Developmental Regulation of Effector and Resident Memory T Cell Generation during Pediatric Viral Respiratory Tract Infection

Viral respiratory tract infections (VRTI) are a major cause of morbidity and mortality in children, especially those <5 y of age (1). VRTI in infants and young children can result in clinically significant lung injury with the timing, inciting pathogen and resultant disease severity associated with alterations in pulmonary function in later life (2, 3). The mechanisms underlying increased susceptibility to VRTI in infants, in particular, remains incompletely understood, with both viral and host factors likely contributing to disease severity. It has generally been considered that overwhelming disease in infants is related to impaired T cell responses, which are required to mediate viral clearance and establish long-term protective immunological memory (1). However, other evidence points to distinct, but not necessarily impaired, responses of infant T cells (4–6). A greater understanding of the developing immune system during active VRTI and its relationship to disease severity is crucial to improving preventative and therapeutic modalities.

Viral clearance requires effective and coordinated responses from subsets of T cells in the barrier and mucosal surfaces during active infection. T cell differentiation to effector cells following naive T cell activation is influenced by the local cytokine environment as well as expression of various transcription factors (7). In particular, the transcription factor T-bet is required for differentiation to Th1 effector cells producing IFN-γ, important for antiviral immunity (8, 9). T-bet also determines effector and memory T cell fate, with high T-bet levels promoting terminal effector generation and lower levels enabling memory T cell generation (6, 10, 11). Subsets of memory T cells include central memory T cell (Tcm) and effector memory T cell (Tem) subsets, which circulate and enter lymphoid and peripheral tissues, respectively, whereas resident memory T cells (Trm) are maintained within diverse tissue sites but do not circulate (12–14). Mouse models of VRTI have demonstrated that Trm are generated in the lung and upper respiratory tract (URT) and can mediate efficacious protective immunity to viral challenge (15–19). Human Trm have been identified in lungs and multiple tissues sites and share phenotypic and transcriptional features with mouse Trm (20–23); however, the role and establishment of human Trm during VRTI has not been well studied.

Infants produce virus-specific CD8⁺ T cells in response to respiratory infection (24); however, mouse models have shown that pathogen-specific T cells generated in infancy do not persist in a manner similar to those generated by adults (6, 11). Developmental differences in early life T cells favor rapid proliferation and terminal differentiation over the generation of long-lived memory responses (6). We previously showed that infant mice show reduced establishment and persistence of lung Trm following respiratory infection or intranasal vaccination compared with adult mice (11). Importantly, decreased Trm generation was intrinsic to infant T cells in mice owing to elevated expression of T-bet (11). Whether similar developmental regulation of Trm generation occurs in human T cells has not previously been assessed.

In this study, we sampled nasopharynx and endotracheal aspirates representing the URT and lower respiratory tract (LRT), along with blood of infants and children with VRTI to investigate the developmental and spatial aspects of the pediatric T cell response to VRTI. We show that local immune responses in both the URT and LRT are distinct from those in blood; URT and LRT contain predominant Tem and terminal effector T cell populations whereas blood mostly contains naive T cells. Disease severity is likewise correlated to T cell subset composition in the URT and LRT, which is further influenced by the viral pathogen. We identify developmental differences in T cell subset differentiation during the early years of childhood, favoring formation of effector memory RA⁺ T cells (Temra) during infancy and Trm generation increasing with age, correlating with increased expression of the transcription factor T-bet in pediatric T cells. Our results reveal important insights into the dynamics of pulmonary T cell responses to VRTI during early life, their association to disease pathogenesis, and intrinsic factors that control developmental regulation of fate determination.

Infants and children (<10 y of age) admitted to the pediatric intensive care unit (PICU) at Morgan Stanley Children’s Hospital/New York Presbyterian Hospital (New York, NY) with acute respiratory failure, defined as treatment with noninvasive (either continuous positive airway pressure or bilevel positive airway pressure) or invasive (endotracheal tube or tracheostomy) mechanical ventilation, were recruited to the study from April 2012 to December 2017 (Table I). Testing for viral pathogens was performed by multiplex PCR by FilmArray (BioFire Diagnostics, Salt Lake City, UT) and included adenovirus, coronavirus (229E, HKU1, NL63, OC43), human metapneumovirus (HMPV), human rhinovirus/enterovirus, influenza A, influenza B, parainfluenza (subtypes 1–4), and respiratory syncytial virus (RSV). Exclusion criteria included primary immunodeficiency, trisomy 21, or patients receiving chemotherapy or immunotherapy.

An additional cohort of healthy control subjects was enrolled (young children <5 y of age and adults [aged 25–35 y]). Children having phlebotomy for routine pediatric screening purposes were consented to donate an additional aliquot of blood for research purposes, and healthy adults were recruited to provide a blood sample (5 ml) to serve as an additional control group. All protocols were approved by the Institutional Review Board of Columbia University Medical Center (New York, NY), and informed consent was obtained from parents of subjects prior to enrollment in the study.

Nasopharyngeal washes of the URT and/or endotracheal tube or tracheostomy aspirates of the LRT were collected daily until resolution of acute respiratory failure or 2 wk, whichever came first. Blood samples were obtained (1 ml for subjects up to 3 kg, then 1 ml/kg up to a maximum of 5 ml) when permitted on the first day of study enrollment and 72 h later or at resolution of acute respiratory failure. URT and LRT samples were obtained during routine care performed by the PICU nurse or respiratory therapist. Samples were suctioned into a sterile sputum trap after instillation of 0.9% saline and passage of a catheter into the nares or artificial airway. Respiratory tract and blood samples (including flow cytometry and cytokine assays) were processed as previously described (25).

Demographic and clinical data were extracted from electronic medical records, and included hospital length of stay, PICU length of stay, duration of mechanical ventilation, and mortality. Significant lung injury was determined to be present when subjects met the criteria for pediatric acute respiratory distress syndrome (PARDS) (26). PARDS is a consensus definition for diagnosing the presence of lung injury in children. In contrast to the adult criteria for acute respiratory distress syndrome (27, 28), PARDS does not necessitate arterial blood sampling, which can be more difficult to obtain in children. The PARDS criteria allow application of oxygen saturation (universally available) in place of blood sampling and also provides the ability to identify PARDS in subjects receiving noninvasive mechanical ventilation in addition to those receiving invasive mechanical ventilation.

The following Abs were used for myeloid cell and lymphocyte surface and intracellular staining and flow cytometric analysis: CD3 (OKT3), CD4 (SK3), CD11c (3.9), CD14 (M5E2), CD16 (3G8), CD24 (ML5), CD45 (HI30), CD45RA (HI100), CD56 (HCD56), CD123 (6H6), CD169 (7-239), and CD206 (15-2), purchased from BioLegend (San Diego, CA); CD45RO (UCHL1), T-bet (eBio4B10), and HLA-DR (LN3), purchased from eBioscience/Thermo Fisher (San Jose, CA); and CD8 (SK1), CD69 (FN50), and CD103 (Ber-ACT8), purchased from BD Biosciences (San Jose, CA). Chemokine quantification was performed using a cytometric bead array, human chemokine kit (BD Biosciences). Stained cellular samples were acquired using an LSR II flow cytometer (BD Biosciences), chemokine samples were acquired using an LSRFortessa (BD Biosciences), and data were analyzed using FCS Express v6 (De Novo Software, Glendale, CA).

Statistical analysis was performed using GraphPad Prism version 7.03 (GraphPad Software, San Diego, CA). A Mann–Whitney test, Fisher exact test, or Wilcoxon matched-pairs signed rank test was used to determine significance between groups. Linear regression testing was used to perform correlation analysis. All hypothesis testing was conducted at the 0.05 level of statistical significance.

A total of 62 pediatric subjects with respiratory failure due to VRTI were recruited to the study. Most (34 of 62) subjects enrolled received only noninvasive ventilation; however, a large number (28 of 62) also received invasive mechanical ventilation (Table I). Most subjects were infants <12 mo of age, with the next largest group between the ages of 1 and 4 y, with a few subjects >4 y old. RSV was the most commonly detected virus followed by rhinovirus/enterovirus and HMPV. Most subjects were infected with only one virus, but several had coinfection with multiple viruses (Supplemental Fig. 1, Table I). The presence of clinically significant acute lung injury, as defined by PARDS (26) (see 2Materials and Methods), was diagnosed in 25% (16 of 62) of enrolled subjects. We analyzed a total of 113 URT samples obtained from nasopharyngeal aspirates, 82 LRT samples obtained from endotracheal tube or tracheostomy aspirates as previously described (25), and 21 blood samples. These samples include 43 paired URT and LRT samples from 15 individuals.

We analyzed the immune cell composition in the URT based on a gating strategy that enabled quantitation of the major innate and adaptive immune cell types (29) (Supplemental Fig. 1). Neutrophils represented the predominant cell type among CD45⁺ cells (Table II), similar to the composition we previously identified in LRT samples (25). Significant proportions of eosinophils and monocytes, along with lower frequency populations of mast cells, T cells, B cells, dendritic cells, and NK cells, were also present in the URT (Table II). We also assessed chemokine content in the URT to determine correlations with cell recruitment. We detected high concentrations of IL-8, a neutrophil-homing chemokine, with significant levels of T cell chemoattractants CXCL9 and CXCL10 in some samples and negligible MCP-1 and CCL5 (Supplemental Fig. 2). These results suggest that elevated levels of specific chemokines promote recruitment of neutrophils and lymphocytes to the URT during VRTI.

We compared the composition of T cells from the URT to those in the LRT and blood among subjects of varying ages receiving mechanical ventilation (Fig. 1). T cells from pediatric blood were predominantly CD4⁺ with a median CD8/CD4 ratio of 0.34 (range, 0.09–1.04) (Figs. 1B, 2A), which is lower than for adult blood (30). In contrast, T cells within the URT and LRT (98 of 195) contained increased frequencies of CD8⁺ T cells, with an increased ratio of CD8/CD4 T cells, often with CD8⁺ T cells predominating over CD4⁺ T cells (Figs. 1B, 2A). These results show altered T cell composition in the URT and LRT compartments compared with blood in infants with VRTI, suggesting a localized response.

In the results above, the T cell composition in the URT more closely resembled that of the LRT compared with blood (Fig. 2A). To directly compare T cell composition in the URT and LRT, we analyzed paired nasopharyngeal and endotracheal tube aspirates obtained from the same patient for each sample day. Linear regression analysis demonstrated a strong correlation between the CD8/CD4 T cell ratio in the URT and LRT for all paired samples (r² = 0.63; p < 0.0001) (Fig. 2B). This strong correlation in T cell composition between these two respiratory sites in children suggests that sampling the URT can provide a representative indication of pulmonary T cell responses.

Lung injury is a severe complication of pediatric VRTI, which we previously found occurred in a significant proportion of infants requiring mechanical ventilation (25). To assess whether T cell composition in the URT or LRT was associated with disease severity, we applied the new pediatric-specific criteria for acute lung injury, PARDS (26). Pediatric subjects with VRTI who developed PARDS had significantly increased peak CD8/CD4 T cell ratios in both the URT and LRT compared with subjects who did not develop PARDS (Fig. 3A). These results show that T cell responses in both the URT and LRT are indicative of disease severity.

We further investigated whether T cell content and/or disease severity was associated with specific viral infections. Increased frequencies of CD8⁺ T cells in the URT and LRT of subjects who developed PARDS occurred mostly in subjects infected with RSV and HMPV (Fig. 3B), which constituted most subjects enrolled (Table I). However, subjects with influenza and coronavirus who developed PARDS also exhibited local accumulation of CD8⁺ T cells (Fig. 3B). This increased CD8⁺ T cell content in the URT and LRT of children with PARDS from VRTI across multiple viral pathogens supports a role for the local immune response in the development of lung injury during VRTI.

We further investigated the T cell subset composition in blood, URT, and LRT based on coordinate expression of CD45RA and CCR7, delineating naive T cells (CCR7⁺/CD45RA⁺), Tcm (CCR7⁺/CD45RA⁻), Tem (CCR7⁻/CD45RA⁻), and Temra (CCR7⁻/CD45RA⁺) (Fig. 4A). In the blood, both CD4⁺ and CD8⁺ T cells were predominantly naive (70 to >90%) with much lower proportions of Tcm, Temra, and Tem (1–15%). In stark contrast, the predominant subset in both the URT and LRT was Tem for both CD4⁺ and CD8⁺ T cells, with very low frequencies of naive T cells (1–31%) (Fig. 4A, 4B). For CD4⁺ T cells, smaller but significant proportions of Tcm were found in both respiratory tract compartments, whereas Temra were found in very low frequencies in most subjects, with some variability (Fig. 4A, 4B). Moreover, for CD8⁺ T cells, significant populations of Temra were observed in both URT and LRT, over a broad range of frequencies (7–81%) (Fig. 4A, 4B). Notably, for both CD4⁺ and CD8⁺ T cell subsets, there were no statistically significant differences in composition between the URT and LRT when comparing paired samples from the same subject (Fig. 4C). These results demonstrate biased localization of Tem and Temra subsets in the respiratory tract during pediatric VRTI.

We asked whether the wide variability we found in subset composition, particularly among CD8⁺ Tem and Temra subsets, was associated with age during the early postnatal period. Plotting T cell subset frequency with age revealed a cluster of higher Temra frequencies during early infancy that exhibited a steep decline after 2 mo of age, suggesting that this response does not wane but rather changes abruptly after the early neonatal period (Supplemental Fig. 3A). Compiling subset data of URT samples from subjects grouped by early neonatal (<2 mo) to later (>2 mo) reveals no significant difference in proportions of CD4⁺ Tcm or Tem (Fig. 4D). However, for CD8⁺ T cells, there was a marked increase in Temra (and compensatory decrease in Tem) in infants <2 mo of age, compared with those >2 mo old (Fig. 4D). A similar trend toward increased Temra was observed for CD4⁺ T cells from young infants, although the overall CD4⁺ Temra frequency was still low (<30%) (Fig. 4D). The inverse relationship of Temra and Tem was correlated for CD8⁺ T cells (Supplemental Fig. 3B), but not for other CD8⁺ or CD4⁺ T cell subsets.

There were no significant differences in local T cell subset composition (CD4⁺ Tem/Tcm/Temra or CD8⁺ Tem/Temra) between subjects who developed PARDS and those who did not (data not shown). These results suggest that T cell differentiation in response to respiratory viral infection in early life is skewed to terminal effector-like (Temra) over memory T cell fate, consistent with results in mouse neonatal T cells (11); however, these early life T cell subset variations are not necessarily related to increased disease severity.

Mouse models have shown that Trm in the airways and lungs mediate optimal protective immunity to respiratory viruses (15–17, 31). We assessed whether Trm were present in the URT and LRT of infants and young children with VRTI, based on expression of cell surface markers known to be associated with human lung Trm, including CD69 and CD103 (23, 32). Expression of CD69 with or without CD103 was found on most of the Tem phenotype CD4⁺ and CD8⁺ T cells from the URT and LRT (data not shown) samples in subjects as young as 6 wk of age, but not on T cells from paired blood samples, which were uniformly CD69⁻ and CD103⁻ (Fig. 5A). CD69 was expressed by the majority of CD4⁺ and CD8⁺ Tem in the URT and LRT, and this fraction did not change significantly with age during the early childhood years (Fig. 5B). In contrast, CD103 was expressed by a higher frequency of CD69⁺CD8⁺ Tem compared with CD69⁺CD4⁺ Tem in the URT and LRT, and age-related effects were observed with CD103 expression (Fig. 5C). Importantly, coexpression of CD69 and CD103, thought to represent the most mature Trm (20, 33), increased with age during the early years of life for both CD4⁺ and CD8⁺T cells in the URT and LRT (Fig. 5C). These results indicate that CD69⁺ memory T cells are present in the respiratory tract during pediatric VRTI throughout the early years of life, but that the frequency of fully differentiated CD103-expressing Trm increases with age during this period.

Our findings that infants infected with VRTI in the early months of life exhibited both increased Temra accumulation and decreased Trm formation in the respiratory tract suggested an altered differentiation pathway in early life T cell responses. We previously showed in an infant mouse model of influenza infection that infant T cells exhibited increased upregulation of the transcription factor T-bet, which inhibited Trm formation (11). We therefore investigated T-bet expression by human infant T cell subsets in blood or respiratory tract to assess whether a similar increase in T-bet expression is observed in human early life T cells. In T cells obtained from healthy adult blood donors (Supplemental Table I), the highest level of T-bet expression was found in CD8⁺ Temra followed by Tem, with low T-bet levels in naive T cells (Fig. 6A) as previously described (8, 9). In pediatric blood, T-bet expression was also higher in Temra and Tem compared with naive T cells, and was further increased in Tem compared with Temra (Fig. 6A). These differences in relative T-bet expression in pediatric versus adult T cell subsets are significant as shown in paired analysis of multiple donors (Fig. 6B, left) and in compiled data showing a higher ratio of T-bet expression in Temra/Tem in adults compared with pediatric samples (right).

We further investigated T-bet expression in respiratory tract samples, identifying similar trends of increased expression of T-bet by pediatric Tem compared with Temra in blood, URT, and LRT samples from an individual subject (Supplemental Fig. 3C). Pediatric URT T cells from infected infants of different ages exhibited higher T-bet expression on Tem compared with Temra, with some overall decrease in T-bet expression with age (Fig. 6C). These data suggest that increased T-bet expression is associated with T cell differentiation by pediatric T cells during early life.

Infants are exposed to a multitude of viral pathogens with varying pathogenicity during their formative years (34–36), with viral clearance ultimately relying on T cell activation and effector differentiation (37). Virus-specific T cells are generated in early life (24), but whether these responses give rise to long-lasting memory or prevent tissue injury remains unknown. In this study, we show that infants and young children mount a robust T cell response in both the URT and LRT, with increased CD8⁺ T cell infiltration associated with lung injury and certain types of viral infections. We demonstrate that during active VRTI in infants and young children, local responses are predominated by Tem whereas those in circulation remain overtly naive. In the airways, Temra predominated in the early neonatal period and decreased thereafter, whereas development of mature Trm occurred later after the first year of life. Importantly, this altered T cell differentiation was associated with increased T-bet expression in respiratory tract T cells during early infancy, and in circulating Tem in the blood. Our results identify developmental regulation of local respiratory T cell responses by subset and localization, with intrinsic alterations in transcriptional regulation of human T cells in early life.

Studies in mice and humans have demonstrated the importance of respiratory Trm in mediating protection from VRTI (15, 38). The establishment of local memory responses in the URT is critical to prevention of infection in the LRT (16), and intranasal vaccination has been shown to induce generation of virus-specific pulmonary Trm, capable of providing protection in mice (31, 39). We found that Trm phenotype cells expressing CD69 are detectable in the respiratory tract in the youngest subjects, as early as 6 wk of age. However, CD103 expression by these CD69⁺ memory T cells, which is a feature of mucosal CD8⁺ Trm, increases with age in the first few years of life, consistent with our previous findings that CD69⁺memory T cells in infant mucosal tissues had reduced CD103 expression compared with mucosal tissues from older children or adults (32). These results suggest that full Trm maturation requires sustained exposure to Ags, multiple prior infections, and/or prolonged exposure to signals from the tissue environment. Lung-specific factors such as those derived from commensal microbial species, cytokines, and chemokines may be differentially present in infant lungs compared with those of older children, with resulting effects on Trm maturation.

Concomitant with reduced Trm in infant URT samples was an increase in Temra frequency, indicating a preponderance of terminal effector cell generation. We likewise show elevated T-bet expression in infants compared with adult T cells in the blood, as well as elevated T-bet expression in infant URT, suggesting that intrinsic properties of infant T cells may impact their differentiation fate. Notably, our results with human pediatric respiratory samples are similar to our previous studies in mice where we found that increased T-bet expression in infant T cells promoted effector T cell generation and recruitment to the lung but inhibited Trm formation during early life infection and intranasal vaccination (11). We propose that targeting T-bet during early life infection or vaccination may be advantageous for promoting establishment of protective, long-lived T cell populations in the respiratory environment.

Studying cellular immune responses in children is limited due to difficulties in obtaining samples from the local site of infection. Mouse models have shown that following viral respiratory infection, T cells in the URT are clonally related to those in the LRT (40). We have demonstrated a correlation between T cell responses in the URT and LRT in both magnitude and subset diversity during VRTI in children. Additionally, the T cell response in these compartments is associated with the development of PARDS in the setting of multiple inciting viral agents. Although we did not quantify viral load in the respiratory tract of enrolled subjects, multiple studies have shown considerable overlap in viral titers and disease severity (41–44), suggesting that host responses play a critical role in determining disease course.

In conclusion, our results provide novel evidence for an association of local immunologic responses to disease severity in pediatric VRTI. These findings have important implications for disease treatment and development of protective responses, including potential noninvasive avenues that could be used for diagnostic, prognostic, or mechanistic investigations. Our results lend further evidence to the need to study localized responses to further the understanding of immune-mediated protection during active viral infection.

We thank the families who participated in this study. We thank the PICU nurses and respiratory therapists for help in the collection of samples. We thank Filip Cvetkovski for critical reading of the manuscript.

The authors have no financial conflicts of interest.