Highly multiplexed immune profiling throughout adulthood reveals kinetics of lymphocyte infiltration in the aging mouse prostate

Identification of distinct immune subsets in mouse prostate using mass cytometry

Prostates from mice of various ages between 1 and 16 months were collected and weighed to confirm the relationship between age and prostate development. We observed a positive correlation between age and wet prostate weight (ρ = 0.6673, P_ρ < 0.001) with a rapid increase in weight in pre-pubescent mice (<3 months old) (Supplementary Figure 1A), consistent with the rapid, androgen-dependent growth of the prostate which occurs postnatally until sexual maturity at 2–3 months of age [24, 25]. After reaching adulthood, the mouse prostate continues to grow at a much slower rate. In post-pubertal adult mice aged 3 to 16 months, we observed a positive correlation between age and prostate weight (ρ = 0.5144, P_ρ < 0.01), with a significant increase in mass at 16 months of age (p < 0.05) (Supplementary Figure 1A). Importantly, age was also positively correlated with the percentage of CD45⁺ immune cells in the prostate as measured by flow cytometry (ρ = 0.7034, P_ρ < 0.01) (Supplementary Figure 1B), consistent with age-related prostatic inflammation. The increase in total immune cells in the aged prostate was confirmed by quantifying the total lymphoid area in aged mouse prostates using immunohistochemistry (Supplementary Figure 1C, 1D).

Next, we used CyTOF to characterize age-related changes in the immune microenvironment of the prostate throughout adulthood. Prostates were isolated from adult mice aged 3, 6, 9, 12, and 16 months (each n = 4), and dissociated cells were stained with a panel of 19 metal-tagged antibodies designed to label a broad variety of immune cell types using CyTOF (Table 1 and Figure 1A). Immune cells from all ages were clustered using the t-distributed stochastic neighbor embedding (t-SNE) algorithm for dimensional reduction and visualization (Figure 1A, 1B and Supplementary Figure 2A). Based on expression of well-defined immune cell lineage markers, the t-SNE plot was separated into 3 broad groups: T cells (expressing CD3e), B cells (expressing B220), and myeloid/NK cells (expressing CD335, F4/80, CD11b, CD11c, Ly6C, or Ly6G) (Figure 1B). We observed a positive correlation between age and the percentage of T cells (ρ = 0.8952, P_ρ < 0.0001) and B cells (ρ = 0.8216, P_ρ < 0.0001) in the adult mouse prostate (Figure 1C, 1D). The enrichment for T and B lymphocytes with age was accompanied by a respective decrease in the percentage of myeloid/NK cells (ρ = –0.9197, P_ρ < 0.0001) in the aging mouse prostate (Figure 1E). These trends are consistent with the age-related increase in T and B lymphocytes and decrease in myeloid and NK cells previously reported in 3- and 24-month-old mouse prostates by our group [12].

To provide more granularity to our analysis we performed unsupervised k-nearest neighbors clustering using the X-Shift algorithm to further subdivide the t-SNE plot into phenotypically-distinct immune cell clusters (Figure 2A and Supplementary Figure 2B). The X-Shift algorithm was chosen because it optimizes the numbers of clusters to prevent over- and under-clustering [26]. We identified 29 distinct immune cell clusters which were then subdivided by cell type. Expression of lineage markers was used to identify 24 of 29 immune cell clusters as T cells (CD3e⁺), B cells (B220⁺ and CD19⁺), NK cells (CD335⁺), and myeloid cells (any combination of F4/80⁺, CD11b⁺, CD11c⁺, or Ly6G⁺) (Figure 2A–2E). The 5 remaining clusters were classified as unknown due to expression of both T cell and myeloid cell markers (U1), lack of a strong signal for any lineage-specific markers (U2–U3, U5) or mild, nonspecific staining for every marker (U4) (Figure 2A).

Surprisingly, there were 6 immune cell clusters with staining for both Ly6C and CD19 (T1, T8, B1, NK1, M4, U2) (Figure 2A). Since CD19 is only expected on B cells, expression by other immune cell lineages is unlikely [27]. Instead, we determined that there was interference between the Ly6C and CD19 channels in our experiment. At high levels of Ly6C, there was overflow of signal into the CD19 channel, resulting in a linearly proportional increase in signal between the two channels (Supplementary Figure 3A). In our panel, Ly6C was conjugated to neodymium-150, and CD19 was conjugated to erbium-166 (Table 1). These two lanthanide isotopes differ by 16 Da, consistent with known effects of oxidation of neodymium-150 by oxygen-16 as the sample is ionized to plasma in the mass cytometer (Supplementary Figure 3B) [28]. We checked every other combination of the 19 markers used and found no other instances of signal spillover (Supplementary Figure 3C). The effects of signal spillover from Ly6C into CD19 taken into account, high Ly6C expression likely caused the T1, T8, NK1, M4, and U2 clusters to appear as CD19⁺. However, expression of the B cell specific marker B220 by cluster B1 indicates that its CD19 signal is likely true [29]. Interestingly, Ly6C is expressed across many different immune lineages but is not expressed on resting B cells [30, 31]; however, Ly6C has been found in some cases to be upregulated in B cells exposed to lipopolysaccharide and type III interferon [32, 33]. Given that CD19 is not expressed in immune cells besides B cells, the artifactual signal interference from Ly6C into CD19 is not likely to have affected the clustering or cell type identification in our analysis.

Aging-related changes of mouse prostatic immune cell populations arise early in adulthood

Having used unsupervised clustering to subdivide immune cells in the mouse prostate into distinct subsets, we next examined how these populations change throughout adulthood. Correlation analysis was performed to determine whether each immune cell cluster was enriched (ρ > 0) or depleted (ρ < 0) with age. We found that 19/29 immune cell clusters were significantly correlated with age (ρ ≠ 0, P_ρ < 0.05). Consistent with the overall increase in T and B lymphocytes in the mouse prostate with age, 8/10 T cell clusters and 2/2 B cell clusters were significantly enriched with age. Similarly, we found that 1/2 NK cell clusters and 5/10 myeloid cell clusters had a significant negative correlation with age (Figure 3A and Supplementary Table 1). Excluding an extreme outlier in the abundance of cluster M2 in one of the 9-month-old mice, we also found a significant negative correlation (P_ρ < 0.05) between M2 abundance in the prostate and age (Supplementary Figure 4).

Principal component analysis was performed to assess the progression of changes in the mouse prostate immune microenvironment. Prostates from 3- and 6-month-old mice clustered together as did those from 12- and 16-month-old mice, demonstrating the similarity in immune microenvironments of prostates in the youngest and oldest age groups (Figure 3B). Prostates from 9-month-old mice clustered between the other ages, demonstrating an intermediate immune phenotype at this age. These results show that the shift in the inflammatory microenvironment of the aging mouse prostate occurs between 6 and 12 months of age.

To validate these changes to the immune microenvironment of the mouse prostate, we performed a separate CyTOF immunophenotyping experiment with a panel of the same markers and mice 4-, 9-, and 15-months-old (each n = 3) (Supplementary Table 2). The 24 known immune cell clusters from the discovery experiment were matched to immune cells from the validation experiment, and Spearman correlation analysis was performed (Supplementary Figure 5 and Supplementary Table 3). Four of the 24 immune cell clusters (T1, T3, B1, and M10) were validated (same sign for ρ and P_ρ < 0.05 in both discovery and validation experiments) (Figure 3C). Fifteen of the 24 immune cell clusters had the same direction of correlation with age in both the discovery and validation experiments but failed to meet the barrier of statistical significance in both experiments. The majority of these were significant in the discovery but not the validation experiments, likely due to the low statistical power of n = 3 in the validation. Overall, there was a statistically significant correlation between the sign of ρ for immune cell clusters in each experiment (p < 0.01, Fisher’s exact test), supporting the reproducibility of our CyTOF immunophenotyping (Figure 3C).

Because CyTOF is not a widely available technique, we sought to provide a more accessible way to study these age-related immune microenvironment changes in the mouse prostate. Guided by the phenotypes of the validated immune cell clusters, a simpler gating scheme was constructed which can identify the 4 validated clusters (T1, T3, B1, and M10) and 5 additional clusters that were significant in the discovery experiment (T2, T7, T10, B2, and M9) using a reduced number of markers (Table 2 and Figure 4A). The reduced panel has fewer markers to be used with more widely accessible 12-color flow cytometry [34]. Importantly, manual gating to reidentify these immune cell clusters in the discovery experiment produced populations with significant correlations with age that are consistent with correlations found using X-Shift clustering (Figure 4B–4D).

Mouse prostate and bladder share features of inflammaging

As part of the discovery experiment, we also performed CyTOF on bladders and kidneys from the same 3- and 16-month-old mice (each n = 4) to compare aging of the prostate to other urogenital tissues with age-related cancer risk [35, 36]. Immune cells from these tissues were included for X-Shift clustering so that we could examine age-related changes to the same 29 immune cell clusters as the prostate (Supplementary Table 1). In the mouse bladder, we identified significant correlations between age and 17/29 immune cell clusters (ρ ≠ 0, P_ρ < 0.05) (Supplementary Figure 6A). Unlike the prostate and bladder, the immune profiles of the 3- and 16-month-old mouse kidneys displayed more similarity, with only 5/29 immune cell clusters significantly correlated with age (Supplementary Figure 6B). Principal component analysis was performed on the 3- and 16-month-old samples from the prostate, bladder, and kidney to assess the similarities in how different urogenital tissues age (Figure 5A). Three-month-old bladder and prostate clustered together as did 16-month-old bladder and prostate, suggesting these tissues are similar in how their immune microenvironments change with age. On the other hand, both ages of kidney clustered together near the 16-month-old bladder, suggesting that the kidney immune microenvironment changes less with age and was most similar to the 16-month-old bladder.

To better understand these relationships, we compared the correlation coefficients (ρ) for each immune cell cluster, irrespective of p-value, in the prostate, bladder, and kidney (Figure 5B). Thirteen of 29 clusters were correlated with age in the same direction in all three tissues (all ρ > 0 or ρ < 0) (T1–T4, T9, B1–B2, NK1, M6, M8–M10, and U5). Among the other immune cell clusters, prostate and bladder shared more similarities to each other than to kidney (Figure 5B–5E and Supplementary Figure 6C). These results demonstrate that different mouse urogenital tissues have distinct inflammatory signatures that change with age and that the aging prostate more closely resembles the aging bladder than the kidney.

Animal work

Male C57BL/6J and C57BL/6N mice from Jackson Laboratories and the UCLA Department of Radiation Oncology’s animal core facility were used in this study. The UCLA Division of Laboratory Animal Medicine (DLAM) bred and maintained the mice. The UCLA Animal Research Committee (ARC) and DLAM approved all animal protocols.

Tissue isolation and dissociation to single cells

Male mice of different ages were euthanized using carbon dioxide asphyxiation, then their urogenital tract and right kidneys were removed. Under a microscope, microdissection was performed to isolate the prostate and bladder from the urogenital tract. Tissues were mechanically dissociated with a razor blade then enzymatically dissociated in RPMI 1640 (Gibco) containing 10% fetal bovine serum (FBS), 1 mg/mL collagenase type I (Gibco), and 0.1 mg/mL DNase (Sigma) at 37°C on a nutating platform for 60–90 minutes. After washing the cell pellets with 1X DPBS (Gibco), cells were resuspended in 37°C TrypLE Express Enzyme, no phenol red (Thermo Fisher Scientific) for 5 minutes before being quenched by adding RPMI 1640. Cells were further disrupted using the shear stress generated by drawing through an 18G syringe. Finally, cells were sequentially passed through 100 μm and 70 μm cell strainers (Corning) to produce a single-cell suspension.

Flow cytometry

Single-cell suspensions of 5 × 10⁴–1 × 10⁶ cells from mouse prostate tissue of different ages (each n = 4) were stained in cell staining buffer comprised of 1X DPBS (Gibco) with 5 g/L protease-free bovine serum albumin (Sigma-Aldrich) and 200 mg/L sodium azide (Sigma-Aldrich). Nonspecific antibody binding by Fc receptors was blocked using TruStain fcX (anti-mouse CD16/32) Antibody (BioLegend) according to the manufacturer’s protocol. Cells were then stained with rat anti-CD49f-PE (BioLegend), rat anti-CD326 (EpCAM)-APC/Cy7 (BioLegend), goat anti-Trop2-APC (R&D Systems), and rat anti-CD45-FITC (BioLegend) for 30 minutes at room temperature. Cells were washed with cell staining buffer then fixed in 1% paraformaldehyde (Electron Microscopy Sciences) for 10–15 minutes at 37°C. After fixation, cells were chilled on ice for 1 minute, washed with cell staining buffer, and stored at 4°C before analysis on a FACSCanto flow cytometer (BD Biosciences).

Immunohistochemistry

Immunohistochemistry was performed as described previously [62] with some modifications. Paraffin sections of the mouse prostate were deparaffinized and underwent epitope unmasking using a heat antigen retrieval step. Section-staining was performed using the manufacturer’s protocol for Mouse and Rabbit Specific HRP/DAB IHC Detection Kit - Micro-polymer (Abcam ab236466). Antibodies used were: rabbit anti-CD45 (Abcam ab208022, 1:1000), anti-CD3 (Abcam ab5690, 1:300), and anti-CD19 (Cell Signaling 90176, 1:800). Hematoxylin and eosin (H&E) staining was performed by UCLA’s Translational Pathology Core Laboratories. Entire areas of an H&E-stained prostate section were captured by Axio Imager M2 (Zeiss) from two separate sections from a prostate, from four animals per age group. The sectional areas of the tertiary lymphoid structures was calculated using Zen 3.5 software (Zeiss).

Antibodies for mass cytometry

Antibody panels were designed using the Maxpar Panel Designer software (Fluidigm) to check for interference between the channels. The discovery and validation experiments used the same list of markers but differed in the metal tags conjugated to the antibodies (Table 1 and Supplementary Table 2). Antibodies were purchased pre-conjugated from Fluidigm or conjugated at the UCLA Jonsson Comprehensive Cancer Center (JCCC) and Center for AIDS Research Flow Cytometry Core Facility using the Maxpar X8 Polymer Chemistry kit (Fluidigm) following the manufacturer’s instructions. Antibody cocktails were prepared by adding 1.1 μL/sample of the conjugated antibody stock to a single tube then diluting to a volume of 50 μL/sample with cell staining buffer comprised of 1X DPBS (Gibco) with 5 g/L protease-free bovine serum albumin (Sigma-Aldrich) and 200 mg/L sodium azide (Sigma-Aldrich).

Cell surface staining for mass cytometry

Single-cell suspensions of 3 × 10⁵ – 1.8 × 10⁶ cells from mouse prostate, bladder, and kidney tissue of different ages were stained in cell staining buffer comprised of 1X DPBS (Gibco) with 5 g/L protease-free bovine serum albumin (Sigma-Aldrich) and 200 mg/L sodium azide (Sigma-Aldrich). Staining was performed in a 96-well V-bottom plate. The discovery experiment was performed using prostates from 3-, 6-, 9-, 12-, and 16-month-old mice; bladders from 3- and 16-month-old mice; and kidneys from 3- and 16-month-old mice (each n = 4). The validation experiment was performed using prostates from 4-, 9-, and 15-month-old mice (each n = 3). To distinguish between live and dead cells, cells were washed with cell staining buffer, resuspended in 200 μL cell staining buffer containing 1 μM Cell-ID Intercalator-103Rh (Fluidigm), and incubated for 15 minutes at 37°C. Staining with rhodium-103 was quenched with 2 mL cell staining buffer. Next, cells were centrifuged at 400 × g for 5 minutes, resuspended in 50 μL cell staining buffer containing 1 μg/mL TruStain fcX (anti-mouse CD16/32) Antibody (BioLegend), and incubated at room temperature for 10 minutes. Next, 50 μL of the diluted antibody cocktail was added to each sample, and cells were incubated at room temperature for 30 minutes. Cells were washed twice with cell staining buffer then resuspended in 200 μL cell intercalation solution comprised of 1X DPBS containing 1.6% paraformaldehyde and 125 nM Cell-ID Intercalator-Ir (Fluidigm). Samples were incubated at 4°C for 12–48 hours then washed sequentially with cell staining buffer and 1X DPBS. To remove clumps of cells, samples were passed through a 40 μm cell strainer (Corning) before a final wash with MilliQ water (Millipore). After the final wash, cells were resuspended in 50 μL MilliQ water for analysis via mass cytometry.

Mass cytometry

Mass cytometry was performed with a Helios mass cytometer (Fluidigm) at the UCLA Jonsson Comprehensive Cancer Center (JCCC) and Center for AIDS Research Flow Cytometry Core Facility. Samples were washed twice with Maxpar cell staining buffer (Fluidigm) and twice with MilliQ water (Millipore). Next, samples were resuspended in 10% EQ Four Element Calibration Beads (Fluidigm) containing natural abundance cerium (140Ce/142Ce), europium (151Eu/153Eu), holmium (165Ho), and lutetium (175Lu/176Lu). Samples were run at an event rate of 300–500 events/second, and data were normalized using bead-based normalization in the CyTOF software.

Mass cytometry data analysis

Analysis of mass cytometry data was performed in FlowJo V10 (FlowJo LLC). Single, live CD45+ immune cells from each sample in the discovery and validation experiments were manually gated. Each sample was given a unique Sample ID for later deconvolution, then all CD45+ cells from samples in the discovery experiment were concatenated into a single .fcs file. The native FlowJo t-distributed stochastic neighbor embedding (t-SNE) function was used to perform dimensional reduction on the single file containing immune cells from all samples in discovery experiment, using all markers besides CD45 and the following settings: Iterations, 3000; Perplexity, 30; Eta (learning rate), 200. Heat maps of marker expression on the t-SNE plot were generated using the FlowJo Color Map Axis function. Unsupervised k-nearest neighbors (KNN) clustering was performed using the X-Shift algorithm plugin for FlowJo [26]. Plugins for FlowJo are found on the FlowJo Exchange and utilize R (R Core Team). X-Shift clustering was performed using all markers besides CD45 and the following settings: number nearest neighbors (K), 54; distance metric, angular; subsampling limit, 42,790; Run ID, auto. Cluster phenotypes were calculated by taking the geometric mean of marker expression by cells in each cluster then transforming using arsinh(x). Immune cell clusters were manually classified as T cells, B cells, NK cells, myeloid cells, and unknown based on expression of the known lineage markers CD3e, CD4, CD8, B220, CD19, CD335, F4/80, CD11b, CD11c, and Ly6G. Phenotypes of immune cell clusters in the discovery experiment were used to manually gate for matched immune cells in the validation experiment. Cluster abundance was calculated by determining the number of cells from each cluster in each sample using the unique Sample ID, then dividing by the total number of CD45+ immune cells in each sample. Age-related immune cell cluster enrichment was analyzed by calculating Z-scores for cluster abundance, using the mean and standard deviation of all replicates and ages for a given cluster. N x N plots were made in FlowJo using the Layout Editor function. Heat maps of cluster phenotype and enrichment were generated in Prism V7 (GraphPad).

Principal component analysis

Principal component analysis (PCA) was performed using immune cell cluster abundance data from the discovery experiment. PCA was performed in R (version 3.6.2) using the singular value decomposition method on scaled matrix data.

Statistical analysis

All statistical tests were performed in Prism V7 (GraphPad). Correlations with age were calculated using the non-parametric Spearman correlation coefficient (ρ) with a two-tailed test for significance (P_ρ). The symbol “P_ρ” was used instead of “p” to denote the p-value for correlation analysis due to the visual similarity to the symbol for the Spearman correlation coefficient “ρ”.

Comparisons between ages were made using the non- parametric Kruskal-Wallis H test followed by Dunn’s multiple comparisons test against the 3-month-old samples, using adjusted p-values to account for multiple comparisons. Kruskal-Wallis and two-way ANOVA were used to evaluate differences in immune lineage markers between immune cell clusters. Fisher’s exact test was used to compare the direction of correlation with age (ρ > 0 or ρ < 0) for immune cell clusters between the discovery and validation experiments. Mann Whitney U test was used to compare immune cell cluster frequencies at 3- and 16-months-old for each tissue. Number of replicates (n) and type of replicate are listed in the figure legends. Error bars represent SD. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001. n.s., not significant, p ≥ 0.05.