Acute Respiratory Illness Is Associated with Memory T Cell Differentiation and Other Immune Cell Changes in an Age-Associated Manner

Respiratory viruses tend to cause repeated acute infections throughout our lifespan. The 20th century has witnessed three major pandemics due to the influenza A virus: the antigenic subtype H1N1 that caused the 1918 pandemic, H2N2 that led to the pandemic in 1957, and H3N2 in 1968. Other outbreaks of H5N1 and H7N9 and influenza B underscore the impact of the influenza virus in humans (1, 2). Other respiratory viruses, such as seasonal coronaviruses, rhinovirus, and respiratory syncytial virus (RSV), are also common and recurrent global pathogens. Repeated infections by these viruses are a consequence of strain variability. Influenza viruses undergo antigenic drift (minor alterations in the surface epitopes of the virus) and antigenic shift (an abrupt alteration on the viral surface, leading to new HA and/or NA proteins). As a consequence, the annual formulation of newer vaccines is a conventional approach to prevent a potential outbreak.

The degree of susceptibility to the infection and the subsequent response of the host to any prophylactic measure are largely determined by age (3). Although there are geographical differences, repeated vaccinations starting in childhood are common in many countries, and these episodic encounters induce acute changes in immune cell phenotypes. In younger individuals, the history of vaccinations and infections is quite limited, as opposed to older people who have likely been exposed to many vaccinations and/or infections across their lifespan. Superimposed on this are the various immune phenotype changes that occur in response to aging, even in the absence of acute infection. These age-based factors that greatly determine the immune system’s response to viral infection therefore need to be accounted for when designing effective vaccines.

Typically, the pathogenesis of influenza, RSV, or coronavirus involves an infection of the upper respiratory tract, with severe cases witnessing a spread to the alveolar regions as well. Infants and elderly adults are particularly susceptible to severe consequences, especially in the presence of underlying comorbidities. Thus, it is not surprising that many seasonal influenza waves have higher mortality rates for people ≥65 y of age (1, 3, 4). However, this trend was defied in the H1N1 pandemics of 1918 and 2009, where the young-adult population (20-44 y of age) witnessed substantial death rates (5). This atypical phenomenon remained elusive for a long time, with several researchers posing different hypotheses. A common theory relied on the lesser “antigenic history” that children have, as compared with adults (6). This lack of exposure may have led to a weaker immune response in the young. Alongside this, it was proposed that the “honeymoon period” of infectious diseases typically protects children (4–14 y of age) from morbidity and mortality (7). Despite these assumptions, the theory does not truly justify the spike in the increase in fatality among young adults. Although another well-followed hypothesis was that young adults may have succumbed to the virus due to an overactive immune response or a cytokine storm (8), a recent finding by Shanks and Brundage (9) proposed T cell dysregulation to be the prime culprit. Previous animal studies had proposed that animals exposed at least once to the H3Nx strain of the influenza A virus were very likely to exhibit an aberrant T cell response to the A/H1N1 strain. This could be attributed to the generation of antigenic peptides that act as TCR antagonists when the host immune system mounts a response against the virus. As a consequence, the host becomes even more susceptible not only to influenza but potentially to lethal bacterial pneumonia (10).

These findings help us understand the complex relationship between age and the immune system in the context of acute respiratory illness. Despite substantial research toward the development of vaccines, there remain significant challenges. While children may mount lower immune responses than adults, older adults could have immunosenescent and immunocompromised systems, factors that may lead to a potential decline in the Ab response to vaccination. When Abs fail to achieve clinical protection against the virus, the host relies on cell-mediated immune mechanisms to clear it. According to previous studies, heterosubtypic immunity (i.e., protection across various influenza A strains) is primarily mediated by CD8⁺ cytotoxic T lymphocytes, at least in mice (11). The establishment of influenza-specific memory relies on the differentiation of naive CD4⁺ and CD8⁺ T cells, as well as memory B and T cells. Although current vaccines (inactivated virus) primarily stimulate B and CD4⁺ T cells, the CD8⁺ cytotoxic immune response by the host depends on the restimulation of immunological memory from prior exposure to the virus (3). Current influenza vaccines are therefore poor inducers of heterosubtypic cell-mediated immunity.

To produce better vaccines, it is imperative to understand the changes in immune cell phenotypes and their functions in response to respiratory viral infection, especially in conjunction with age. In this regard, previous studies have made observations on certain cell populations in a limited range of individuals. For example, one study showed increases in central and effector memory T cell populations, with a concomitant decrease in naive T cells, in hospitalized adults with acute influenza (12). Another study showed decreases in total T and NK cell counts in children hospitalized with acute influenza (13). Still, studies analyzing the full breadth of immune cell subsets with acute infection, across a range of ages (and in nonhospitalized individuals), are lacking.

Detailed immune cell phenotyping can be performed using high-parameter flow cytometry or cytometry by time of flight (CyTOF) to quantify alterations in immune cell phenotypes. In this study, we performed CyTOF immune phenotyping on the peripheral blood of 22 individuals presenting with symptoms of acute respiratory infection, spanning a range from 1 to 74 y of age. We similarly phenotyped a set of 22 age-matched noninfected controls. We sought to address two major questions. 1) How do individuals with acute infection differ from noninfected controls at young versus older ages? 2) How do immune phenotypes differ with age in infected versus uninfected individuals? Our findings suggest that infection-related changes are mostly seen in younger individuals and that they mimic changes generally attributed to aging, including increased effector differentiation of T cells.

SLVP022 was a study of volunteers aged 1–74 y presenting to an outpatient clinic at Stanford University with a diagnosis of influenza-like illness as defined by the Centers for Disease Control and Prevention: fever (temperature of ≥100°F and a cough and/or sore throat in the absence of a known cause other than influenza. Of participants, 79% were positive for influenza A or B, with 8% having other respiratory viruses, and 13% were negative for any of the viruses tested (although these could be influenza that was not detected due to low viral load or strain heterogeneity). Age-matched controls were chosen from studies SLVP015 (NCT01827462), SLVP029 (NCT03028974), SLVP030, and SLVP033 (NCT03312699), all of which consisted of generally healthy individuals undergoing influenza vaccination, but with blood samples for this study drawn prior to vaccination. Some of the older controls had underlying chronic conditions such as hypertension or nonmelanoma skin cancer. All studies were performed under Stanford University Institutional Review Board approval.

Heparinized whole blood was used to isolate PBMCs by the Stanford Clinical and Translational Research Unit using standard protocols. Briefly, blood was subjected to density gradient separation using Ficoll-Hypaque, washed with PBS, and cryopreserved in FBS containing 10% DMSO. Samples were placed at −80°C for 24 h, then transferred to liquid nitrogen for long-term storage.

This assay was performed in the Human Immune Monitoring Center at Stanford University. PBMCs were thawed in warm RPMI 1640 + FBS media containing Benzonase, washed twice, and resuspended in CyFACS buffer (PBS supplemented with 2% BSA, 2 mM EDTA, and 0.1% sodium azide), after which viable cells were counted by a Vi-CELL cell analyzer (Beckman Coulter, Indianapolis, IN). Cells were added to a U-bottom microtiter plate at 1.5 million viable cells per well and washed once by pelleting and resuspension in fresh CyFACS buffer. The cells were stained for 60 min at room temperature with 50 μl of an Ab–polymer conjugate (Supplemental Table I). All Abs were from purified unconjugated, carrier protein–free stocks from BD Biosciences (San Diego, CA), BioLegend (San Diego, CA), or R&D Systems (Minneapolis, MN). The polymer and metal isotopes were from Standard BioTools (South San Francisco, CA). The cells were washed twice by pelleting and resuspension with 500 μl of FACS buffer. The cells were resuspended in 100 μl of PBS buffer containing 2 μg/ml Live-Dead (DOTA-maleimide [Macrocyclics, Plano, TX] containing natural-abundance indium). The cells were washed twice by pelleting and resuspension with 500 μl of PBS. The cells were resuspended in 100 μl of 2% paraformaldehyde in PBS and placed at 4°C overnight. The next day, 500 μl of CyFACS was added and the cells were pelleted. The cells were resuspended in 100 μl of eBioscience permeabilization buffer (1× in PBS) (Thermo Fisher Scientific, Waltham, MA) and placed on ice for 45 min before washing once with 500 μl of CyFACS. The cells were resuspended in 100 μl of iridium-containing DNA intercalator (1:2000 dilution in 1× PBS; Standard BioTools) and incubated at room temperature for 20 min. The cells were washed once with 500 μl of CyFACS and twice in 500 μl of Milli-Q water. The cells were diluted to 800,000/ml in Milli-Q water containing a 10× dilution of EQ normalization beads (Standard BioTools) before injection into the CyTOF (Standard BioTools). Data analysis was performed using FlowJo v10 (BD Biosciences) by gating on cells based on the iridium signal, then intact cells based on both of the iridium isotopes from the intercalator, then on singlets by Ir191 versus cell length, then on live cells (indium-Live-Dead minus population), followed by cell subset–specific gating (Supplemental Fig. 1). Cytobank (Beckman Coulter) was used for uniform manifold approximation and projection (UMAP) analysis, gating only on live intact singlet cells.

Except for the UMAP analysis, we used percent of parent frequencies from FlowJo for all comparisons. Because the data were not paired or normally distributed, the Mann–Whitney U test, an unpaired and nonparametric test for median comparisons, was used. This allowed us to analyze the differences between the median frequencies of noninfected and infected datasets for young (1–19 y of age) and old (30–74 y of age) groups individually. Statistical significance was defined as p < 0.05 (confidence interval [CI] = 95%).

Because a shared control sample was included in each batch of CyTOF samples, we compared the controls across all the batches to eliminate any further bias owing to batch-to-batch variability. For this, a Mann–Whitney U test was used to compare across these controls, and only the cell types exhibiting a nonsignificant difference among controls were used for further analysis. Out of the 99 cell types (in young and old groups), 46 cell types showed a probable difference in trends between infected and noninfected groups (through linear regression). Of these, 20 were eliminated owing to variability in controls across batches. The remaining 26 cell types that exhibited a nonsignificant difference in controls were used for further statistical analyses, thus reducing the risk of artifactual findings due to technical differences in the datasets.

After filtering the datasets based on the lack of significant differences in the controls, the Mann–Whitney U test was used to compare median frequencies for noninfected versus infected cohorts in young and old age groups separately. The reverse was also implemented, where the median frequencies for young versus old age groups were compared in noninfected and infected cohorts separately. Scatter plots were used to graphically demonstrate the outcomes, and the resultant p values were corrected for multiple comparisons to control the false discovery rate at 5% (14). Moreover, linear regressions of median frequencies against age were plotted for noninfected versus infected cohorts, and a linear regression model was used to compare their slopes and intercepts (at age 0). Statistical significance was defined at p < 0.05 (CI = 95%).

Based on the age distribution of our study subjects, they were divided into young (1–19 y of age) and old (30–74 y of age) groups. We then looked for differences in the immune cell frequencies between infected and noninfected (healthy) individuals for both age groups separately. Interestingly, these differences were observed to be significant only in the young age group (Fig. 1). All the corresponding immune cell types showed insignificant differences between infected and noninfected old individuals (Supplemental Fig. 2).

Among the cell types differing in young infected versus healthy individuals were total lymphocytes, whose median frequency was significantly lower in the infected group (p = 0.048). Within the lymphocyte lineage, the median frequency of B cells for the young group was significantly lower in infected individuals (p = 0.022). We also observed a reduction in the median frequency of CD94⁺ NK cells for the young infected group (p = 0.048). All p values were corrected for multiple comparisons as described in Materials and Methods.

Within the T cell lineage, we observed a significant reduction in the median frequency of naive CD4⁺ T cells in the young infected groups (p = 0.0034). Similarly, a significant decrease in frequency was also seen for CD4⁺ T cell types that express CD27 (p = 0.0041) and CD28 (p = 0.045) receptors. Conversely, an increase was observed in the frequency of effector memory CD4⁺ T cells (p = 0.0034). We also identified a significantly higher frequency of non–T follicular helper (TFH) CD4⁺ T cells (p = 0.048), defined as non-naive CD4⁺ T cells that do not express the CXCR5 marker. Finally, we saw a significantly lower frequency of regulatory T cells in young infected individuals (p = 0.022).

As observed for the CD4⁺ T cells, young infected individuals had significantly lower frequencies of CD8⁺ T cells that expressed the early differentiation marker CD27 (p = 0.045). Also, the frequency of effector memory CD8⁺ T cells was significantly higher in the young infected individuals (p = 0.0034).

In contrast to lymphocytes, the median frequency of peripheral blood monocytes was significantly higher in the young infected individuals (p = 0.048).

We sought to corroborate these results using an automated clustering algorithm, as opposed to manual gating, which could be subject to bias. Automated clustering using UMAP (15) showed decreased intensity of CD20 (B cell), CD56 (NK cell), and CD27 (naive and central memory T cell) staining in the young infected versus young uninfected groups. This decrease was not seen in the old infected versus old uninfected groups (Fig. 2). These observations echo the major findings from manual gating (Fig. 1).

To assess the effects of infection using age as a continuous rather than a categorical variable, we performed linear regression analysis for each cell type against the age spectrum (Fig. 3). The slopes of the regression lines for noninfected and infected individuals and their intercepts (at age 0) were compared using a nonlinear regression model. Although we did not observe any cell types showing significant differences between the slopes of noninfected (healthy) versus infected cohorts (Supplemental Fig. 3), we were able to identify differences in intercept (at age 0) for some of the cell types (Fig. 3). This suggests to us that differences with infection were greatest at the youngest ages. Most of the age regression plot lines converged with an increase in age, thus visibly demonstrating the nonsignificant changes in immune cell phenotypes with infection in older individuals.

Cell types with significantly different intercepts included many of the same or similar subsets identified by the analysis in Fig. 1. These included a significantly lower age = 0 intercept in infected individuals for total lymphocytes (p = 0.0039), B cells (p = 0.0038), naive CD4⁺ T cells (p = 0.0001), CD4⁺CD27⁺ T cells (p = 0.0058), and CD8⁺ T cells expressing CD27 (p = 0.0073) and CD28 (p = 0.0036). Moreover, in concordance with the Fig. 1 findings, we saw significantly higher age = 0 intercepts for effector memory CD4⁺ T cells (p = 0.0001), effector memory CD8⁺ T cells (p = 0.0001), and monocytes (p = 0.0039).

The regression analyses also identified a few subsets with significantly different age = 0 intercepts that were not identified by the categorical age analysis in Fig. 1. These included significantly higher intercepts for infected individuals for both CD4⁺ and CD8⁺ Th2 TFH cells (defined as CXCR5⁺CCR6⁻CXCR3⁻) (p = 0.0001 and p = 0.0058, respectively).

Finally, we looked for differences in the immune cell frequencies between the young and old groups, separately for infected and noninfected (healthy) individuals. This helped us observe the cell types that showed a significant rise or fall with age. Of the four such observed cell types, all belonged to the noninfected (healthy) cohort (Fig. 4). The cell types showing a nonsignificant difference between young and old groups are included in Supplemental Fig. 4.

We observed a significant reduction in the median frequency of naive CD4⁺ T cells in the older individuals as compared with the younger individuals (p = 0.02). We also saw a reduced frequency of CD27⁺CD8⁺ T cells in older individuals (p = 0.038) and a concomitant increased frequency of effector memory CD8⁺ T cells (p = 0.0012). Finally, we observed a significantly higher frequency of NKT cells in older individuals (p = 0.0099).

Our study highlights the cell subsets whose frequencies differ between individuals with an acute respiratory infection and noninfected (healthy) controls at various ages. We found that all significant differences were within the younger (1–19 y of age) population. Because influenza and other respiratory virus infections accumulate with age, this suggests that earlier encounters with such viruses preferentially alter immune cell phenotypes. These cell types are then less affected as individuals age, having encountered respiratory viruses and/or vaccines many times in the past. Essentially, their immune systems no longer change as much upon infection.

We used both categorical analyses of age (younger versus older) as well as regression analyses, for which age served as a continuous variable. All of the significant differences in the regression analyses were differences of age 0 intercept (not slope), which we interpret as consistent with the differences between infected and noninfected being greatest at the youngest ages. Most of the cell subsets with significant differences were the same in either analysis. This included total lymphocytes, for which we observed lower frequencies in young-infected compared with noninfected individuals. Lymphopenia has frequently been observed in clinical studies of acute influenza patients (16).

Most of the differences seen with infection involved T cell subsets, especially CD4⁺ T cell subsets. They consisted of generally lower frequencies of naive and central memory T cells (CD27⁺ and CD28⁺), and higher frequencies of effector memory T cells. This suggests a differentiation toward more effector-like T cell populations with acute infection, which is not unexpected given observations of T cell differentiation in viral infection models (17, 18). It also fits with a previous study showing increases in memory T cell subsets and decrease in naive T cells in hospitalized adults with influenza (12). The reduced frequency of regulatory T cells in young infected individuals could indicate a shift from an anti-inflammatory to an inflammatory response to infection. Finally, increased TFH subsets in younger infected individuals suggest a differentiation toward cells that facilitate Ab production. TFH cells have been shown to correlate with Ab production in influenza infection (19). Again, all of these differences were seen in younger individuals, suggesting that early encounters with respiratory infection provoke greater changes than infections occurring later in life.

In addition to T cell differences, young infected individuals showed lower frequencies of B cells, suggesting possible migration of these cells out of the blood. Alternately or in addition, this finding could reflect increases in other non–T cell lymphocyte populations because the effect was seen in the parent population of B cells. However, a previous study in hospitalized children with influenza also showed decreases in B cells as well as NK cells (13).

CD94 receptor expression on NK cells allows for the regulation of effector functions as well as the survival of NK cells, thereby being a major contributor to adaptive immunity. CD94⁺ NK cells were significantly lower in infected individuals, perhaps due to their trafficking out of the blood and/or decreases in expression of this receptor as a direct or indirect consequence of infection. Other studies have also observed increased apoptosis of CD94⁺ NK cells as a consequence of NK cell targeting by influenza virus (20), suggesting another possible reason for this finding.

Not surprisingly, monocytes were higher in infected individuals, probably due to the innate immune response and activation of these cells by proinflammatory cytokines associated with acute infection (21). Because we performed relative frequency analyses (using total live cells as the denominator), increases in monocytes and decreases in certain lymphocyte subsets such as B cells and CD94⁺ NK cells could contribute to the observed decrease in total lymphocytes in young infected individuals.

We also looked at age-related differences in noninfected and infected individuals and found these only in the noninfected cohort. This is perhaps understandable in the sense that the differences seen with infection in young individuals were similar to those typically associated with aging, namely a decrease in naive and an increase in effector-differentiated T cells (22) and loss of B cells (23). Indeed, we saw lower levels of naive CD4⁺ and effector memory (CD27⁺) CD8⁺ T cells in the older cohort, as well as higher levels of effector memory CD8⁺ T cells. We also saw higher NKT cell frequencies in the older cohort, a finding for which we are unaware of previous reports in the literature. We likely did not recapitulate a larger set of aging-associated phenotypes because our older cohort was 30–74 y of age, and many age-related changes have been demonstrated only with more advanced age (e.g., 60–90 y of age) (24).

There are several limitations to our study. Most importantly, it was a cross-sectional analysis with a relatively small number of participants. It would be more powerful (though much more difficult) to study the same individuals before, during, and after respiratory infection. We also did not have an even distribution of age groups across the lifespan, so we performed most of our comparisons between two groups, young (defined as 1–19 y of age) and older (defined as 30–74 y of age). We did perform linear regression across the age continuum for infected and noninfected groups, but this analysis is limited by the number of individuals in any particular age group. A larger study, including significant numbers of individuals at all age groups, including extremes of old and young, may reveal more specific age-related differences.

Overall, our study suggests that T cell differentiation is associated with an acute respiratory infection, but preferentially in younger individuals. Understanding changes in immune phenotypes with infection can eventually help define correlates of protection for vaccines and may help us to better understand the age-related differences in response to vaccines and infections, including deleterious responses in particular age groups.

The authors have no financial conflicts of interest.

We thank Michael Leipold for performing CyTOF assays and providing data access, and Harry Greenberg for assistance in generating specimens.