Background
While the biological process of aging is inevitable, some people remain healthy until an advanced age while others suffer from age-related diseases early in life. The reasons for these vastly different aging patterns are still poorly understood. To protect the body from damage normally associated with aging, a balanced immune system is needed [
1], with different subpopulations of immune cells working in close harmony. One telling sign of a disturbed immune balance is a state of chronic low-grade inflammation, which may be revealed by measuring biomarkers such as C-reactive protein [
1‐
3]. It remains a challenge, however, to identify other biomarkers of the immune system that signal or explain differences in aging patterns.
Both innate and adaptive immune cell lineages are essential for a proper functioning of the immune system. It is well-established that the composition of the immune cell repertoire—that is the relative and absolute abundances of the various subpopulations of immune cells—changes with age [
4]. However, it is largely unknown how changes in the immune cellular composition differ between individuals who age in good health from those who become frail at a relatively early age. Several studies have been performed on this subject, but some of these were done in the extreme elderly of 85 years old or above [
5], while others focused on a limited set of lymphocyte phenotypes [
6‐
8]. Importantly, few studies have been performed enumerating a comprehensive set of immune phenotypes in freshly drawn whole blood samples.
Recently, we investigated associations between frailty and absolute numbers of the major immune cell populations in fresh whole blood. In that study, we detected associations between frailty and numbers of granulocytes and monocytes [
3]. However, the complex functioning of the immune system involves a range of specialized functions that are mediated by various myeloid cells and lymphocytes. It is therefore to be expected that shifts in the major immune cell populations are accompanied by less visible, yet potentially important, changes in their subpopulations that are functionally distinct. This hypothesis prompted us to extend our previous study with an in-depth analysis of various immune cell subpopulations in relation to frailty.
Since men and women tend to age differently [
9], we deemed it important to consider differences between the sexes. Furthermore, cytomegalovirus (CMV) serostatus should be taken into account, because chronic CMV infection is well known to impact numbers of immune cell subpopulations [
10,
11].
Thus, our main goal in this exploratory study was to investigate how general health, expressed in terms of a frailty index, is associated with various immune cell subpopulations in peripheral blood of older adults. Our secondary goal was to explore how these cell populations differ between the sexes or according to CMV serostatus. In order to do so, we employed extensive immune profiling, enumerating 37 immune cell subpopulations and quantifying the expression of several surface markers on immune cells in fresh whole blood samples from 289 older people aged between 60-87 years that were selected from among the participants in an ongoing cohort study in the Netherlands [
12,
13].
Discussion
This study shows that frailty in a 60-87-year-old population was associated with higher absolute numbers of neutrophils in men and women, and also with several sex-specific changes in the immune cellular profile. To the best of our knowledge, this is the first study in which frailty was separately investigated in men and women for its relations with a wide variety of immune cell subpopulations. Frailer women demonstrated higher numbers of total- and CD16- monocytes, lower numbers of CD56+ T cells, and lower numbers of late differentiated CD4+ TemRA T cells. These findings, obtained from association studies, were confirmed in a random forest prediction analysis. In men, the predictive value of the random forest analysis was too low to warrant substantial additional conclusions on top of the ones from the association study. The highest ranking subpopulation based on variable importance were neutrophils, in line with the association study. Based on variable importance, also CD16+ monocytes may play a role in prediction of frailty, but since this subpopulation was not found in the association study and the predictive accuracy was low, the relationship between CD16+ monocytes and frailty in men is unclear. In women, however, the predictive value of the random forest analysis was higher than in men and additional potentially relevant relationships were found with, respectively, total numbers of T cells and lymphocytes.
The associations found between higher neutrophil numbers and frailty are in line with previous reports [
3,
5,
40]. Among the entire immune cellular profile we investigated, neutrophil numbers were most strongly associated with frailty and the association was found in both men and women. In a previous study it was shown that elevated neutrophil levels in humans are related to negative health outcomes [
41], which corresponds to our results, since frail people are more prone to adverse outcomes like early mortality. The association of total monocyte numbers with frailty in women was also in line with other studies [
3,
40]. We have now shown that the relationship of higher monocyte numbers with frailty in women was due to higher numbers of CD16
- classical monocytes. The classical CD16
- monocytes, having clear phagocytic capacity, further differentiate into macrophages [
42], and are therefore involved in the defense against pathogens and in inflammation. Classical monocytes have a short lifespan of approximately one day in the circulation [
43]. Thus, an increase in this population is likely due to an increased production by myeloid progenitor cells rather than due to reduced clearance.
Noteworthy is that the clearest associations with frailty found in our study were with myeloid cells. In a murine study, it was argued that both higher myeloid cell abundancies and lower vaccine responses in older mice could be the result of intrinsic differences in hematopoietic stem cells with age [
44]. Other studies suggested that the myeloid bias of hematopoietic stem cell differentiation seen in old age is due to a chronic low-grade inflammation [
45]. This stronger tendency of hematopoietic stem cells to differentiate in the myeloid direction in an environment of low-grade inflammation is possibly driven by plasma cells in the bone marrow that can produce pro-inflammatory cytokines which also influence myelopoiesis [
46]. Such a chronic low-grade inflammation is often seen in frail people [
47], with a role for the IL-6 pathway, since consistent associations have been reported between frailty and higher CRP levels [
47]. Parallel to the more pronounced myeloid skewing with frailty in women, the association of chronic low-grade inflammation with frailty is also more pronounced in women than in men [
3]. However, many other factors may impact cellular numbers in peripheral blood, like variations in extravascular homing, or differential apoptosis [
48], which could explain the great variability between participants seen in our study population.
The present study also shows that several immune cell subpopulations differ in abundancy between the sexes, with higher numbers of most T cell subpopulations, especially CD4 T helper cells, and higher numbers of B cells, in women. These findings are in line with previous studies [
25]. We also found sex-specific associations between frailty and lower numbers of several CD4 T cell subpopulations in women but not men, which has not been reported earlier. These results, together with the stronger associations between frailty and myeloid cell numbers in women, could suggest that the skewing of hematopoietic stem cells with frailty towards production of myeloid cells is more pronounced in women than in men. Immunological homeostasis in women may involve a different balance between CD4 and CD8 T cells than in men, with estrogen levels known to be involved in lymphocyte development and in particular CD4 T cell proliferation [
49,
50]. In addition, immune function differs between the sexes, with women showing stronger responses to antigens and a stronger tendency to develop autoimmune diseases [
49,
51]. Testosterone levels are also thought to impact this process, with testosterone possibly having an anti-inflammatory role [
52]. The sex-specific findings in our study could also be of interest in relation to the so-called sex-frailty paradox, namely that women generally tend to be frailer, yet live longer than men [
51]. In our study, frailer women had lower numbers of CD56
+ T cells, which are known for their cytotoxic capacity [
53]. While the relationship of frailty with CD56
+ T cells has not been described previously, in one study NK cell markers (CD16 and CD56) were shown to be more highly expressed by T cells from people with better cognitive and physical functioning [
54], which is in line with our findings. CD4
+ TemRA T cell numbers were also found to be lower in frailer women. Since this was the weakest association in women, it would require confirmation in future studies. In fact, we expected to observe a negative correlation, i.e. meaning an increase of T cell memory numbers with frailty, instead of a lower number. It is known that the balance between naïve and memory T cells in peripheral blood changes with advancing age, mainly due to a decline in numbers of naïve (CD8
+) T cells [
11,
55,
56], which was also observed in our study (data not shown). Only few studies reported relationships between T cell populations and frailty and, to our knowledge, no associations were reported with late differentiated CD4
+ TemRA T cells. In one study, frailty was related to lower naïve CD4
+ T cells [
8], in another one with lower CD4
+ central memory (CD27
+CD45RA
-) T cells [
6], while in a third study no association was found between T cell subpopulations and frailty [
5]. Reasons for these heterogeneous results could arise from differences in gating strategies, not separating results for men and women, or not adjusting for CMV serostatus. Moreover, all previous studies involving memory T cell subpopulations were performed with relative values (percentages) rather than absolute cell numbers. Nevertheless, we found a relationship between frailty and lower numbers of total T cells in women. This relationship showed possible interactions with age and with numbers of CD56
+ T cells, which could explain why it was detected in the random forest analysis but not in the association study.
In our study we confirmed previously reported associations of CMV seropositivity with the late-stage memory T cell population [
11,
25]. We also observed that CMV seropositive men had lower classical monocyte numbers. In addition, their classical monocytes seemed to have higher CD38 expression. It is thought that CD38 can be upregulated in monocytes and macrophages in inflammatory conditions [
21]. Further research is needed to confirm this association and to answer the question why it was only found in men. Furthermore, we observed that CMV serostatus was not important in our frailty prediction model. Conflicting evidence has been reported on whether there is a positive relationship between CMV serostatus and frailty, with some studies showing an association [
57,
58] while others do not [
5,
59]. These conflicting results might be explained by differences instruments used to evaluate frailty (frailty phenotype or frailty index), or differences in study population, with some studies being restricted to only one of the sexes [
57] or to extreme elderly [
5,
59].
A major strength of this study is the use of fresh whole blood samples to measure an extensive set of immune cell subpopulations. This made it possible to enumerate cell populations like neutrophils that cannot be quantified when using cryopreserved PBMCs. In addition, we were able to relate absolute numbers of 37 immune cell subpopulations to frailty. This is important as these are needed to correctly interpret relative values (e.g. percentages). For example, a change in relative values can be the result of a change in either the numerator or the denominator. This issue can be solved only with information on absolute numbers. Our study showed that, when we performed an analysis with subpopulations of B cells, T cells, NK cells, and monocytes expressed as percentages, this analysis did not give additional insight in how frailty relates to the immune cellular profile. Another strength of our study is the use of a robust statistical analysis framework with two different analytical approaches that complement each other to find associations with frailty while correcting for multiple testing. The analysis framework that we used performs well when there is a high number of explanatory variables and strong multicollinearity, as was the case in our study. In particular, the random forest algorithm is known for its accuracy in complex analyses when many explanatory variables are involved [
60‐
62]. This algorithm made it possible to investigate how frailty is related to all cellular subpopulations together, i.e. the ‘cellular immune profile’. Of note is that we also used a quantitative measure of frailty that was defined in a consistent manner and implemented in a richly documented population, allowing the inclusion of 36 health characteristics. This made it possible to study the relation between frailty and immune cell subpopulations in a quantified manner. Lastly, our cohort consisted of a sample from the general community-dwelling population, thus allowing a window on aging of the immune system in a non-clinical setting. On the other hand, due to the nature of using fresh whole blood samples and our aim to quantify a large number of different immune cell subpopulations, we were restricted in the coverage of our antibody panel and therefore not all the subpopulations could be further classified with additional phenotypic markers. For example, to define regulatory T cells we used CD45RA and CD25 as was done previously [
23‐
25], but not FoxP3 or CD127. In addition, it appeared, as already expected, that the variability in numbers of immune cell subpopulations is very large in older adults. This may have contributed to the low predictive accuracy observed in the prediction analysis. Furthermore, there is no consensus yet on which instrument is the best to investigate frailty, and different instruments might yield different results. Since we were interested in a measure that approximates general health, we used a Rockwood frailty index [
16] based on 36 deficits rather than a frailty instrument with a less broad definition like the Fried frailty phenotype [
63].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.