength was 1X, and lysis time was 4min. Multiome ATAC libraries (50x8x24x9) and GEX libraries (28x8x24x9) were sequenced at a depth of 2550,000 read pairs per cell on either the Illumina NextSeq2000P2 or Illumina NovaSeq SP.

The expression of Type 1 and Type 2 lineage-defining and lineage-associated genes was assessed in ex vivo expanded ILC2s and pcILC2s via quantitative real-time polymerase chain reaction (qRT-PCR) was performed with the RT2 Profiler PCR Array system (Qiagen, Hilden, Germany). Briefly, ILC2 and pcILC2 cells were cultured as described above and RNA was extracted after 6 days of expansion with the RNeasy Mini Kit (Qiagen). cDNA synthesis and genomic DNA elimination were performed with the RT2 First Strand Kit (Qiagen). cDNA was applied to the RT Profiler PCR Array Mouse Th1 & Th2 Response kit (PAMM-034Z; Qiagen) and PCR amplification was performed using RT2 SYBR Green qPCR Mastermix (Qiagen). Raw CT values were uploaded to Qiagens web-based software GeneGlobe, wherenormalization was performed by comparing to the internal housekeeping gene panel, and relative mRNA expression in the ILC2s and pcILC2s was calculated via the CT method.

Adapter sequences were removed from reads using cutadapt (v. 1.12). Reads were quality filtered using FASTX-ToolKit (v0.0.12) passing options Q 33, -p 90, and q 20. Reads were aligned to the mm10 genome using STAR (v2.5.2b) with the options: --outFilterScoreMin 1, --outFilterMultimapNmax 1 --outFilterMismatchNmax 2, --chimJunctionOverhangMin 15, --outSAMtype BAM Unsorted, --outFilterType BySJout, --chimSegmentMin 1. Read per million (RPM) normalized H3K4me3 signal at gene (RefSeq) promoters was calculated using deepTools (v.3.5.2), with the promoter region being defined as +300bp from mm10 RefSeq TSS to 500bp. Published H3K4me3 fastq files were downloaded from a public repository (GEO: GSE85156)8 and were aligned to the mm10 genome as described above. H3K4me3 peaks were identified using MACS2 (v.2.1.2, using default parameters)9 for all subpopulations of ILCs (two replicates each). Differential H3K4me3 marked promoters were identified through the application of DEseq2 (Likelihood ratio test, v.1.40.2)31 on the union set of peaks. Multiple testing was mitigated using the Benjamini and Hochberg method. ILC1, ILC2, and ILC3-specific H3K4me3 marked promoters were identified by hierarchical clustering of differential promoters. Representative tracks were created by passing RPM normalized bigwig files to the UCSC Genome Browser.

Data from Bruce et al. J Clin Invest, 127(5):1813-1825. doi: 10.1172/JCI91816, PMID: 28375154. FASTQ files were downloaded from GEO (GSE95811)4. Adapter sequences were removed using (cutadapt v1.12). Reads were quality filtered using fastq quality filter in FASTX-Toolkit (v0.0.12) with options -Q 33, -p 90, and q 20. Reads were aligned to mm10 genome using STAR (v2.5.2b) with the following options: --quantMode TranscriptomeSAM, --outFilterMismatchNmax 2, --alignIntronMax 1000000, --alignIntronMin 20, --chimSegmentMin 15, --chimJunctionOverhangMin 15, --outSAMtype BAM Unsorted, --outFilterType BySJout, --outFilterScoreMin 1. Transcript-level expression values (TPM) were estimated using Salmon (v0.11.3) for RefSeq transcripts. TPM data was collapsed into gene-level expression values using a tximport. Refseq was used for gene annotation.

Sequence adapters were trimmed using cutadapt (v. 1.12) with the following options -a CTGTCTCTTATA -A CTGTCTCTTATA and --minimum-length 36 and set for paired-end sequence data. Reads were then quality filtered using fastq_quality_filter in FASTX-Toolkit (v. 0.0.14), with the options -Q 33, -p 90, and -q 20. PCR duplicates were restricted by limiting the same sequence to a max of 5 copies. After this duplicate threshold was reached, the remaining reads were excluded (In house scripts). As one read of a read-pair may have been excluded during filtering, fastq files were synced following the filtering step to ensure accurate alignment. Reads were aligned to either mm10 or hg38 using STAR (v2.5.2b) with options chimSegmentMin 15, --outFilterMismatchNmax 2, --chimJunctionOverhangMin 15, --outSAMtype BAM Unsorted, --outFilterScoreMin 1, --outFilterType BySJout and --outFilterMultimapNmax 1. Samtools (v. 1.3.1) and bedtools (v. 2.26) were then used to create bigwig files. These files only include fragment data for the first 5bp from the 5 end and the last 4bp from the 3 end of fragments. Regions of chromatin accessibility (peaks) were defined using MACS (v. 2.1.2) with options nomodel, --shift 0, --extsize 5. Sites of differential chromatin accessibility were identified using DESeq2 (v. 1.40.2) with default parameters using a union set of peaks. Motif analysis was performed using findMotifsGenome.pl from HOMER (v. 4.11.1). Representative tracks of normalized ATAC signal were generated by visualizing bigwig files on the UCSC genome browser.

From Gury-BenAri et al. Cell, 2016, 166(5):1231-1246.e13, doi: 10.1016/j.cell.2016.07.043, PMID: 27545347. The UMI count table was downloaded from GEO (GSE85152). The count table was calculated as previously described8. The first gene name was selected to represent features when more than one possible annotation was available. Cells with less than 200 UMIs were discarded from downstream analysis.

Paired-end FASTQ files were pre-processed and aligned to the mouse reference genome (GRCm38/mm10) using cellranger-arc count (v. 2.0.0) with default parameters. Next, we selected nuclei with a total RNA read count <25,000, total RNA read count >1000, total ATAC read count <70,000, total ATAC read count >5000, and mitochondrial counts <20% for downstream analysis. We then integrated the datasets using RNA data only or with both RNA and ATAC. Integration using RNA was accomplished using the IntegrateData function within Seurat passing default parameters (v. 4.3.0.1)10. The dimensionality of the data was reduced using principal component analysis (PCA), and Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP). Cells were then clustered using the FindClusters function within Seurat. Prior to the integration of multi-modality data, the RNA and ATAC count matrices for each sample were merged. The merge function within Seurat was used to concatenate the RNA count matrices. To create the merged ATAC count matrix, we first identified a union set of peaks across all samples and calculated the number of ATAC fragments overlapping those shared genomic coordinates for each sample using FeatureMatrix (Signac v. 1.10.0). The RNA and ATAC count matrices were then merged by sample, and the data were integrated using the Weighted nearest neighbor analysis (Seurat) to develop shared inference32. Differentially expressed genes were identified using a Wilcoxon rank sum test as implemented in the package presto (v. 1.0.0) or with the FindMarkers function in Seurat. Heatmap of differentially expressed genes depicts mean-centered and normalized RNA abundance. Normalized (SCTransform) and uncorrected (pre-integration) RNA counts were used for differential testing. A 2-cell cluster (pre- and post-transplant mouse data, AAGGATTAGCTCATAA-1_2, AGTGGACAGCTATTAG-1_2) was identified and excluded from downstream analysis. To determine differential regions of chromatin accessibility, we also used a Wilcoxon rank sum test (presto, v. 1.0.0) on TF-IDF normalized ATAC counts. Heatmap of differential regions of chromatin accessibility depicts mean-centered and normalized ATAC signal. Per-cell accessibility for motifs (motif enrichment scores) was calculated using the package chromVAR (v. 1.22.1). The JASPER 2020, CORE, vertebrate motif database was used during motif identification. To identify putative TF regulators for each cluster of cells, the AUC statistics calculated by presto during the differential expression and differential motif enrichment analysis (Wilcox rank sum test (p adj <0.05, RNA.logFC> 0 and motif.p adj <0.05, motif.logFC> 0)) were averaged, and the TFs with the highest value were considered top candidates. Average per-cell chromatin accessibility (ATAC signal) across genes previously associated with ILC1 and ILC2 isolated from small intestinal lamina propria of mice8 was calculated using the FeatureMatrix from the Signac Package. For ontology analysis, we used Enrichr33 and tested for significant overlap between genes associated with post-transplant cells and gene sets found in the mouse gene atlas database. The average marker expression of ILC1, ILC2, ILC3, and NK cells was defined by averaging the normalized gene expression for genes previously associated with each cell type11. Classification of post-transplant cells as ILC1, ILC2, or ILC3 was based on lineage defining genes previously associated with each type of ILC8. Seurat was used to integrate post-transplant single nucleus RNA data with previously published8 MARS-seq RNA abundance data as previously described10. Cells were classified by calculating the average rank of genes associated with each type of ILC and labeling each cell with the cell type with the highest rank. Undetermined signifies a tie in ranking. Upset plots were created using UpSetR (v. 1.4.0).

Paired-end FASTQ files were pre-processed and aligned to the mouse (GRCm38/mm10) or human reference genome (GRCh38/hg38) using cellranger-atac count (v. 2.0.0) with default parameters. To compare the average normalized ATAC signal across each patient sample at hILC2 sites of CA (defied with bulk ATAC data), we first calculated the number of ATAC fragments overlapping the 7456 associated sites of CA using FeatureMatrix (Seurat, v. 4.3.0.1), respectively. We then merged these count matrices and normalized the data using TF-IDF as implemented in Seurat (v. 4.3.0.1). The same operations were performed for the pc-hILC2 associated sites of CA (n=5,051) as well as random sites (n=7,456) of chromatin accessibility shared between hILC2 and pc-hILC2s. Samples were grouped into six patient conditions, four according to transplant status and aGVHD, two healthy donors, and expanded ILC2s derived from an additional single healthy donor. Differential ATAC was tested using patient condition as a categorical variable in a linear model. The model was fit in R and the coefficients were evaluated to highlight different effects for patient conditions. We then grouped the sample into four categories: healthy donor, ILC2, pre- and post-transplant samples (independent of aGVHD) and repeated the above analysis. For both tests, we assessed significance using p<0.05. Identification of genes with differential ATAC signals between pre- and post-transplant patients was performed using the sum of ATAC signals at each gene for all patient samples (calculated with FeatureMatrix). The 6 pre-transplant and 6 post-transplant samples were then analyzed as biological replicates of each condition, and differential analysis was performed using DESeq2. Ontology analyses were performed using g:profiler (v. 0.2.2)34. For differential analysis of aggregate ATAC signal for each gene average ATAC signal was defined as the total normalized ATAC signal over all gene bodies plus 2kb upstream for all pre- (n=6) and post-transplant (n=6) patient samples. Genes with differential ATAC signals were then identified using DESeq2.

Survival differences were evaluated using a Mantel-Cox log-rank test. Survival curves were generated using the Kaplan-Meier method. Differences in GVHD clinical and pathology scores were determined using 2-way ANOVA, with Bonferroni correction for repeated measures of multiple comparisons. Statistical analysis of ATAC-seq and RNA-seq data is described above. Unless otherwise noted in the figure legends, all other continuous variables were compared using a 2-tailed Students t test with Welchs correction. A P-value of 0.05 was considered statistically significant.

Further information on research design is available in theNature Portfolio Reporting Summary linked to this article.

Read more:
Type II innate lymphoid cell plasticity contributes to impaired reconstitution after allogeneic hematopoietic stem cell transplantation - Nature.com

Comments are closed.