Reduced generation of lung tissue–resident memory T cells during infancy

Infants exhibit increased morbidity and mortality after respiratory infections and experience more repeat infections compared with older children and adults, suggesting impaired protective immunity. The worse outcome for infants in response to infection and their limited or delayed response to vaccines (Siegrist, 2007) have been attributed to the immaturity of immune responses and to T lymphocytes, in particular, which coordinate adaptive immunity (PrabhuDas et al., 2011). Although differences in T cell subset composition and cytokine profile between infant and adult T cells have been described (Lewis et al., 1991; Gibbons et al., 2014; Thome et al., 2016), the basic mechanisms underlying the regulation of infant T cell responses, including their functional differentiation, localization, and maintenance in response to infection remain undefined. There is a critical need for new insights into infant immune responses to both promote protection in response to infection and maximize efficacy of the multiple vaccines administered in early life.

Effective clearance of respiratory pathogens is coupled to establishment of lung-localized effector and memory T cells. In adult mouse models, lung-localized Th1 effector cells producing IFN-γ are important for directing clearance of primary influenza infection (Graham et al., 1993, 1994). We previously showed that populations of CD4⁺ and CD8⁺ lung tissue–resident memory T cells (TRM) are generated in response to influenza infection or i.n. administration of live-attenuated influenza vaccine (LAIV) in mice and that these cells mediate rapid, in situ protective responses to secondary viral challenge (Teijaro et al., 2011; Turner et al., 2014; Zens et al., 2016). In humans, influenza-specific CD4⁺ and CD8⁺ T cells with TRM phenotypes have been identified within lung tissue (de Bree et al., 2005; Purwar et al., 2011; Turner et al., 2014), and TRM-phenotype cells comprise the majority of memory T cells in diverse human tissues (Sathaliyawala et al., 2013; Thome et al., 2014). The robust protection mediated by TRM in the lungs and their predominance within multiple tissue sites (Masopust et al., 2001; Wakim et al., 2010; Jiang et al., 2012; Shin and Iwasaki, 2012) suggests that TRMs are an important target for promoting antiviral immunity by vaccines and immunotherapies.

The generation of tissue-localized T cell responses within the lung or other sites and the extent to which protective T cell memory and TRMs can be established during infancy have not been well studied. In contrast to adults, most peripheral T cells are naive in early life (Thome et al., 2016) and have distinct patterns of homing receptor expression (Grindebacke et al., 2009; Crespo et al., 2012). Neonatal and infant T cells also exhibit differences in cytokine expression and differentiation after in vitro activation or infection, compared with their adult counterparts (Lewis et al., 1986, 1991; Gibbons et al., 2014; Smith et al., 2014). How such differences affect protection and the generation of lasting T cell memory after infection or vaccination is not known.

We hypothesized that reduced protection after infection and decreased vaccine responses observed during infancy could be due to impaired tissue localization of effector T cell responses and/or the establishment of persisting TRM. Using an infant mouse model of influenza infection and vaccination, we found that infants mounted robust, primary, lung-localized CD4⁺ and CD8⁺ T cell responses to virus infection and LAIV. However, these cells were inefficiently maintained long term as TRM. In reciprocal transfers, we observed reduced lung TRM establishment after infection by infant, compared with adult, CD4⁺ T cells in either adult or infant hosts, suggesting T cell–intrinsic differences, rather than the lung environment mediating the distinct infant immune responses. We found distinct transcriptional profiles for infant, compared with adult, T cells after short-term activation in vitro and during the acute response to infection in the lung in vivo, with enhanced expression of T-bet and T-bet–regulated genes in both conditions. Importantly, reducing T-bet expression during infection promoted lung TRM establishment to near-adult levels, indicating that altered differentiation of infant T cells after infection affects TRM formation. These findings provide mechanistic insight into the process of TRM establishment with important implications for promoting vaccine-mediated protection during early life.

We used an infant mouse model to assess tissue-localized immune responses after influenza infection. To accurately recapitulate the immune environment of humans during early life, we elected to use 2-wk-old mice for two reasons: (1) unlike humans, mice are profoundly lymphopenic at birth; and (2) a significant fraction of peripheral T cells in the first week of mouse life have a memory phenotype because of recent homeostatic expansion (Garcia et al., 2000; Le Campion et al., 2002). We assessed T cell populations in the spleen and lungs of mice at different postnatal times and found that, at 2 wk old, the peripheral T cell compartment begins to resemble that of humans in early life (Thome et al., 2016), with increased T cell numbers in the lungs and spleen (Fig. 1 A) exhibiting predominant, naive phenotypes (not depicted).

We infected 2 wk-old, infant mice and 8–16-wk-old adult mice, for comparison, with weight-adjusted, sublethal doses of influenza virus. Throughout the course of the infection, infant mice had similar lung viral burdens as adults, with peak viral loads between days 3 and 5, and the virus was effectively cleared by day 10. However, infants exhibited a slight delay in complete viral clearance (D9 vs. D11) relative to adults (Fig. 1 B). The neutralizing antibody (Ab) titer on day 10 after infection was reduced in infants compared with adults (Fig. 1 C), coincident with the delayed viral clearance.

The primary T cell response in the lung at days 10–15 after infection was similar in infant and adult mice as that assessed by phenotype and by the frequencies of influenza-specific T cells. Overall, there were similar high frequencies (60–80%) of CD4⁺ and CD8⁺ T cells exhibiting an effector phenotype (CD44⁺CD62L^lo) in infant and adult lungs, compared with lower frequencies of effector cells in the spleen (30–40%; Fig. 1 D). Moreover, frequencies and total numbers of lung-localized CD8⁺ T cells specific for influenza virus–immunodominant epitopes, including hemagglutinin (HA), polymerase basic protein 1 (PB1), and nucleoprotein (NP), were similar in infant and adult mice 10–15 d after infection and greatly exceeded the frequencies in adult and infant spleens (Fig. 2 A). Together, these results reveal a robust, primary, influenza-specific T cell response in the lungs of infant mice that is comparable to adult mice, indicating that infant mice are fully competent to mobilize T cell responses to the site of respiratory tract infection.

Although peak T cell responses (D7–15) were similar between infant and adult mice, infant mice displayed a marked reduction in the percentage of lung virus–specific T cells, relative to adults, during the contraction phase of the response that occurs at early times after viral clearance (D16-21; Fig. 2 B). At longer times after infection (6–8 wk), when persisting cells bear phenotypic and functional features of memory T cells (Jelley-Gibbs et al., 2005; Teijaro et al., 2011; Turner et al., 2014), there were reduced frequencies and total numbers of NP-specific CD8⁺ T cells in mice infected initially as infants, compared with those infected as adults, with biased persistence in lung compared with spleen (Fig. 2, C and D). Taken together, these data demonstrate that infant mice generate lung-localized, virus-specific T cells during acute infection, which correlates with lung viral clearance, but show reduced persistence of virus-specific T cells in the lung at longer times after viral clearance, compared with persisting memory T cells observed in adult mice.

Because of the importance of TRM in long-term protection from repeated influenza infection (Teijaro et al., 2011; Zens et al., 2016), we investigated the generation of persisting lung TRM in mice previously infected as infants (at 2 wk old) compared with those previously infected as adults (at 10–16 wk old). To distinguish circulating T cells from tissue-localized TRM, we used a well-validated, i.v. Ab-labeling technique involving i.v. infusion of fluorescently conjugated anti–T cell antibodies to distinguish between circulating and tissue-resident T cells with the circulating T cells being labeled by i.v. Ab, whereas T cells retained within tissues are protected from Ab binding (Anderson et al., 2012; Turner et al., 2014). “Labeled” T cells, which bind i.v. Ab, and “protected” T cells, which do not, exhibit phenotypic markers of circulating T cells and TRM cells, respectively, and are localized to distinct niches of the lung (Anderson et al., 2012; Turner et al., 2014).

In mice previously infected as adults, most lung CD4⁺ and CD8⁺T cells (>75%) were protected from i.v. Ab labeling 6 wk after infection, consistent with our previous findings (Turner et al., 2014; Zens et al., 2016). In contrast, in mice previously infected as infants (at 2 wk old), the proportion of protected lung T cells at 6 wk after infection was greatly reduced compared with mice infected as adults (35–45%, on average) and was similar to that observed in uninfected controls (Fig. 3 A). Furthermore, although most NP-specific CD8⁺ T cells in the lungs of mice infected as adults were protected from i.v. labeling, consistent with localization in the tissue, a significant fraction (25–30%) of NP-specific lung CD8⁺ T cells in mice previously infected as infants were labeled by i.v. Ab (Fig. 3 B), indicating these cells were circulating. These results indicate that infection during infancy reduces the persistence of total and virus-specific memory T cells in the protected niche of the lung.

We further investigated whether protected lung T cells in infants expressed the canonical TRM marker CD69 on the cell surface, along with CD11a for CD4⁺ TRM and CD103 for CD8⁺ TRM (Mueller et al., 2013; Turner and Farber, 2014). Consistent with previous studies, a significant fraction of protected CD4⁺ and CD8⁺ T cells in adults were CD69⁺CD11a⁺ (>60% of CD4⁺) or CD69⁺CD103⁺ (∼50% of CD8⁺). By contrast, in mice previously infected as infants, <50% of protected CD4⁺T cells and 25% of protected CD8⁺T cells expressed those TRM phenotypes (Fig. 3 C). Expression of TRM markers in previously infected infants was, however, increased compared with uninfected controls (Fig. 3 C), suggesting that TRM generation during infancy is greatly reduced but not completely inhibited. Overall, these findings demonstrate reduced establishment of canonical lung TRM after influenza infection during infancy.

To assess whether reduced TRM formation in mice infected initially as infants resulted in decreased protection against heterosubtypic strains later in life, we challenged infant and adult mice infected with PR8 influenza 6 wk previously and naive adults as controls, with the heterosubtypic H3N2 viral strain X31. To further investigate TRM-mediated protection, independent from that mediated by circulating T cells, we treated cohorts of mice throughout infection with the sphingosine 1-phosphate receptor-1 agonist FTY720, which sequesters circulating T cells within the secondary lymphoid tissues and depletes circulating T cells (Chiba et al., 1998; Pinschewer et al., 2000), whereas lung TRM are maintained (Anderson et al., 2014; Turner et al., 2014; Zens et al., 2016).

In PBS-treated cohorts, heterosubtypic challenge of mice previously infected as adults or infants resulted in minimal weight-loss morbidity, compared with primary X31 infection of naive mice, which resulted in significant weight loss during the course of infection (Fig. 4 A). However, mice previously infected as adults exhibited enhanced viral clearance from the lungs by day 5 after challenge compared with mice previously infected as infants although titers in this group were reduced, compared with primary infection controls (Fig. 4 B). These results indicate that infection during infancy results in immunity to heterosubtypic infection, which is reduced in efficacy compared with infections occurring later in life.

In FTY720-treated groups, mice previously infected as infants lost significant weight after heterosubtypic challenge, which was similar to the weight loss of primary infection controls and significantly greater than that observed in mice previously infected as adults (Fig. 4 A). Moreover, the lung viral burden 5 d after challenge was similarly increased in naive mice and mice previously infected as infants, compared with mice previously infected as adults (Fig. 4 B). These results demonstrate that FTY720 treatment (reducing circulating T cell responses) reduced the rapid viral clearance and early protection in mice previously infected as infants, whereas the early and enhanced protective response in mice previously infected as adults was preserved. These findings suggest impaired establishment of protective lung TRM after infection in infancy.

We previously showed in adult mice that, of the current influenza vaccines, LAIV administered i.n. and inactivated influenza virus (IIV) administered by injection (s.c. or i.p.), only i.n. LAIV promoted generation of lung TRM protective against heterosubtypic infection (Zens et al., 2016). We investigated whether reduced TRM generation in infant, compared with adult, mice was specific to influenza infection or was a generalized defect in overall immune and vaccine responses. After immunization of infant and adult mice i.n. with LAIV or i.p. with IIV, we analyzed lung T cell populations 10 d or 6 wk after vaccination and protection to heterosubtypic challenge. As we previously reported, adult mice exhibited robust primary lung CD4⁺ and CD8⁺ effector responses to i.n. LAIV and minimal lung T cell responses to IIV (Fig. 5 A). Infants exhibited similar primary vaccine responses to adults with higher percentages of lung CD4⁺ and CD8⁺ effector cells were observed after vaccination with LAIV, compared with IIV (Fig. 5 A). Thus, infants mount comparable primary lung T cell responses as adults to the vaccine formulations.

At longer times after vaccination, adult mice generated significant frequencies of lung CD4⁺ and CD8⁺ TRM (as assessed by CD69 and CD103 expression) after vaccination with LAIV but not IIV (Zens et al., 2016; Fig. 5 B). In contrast, infant mice vaccinated with LAIV or IIV had only low frequencies of persisting lung TRM (Fig. 5 B). TRM-mediated protection to heterosubtypic challenge with PR8 influenza was also assessed in vaccinated mice in the presence of FTY720, as shown in Fig. 4. Infant mice vaccinated with LAIV or IIV and adult mice vaccinated with IIV all lost significantly more weight throughout the course of infection, compared with adult mice vaccinated with LAIV who did not exhibit as extensive weight loss (Fig. 5 C). In terms of lung viral clearance, adults and infants vaccinated with IIV had similarly high lung viral burdens 7 d after infection, whereas adults and infants vaccinated with LAIV had reduced lung viral titers at that time point (Fig. 5 D). These results demonstrate that, although vaccination with LAIV during infancy elicits reduced establishment of lung TRM compared with adults, as observed with influenza infection, some degree of protective lung immune responses was generated by LAIV administration during infancy.

We investigated whether the reduced generation of lung TRM from influenza infection during infancy was due to the distinct lung environment of infants and/or to intrinsic differences between infant and adult T cells. We focused our mechanistic studies on CD4⁺ T cells because of their known requirement in generation of memory CD8⁺ T cells and lung CD8⁺ TRM formation (Sun and Bevan, 2003; Bevan, 2004; Williams et al., 2006; Laidlaw et al., 2014). We performed reciprocal adoptive transfers of infant- or adult-derived influenza HA-specific CD4⁺ T cells (HA T cells) to congenic infant or adult recipients to generate four groups: (1) adult HA–adult host (adult to adult), (2) adult HA–infant host (adult to infant), (3) infant HA–adult host (infant to adult), and (4) infant HA–infant host (infant to infant). The resultant groups were subsequently infected with PR8 influenza; primary lung T cell responses were assessed on day 13, and persistence as memory T cells was assessed 6 wk later.

Early after transfer and infection, there were comparable frequencies of infant- and adult-derived HA T cells in both the lungs and mediastinal LN of adult recipients (Fig. 6 A), suggesting that both adult and infant T cells survived adoptive transfer to comparable extents. Similarly, during the peak primary T cell response (day 13) similar percentages of HA T cells were detected in the lungs in all recipient mice (Fig. 6, B and C). At longer times after infection, there were comparable low, but discernable, frequencies of adult HA T cells in the lungs of both adult and infant recipients; however, we did not detect significant populations of persisting, infant-derived HA T cells in the lungs of adult or infant hosts (Fig. 6, B and C). These results demonstrate that, although infant cells migrate to the lung and expand in response to influenza infection, similar to adult cells, they are not efficiently maintained as lung memory T cells regardless of the host environment, indicating a cell-intrinsic basis for reduced TRM generation in infants.

We hypothesized that distinct gene expression patterns between adult and infant T cells during the primary response could drive T cell–intrinsic differences in TRM formation. We transferred infant or adult HA T cells to adult congenic hosts, infected them with PR8 influenza as shown in Fig. 6, and sorted infant or adult HA-specific CD4⁺ T cells (referred to as “HA effectors”) from host lungs 13 d later for whole transcriptome profiling by RNA sequencing.

A total of 634 genes were significantly differentially expressed (P < 0.05) between infant and adult HA effectors (Fig. 7 A and Tables S1 and S2). Among those, several well-established molecules important for lung tissue homing and retention, including CCR2, CCR5, CXCR6, CCR8, and Itga1 (VLA-1; Dawson et al., 2000; Ray et al., 2004; Richter et al., 2007; Kohlmeier et al., 2008; Morgan et al., 2008; Lee et al., 2011), were consistently up-regulated by infant HA effectors relative to those of adults (Table S1). Expression of transcription factors typically associated with effector T cell differentiation and function, including Blimp1, Id2, and T-bet (Intlekofer et al., 2005; Cannarile et al., 2006; Ji et al., 2011; Yang et al., 2011), were also increased in infant, compared with adult, lung HA effectors (Table S1). Conversely, expression of transcription factors associated with memory T cell generation, including Eomes, Bcl6, and Id3 (Johnston et al., 2009; Banerjee et al., 2010; Ji et al., 2011; Yang et al., 2011) were down-regulated in infant, compared with adult, HA effectors (Table S2).

Ingenuity Pathway Analysis (IPA; Krämer et al., 2014) revealed STAT4 and its major target T-bet (Thieu et al., 2008) as predicted, upstream, positive transcriptional regulators of the genes differentially expressed between infant and adult HA effectors (Fig. 7 B). To assess whether either of these or other pathways were up-regulated in infant T cells early after activation, we performed whole transcriptome profiling of infant and adult polyclonal naive (CD44^lo) CD4⁺T cells after short-term activation in vitro with anti–CD3/CD28 Abs for 24 h (Tables S3 and S4). Similar to the differential gene expression data of lung-homing effector cells, T-bet was among the top predicted, upstream regulators of short-term, activated, infant CD4⁺ T cells by IPA and the only transcriptional regulator present in both data sets with a similar direction of expression (Fig. 7, B and C). Gene Ontology (GO) analysis demonstrated an enrichment for cellular processes associated with cellular replication and proliferation by short-term–activated infant, compared with adult, CD4⁺ T cells (Table S5). These results suggest enhanced proliferation and differentiation of infant T cells relative to adult cells initiated in the early phases of activation.

To further evaluate a role for T-bet in the regulation of this differentially expressed gene profile, we compared genes differentially expressed between infant and adult HA effectors (634 total) to a list of genes found to be significantly positively or negatively regulated by T-bet in an independent study (430 total; Zhu et al., 2012). There was a substantial degree of overlap between the infant, versus adult, HA effectors gene set and T-bet–regulated genes, with 12% of genes represented in the T-bet–regulated set (Fig. 7, A and D; and Tables S1 and S2). (A similar proportion (∼10%) of genes differentially expressed between in vitro–activated infant and adult CD4⁺ T cells were present in the T-bet–regulated set [Tables S3 and S4]). Gene set enrichment analysis (Subramanian et al., 2005) was applied to the ranked genes to determine whether the direction of gene expression in infant HA effectors was consistent with up-regulated T-bet expression. GSEA revealed a strong correlation (P < 0.0001) between genes up-regulated in infant HA effectors with genes known to be positively regulated by T-bet; genes known to be negatively regulated by T-bet were also down-regulated in infant HA effectors, relative to adult (Fig. 7 D). Together these results provide evidence that enhanced T-bet expression may be a key regulator of the distinct differentiation of infant, compared with adult, effector T cells.

We evaluated T-bet expression by flow cytometry in infant and adult CD4⁺ T cells before and after activation both in vitro and in vivo. Activation of mouse, infant, naive (CD44^lo) CD4⁺ T cells in vitro resulted in augmented T-bet expression with increased kinetics compared with adult, naive CD4⁺T cells, although levels were comparable by 72 h after stimulation (Fig. 8 A). Similarly, in humans, a greater percentage of infant cells sustained high levels of T-bet relative to adult-derived cells (Fig. 8 B). In vivo, we assessed T-bet expression by infant and adult HA T cells adoptively transferred to infant or adult congenic hosts, which were subsequently infected with influenza. At 13 d after infection, a greater percentage of infant cells expressed T-bet compared with adult cells. This enhanced expression of T-bet was further independent of the host environment with comparable percentages of T-bet^hi, infant, HA-specific T cells in both infant and adult hosts (Fig. 8 C), suggesting that enhanced T-bet expression occurs in a cell-intrinsic manner.

For CD8⁺ T cells, expression of T-bet is inversely correlated with the expression of CD127 (Intlekofer et al., 2007; Joshi et al., 2007; Knox et al., 2014), the α subunit of the IL-7 receptor, a well-described T cell survival factor (Schluns et al., 2000; Kaech et al., 2003), with high levels of T-bet driving the generation of terminally differentiated, short-lived effector cells (Intlekofer et al., 2007; Joshi et al., 2007). We therefore measured CD127 expression during the course of influenza infection in both adult and infant mice. At 6, 12, and 21 d after infection, CD127 expression levels were significantly higher in adult lung CD4⁺ T cells compared with infant cells (Fig. 8 D), suggesting that enhanced T-bet expression in infant T cells might be associated with reduced capacity to respond to survival factors.

Our results suggested that increased T-bet expression by infant T cells after activation could bias toward generation of short-lived effector cells, inhibiting development of memory T cells and persistence of lung-homing effector cells as TRMs. To test whether specific reduction of T-bet expression by infant T cells could enable generation of lung TRM, we used infant mice heterozygous for T-bet to effectively “modulate” T-bet expression. Because T-bet expression is codominant, deletion of a single allele reduces expression on a per-cell basis. We infected infant T-bet^+/+ (WT) and T-bet^+/− mice, along with T-bet^+/− and WT adult controls, with PR8 influenza and assessed lung TRM generation 6–8 wk later.

At 6 wk after infection, T-bet^+/− mice previously infected as infants had increased numbers of protected lung CD4⁺ and CD8⁺ T cells, compared with WT mice infected as infants, and similar numbers of protected lung T cell as WT and T-bet^+/− mice infected as adults (Fig. 9 A). Similarly, in T-bet^+/− mice previously infected as infants, most protected lung CD4⁺ (∼50–60%) and CD8⁺ T cells (40–50%) expressed canonical CD4⁺ and CD8⁺ TRM markers, similar to WT and T-bet^+/− mice previously infected as adults (Fig. 9 B). By contrast, only 40% of protected CD4⁺ T cells and 20% of protected CD8⁺ T cells in WT mice infected as infants expressed those markers (Fig. 9 B). In addition, we confirmed that T-bet expression was reduced in both naive CD4⁺ and CD8⁺ T cells and TRMs in T-bet^+/− infant memory mice compared with WT adult memory mice (Fig. 9 C). These results demonstrate that reducing T-bet expression in infant T cells results in TRM establishment similar to that in WT adults, suggesting that dysregulation of T-bet by infants during the primary response to infection results in the reduced generation of lung TRM.

Infants are highly susceptible to respiratory viral pathogens; however, the state of infant lung immune responses during homeostasis and infection is poorly understood. Using an infant mouse model that recapitulates the state of human early life immunity, we provide mechanistic insights into the regulation, differentiation, and maintenance of infant lung T cell responses during respiratory virus infection and vaccination. We demonstrated that influenza infection (or i.n. vaccination) of infant mice results in effective mobilization of CD4⁺ and CD8⁺ T cell responses in the lung and viral clearance but inefficient establishment of lung TRMs and in situ protective immunity. Reduced TRM differentiation by infant lung-homing T cells was due to a distinct transcriptional profile indicative of enhanced proliferation and increased expression of T-bet–regulated genes. This enhanced expression of T-bet correlated with reduced expression of the survival factor CD127 by infant T cells. Significantly, down-modulation of T-bet expression augmented TRM development in infant mice. Together, our findings suggest that infant T cells are intrinsically programmed for short-term pathogen clearance rather than long-term maintenance and that targeting a key regulator can enable long-term immunity at this critical life stage.

We demonstrated that infant mice generate influenza-specific, lung-homing T cell responses after respiratory infection, which are qualitatively and quantitatively similar to responses observed in adult mice. That result was confirmed in different strains of mice for T cells specific for multiple influenza epitopes and is consistent with studies in humans showing mobilization of T cells with effector phenotypes in the airways of infants with respiratory infections (Connors et al., 2016). Previous studies of respiratory infection in neonatal mice (<1 wk old) have reported delayed and quantitatively reduced CD8⁺ T cell responses with altered homing patterns relative to adults (You et al., 2008; Lines et al., 2010). However, neonatal mice are profoundly lymphopenic, resulting in extensive T cell homeostatic expansion in the early neonatal period (Min et al., 2003; Schönland et al., 2003), which can affect T cell differentiation and function (Surh and Sprent, 2000, 2008). Mice at 2 wk of age were chosen specifically to be representative of human infants who are born replete with a full T cell complement.

Developing protective immunity to respiratory infections is coupled to the generation of lung TRMs (Teijaro et al., 2011; Sakai et al., 2014; Turner et al., 2014; Wu et al., 2014), which can also be generated by i.n. vaccination with bacille Calmette–Guérin or LAIV (Perdomo et al., 2016; Zens et al., 2016). We showed that, in multiple mouse strains and with different viral subtypes, infection or vaccination during infancy results in impaired TRM formation, as assessed by tissue localization and phenotype. Our findings are consistent with recent results using human organ-donor samples showing reduced CD103 expression by memory CD8⁺ T cells isolated from infant lungs and intestines compared with memory CD8⁺ T cells in adult mucosal sites (Thome et al., 2016). In response to heterosubtypic challenge, mice previously infected during infancy were less protected with reduced viral clearance compared with previously infected adults—particularly in FTY720-treated mice—when protection was limited to TRM. In untreated mice previously infected as infants, there was enhanced viral clearance to heterosubtypic challenge compared with naive mice, suggesting that infants may generate some circulating protective memory T cells, albeit in reduced numbers, compared with adults as also shown in an systemic viral infection model (Smith et al., 2014).

Reduced TRM generation during infant infection was intrinsic to infant T cells and was not a function of the infant lung environment as determined by adoptive transfer studies. The distinct transcriptional profile of infant (compared with adult) T cells migrating to the lung during infection included genes associated with migration, homing, differentiation, and cytokine responses, suggesting a distinct differentiation program for infant T cells. Notably, the gene expression pattern was enriched for T-bet–regulated genes, a master regulator for Th1 responses and effector cell differentiation (Szabo et al., 2000, 2002). Similarly, at very early times (24 h) after stimulation, infant T cells demonstrated a gene expression pattern enriched for molecules involved in cell cycle and proliferation consistent with enhanced T-bet expression. T-bet levels have been shown to control memory T cell generation with high levels of T-bet leading to terminal differentiation and cell death (Joshi et al., 2007) and intermediate expression of T-bet promoting TRM establishment (Laidlaw et al., 2014; Mackay et al., 2015).

We found that infant T cells in mice and humans exhibit greater up-regulation of T-bet expression after activation compared with adult T cells, suggesting that infant T cells are intrinsically programmed for terminal effector development. Our results are consistent with findings that mouse, neonatal CD8⁺ T cells exhibit increased proliferative expansion when stimulated (Smith et al., 2014), as do human cord-blood T cells (Schönland et al., 2003; Schüler et al., 2004)—features consistent with progression toward a terminal differentiation program. Infection of T-bet^+/− infant mice expressing intermediate levels of T-bet, promoted TRM establishment at levels similar to that observed in both WT and T-bet^+/− adults, suggesting a key role for that regulator in determining infant T cell fate.

Although the mechanisms underlying increased T-bet expression in infant T cells remain to be determined, the homeostatic expansion of T cells that occurs in early life (postnatally in mice and during fetal development in humans; Haynes et al., 1988) may be a contributing factor. In mice, homeostatic proliferation depends on mTOR signaling (Li et al., 2011), which, in turn, drives T-bet expression (Rao et al., 2010). The increased propensity for T-bet up-regulation may be a developmental adaptation for early life T cell responses. A major function of T-bet is to promote IFN-γ expression by binding to and remodeling the IFN-γ promoter locus (Mullen et al., 2001), which in neonatal T cells is hypermethylated and, therefore, less accessible to transcriptional up-regulation (White et al., 2002). Increased T-bet expression may be necessary to overcome transcriptional inhibition for promoting IFN-γ production (even at a low level) necessary for pathogen clearance to survive the critical period of early life, when generation of long-term protection is less pressing.

Our results have important implications for vaccine optimization in the early life period, to limit the global burden of disease. Although we demonstrate reduced generation of LAIV vaccine responses in infant, compared with adult, mice, particularly when assessing TRM formation, LAIV-vaccinated infants are better protected from heterosubtypic challenge than are IIV-vaccinated mice, suggesting that tissue-targeting is an optimal strategy for promoting early T cell responses. In humans, LAIV has previously shown to exhibit higher efficacy in children compared with adults (Belshe et al., 2007), and conversely, oral administration of vaccines during infancy has also proven efficacious. We propose that a combination of tissue targeting with modulation for promoting TRM responses may not only fortify in situ immunity during this vulnerable period but also generate long-term protective responses persisting throughout childhood and beyond.

C57BL/6 or BALB/c mice were purchased (Charles River) or bred and maintained under specific pathogen–free conditions in a BSL2 biocontainment room within Columbia University Medical Center (CUMC) animal facilities. TS1 (HA TCR) mice (Kirberg et al., 1994) were bred and maintained within CUMC animal facilities. T-bet heterozygous mice were generated by crossing male Tbx21^−/− mice (Intlekofer et al., 2005), provided by S. Reiner (Columbia Universtiy, New York, NY), with WT C57BL/6 females. To generate littermate controls, T-bet heterozygous males and females were crossed, and offspring were screened by PCR of genomic DNA extracted from ear punches using the Clontech Terra PCR Direct Polymerase Mix kit (Takara Bio Inc.) with the following primers: (1) forward, 5′-TATGATTACACTGCAGCTGTCTTCAG-3′; (2) WT reverse, 5′-CAGGAATGGGAACATTCGCCTGTG-3′; and (3) null reverse, 5′-CTCTGCCTCCCATCTCTTAGGAGC-3′. For all studies, infant mice were used between 10–14 d after birth, and adult mice were used between 8 and 20 wk of age, with both genders used. All animal studies and procedures were conducted according to the National Institutes of Health guidelines for the care and use of laboratory animals and were approved by the CUMC Institutional Animal Care and use Committee.

Human splenic tissues were obtained from deceased (brain dead) organ donors at the time of organ acquisition for life-saving clinical transplantation through an approved protocol and material transfer agreement with LiveOnNY. Organ donors were free of chronic disease and cancer and were negative for HIV and hepatitis B and C. The study did not qualify as human subjects research, as confirmed by the Columbia University institutional review board because tissue samples were obtained from deceased individuals. Tissues were obtained and processed for single-cell suspensions, as previously described (Sathaliyawala et al., 2013; Thome et al., 2014, 2016).

FluMist Quadrivalent, 2015–2016, live-attenuated vaccine (MedImmune) and Fluzone Quadrivalent, 2015–2016, inactivated virus vaccine (Sanofi Pasteur), both containing the four strains: (1) A/California/7/2009 (H1N1) pdm09, (2) A/Switzerland/9715293/2013 (H3N2), (3) B/Phuket/3073/2013, and (4) B/Brisbane/60/2008, were purchased from Moore Medical. Anti–CD3e (clone 145-2C11) and anti–CD28 (clone 37.51) mAbs were purchased as azide-free, low-endotoxin formulations from BD for use in T cell–activation assays. Fluorescently conjugated Abs for cell-surface phenotypic staining, cell sorting, and intracellular staining were purchased from BD, BioLegend, and eBioscience and are listed, including target, clone, color, and manufacturer, in Table S6. The influenza nucleoprotein-specific (NP_366–374 ASNENMDAM) MHCI H-2D^b tetramer; the influenza, RNA-dependent RNA polymerase PB1 subunit-specific (PB1_703-711 SSYRRPVGI) MHCI H-2K^b tetramer; and the influenza HA-specific (HA_533–541 IYSTVASSL) MHCI H-2K^d tetramer, all conjugated to PE, were purchased from MBL International. FTY720 was purchased from Cayman Chemical.

Mice were anesthetized by inhalation of isoflurane (5% induction/2% maintenance), delivered through a vaporizer within the CUMC animal facilities, and infected i.n. with weight-adjusted doses of PR8 (A/Puerto Rico/8/34 [H1N1]) or X31 (A/Hong Kong/1/68-x31 [H3N2]) influenza viruses. Appropriate viral doses were determined based on viral strain, mouse age and strain, and primary versus heterosubtypic infection, as listed in Table S7. Morbidity after infection was monitored by daily examination and weight assessment. Mice were euthanized when weight loss exceeded 30% of starting weight.

For vaccinations, adult and infant mice were administered 20 μl of FluMist (one tenth of the adult human dose) i.n. for a final dose of 10^5.5–6.5 focus forming units/viral strain/mouse, or 50 µl of Fluzone (one tenth of the adult human dose) i.p. for a final dose of 1.5 µg HA/viral strain/animal.

Mice were treated daily with 1 mg/kg FTY720 (25 μg/mouse) or PBS i.p., beginning 2–3 d before infection and continuing daily throughout infection.

For in vivo antibody labeling, mice were given 3 µg fluorochrome-conjugated anti–Thy1.2 or anti–CD45 Ab (Table S6) by i.v. injection 10 min before tissue harvest. Lungs were then perfused with 15–20 ml PBS via instillation of the right ventricle of the heart. Lungs and spleens were collected and lymphocytes isolated, as previously described. (Turner et al., 2014). Cells were stained with fluorochrome-conjugated Abs. To detect influenza-specific T cells, cells were incubated for an additional 30 min at room temperature (19–25°C) in the dark with a tetramer reagent and were analyzed using an LSRII flow cytometer (BD), and data were assessed using FlowJo software (Tree Star).

Influenza titers in lung homogenates were measured by a 50% tissue culture infective dose (TCID₅₀) assay, as previously described (Teijaro et al., 2010), with titers expressed as the reciprocal of the dilution of lung extract that corresponded to 50% virus growth in Madine–Darby canine kidney cells, as calculated by the Reed–Muench method.

Serum-neutralizing Ab titers were measured by hemagglutination inhibition assay, as previously described (Teijaro et al., 2010). In brief, serum was collected by cardiac puncture, then treated overnight with Receptor Destroying Enzyme (Denka Seiken). Treated serum was serially diluted, incubated with PR8 or X31 whole virus particles as HA sources, then incubated with 0.5% chicken red blood cells (Lampire, Biological Laboratories). Titers were expressed as the reciprocal of the last dilution of serum that completely inhibited hemagglutination.

Naive (CD44^lo) CD4⁺ T cells were enriched from mouse splenocytes using MagniSort CD4⁺ naive T cell–negative selection kits (eBioscience). Human naive (CD45RO⁻) CD4⁺ T cells were purified from spleen using Mojosort naive T cell negative isolation kit (BioLegend). Purified, naive T cells (2 × 10⁵) were activated for 24–72 h at 37°C in 96-well, flat-bottomed plates coated with a solution of 5 µg/ml anti-CD3 and 5 µg/ml anti-CD28. For intracellular staining, cells were first surface stained with Ab reagents and then fixed for 30 min at room temperature (19–25°C) in the dark with FoxP3 fixation/permeabilization buffer (eBioscience), washed with permeabilization buffer (eBioscience), and stained with fluorescently conjugated Abs to the transcription factor T-bet (Table S6) diluted in permeabilization buffer. Staining was performed for 30 min at room temperature (19–25°C) in the dark, followed by a minimum of 30 min at 4°C protected from light, before washing with FACS buffer.

Naive CD4⁺ T cells were isolated from the spleens of infant (10–14 d old) or adult (8–16 wk old) TS1 mice using the MagniSort CD4⁺, naive T cell–enrichment kit (eBioscience), and 1.5–5 × 10⁵ purified cells were transferred by i.p. injection into naive, congenic (Thy1.2⁺) infant or adult recipients. The next day, recipient mice were infected i.n. with 10–13 TCID₅₀ PR8 influenza/g body weight (250–325 TCID₅₀ for adults and 70–91 TCID₅₀ for infants). For whole-transcriptome profiling studies, lungs were harvested 13 d after infection and processed as described for FACS analysis, and single-cell suspensions were stained with fluorescently conjugated Abs (Table S6). Transferred (Thy1.1⁺) T cells were sorted with a BD Influx cell sorter, and sorted T cells were immediately pelleted and flash-frozen at −80°C until RNA extraction.

RNA from sorted cells was isolated using the QIAGEN RNeasy Mini Kit, flash frozen, and stored at −80°C before quality analysis and library generation. Sample RNA integrity and concentration were assessed using an Agilent 2100 Bioanalyzer Instrument (Agilent Technologies), and all samples had RNA integrity number (RIN) values of ≥8. Library preparation and RNA sequencing were performed by the Columbia Genome Center using an Illumina HiSeq 2500 Instrument. RNA samples were sequenced in a single-read manner with a sequencing depth of 30 million reads. Mapping of reads to a mouse reference genome was performed with TopHat (Trapnell et al., 2009), with downstream estimation of gene abundance using cuff links (Trapnell et al., 2010). GEO accession number for RNA sequencing data: GSE101696.

Differential gene expression was determined using the R-based package DESeq2 (Love et al., 2014) available from Bioconductor (Huber et al., 2015). Gene expression heat maps were generated using Morpheus (http://software.broadinstitute.org/morpheus); with a p-value cutoff of 0.05 and sorting genes by log fold change (logFC). IPA software (QIAGEN) was used to predict upstream transcriptional regulators using all differentially expressed genes with a p-value cutoff of 0.10. Transcriptional regulators were determined by selecting “transcriptional regulators” from the upstream regulators tab. Differentially expressed gene data were assessed for T-bet regulation by GSEA. A ranking statistic was computed for each gene that passed the default, independent, filtering criteria (Bourgon et al., 2010) of DESeq2 (α = 0.1) by multiplying the log of its p-value with the sign of its logFC, and that ranked list was tested for enrichment of previously published T-bet–regulated gene sets (Zhu et al., 2012) by applying GSEA-preranked (Mootha et al., 2003) using a weighted scoring scheme and 1,000 permutations. GO data were evaluated using the protein–protein interaction database Search Tool for the Retrieval of Interacting Genes/Proteins (STRING; Szklarczyk et al., 2015) using the top 200 genes differentially expressed between infant and adult samples by logFC (P < 0.05).

Flow cytometry, weight loss, lung viral titer, and neutralizing Ab data were compiled and statistical analyses were performed using Prism software (GraphPad Software). Results are expressed as the mean value from individual groups ± SEM, unless otherwise designated, indicated by error bars. Significance between experimental groups was determined by one- or two-way ANOVA or the Student’s t test and corrected for multiple comparisons, as indicated, assuming a normal distribution for all groups.

Table S1 shows up-regulated genes in infant versus adult primary HA effectors. Table S2 shows down-regulated genes in infant versus adult primary HA effectors. Table S3 shows up-regulated genes in 24-h–stimulated infant CD4⁺ T cells. Table S4 shows down-regulated genes in 24-h–stimulated infant CD4⁺ T cells. Table S5 provides GO terms for differentially expressed genes in 24-h–stimulated infant and adult CD4⁺ T cells. Table S6 provides fluorophore-conjugated Abs used for flow cytometry. Table S7 provides the doses of influenza strain used in this study.

We thank Esi Lamouse-Smith for assistance with influenza vaccines and Tomer Granot and Daniel Paik for critical reading of this manuscript.

This study was supported by National Institutes of Health grant R01 AI100119 awarded to D.L. Farber. K.D. Zens and M. Miron were supported by National Institutes of Health grant T32 AI106711 (principal investigator, D. Fidock). Research reported here was performed in the Columbia Center for Translational Immunology Flow Cytometry Core, supported in part by National Institutes of Health grants S10RR027050 and S10OD020056. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

The authors declare no competing financial interests.

Author contributions: K.D. Zens and D.L. Farber designed the research studies. K.D. Zens, J.K. Chen, R.S. Guyer, F. Cvetkovski, and M. Miron conducted experiments and analyzed data. F.L. Wu analyzed RNASeq data. K.D. Zens and D.L. Farber wrote the manuscript.