The animals used in the study were Indian-origin rhesus macaques (Macaca mulatta). All procedures were carried out in accordance with institutional and national guidelines and laws governing research in animals including the Animal Welfare Act. The animal protocols were reviewed and approved by the NIAID and NHLBI Animal Care and Use Committees. The National Institutes of Health (NIH) and Bioqual Inc. (where some animals were housed) are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care and are in full compliance with the Animal Welfare Act and Public Health Service Policy on Humane Care and Use of Laboratory Animals. Sedation and euthanization followed previously described protocols (38).

For SIVS infusions, purified anti-CD45 antibody (mouse IgG1κ) was developed in house and conjugated to the Alexa Fluor dyes (Thermo Fisher) shown in Figure 2A using N-hydroxysuccinimide (NHS) chemistry. Conjugated antibodies were diluted in sterile Hanks’ buffered salt solution. For all infusions, 5 mL of antibody solution was administered over 30 seconds via a catheter placed in the saphenous vein. Animals DGA0 and DGF5 received antibodies at doses of 10 and 100 ug/kg respectively, and animals G45T, ZJ31, HAXR, TID, and K628 received antibodies at a dose of 30 ug/kg. All infusion reagents were tested for endotoxin and were found to be negative. Blood draws were taken from the femoral artery, and lymph node biopsies were taken from the specified sites (axillary, mesenteric, or inguinal).

PBMCs isolation was performed using Ficoll-Paque PLUS (GE Healthcare) gradient separation and standard procedures. Lymph nodes and spleen tissues were mechanically disrupted and filtered through a 70-μm cell strainer. For spleen, the single-cell suspension was enriched for lymphocytes using Ficoll-Paque PLUS gradient separation. Lung tissue was processed using gentleMACS C Tubes and gentleMACS Dissociator (Miltenyi Biotec) and digested using type I collagenase (Gibco) and deoxyribonuclease (DNase) I (Roche) shaking at 37°C for 1 hour. Red blood lysis was performed with diluted 10× RBC Lysis Buffer (BioLegend). Cells were washed with phosphate buffered saline (PBS) and filtered through a 70-μm cell strainer. Jejunum and liver tissues were digested with the collagenase/DNase solution for 1-hour shaking at 37°C and filtered through 70-μm filters. Cells were either washed and immediately analyzed or were cryopreserved in 10% dimethyl sulfoxide (DMSO) in fetal bovine serum (FBS) and stored in liquid nitrogen.

Cryopreserved cells were thawed for staining and flow cytometric analysis. Cells were split into technical replicates and stained separately with antibody panel #1 and panel #2 (Table S1). For panel #1, intracellular Ki-67 staining was performed after surface markers staining using the intracellular staining protocol from the eBioscience Foxp3/Transcription Factor Staining kit. Flow cytometry was performed using the BD LSRFortessa™ Cell Analyzer.

The mouse anti-human CD16 monoclonal NK depleting antibody (clone 3G8) was obtained from the Nonhuman Primate Reagent Resource (http://www.nhpreagents.org) and used for NK depletion as described in previous studies (40). Animals HAXR and K628 in this study received a single 50 mg/kg intravenous dose of anti-CD16 depleting antibody as detailed in Figure 5A. The gating scheme utilized to follow CD16 depletion and technical validation for SIVS staining is shown in Figure S7.

All flow cytometry data was analyzed using FlowJo v10. The SIVS proportions for T cells and each NK cell subset were exported and analyzed using R v4.0.1 (41). The two different antibody panels were used as technical replicates for quantifying SIVS proportions. The networkD3 package v0.4 (42) was used to create Sankey diagrams such as in Figure 2B. Wilcoxon tests were performed between NK cell subsets using the ggpubr package v0.4.0 (43) with each animal as a biological replicate. Other packages used included ggplot2, dplyr, plyr, reshape2, and RColorBrewer. The line plot in Figure 5C was created using GraphPad Prism v8 (GraphPad Software).

We measured the overall frequency of NK cells within lymphocytes and of CD56+, DN, and CD16+ NK cell subsets within the NK cell population in blood and six tissues from 20 healthy RMs. NK cells were defined as negative for the lineage markers CD3, CD14, and CD20 and positive for the NK cell marker NKG2 via an antibody which binds to both NKG2A and NKG2C, surface receptors present on mature NK cells in non-human primates (14). Representative gating is shown in Figure 1A. Three NK cell subsets were defined based on expression levels of CD56 and CD16, which varied in distribution across blood and tissue NK cells (representative distributions for each tissue shown in Figure 1B).

NK cells made up 2-7% of lymphocytes in peripheral PB, BM, spleen, liver, jejunum, and lung tissues and less than 1% of lymphocytes in LNs (Figure 1C). PB, BM, and lung contained predominantly CD16+ NK cells with fewer DN NK cells and very few CD56+ NK cells (Figure 1D). Spleen and liver showed nearly equal proportions of DN and CD16+ NK cells with very few CD56+ NK cells. In contrast, LN samples showed primarily DN and CD56+ NK cells with very few CD16+ NK cells. In a single jejunum sample, the majority of NK cells were DN, with some CD16+ and very few CD56+ NK cells. These results align with recently reported data showing tissue-specific distribution and phenotypic characteristics of NK cells in humans (8) and non-human primates (14).

We administered anti-CD45 SIVS antibody infusions to five healthy RMs (Figure 2A). Animal ZJ31 received a single infusion 5 minutes before time zero when final blood sampling and tissue harvesting was performed. Animals DGA0 and DGF5 received SIVS antibodies 6 hours, 2 hours, and 5 minutes before time zero, and animal G45T and TID received a longer time course of SIVS antibodies 48 hours, 24 hours, 6 hours, and 5 minutes before time zero.

We stained PBMCs for different NK cell surface markers and quantified the SIVS antibody staining within CD56+, DN, and CD16+ NK cells and T cells (Figure 2B). We first validated that there is no background signal for the SIVS antibodies prior to SIVS infusions (Figure 2B, lower left). After defining positive and negative gates for each SIVS marker in each cell population, we analyzed the data for each animal’s NK cell populations individually using Sankey diagrams, which shows the percentage of cells positive and negative for each individual timed SIVS infusion over subsequent time points. For animals G45T and TID, receiving SIVS infusions over 48 hours, the majority of CD56+ NK cells recovered from PB at final collection were negative for the -6 hours infusion, indicating that they had recently emigrated from tissue to PB within the past 6 hours (Figure 2C). There was substantial trafficking of CD56+ NK cells between the vasculature and tissue compartments indicated by the fraction of cells changing from positive to negative or vice versa between subsequent SIVS infusions (gray flow lines). For DN NK cells, more cells were positive for the -6 hours antibody than for CD56+ NK cells, indicating PB localization at that time period, although approximately 30% of DN NK cells were recent emigrants from tissue within the last 6 hours before necropsy. There was some trafficking between tissue and vasculature for DN cells, though a considerable fraction (25-50%) of DN NK cells were retained in the vasculature for the full time course of SIVS infusions. For CD16+ NK cells, the majority of cells were positive for all SIVS infusions, indicating sustained residence in the vasculature for the 48 hour time course. For the animals DGA0 and DGF5 receiving shorter SIVS time courses, similar patterns in trafficking kinetics between NK subsets were found, with CD16+ NK cells having the longest retention time in blood, followed by DN NK cells, and CD56+ NK cells showing the most trafficking between blood and tissue (Figure S1A). These data, consistent across four animals, indicate that distinct NK cell subsets have differences in trafficking rates between PB and tissue at steady state.

We compared the percentage of cells which were “Continuously Circulating” (CC) i.e., retained in the vasculature, for each NK cell subset for each animal and for T cells. When considering timepoints going back to 6 hours before necropsy, all four animals had the same pattern with CD56+ NK cells having the lowest proportion of CC cells, CD16+ NK cells having the highest proportion, and DN NK cells intermediate between the two (Figure 2D). When considering all time points through -48 hours for animals G45T and TID, the same pattern held, indicating sustained residence in blood for the full time course of experiment i.e., 48 hours. We compared the fraction of CC cells between NK subsets using a Wilcoxon test with the four animals as biological replicates and found that DN and CD16+ NK cells had a higher percentage of continuously circulating cells than CD56+ NK cells (Figure 2E). Taken together, these results indicate that CD16+ and DN NK cells, which make up on average >95% of PB NK cells in RM (Figure 1D), are circulating continuously with an average residence time in the vasculature greater than six hours. Whereas the minor population of CD56+ NK cells in blood have a significantly shorter residence time and more commonly traffic between blood and tissue.

We determined the positional history of lymphocytes isolated from LN samples at tissue harvest. Across multiple LN samples from three animals, CD16+ NK cells showed the highest percentage of IVas+ stained by the SIVS infusion given 5 minutes before LN biopsy (Figure 3A). This difference between CD16+ and both CD56+ and DN NK cells was statistically significant in axillary lymph node (AxLN) samples from the four animals analyzed (Figure 3B). Though IVas+ NK cells in the LN could represent blood contamination, the NK cell phenotype of the LN IVas+ compartment more closely resembled the LN IVas- compartment than the blood compartment in most animals (Figures S2B–D). IVas+ cells in the LN have been shown to be closer to blood vessels (38) so our results indicate that CD16+ NK cells, which make up on average less than 10% of cells in LN (Figure 1D), may be preferentially localized to the cortex of the LN closer to the blood vessels. For assessing tissue localization and cell entry rates into LN, we considered only IVas- cells.

The IVas- NK cells for all three subsets showed high percentage of Tissue Localized (TL) cells, defined as negative for all SIVS antibody infusions (Figure 3C). For all LN samples tested, DN NK cells had the highest fraction of TL cells, though the difference was not statistically significant when comparing the AxLN samples from the four animals (Figure 3D). We next assessed the entry rate into LN, defined as the percentage of IVas- cells which were positive for the -6 hour infusion divided by six hours to give an entry rate in percentage of cells per hour. For the AxLN samples from all four animals, DN NK cells had the slowest rate of entry into the LN (Figure 3E). However, this difference was not statistically significant with our sample size of four AxLN samples (Figure 3F). The high variation in LN entry rate of CD16+ NK cells is likely due to the lower cell number, considering it is a minor population in LN and for some animals, a large percentage of CD16+ NK cells are IVas+ so were excluded from entry rate calculations. Taken together, these results indicate that for the two major NK cell populations in the LN, CD56+ and DN NK cells, the majority of cells are localized to the LN with a residence time greater than six hours, and that the trafficking rate from blood into the LN is less than 1% of LN cells per hour.

We repeated the same analysis on spleen samples from animals DGF5, G45T, and ZJ31 (Figure S3). We found high percentages of IVas+ cells across all three NK cell subsets and T cells (Figure S3A) which was expected given that spleen capillaries harbor discontinuous endothelial cells and as such are termed sinusoidal capillaries (44). Among the IVas- cells in the spleen, CD16+ NK cells had the lowest fraction of TL cells (Figure S3B) and the highest entry rate into the spleen (Figure S3C) out of the three NK subsets. Comparison of the NK phenotype between PB, spleen IVas+ cells, and spleen IVas- cells supported the notion that the IVas+ cells may occupy a distinct perivascular niche within the spleen, since they did not match the phenotype of PB NK cells (Figures S3D, E).

We analyzed NK cell trafficking kinetics into liver, lung, and jejunum, though we had fewer biological replicates for these tissues. In our liver sample, most NK cells were IVas+, which was expected given that the liver contains primarily sinusoidal capillaries. The liver IVas+ cells had a distinct NK phenotype from blood cells in the same animal (Figure S4A), indicating that a large fraction of NK cells in liver may occupy a perivascular niche, as previously reported for T cells in the liver (38). Within our lung sample, the majority of DN and CD16 NK cells were IVas+, while the majority of CD56 NK cells were IVas-, indicating a possible difference in localization within lung tissue (Figure S4B). The majority of IVas- CD16+ NK cells were Recent Immigrants (RI) to the lung (defined as IVas- but SIVS+ for one of the previous time points), while the majority of IVas- CD56 and DN NK cells were TL. Again, the IVas+ lung cells differed in NK phenotype from the PB sample from the same animal, suggesting the presence of a perivascular niche. Within our jejunum sample, there were very few CD56+ NK cells, and almost all NK cells were IVas- (Figure S4C). Of the IVas- NK cells, the majority of DN NK cells were TL within the jejunum while the majority of CD16+ NK cells were RI. Our results suggest that different NK cell subsets may exhibit different localization patterns and residence times within lymphoid and non-lymphoid tissues.

We analyzed the surface levels of six markers associated with NK cell trafficking and/or activation (CD49a, CD161, NKp46, CCR7, CD69, and CXCR3) as well as the intracellular proliferation marker KI-67 across LN samples. We compared the percentage of cells positive for each marker between recent immigrant (RI) and tissue-localized (TL) cells from each NK cell subset and found that the percentage of CD69+ cells was higher in TL than RI NK cells (Figure 4A). This difference was statistically significant for the two major populations of LN NK cells, CD56+ and DN NK cells, (n=4 AxLN samples) (Figure 4B). Additionally, we quantified the percentage of LN NK cells which were TL within the CD69- and CD69+ fraction of each NK cell subset (Figure 4C). Our results indicate that CD69 is a good marker for NK cell retention in the LN, with >99% of CD69+ DN NK cells, 94-100% of CD69+ CD56+ NK cells, and 84-97% of CD69+ CD16+ NK cells showing sustained residence in the LN for the full extent of our experiment (at least 6 hours, and up to 48 hours). Taken together, these results support the use of CD69 as a tissue residence marker for NK cells in the LN. The other five surface markers and KI-67 did not show any statistically significant difference between RI and TL NK cells in the LN (Figure S5).

To investigate NK cell dynamics during NK regeneration, we transiently depleted mature CD16+ NK cells in two animals and then performed a SIVS experiment at the peak of NK regeneration as assessed by maximal cell cycling in the PB. As previously shown, CD16 depletion can efficiently deplete CD16+ NK cell cells (35, 40), and clonal patterns suggest that CD16+ NK cell regenerate from a proliferating CD16^dim subset (35). After collecting baseline PB and LN samples, we administered a single dose of anti-CD16 depleting antibody (Figure 5A). We collected PB daily and monitored the NK cell numbers and proliferation rates to select a time point with high proliferation to start the SIVS infusions. Following CD16+ NK depletion, the proportion of NKG2+ cells decreased sharply and then began to rebound after three days (Figure 5B). The surface levels of CD16 decreased, and we defined both CD16^dim and CD16^bright populations of CD16+ NK cells. The DN NK cells increased in proliferation at day three after antibody administration, followed by an increase in the proliferation of the CD16^dim and CD16^bright NK cells respectively (Figure 5C). There was no increase in proliferation within the CD56+ NK cell subset in blood in animal HAXR and a very mild increase in animal K628 (Figure 5C). We checked the levels of the NK cell degranulation marker CD107a and early activation marker CD69 surface levels throughout the CD16 depletion (Figure S6A). We found very low numbers of CD107a+ NK cells throughout the time course and a transient increase in CD69+ percentage after antibody administrations, which returned to approximately baseline levels by day nine.

We began SIVS antibody infusion nine days after anti-CD16 administration, based on a high proliferation rate at that time point of DN and CD16^dim NK cells, which represent putative precursors to CD16+ NK cells (35). We administered SIVS infusions at 48 hours, 24 hours, 6 hours, and 5 minutes before PB and tissue collection in animal HAXR (Figure 5A). We attempted to follow the same SIVS time course for animal K628, but due to extravascular extravasation of the -48 hour SIVS infusion, only the -24 hour, -6 hour, and -5 minute SIVS infusions were interpretable. During recovery from CD16 depletion, CD56+ NK cells showed a similar dynamic in terms of trafficking between blood and tissue (Figure 5D) as compared to steady state (Figure 2C). Despite the fact that CD16^bright NK cells were being actively regenerated at the time of SIVS infusions, the retention time of DN, CD16^dim, and CD16^bright NK cells in the blood was very high, with all three subsets having high SIVS+ percentage at each timepoint back to 48 hours (HAXR) or 24hr (K628) before tissue collection. The fraction of cells that were continuously circulating was low for CD56+ cells and high for CD16+ (CD16^dim and CD16^bright), similar to what was observed in steady state animals with the longer SIVS time course (Figure 5E). DN NK cells showed a higher percentage of continuously circulating cells in CD16 depletion animals as compared to steady state animals. This indicates that the regeneration of mature CD16^bright NK cells may be taking place peripherally within blood or from cells in the perivascular space of tissue, which may also be tagged by the infused antibody, not recruited from tissue resident NK cell populations.

Within the LN, the CD16 depletion animal shared the same pattern of IVas+ frequencies as the steady state animals, with CD16^dim and CD16^bright NK cells having the highest IVas+ frequency. The IVas+ percentage in these two subsets was higher than in CD16+ NK cells in steady state animals (Figure S6B). Of the IVas- cells in LN, there was a decrease in the proportion of tissue localized CD56+ NK cells and an increase in the LN entry rate of multiple NK subsets as compared to steady state animals (Figures S6C,D). This indicates that NK cells may be trafficking to and from the LN more dynamically during regeneration of the CD16 NK cell compartment.