Immunophenotyping of mononuclear cells (MNCs) using flow cytometry was performed on 50 umbilical cord blood (UCB) and 22 adult peripheral blood (PB) samples. Additional analysis with 3 UCB and 3 PB samples was performed using RNA-sequencing for a more detailed characterization. Similar studies performed by D’Arena G et al. (6) and Mantri S. et al. (15), as well as resource availability (laboratory space, finances, and time), were used as a reference to determine sample size.

UCB MNCs were isolated from the UCB of term deliveries (≥34 weeks of gestation) at Lucile Packard Children’s Hospital-Stanford. UCB collections were performed through the Binns Program for Cord Blood Research with donor consent and institutional review board approval. Within 24 hours of collection, MNCs of UCB were obtained by density gradient separation of whole blood (Ficoll Paque Plus, GE Healthcare; 400g, room temperature, 30 minutes, deceleration off), followed by ammonium chloride red blood cell lysis (9:1 NH4Cl lysis buffer to cell suspension). Some UCB samples underwent additional processing to isolate CD34-negative MNCs by labeling with the human CD34 Microbead Kit Ultrapure according to manufacturer protocol (Miltenyi Biotec, San Diego, CA, USA) (14). Adult peripheral blood (PB) was collected from the Stanford Blood Center (SBC). Within 24 hours of collection, whole blood was diluted by 1:1 ratio with MACs buffer (1x PBS, EDTA, FBS). MNCs of PB were then obtained by density gradient separation (Ficoll Paque Plus, GE Healthcare; 400g, room temperature, 30 minutes, deceleration off), followed by ammonium chloride red blood cell lysis (9:1 NH4Cl lysis buffer to cell suspension). While our initial aims were to capture mononuclear cells, multinucleated cells such as Neutrophils found in the MNC fraction were also included as part of the study.

Following MNC isolation, 1 million cells per sample were stained with antibodies against markers at optimal concentrations according to manufacturer protocol ( Supplementary Table 1 ). At least 100,000 events were acquired on a Beckman Coulter CytoFLEX flow cytometer and analyzed using FlowJo software (version 10.8.1) ( Supplementary Figure 1 ). The study began with two panels to capture broad immune MNC subsets. They were updated to include new markers and a third panel to identify naive B cells (IgD), monocytes and granulocytes (CD66b), and TR1 cells (CD19b, LAG3). While our flow dataset has a total of 22 PB donors and 50 UCB donors, Figures 1D, E , 2H , 3A, B , have 14 PB donors and 24 UCB donors, reflecting the donors sampled after the panels were updated.

Libraries were prepared using Chromium Next GEM Single Cell 3′ Reagent Kits v3.1 single index kit according to the manufacturer’s protocol (10x Genomics), targeting 10,000 cells per sample. 12 cycles of cDNA amplification were done for all samples. Individual libraries were quality checked on an Agilent 4200 Tapestation 827 using D5000 screen tape. Next, KAPA library quantification kit (#KK4923) was used for qPCR on a BioRad CFX96 RT PCR thermal cycler. Single index libraries were sequenced on.

10x Genomics Cell Ranger was used to filter and align FASTQ files. Data analysis was performed in an R environment with Seurat (16) (https://satijalab.org/seurat/). Data were first filtered for quality based on number of unique features as well as mitochondrial counts. QC criteria were 200<nFeature<95th percentile and % Mitochondrial reads<98th percentile per cell for each donor. Log-normalization was performed on each cell to normalize feature expression measurements by total expression. These were performed according to donor type (UCB or PB) separately ( Supplementary Table 4 ). The two Seurat objects were then merged for scaling and clustering ( Figure 4B ). Cell clusters were labeled based on lineage specifying genes ( Supplementary Table 5 ) and clusters that were attributed to a single donor were specified.

Clustering in Seurat is based on finding a subset of features exhibiting high variability in expression in the dataset. These were identified using the ‘VariableFeatures’ function. The counts were then scaled before performing dimensional reduction using Principal Component Analysis (PCA). We then used Uniform Manifold Approximation and Projection (UMAP) to project cells in two dimensions, with similar cells being plotted closer together. Known hematopoietic lineage markers were projected onto cell populations using ‘FeaturePlot’ to identify hematopoietic and immune cell types that could be compared with flow cytometry data. Elbow plots reflecting the standard deviations of the principle components were used to identify the number of significant dimensions that were used for clustering.

Statistical analysis was performed by GraphPad Prism (Version 9.4.0) and SAS^® Studio (Release 3.81; Enterprise Edition). Data reported in the figures reflect the mean with a 95% Confidence Interval and we report mean and Standard Deviation (SD) in the text, assessing the heterogeneity within UCB MNCs samples and their specific immune subtypes. The relationship between UCB and PB MNCs was assessed using a parametric unpaired t-test with Welch’s correction or a nonparametric Mann-Whitney test based on Shapiro-Wilk normality test results. Detailed statistical results can be found in Supplementary Tables 2, 3 . P values reported are as follows: *: p-value<0.05, **: p-value:<0.01 ***: p-value=<0.001 ****: p-value<0.0001.

UCB was obtained through the collaboration between the Binns Program for Cord Blood Research and Lucile Packard Children’s Hospital. PB from healthy adults was obtained through the Stanford Blood Center. The collected whole blood samples were then processed through density gradient separation to isolate MNCs. For flow cytometry, MNCs were stained with antibodies against immunologic markers of interest, and data was collected with CytoFLEX and analyzed with FlowJo. For RNA-seq, the samples went through library prep (10x genomics), QC, sequencing (Novaseq S1), and analyzed with Seurat (Figure 5).

MNCs were stained and analyzed to describe the populations’ major immune subpopulations, including T lymphocytes, monocytes, granulocytes, natural killer (NK) cells, and B lymphocytes ( Table 1 , Figure 1 ). The largest mean proportion of MNCs in UCB in Panel 1 was identified as T lymphocytes (CD3+) at an average of 42.89% (SD: 10.37) of the live cells ( Figure 1A ). The smallest proportion of MNCs in UCB was, on average, 8.07% (SD: 3.17) B lymphocytes (CD19+) ( Figure 1B ). UCB composition was, on average, 19.85% monocytes (SD: 6.38) and 10.57% NK cells (SD: 5.44) ( Figures 1C, D ). Additionally, we found UCB to have an average of 15.40% granulocytes (SD: 12.67) ( Figure 1E ). The remainder of the UCB and PB descriptive statistics can be found in Supplementary Table 2 . The immune composition of UCB compared to PB did not significantly differ for T lymphocytes, NK cells, monocytes, and granulocytes (p>0.05). However, UCB had a 2.13% higher mean proportion of B lymphocytes compared to PB (p=0.0125) ( Supplementary Table C ).

CD3+ T lymphocyte populations ( Figure 2 ) make up a large percentage of the live mononuclear cells seen in umbilical cord blood and adult peripheral blood. In UCB and PB, the T lymphocyte population is mostly composed of CD4 T cells and CD8 T cells ( Figure 2 ). On average, the T lymphocyte cells of UCB were 71.95% CD4 T cells (SD: 6.03) and 25.33% CD8 T cells (SD: 6.04) ( Figures 2A, B ). UCB had a mean CD4 T cell proportion that was 6.73% higher than the proportion for PB (p=0.0265) ( Supplementary Table 3 ). There was no significant difference in CD8 T cells between UCB and PB (p>0.05) ( Supplementary Table 3 ).

Looking closer at the subtypes of CD4 T cells, a large proportion were CD4 CD45RA naive T cells in UCB. Out of CD4 T cells, 87.79% were CD4 naive T cells (SD: 13.35), ( Figure 2C ). UCB had a mean CD4 naive T cell count that was double the mean count in PB (p<0.0001) ( Supplementary Table 3 ). Within the CD4+ naive T cell population, there was also a significantly higher average cell count of recent thymic emigrants (RTE) cells in UCB compared to PB (p<0.0001) ( Supplementary Table 3 ; Figure 2D ). UCB had an average of 79.67% RTE subpopulation (SD: 12.14) within the CD4 CD45RA naive T cells. Comparatively, the mean percentage of RTE cells within the CD4 naive T cells in the PB samples was 23.39% lower ( Supplementary Table 3 ).

From the CD4 T cells, there were also small subpopulations of Regulatory T (Treg) and Type 1 regulatory cells (TR1) in both UCB and PB ( Figures 2G, H ). The UCB samples had an average of 3.71% (SD: 1.69) and 0.27% (SD: 0.37) from the CD4 T cell population identified as Treg ( Figure 2G ) and TR1 cells ( Figure 2H ), respectively. No significant difference was found in the mean subpopulation percentages for TR1 or Treg cells between UCB and PB (p>0.20) ( Supplementary Table 3 ). The geometric mean fluorescence intensity of HLA-DR on CD3 T cells was significantly lower on UCB cells compared to PBMCs (p=0.077).

When analyzing the subpopulations of CD8 T cells, the majority were CD8 CD45RA naive T cells in UCB and PB ( Figure 2E ). On average, CD8 T cells in UCB had 94.87% (SD: 6.52) cells identified as CD45RA naive T cells. Similar to the CD4 naive T cells, UCB had a 33.39% higher proportion of the CD8 T cells as naive compared to PB (p<0.0001) ( Supplementary Table 3 ). Of these CD8 CD45RA naive T cells, an average of 99.53% (SD: 0.55) were recent thymic emigrants (RTE) in UCB (Figure 2F). There was a 15.23% higher proportion of RTE cells in UCB CD8 naive T cells compared to PB CD8 naive T cells (p=0.0002) ( Supplementary Table 3 ).

Mononuclear cells were also stained and analyzed to characterize natural killer (NK) cells and B cells ( Figure 3 ). Mature NK cells were gated from the CD3- CD19- CD56+ population ( Supplementary Figure 1 ) (6, 17). On average, 73.32% of the CD56+ cells (SD: 15.55) were mature NK cells in the UCB samples ( Figure 3C ). The difference in percentage of mature NK cells between UCB and PB samples was not significant (p>0.05) ( Supplementary Table 3 ).

Lastly, we analyzed the mature vs naive subtypes of B lymphocytes ( Figures 3A, B ). From the CD19+ B lymphocyte population, UCB had a dramatically larger mean proportion of naive B cells compared to its mean proportion of memory B cells—80.12% (SD: 10.40) versus 3.19% (SD: 2.23), respectively ( Figures 3A, B ). Meanwhile, the B lymphocytes in PB samples were, on average, 53.24% naive B cells (SD: 13.71) and 24.54% memory B cells (SD: 8.22) ( Figures 3A, B ). From their B lymphocyte populations, UCB had a mean proportion of memory B cells nearly eight times smaller compared to PB (p<0.0001) ( Supplementary Table 3 ). Additionally, the proportion of naive B cells from the B lymphocytes was an average of 26.88% higher in UCB than in PB (p<0.0001) ( Supplementary Table 3 ).

In addition to analyzing the merged datasets, we clustered the UCB and PB data separately resulting in 15 clusters for PB donors ( Figure 6A ) and 20 clusters for UCB donors ( Figure 6B ). These were annotated based on lineage-specifying genes ( Supplementary Figure 3 and Supplementary Table 5 ); clusters without a clear cell lineage were labeled ambiguous. The cell type proportions derived from our scRNA-seq analysis for the two UCB donors aligned well with the proportions derived from our flow cytometry data ( Figure 6D , Table 2 ). Cell identities were obtained from flow through the markers CD13 (Monocytic), CD56 (NK), CD19 (B cell) and CD3 (T lymphocyte) from Panel 1. The proportion of CD4+ and CD8+ cells out of total CD3+ cells were obtained from Panel 2 out of total CD3 ( Figure 6C ). The data from flow cytometry contained more cells with an unknown identity because of the limited number of markers we could use to classify them. Both datasets showed a smaller proportion of B and NK cells, but a larger proportion of Granulocytes in UCB D5 compared to UCB D4. For example, UCB donor 5 had a smaller proportion of NK cells in both datasets (5.6% by flow, 7.1% by scRNA-seq) compared to UCB donor 4 (16.2% by flow, 16.7% by scRNA-seq) ( Figure 6D , Table 2 ). The proportions of CD4, CD8 and NK cells were most similar across datasets. Using the Chi-squared test, we did not detect a significant difference in the cell type proportions identified by flow and scRNA-seq in both UCB donor 4 (χ² = 0.152, df=6, p=0.999) and UCB donor 5 (χ² = 0.128, df=6, p=0.999).