Indo-1

This work extended the available six-cell reference library using twelve cell subtypes for deconvolution of blood cell proportions using the EPIC array, as well as a legacy library for the 450k platform. Using cytometric and magnetic-sorted, flow confirmed neutrophils, eosinophils, basophils, B cells (naïve and memory), monocytes, NK cells, CD4 + T cells (naïve, memory and T regulatory cells), and CD8 + T cells (naïve and memory), DNA methylation was measured with the 850 K/EPIC DNA methylation array. We applied the IDOL method to identify optimal Leukocyte -Differentially Methylated Regions (L-DMR) libraries using a testing set of six artificial mixtures containing the 12 cell types. Artificial mixtures (also referred to as reconstructions) consist of DNA from purified isolated cell types, representing mock blood samples of known, predefined cell proportions. Six additional testing artificial mixtures were employed to corroborate the performance, and 12 independent artificial mixtures derived from a set of 12 isolated cell types not utilized in the training or testing, were used to validate the results. We compared the performance of cell estimates obtained applying our previously developed six model cell type (available in Bioconductor as“FlowSorted. Blood.EPIC”) and optimized an additional L-DMR IDOL library limiting the probes to those available in the older 450 K array, and again compared the performance using the training, testing, and validation datasets.
The DNA used to generate mixtures were derived from four MACS-isolated and FACS-verified purity cell subtypes from the myeloid lineage [neutrophils (Neu), eosinophils (Eos), basophils (Bas), and monocytes (Mono)], and eight MACS-isolated and FACS-verified purity cell subtypes from the lymphoid lineage [B lymphocytes naïve (Bnv), B lymphocytes memory (Bmem), T-helper lymphocytes naïve (CD4nv), T-helper lymphocytes memory (CD4mem), T regulatory cells (Treg), T-cytotoxic lymphocytes naïve (CD8nv), T-cytotoxic lymphocytes memory (CD8mem), and natural killer lymphocytes (NK) cells] were purchased from AllCells® corporation (Alameda, CA, USA), StemExpress (Folsom, CA), and STEMCELL Technologies (Vancouver, BC, Canada). Cells were isolated from 41 males and 15 females, all anonymous healthy donors. The donors had a mean age of 32.2 years (sd = 12.2, range 19–58 years) and an average weight of 85.1 kg (range 57–136 Kg). Donors identified themselves from multiple ethnicities, including mixed ethnicities, and were categorized broadly into four groups (African-Americans, East-Asian, Indo-European, multiple/admixed). They were negative for Human Immunodeficiency Virus-HIV, Hepatitis B Virus-HBV, and Hepatitis C Virus-HCV. Women were not pregnant at the time of sample collection, and samples were collected from donors with no history of heart, lung, kidney disease, asthma, blood disorders, autoimmune disorders, cancer, or diabetes. All donors provided written informed consent before donation. The data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus (Salas et al., 2021) and are accessible through GEO Series accession number GSE167998. Isolation protocols are available through the commercial websites of AllCells, StemExpress, and STEMCELL Technologies. In brief, cells were selected using immunomagnetic labeling through the vendors’ specific protocols (see, Supplementary Table 1 for details). Recovered cells were confirmed using flow-sorting. Twenty-four artificial mixtures were determined by randomly generating proportions from a twelve-component Dirichlet distribution. Each mixture of 1.2 μg total DNA was generated from isolated cell DNA using the proportions in Supplementary Table 3. The isolated cell DNA and those of the artificial mixtures were bisulfite converted and processed according to the Illumina protocols at the Vincent J. Coates Genomics Sequencing Laboratory at UC Berkeley, Avera, or Diagenode. Samples were randomized prior to loading onto microarray chips. The EPIC methylation array raw idat files were pre-processed using minfi, EnMIX, and SeSaMe for quality control using R v.4.0.2 and 4.1.0^{53 (link)–55 (link)}. To assess data quality, we used an out-of-band detection P-value of 0.05, three standard deviations of the mean bisulfite conversion control probe fluorescence signal intensity, and a minimum of three beads per probe. To ascertain the highest purity of the samples included in our library, in addition to the information obtained through the FCM confirmation, we projected back the proportions, thus “purity,” of the cells, using Jaffe’s procedure¹¹ . The corresponding cell-type proportion was retrieved and designated as “DNA methylation purity.” Only samples with DNA methylation purity levels higher than 85% (range 85.7 to 100%) were included in the library. Subsequently, a stringent out-of-band p-detection value (pOOBHA) > 0.05 was applied and set those that could not be distinguished from the background probes as “missing values.” Only those probes with complete information for all the samples were selected for the library. No imputation was performed in this context as the signals could be dependent on the specific cell type. Additionally, all non-CpG (referred to as CpH) probes were filtered in light of the minimal variation and all CpH beta values were under 0.08. Finally, as both female and male samples were present, we discarded probes tracking the X and Y chromosomes. According to Zhou et al., those that showed known polymorphisms or cross-reactivity were also excluded^{56 (link)}. Our set for library discovery included 675,992 complete high-quality probes. The EPIC IDOL-Ext L-DMR library is available as an R library “FlowSorted.BloodExtended.EPIC” please contact the Office of Technology Transfer Technology.Transfer@dartmouth.edu for a free Academic license (for license instructions, please refer to https://github.com/immunomethylomics/FlowSorted.BloodExtended.EPIC). The extended blood deconvolution can be performed using the FlowSorted.Blood.EPIC Bioconductor library and we recommend using the minfi noob background correction for the target dataset. The package contains an RGChannelSet R object processed using SeSaMe in which probes showing channel switching were corrected and SNPs derived from Infinium Type I probes were added, using the total signal intensities, to the control for genetic ancestry. The object is unfiltered and contains 56 samples and the 12 artificial mixtures information on 1,008,711 probes corresponding to 866,091 sites (CpGs and CpHs) using the latest annotation released by Illumina (MethylationEPIC_v-1-0_B5). The reader needs to note that the cells were purified using an immunomagnetic procedure; the name “FlowSorted” is kept for historical reasons and downstream integration with previous minfi pipelines and similar algorithms.

LAD2 (2×10⁶ cells) in calcium-free medium (HBSS containing 0.5% BSA) were loaded with 4 μM Indo-1 (Thermo Scientific) for 45 min at 37°C. Cells were centrifugated at 400×g for 5 min and resuspended in loading media (HBSS containing 1 mM calcium, 1 mM magnesium, and 0.5%BSA). The samples were run at a low speed on an LSR Fortessa X20 (BD Biosciences). After collecting baseline data for 60 s, cells were loaded with different stimulations. The change in intracellular calcium flux was calculated as the intensity of Indo-1 (violet)/Indo-1 (blue) fold change after stimulation compared with baseline data.

5x10⁵ cells per sample were resuspended in 1% FBS RPMI and incubated for 30 min at 37°C with 5 μg/ml of Indo-1 and 0.5 μg/ml of pluronic F-127 (Molecular Probes). After washing, cells were kept in the dark on ice. Right before measurement, cells were pre-warmed 5 min at 37°C. The Ca²⁺ response was induced by addition of 0.5 μg/ml anti-CD3ε (UCHT1) after recording 60 s of baseline. The change of the ratio Indo-bound versus Indo-unbound was measured with a MaxQuantX Flow Cytometer (Miltenyi Biotech). Data were normalized to the baseline with FlowJo software version 9.3.2.