Changes in blood lymphocyte numbers with age in vivo and their association with the levels of cytokines/cytokine receptors

We performed a longitudinal study of T- and B-cells and their subsets, and NK cells in peripheral blood of 165 BLSA participants at first visit and 5-year follow-up under the NIH IRB- approved protocol (GRC98-12-28-01) and performed in accordance with the Declaration of Helsinki. Demographic characterization of these participants was summarized in Additional file 1: Table S1. At each visit, blood cell counts were measured by standard complete blood cell counts by Coulter Counter and PBMCs were isolated from 50 mL blood drawn from participants under fasting condition and cryopreserved in liquid nitrogen. Two to five cryopreserved PBMCs from each subject with an average of 5.6 years apart were used in the experiments. PBMCs from all time points were thawed and counted on the day of the experiment. The recovery of frozen PBMC was 77 % ± 0.3 % (mean ± SEM). Complete blood cell counts was combined with flow cytometry analysis (see gating strategy in Additional file 1: Figure S1) to obtain the cell count and to estimate rate changes for different cell populations.

Antibodies used for flow cytometry analysis included: CD2-Tri-Color (TC); CD4-Phycoerythrin (PE) and CD4Allophycocyanin (APC); CD28-Fluorescein Isothiocyanate (FITC); CD8-TC; CD19-APC; CD45RA-APC; CD16-FITC from Life Technologies (Grand Island, NY); CD14-PE; CD27-PE; and IgM-FITC from BD Biosciences (San Jose, CA). Freshly thawed PBMCs from each visit were stained with three to five antibodies: T cells (CD2, CD4, CD8, CD45RA, and CD28); B cells (CD19, IgM, and CD27); NK cells (CD16). The data were collected on a BD FACSCalibur or BD FACSCanto II, and analyzed by Cell-Quest (BD Biosciences) and FlowJo. The gating strategies were presented in Additional file 1: Figure S1.

Fasting blood was collected at each visit for measurement of complete blood cell count and other routine blood chemistry using the standard method under the BLSA protocol. Sera were isolated from blood and stored in a -80 °C freezer prior to cytokine (IL) measurement using a custom-made multiplex assay (BioRad Luminex Assays) according to the manufacturer’s instruction. CMV IgG was measured from sera of 120 subjects (117 of them had two time points and 3 had single time points) using the ELISA kit (Abcam, # ab108639) according to the manufacturer’s instruction.

Figures were plotted as scatterplots with a linear regression line. The regression lines for rate and percentage rate of change by age were analyzed using linear regression. Regressions of number of cells and percent of cells were tested using mixed effects linear regression on age with a random effect for subject to address the within-subject correlation with the repeated measurements. The inclusion of the time difference between the measurements did not affect the assessment. For the regression models, all tests were performed with a p < 0.05. Correlations were calculated between pairs of variables, and after adjusting by FDR [25], p value less or equal to 0.005 is considered as significant. To address the question whether rapid rates of change are present across more cell types than expected, rates for each cell type were dichotomized and summed to identify subjects in the tertile with the greatest rate of cell loss (except for NK rates which were the highest tertile). The summed score was compared to the expected binomial distribution by chi-squared test. Assuming a binomial distribution, the probability coefficient was estimated from the data using the Bayesian modeling program rstan (http://​mc-stan.​org/​rstan.​html). A single sample proportion test was used to test whether the proportion of subject with summed score of 4 or more was greater in the data as compared to the expected summed score of 4 or more from a binomial distribution with probability of 0.33.

All statistical analyses were done using R version 2.12.1 (http://​www.​r-project.​org).

In agreement with previous reports, at baseline older age was associated with slightly lower number of CD4 ⁺ T (Additional file 1: Figure S2a) [26, 27]. Rates of change of CD4 ⁺ T cell, reported as cell number per μl blood per year, were quite heterogeneous across study subjects (ranging from -120 to +170 cells/μl/yr and an average of 9.8 cells/μl/yr) and were relatively stable at different ages (Fig. 1a). Next, we examined rates of change in three subsets of CD4 ⁺ T cells, including naïve (CD45RA⁺CD28⁺), Treg (CD25⁺Foxp3⁺), and CD28^- cells. Similar to previous reports, older age was associated with fewer naïve CD4 ⁺ T cells (Additional file 1: Figure S2b) [28, 29]. Similarly to the total CD4 ⁺ T cells, there were remarkable individual variability in rates of change in naïve CD4 ⁺ T cells (ranging from -80 to +108 cells/μl/yr and average 4.3 cells/μl/yr) but the average trend of change with age was surprisingly flat, suggesting that naïve phenotype CD4 ⁺ T cells were well maintained throughout the adult life span in the absence of apparent new genesis from the thymus (Fig. 1b). In agreement with previous reports [14, 30], Treg (CD4⁺CD25⁺FOXP3⁺) tended to be more numerous in older than in younger individuals (Additional file 1: Figure S2c). However, the degree of individual variations in the rates of change in Treg (ranging from -4 to +10 cells/μl/yr and average 1.4 cells/μl/yr) was smaller than that of naïve cells but similarly stable in the trend of change with age (Fig. 1c). We observed a similar age-related increase of CD4 ⁺ CD28^- T cells from ages 20 to 90 (Additional file 1: Figure S2d). Again, the rate of change in CD4 ⁺ CD28^- T cells showed individual variations and the trend of rate of change with age was relatively stable (ranging from -23 to +60 cells/μl/yr and average 1.6 cells/μl/yr) (Fig. 1d). Together, these results showed a great degree of individual variation in the rates of change in CD4 ⁺ T cells and their subsets among study subjects. Overall there was a rather stable trend in the rates of changes in CD4 ⁺ T cells and their subsets over the adult life time.

In CD8 ⁺ T cells, we observed a similar age related reduction in the number of cells/μl blood in our study cohort of cross-sectional analysis as previously reported [31‐33] (Additional file 1: Figure S3a). The rate of change in CD8 ⁺ T cells (ranging from -163 to +69 cells/μl/yr and average -1.3 cells/μl/yr) showed a comparable degree of variation as observed in CD4 ⁺ T cells and the overall rate of change of CD8 ⁺ T cells was remarkably stable throughout the adult lifetime (Fig. 2a). Decrease of naïve CD8 ⁺ T cells in blood with age was observed in this study cohort (Additional file 1: Figure S3b) as well as in previous reports [5, 12, 13]. The rates of naïve CD8 ⁺ T cells (ranging from -34 to +43 cells/μl/yr and average -1.8 cells/μl/yr) also displayed a high degree of individual variations with no significant change of the average rates at different ages (Fig. 2b). The increase of CD8 ⁺ CD28^- T cells with age is considered as a signature of T cell aging and was observed in this study (Additional file 1: Figure S3c). Interestingly, while the rates of change with age of CD8 ⁺ CD28^- T cells showed individual variations (ranging from -121 to +53 cells/μl/yr and average 0.9 cells/μl/yr), the average trend was not significantly changed with age, suggesting that the increase of CD8 ⁺ CD28^- T cells was accumulated through a relatively constant rate over the adult lifetime (Fig. 2c). Collectively, the individual variations in the rates of changes in CD8 ⁺ T cells and their subsets were similar to those of CD4 ⁺ T cells; the overall trend of the rates of CD8 ⁺ T cells and their subsets were largely stable over the adult life. Previous studies suggest that a reduction of the CD4/CD8 ratio with increasing age is an “Immune Risk Profile (IRP)” [23, 24, 34, 35]. Here, we evaluated whether the CD4 ⁺ /CD8 ⁺ ratio reduced with age along with the number of cells in the blood, and found that the average ratios of CD4/CD8 T cells were relatively stable with age (mean ± SD = 3.4 ± 2.5) (Fig. 2d). Thus, a reduction in CD4 ⁺ /CD8 ⁺ ratio was not observed as a general trend of age in this study cohort.

B cells were defined based by the expression of CD19 and naïve and memory B cells were defined based by CD19⁺IgM⁺CD27^- and CD19⁺CD27⁺, respectively (Additional file 1: Figure S1). We observed a reduction of total B cells as well as naïve B cells with age in our study cohort from the cross-sectional analysis (Additional file 1: Figure S4a-b), and a slight increase in memory B cells with age (Additional file 1: Figure S4c). The average rate of change in B cells was -6.6 cells/μl/yr, and the trend of the rate did not differ across the adult age span (Fig. 3a). Similarly, the average rate of change in naïve B cells was -5.5 cells/μl/yr, and the trend of the rate across the adult age span was flat (Fig. 3b). The average rate of change of memory B cells was -0.1 cells/μl/yr across the age span (Fig. 3c). Overall, these findings showed that the average loss of B cells and naïve B cells with age in vivo did not alter with the subject’s age.

Increase of NK cells in peripheral blood with aging has been reported [4, 21, 36] and was also observed in this study (Additional file 1: Figure S5). Here, we analyzed the major population of NK cells defined by CD14^-CD16⁺ (Additional file 1: Figure S1), which is composed of around 90 % of NK cells (Le Garff-Tavernier et al., 2011). The rate of change in NK cells displayed wide individual variation among the study subjects (from -180 to 100 cells/μl/yr) (Fig. 4). Among the average rates of T (CD4 and CD8) and B cells, the average rate of NK cells of all subjects was the highest (25.3 cells/μl/yr) with no change in the rate with age (Fig. 4).

To determine whether the rates of change in subsets of lymphocytes correlated with each other and with some physiological measurements such as levels of cytokines and soluble cytokine receptors in blood and other physiological parameters, we compared the rate of change of each type of lymphocyte subset with the percents and counts of lymphocytes and a panel of cytokines and physiological parameters (Additional file 1: Table S2) using Pearson’s partial correlation coefficient adjusted for age (Fig. 5). Lymphocyte count was positively correlated with body fat and LDL in sera (Fig. 5). The rate of change in CD4⁺ T cells showed a positive correlation with the rates of change for naïve CD4⁺ and naïve CD8⁺ T cells, total CD8⁺ T cells, CD4⁺CD28^- cells and CD8⁺CD28^- cells, but a negative correlation with numbers of CD4⁺ cells and naïve CD4⁺ cells (Fig. 5).

The rate of naïve CD4⁺ T cells was positively correlated with the rates of change of CD4⁺, CD8⁺ and naïve CD8⁺ T cells but negatively correlated with numbers of CD4⁺ and naïve CD4⁺, and naïve CD8⁺ T cells (Fig. 5). The rate of Treg cells was positively correlated with numbers of CD4⁺ and naïve CD4⁺ T cells, and rate of CD4⁺CD28^- cells (Fig. 5). To determine whether there is a tendency for subjects who show rapid rates of change in one cell type to show similar changes in other cell types, we dichotomized the rates for each cell type to identify subjects in the tertile with the greatest rate of cell loss (except for NK rates which were the highest tertile) and the seven dichotomized rates were summed (total score 0 to 7). If the cell rates are independent, the distribution should be binomial with probability 0.33, which was not the case by chi-squared test (p < 0.0001). The probability which fits the summed score was 0.38 that is significantly different that the 0.33 used in creating the score (p < 0.02). If the cell rates occur together, the expectation would be for an excessive proportion of subjects to have high summed scores. The expected probability of a score of 4 or more is 0.17, while the observed proportion is 0.23 (p = 0.03). The high proportion of high counts argues that there is evidence that rapid rates tend to co-occur among cell types far more often than would be expected by chance (i.e. being independent). Finally, the rate of CD4⁺CD28^- cells was positively correlated with the rates of CD4⁺, Treg and CD8⁺CD28^- T cells (Fig. 5).

The rate of CD8⁺ T cells showed a positive correlation with the rates of CD8⁺CD28^-, naïve CD8⁺, total CD4⁺, naïve CD4⁺ T cells and serum level of IL-15 and a negative correlation with number of CD8⁺ T cells, CD8⁺CD28^- T cells, and rate of naïve B cells (Fig. 5). The rate of change of naïve CD8⁺ T cells was positively correlated with rates of change of CD4⁺, CD8⁺, and naïve CD4⁺ T cells but negatively correlated with the numbers of naïve CD8⁺ T cells (Fig. 5). Finally, the rate of change of CD8⁺CD28^- T cells was positively correlated with the rates of CD8⁺, CD4⁺, CD4⁺CD28^- and serum level of IL-15 but negatively corrected with numbers of CD8⁺ CD8⁺CD28^- T cells, and rate of naïve B cells (Fig. 5). The ratio of CD4⁺/CD8⁺ T cell number was positively correlated with numbers of CD4⁺, Treg, and naïve CD4⁺ but negatively numbers of CD8⁺, CD8⁺CD28^-, and naïve CD8 T cells, and CMV seropositivity defined by the presence of anti-CMV IgG antibodies in sera (Fig. 5). The association CMV seropositivity with the low ratio of CD4⁺/CD8⁺ found here is in agreement with the previous report [37].

The rate of B cells was positively correlated with serum levels of soluble TNFRI and TNFRII, and rates of naïve and memory B cells but was negatively correlated with the numbers of B cells and naïve B cells (Fig. 5). The rate of naïve B cells was positively correlated with rate of B cells but negatively correlated with numbers of naïve B, naïve B cells, rates of CD8⁺CD28^- and CD8⁺ T cells (Fig. 5). The rate of change of memory B cells was positively correlated with serum levels of soluble TNFRI and numbers of NK cells (Fig. 5). Finally, the rate of change of NK cells was negatively correlated with the serum level of triglyceride (Fig. 5).