Ontogeny of human mucosal-associated invariant T cells and related T cell subsets

Mucosal-associated invariant T (MAIT) cells are nonconventional CD3⁺ CD4⁻ CD161^high T lymphocytes, which express a semi-invariant TCR (Vα7.2-Jα33/20/12 in humans, Vα19-Jα33 in mice, combined with a restricted set of Vβ chains; Tilloy et al., 1999; Treiner et al., 2003; Reantragoon et al., 2013; Lepore et al., 2014). MAIT TCRs recognize microbial-derived riboflavin (vitamin B2) biosynthesis intermediate derivatives, such as 5-(2-oxopropylideneamino)-6-d-ribitylaminouracil (5-OP-RU), presented by the monomorphic MHC class I-related molecule (MR1; Treiner et al., 2003; Kjer-Nielsen et al., 2012; Corbett et al., 2014). MAIT cells are preferentially localized in mucosal tissues, including gut and lung, and the liver and represent the most abundant innate-like T cell population in human peripheral blood, comprising up to 10% of the entire T cell population (Martin et al., 2009; Dusseaux et al., 2011). This compares with just 0.1% for natural killer T (NKT) cells, another population of semi-invariant innate-like T cells recognizing glycolipids presented by CD1d. Upon recognition of microbial antigens, MAIT cells display immediate effector responses by secreting inflammatory cytokines and mediating cytotoxicity against bacterially infected cells (Gold et al., 2010; Dusseaux et al., 2011; Le Bourhis et al., 2013; Kurioka et al., 2015; Dias et al., 2017). Thus, MAIT cells have emerged as potentially crucial for antimicrobial defense (Le Bourhis et al., 2010; Georgel et al., 2011; Meierovics et al., 2013; Leung et al., 2014; Smith et al., 2014; Booth et al., 2015; Meierovics and Cowley, 2016; Chen et al., 2017). In addition to microbial products derived from vitamin B2 synthesis, other MR1-binding ligands have been identified, including the nonstimulatory folic acid derivative 6-formyl-pterin (6-FP; Kjer-Nielsen et al., 2012), and various activating and nonactivating drugs and drug-like molecules (Keller et al., 2017b), but the clinical relevance of these ligands is yet to be elucidated. Finally, MAIT cells can respond to a combination of cytokines, such as IL-12 and IL-18, in an MR1-independent fashion (Ussher et al., 2014; Slichter et al., 2016), further extending their potential participation in a wide array of inflammatory conditions (Loh et al., 2016; van Wilgenburg et al., 2016).

At birth, adaptive immunity is naive in the absence of in utero exposure to antigens. Maturation of the immune response occurs gradually after birth in response to antigenic stimulation from the environment (Adkins et al., 2004; Levy, 2007). In the absence of a fully developed adaptive immunity, newborns are heavily dependent on innate immunity for the control and prevention of infections during the first months of life (Kollmann et al., 2017). Preterm neonates suffer a high frequency and severity of microbial infections, many of them occurring spontaneously across epithelial barriers because of the immaturity of the immune system. Because MAIT cells represent a large pool of T cells able to rapidly respond to a wide range of microorganisms, they might be crucial for newborn immunity before the maturation of the specific and long-term memory adaptive immunity.

How and when human MAIT cells develop and differentiate after birth remains, however, little explored. MAIT cells represent only a very small fraction of cord blood T cells but, in contrast, are predominant in adult blood (Martin et al., 2009; Dusseaux et al., 2011; Walker et al., 2012), indicating that thymopoiesis is complemented by an important postnatal peripheral expansion. Using MR1:5-OP-RU tetramers, Koay et al. (2016) recently delineated a three-stage developmental pathway for mouse and human MAIT populations. Immature stage 1 and stage 2 MAIT cells (tetramer^pos Vα7.2⁺ CD161⁻ in humans) predominate in the thymus but represent minor subsets in periphery, where mature stage 3 MAIT cells (tetramer^pos Vα7.2⁺ CD161^high) are largely predominant. In mice, MAIT cell maturation requires the promyelocytic leukemia zinc finger (PLZF) transcription factor and commensal microbiota (Martin et al., 2009; Koay et al., 2016). However, studies in mice are not really contributive to understand the mechanisms driving postnatal MAIT development in the human, because of fundamental differences regarding the maturity of the immune system at birth and the frequency of MAIT cells (Cui et al., 2015; Rahimpour et al., 2015). Owing to the difficulty of performing longitudinal studies in infants, only a couple cross-sectional studies have been conducted on small cohorts, showing that MAIT frequency gradually increased with age in the peripheral blood (Dusseaux et al., 2011; Koay et al., 2016). However, comprehensive analysis across the ages is lacking. In addition, it can be discussed whether MAIT cells are defined as Vα7.2⁺ CD161^high cells selected by the restricting MR1 molecule expressed by CD4⁺CD8⁺ thymocytes leading to PLZF expression or only as T cells binding to the MR1:5-OP-RU tetramer (Keller et al., 2017a; Legoux et al., 2017).

Here, we analyzed the postnatal expansion and maturation of Vα7.2⁺ CD161^high CD4⁻ T cells in large series of newborns, according to gestational age and environmental factors. We also studied their recovery in children after unrelated cord blood transplantation, another clinical situation that heavily depends on innate immunity for the control and prevention of infections. In both clinical settings, we show that the extrathymic development of Vα7.2⁺ CD161^high T cells follows a stepwise program, with an early maturation step contrasting with a very slowly progressive expansion. This postnatal developmental scheme is not observed in the Vα7.2⁺ CD161^high MR1:5-OP-RU tetramer^neg and Vα7.2⁻ CD161^high populations, which are both much more abundant in the cord blood than Vα7.2⁺ CD161^high MR1:5-OP-RU tetramer^pos T cells. Strikingly, MR1:5-OP-RU tetramer^pos cells very rapidly predominate among Vα7.2⁺ CD161^high T cells after birth, becoming the vast majority by 6 mo of age. Molecular analysis of the TCR repertoire suggests that only a few clones found in cord blood naive Vα7.2⁺ CD161^high T cells expand after birth to reach high clonal size in adult blood, probably because of their unique ability to recognize MR1-restricted microbial antigens.

In humans, until MR1:5-OP-RU tetramers were available, MAIT cells were identified as T cells expressing the Vα7.2 TCRα segment and high level of CD161 among CD3⁺ γδ⁻ CD4⁻ cells. In adult blood, this population fully overlaps with the population labeled by MR1:5-OP-RU tetramers (Reantragoon et al., 2013). In cord blood however, the frequency of tetramer^pos Vα7.2⁺ CD161^high T cells is much lower than in adult blood (Koay et al., 2016), but the exact correspondence between Vα7.2⁺ CD161^high cells and MR1:5-OP-RU tetramer^pos cells remains unclear. As MR1 tetramers were not available when starting the present study, we analyzed the presence of the Vα7.2-Jα33 rearrangement (TRAV1-2 TRAJ33) by quantitative PCR (qPCR) in sorted Vα7.2⁺ CD161^high and Vα7.2⁺ CD161⁻ populations from cord blood and adult blood CD4⁻ T cells (Fig. 1 A). The Vα7.2-Jα33 rearrangement was 3.4 times less represented in the Vα7.2⁺ CD161^high fraction from cord blood compared with adult blood. Of note, this rearrangement was still overrepresented (3.6 times) in the cord blood Vα7.2⁺ CD161^high subset compared with the CD161⁻ one. Thus, as in adult blood, the Vα7.2⁺ CD161^high population from cord blood, but not the CD161⁻ one, encompasses T cells bearing the Vα7.2-Jα33 rearrangement characteristic of MAIT cells.

These results were confirmed and extended by using deep sequencing of the TRAV1-2 gene and quantification of the canonical MAIT TCR (TRAV1-2 TRAJ33/20/12 and 12–amino acid–long CDR3) in sorted Vα7.2⁺ CD161^high and Vα7.2⁺ CD161⁻ populations from cord blood and adult blood (Fig. 1 B, left). Although canonical MAIT TCRs were 5.7 times less frequent in the Vα7.2⁺ CD161^high fraction from cord blood compared with adult blood (mean ±16.6% and 95.7%, respectively), they were still identified at a much higher frequency than in the Vα7.2⁺ CD161⁻ population, both in cord blood and adult blood. Moreover, we found that the cord blood Vα7.2⁺ CD161^high population was not enriched for another recently described conserved Vα7.2⁺ T cell population, the CD1b-reactive germline-encoded mycolyl lipid-reactive (GEM) T cells characterized by a TRAV1-2 TRAJ9 and 13–amino acid–long CDR3 TCR (Van Rhijn et al., 2013; Fig. 1 B, right).

Altogether, these results indicate that the Vα7.2 and CD161 staining allows identification of circulating stage 3 MAIT cells at birth. However, the Vα7.2⁺ CD161^high fraction in cord blood may also encompass other T cells probably sharing a common developmental pathway (as discussed in Legoux et al. [2017]).

We determined the frequency and absolute numbers of peripheral blood Vα7.2⁺ CD161^high CD3⁺ CD4⁻ T cells in a cohort of 153 neonates (born between 24 and 42 wk postconception) admitted at birth in the Neonatology Department. In line with studies in cord blood, the frequency of Vα7.2⁺ CD161^high T cells in newborns was 30 times lower than in healthy adults (mean proportion among CD3⁺ cells 0.07% and 2.16%, respectively; Fig. 1, C and D). Of note, significant differences were observed when newborns were divided according to gestational age, with extremely preterm babies displaying higher Vα7.2⁺ CD161^high T cell frequencies than term babies (mean in group 1: 0.10%; group 2: 0.08%; group 3: 0.07%; group 4: 0.04%; P for trend = 0.0002; Fig. 1 E). This inverse relationship with gestational age over the 24-to 42-wk span was not observed for conventional αβ T cells (both CD4 and CD8; Fig. 1 E, bottom). The frequency of Vα7.2⁺ CD161^high T cells in neonates was not correlated to that of their mother on the day of delivery (Fig. 1 F).

CD161 expression in CD8 T cells is a marker of type-17 differentiation program and is associated with specific tissue homing properties (Billerbeck et al., 2010; Maggi et al., 2010). Although the CD161^high CD8 T cell population is dominated by Vα7.2⁺ MAIT cells in adult blood, the opposite has been observed in cord blood where Vα7.2⁻ cells are predominant (Walker et al., 2012). In neonates, Vα7.2⁻ CD161^high T cells were 10 times more abundant than Vα7.2⁺ CD161^high T cells, and their frequency was inversely correlated with gestational age (Fig. 1 E). Moreover, Vα7.2⁺ and Vα7.2⁻ CD161^high T cell frequencies were strongly correlated (r = 0.80, P < 0.0001), supporting the hypothesis that these two populations have a common prenatal developmental program (Walker et al., 2012; Legoux et al., 2017; Fig. 1 G). NKT cell frequencies were also negatively related to gestational age, but their correlation with Vα7.2⁺ and Vα7.2⁻ CD161^high T cell frequencies was much weaker (Fig. 1, E and G).

In 21 twin pairs the percentages of Vα7.2⁺ CD161^high T cells were highly correlated between monozygotic, and strikingly also dizygotic twins, at variance with the large variability observed among unrelated neonates. The same correlation was observed for Vα7.2⁻ CD161^high T cells and NKT cells (Fig. 1 H). This suggests that common prenatal environmental factors may control the development of innate-like T cell populations.

Collectively, our results indicate that fetal Vα7.2⁺ CD161^high T cells undergo an early wave of thymic development, in agreement with their presence in the thymus of aborted second-trimester fetuses (Leeansyah et al., 2014). Their decline with gestational age, also observed for Vα7.2⁻ CD161^high and NKT cells, may be related to their recruitment to mucosal tissues at the end of gestation, although there was no correlation between gestational age and Vα7.2⁺ CD161^high T cell frequency in the fetal tissues examined in a previous study (Leeansyah et al., 2014). More likely, this decreased frequency is due to the dilution of innate-like T cells by conventional T cells seeding the periphery, as suggested by the concomitant increase of conventional CD4 and CD8 T lymphocyte absolute numbers (Fig. S1 A).

The low frequency of MAIT cells at birth may be related to the absence of microbial-derived MR1-ligands because of the sheltered intrauterine life. The acute transition from the normally sterile intrauterine environment to the postbirth antigen-rich environment (85% of the bacteria colonizing the gut synthesize riboflavin; Mondot et al., 2016), should lead to an expansion of MAIT cells and acquisition of memory features in the first weeks of life. Because the number of infants still hospitalized after 1 mo of life was relatively small, we extended our study to 132 additional neonates with gestational age distribution similar to that of the first cohort. As Vα7.2⁺ CD161^high T cell frequencies in this replication cohort entirely confirmed those from the first one, the two cohorts were combined hereafter (Fig. S1 B). Vα7.2⁺ CD161^high T cell numbers only slightly increased during the first 2 mo of life, remaining much scarcer than in adult blood (Fig. 2 A). Vα7.2⁻ CD161^high T and NKT cells did not significantly increase over time, as expected because their numbers in adults are quite similar to those in neonates.

Upon delivery, neonates are rapidly exposed to and colonized by maternal microorganisms whose composition is determined by the type of delivery (vaginal or C-section; Dominguez-Bello et al., 2010). Because microbial colonization plays a major role in MAIT expansion in mice (Martin et al., 2009), we wondered if the mode of delivery had an impact on Vα7.2⁺ CD161^high T cell numbers 1 mo after birth (Fig. 2 B). There was no difference in Vα7.2⁺ CD161^high (as well as Vα7.2⁻ CD161^high and NKT) levels between vaginally and C-section–delivered infants, indicating that the type of microbiota colonizing the baby at birth does not influence early postnatal development of these T cells.

Nonmicrobial environmental factors, such as prolonged exposure to corticosteroid (CS) or vitamin D levels, were previously described as influencing MAIT cell frequencies in adults (Hinks et al., 2015, 2016). Because antenatal CS therapy is recommended in women at risk of imminent preterm birth, we compared Vα7.2⁺ CD161^high T cell frequencies in neonates and in their mothers between those receiving or not CS at least 48 h before delivery. There was no effect of such short-term CS treatment either in neonates or in their mothers (Fig. 2 C). Because serum vitamin D levels were not available, we sought to identify whether there was a seasonal pattern in Vα7.2⁺ CD161^high T cell frequencies, given the known seasonal variation in vitamin D levels. There was no association between Vα7.2⁺ CD161^high T cell frequencies and the month in which neonates were born, suggesting the lack of relation with sunlight exposure of the mother (Fig. 2 D).

In cord blood, MAIT cells are naive (CD45RA⁺/RO⁻) and express the CD8αβ heterodimer, whereas they are mostly memory and express the CD8αα homodimer in adults (Martin et al., 2009; Dusseaux et al., 2011; Walker et al., 2012; Koay et al., 2016). As expected, the vast majority of Vα7.2⁺ CD161^high T cells in neonates displayed a naive phenotype and expressed the CD8β chain, regardless of gestational age. Surprisingly, even though Vα7.2⁺ CD161^high T cells did not expand much in the early postnatal life, they very rapidly acquired a memory phenotype (Fig. 2 E). As early as 2 mo of age, almost 40% of Vα7.2⁺ CD161^high T cells were already CD45RO or CD8αα/αβ^low, in sharp contrast to conventional T cells, which were nearly all still naive (mean 5.2% CD45RO⁺; P < 0.0001).

Because we could not follow up the development of Vα7.2⁺ CD161^high T cells in neonates after their discharge (i.e., at a maximum of 2 mo of life), we performed a cross-sectional analysis in 79 healthy children aged 10 mo to 17 yr and 50 adults aged 18–50 yr. Vα7.2⁺ CD161^high T cell frequencies slowly increased during infancy to reach a plateau around 6 yr of age, extending previous studies in smaller cohorts (Dusseaux et al., 2011; Koay et al., 2016). In contrast, Vα7.2⁻ CD161^high T cell percentages remained stable over time (Fig. 2 F).

Collectively, these results indicate that maturation of Vα7.2⁺ CD161^high T cells rapidly occurs after birth, whereas their gradual expansion with age is a very slow process.

The progressive maturation of the neonatal immune system after birth shares some similarities with the progressive recovery of the immune cell subsets after allogeneic hematopoietic stem cell transplantation (HSCT) in children. These two clinical settings are characterized by a delayed development of the long-term protective adaptive immunity, associated with high predisposition to severe microbial infections. To study Vα7.2⁺ CD161^high T cell recovery after HSCT, we took advantage of a longitudinal study in 17 children who received unrelated cord blood (UCB) transplantation. Vα7.2⁺ CD161^high T cells were almost undetectable in the circulation 1 mo after transplantation, indicating that the host’s MAIT cells were not resistant to the myeloablative conditioning regimen given before HSCT, despite their expression of the multidrug efflux protein ABCB1 (Dusseaux et al., 2011). Moreover, Vα7.2⁺ CD161^high T cell numbers remained extremely low during the 12-mo study period, whereas Vα7.2⁻ CD161^high cells, conventional T cells, and B cells progressively increased to reach normal values 9–12 mo after transplantation (Fig. 3 A). Of note, 80% of circulating Vα7.2⁺ CD161^high T cells displayed a naive phenotype 1 mo after HSCT, supporting that they represented donor (cord blood)-derived cells transferred with the graft and not host’s residual cells. Thereafter, they rapidly acquired memory characteristics, as observed after birth. A rebound in naive Vα7.2⁺ CD161^high T cells occurred between 6 and 12 mo after HSCT (mean 42.5% vs. 75.6% CD45RO⁻; P = 0.06), suggesting the emergence of new HSC-derived naive cells recently exiting the thymus (Fig. 3 B). Importantly, the reconstitution profile of Vα7.2⁺ CD161^high T cells was not influenced by the occurrence of acute (<100 d after transplant) graft-versus-host disease (aGVHD) or severe microbial infection (Fig. 3 C).

To check whether the lack of recovery of circulating Vα7.2⁺ CD161^high T cells could be related to their attraction to inflammatory tissues early after HSCT, we performed immunohistochemical analysis on biopsy samples. In the absence of available MR1 tetramers for in situ staining, we relied on Vα7.2 staining (which also detects other Vα7.2⁺ conventional T cells). In control samples, few Vα7.2⁺ cells were observed in the normal lamina propria, mostly at the bases of the villi and more occasionally at the villus tip and within the epithelium, as reported (Reantragoon et al., 2013; Fig. 3 D). Vα7.2⁺ T cells were not found in intestinal lymphoid structures (mesenteric lymph nodes, appendix), except for rare cells in the mantle region of intestinal lymphoid follicles. We next analyzed intestinal biopsies taken for diagnostic purpose (aGVHD vs. viral infection occurring within 3 mo after transplantation). Intestinal aGVHD was assessed by the presence of apoptotic epithelial cells (determined by caspase 3 staining; not depicted). Although large CD3 and CD8 T cell infiltrates were evidenced, no Vα7.2⁺ lymphocytes were detected in all aGVHD samples examined, whatever the anatomical location (stomach, ileum, colon, rectum) or the intensity/severity of the immune process. Vα7.2⁺ cells were also undetectable in biopsies from HSCT recipients with documented viral infection or reactivation (CMV, adenovirus, HHV6; Fig. 3 E). Thus, the low recovery of circulating Vα7.2⁺ CD161^high T cells after HSCT is not related to their rapid trafficking to mucosal tissues in response to local inflammatory signals.

To extend our findings beyond 1 yr after transplant, we performed a cross-sectional analysis in 20 additional UCB transplant recipients sampled at the time of a routine hospital visit 1–6 yr after HSCT. Similar to what we observed during infancy, Vα7.2⁺ CD161^high T cell frequencies and absolute numbers progressively increased to reach values observed in healthy, age-matched controls around 5 yr after HSCT (Fig. 3 F).

Collectively, these results indicate that, although effective thymic output may ensure an export of some new naive Vα7.2⁺ CD161^high T cells after UCB transplantation, it does not contribute much to their peripheral expansion, which requires other factors. In mice, the presence of B cells was shown to be necessary for MAIT cell expansion (Martin et al., 2009). As the B cell compartment fully recovered as early as 3 mo after UCB transplantation, it is unlikely that the lack of B cells is responsible for the absence of MAIT expansion in the periphery in this setting.

Infection is particularly prevalent in early preterm births: chorioamnionitis is implicated in >85% of deliveries occurring before 28 wk of gestation, but the initiating infection may be subclinical (Rubens et al., 2014). We analyzed the potential relationship between microbial infection and peripheral Vα7.2⁺ CD161^high T cell numbers in 51 neonates with early-onset infection. Of note, when infection was microbiologically documented (85% of cases), it usually corresponded to known riboflavin-producing pathogens (Escherichia coli, Candida albicans, Pseudomonas aeruginosa, group B Streptococcus). Although relative frequencies of Vα7.2⁺ CD161^high T cells at birth were slightly lower in newborns with early-onset infection than in those without infection, their absolute numbers were not different (Fig. 4 A). Moreover, Vα7.2⁺ CD161^high T cell frequencies showed a similar inverse relationship with gestational age in uninfected and infected neonates (Fig. 4 B) and a similar increase during the first 2 mo of life (Fig. 4 C). There was no difference in the Vα7.2⁺ CD161^high T cell maturation profile (CD45RO or CD8β) according to the presence or absence of infection (not depicted).

Extremely preterm neonates in the intensive care unit (ICU) are at high risk of developing infections related to invasive aspects of intensive care. Adult patients with severe nonstreptococcal bacterial infections display an early decrease in MAIT cell count, and the persistence of MAIT depletion is associated with further development of ICU-acquired infections (Grimaldi et al., 2014). We thus analyzed Vα7.2⁺ CD161^high T cell numbers according to the presence or absence of infection acquired during the first 3 wk of life but found no difference (mean absolute number ± SEM at day 60: 3.1 ± 0.8/µl and 2.9 ± 1.5/µl, respectively; NS).

Collectively, these results indicate that neonatal infection is not associated with any changes in peripheral Vα7.2⁺ CD161^high T cell levels or phenotype. In particular, early-onset infection with riboflavin-producing pathogens is not sufficient to induce significant expansion of circulating Vα7.2⁺ CD161^high T cells in the postnatal period.

In fetal and adult tissues, a large proportion of T cells (either Vα7.2⁺ or Vα7.2⁻) express high levels of CD161 (Leeansyah et al., 2014; Fergusson et al., 2016), making the identification of MAIT cells difficult in tissues without RT-qPCR or MR1-tetramer staining. Still, the dynamic of Vα7.2⁺ CD161^high T cell accumulation in tissues after birth is not known. We analyzed the frequencies of Vα7.2⁺ and Vα7.2⁻ CD161^high T cells among mononuclear cells isolated from surgically resected intestinal tissues of different anatomical locations (ileum, colon, rectum) in 16 children aged 1 mo to 16 yr (Fig. 4 D). Whatever the child’s age was, Vα7.2⁺ CD161^high T cells were relatively rare within normal intestine (mean percentage of CD3 T cells ± SEM: 0.27 ± 0.06%; n = 8 patients), i.e., at a proportion similar to those reported in fetuses and adults (Leeansyah et al., 2014; Serriari et al., 2014; Sundström et al., 2015; Fergusson et al., 2016). They were barely detectable within appendixes (0.07 ± 0.03%; n = 5 patients). We also analyzed intestinal samples from three neonates with necrotizing enterocolitis, a devastating hyperinflammatory disease that mostly affects preterm newborns and likely results from inappropriate response to gut microbiota (Neu and Walker, 2011). Even in this inflammatory setting, frequencies of Vα7.2⁺ CD161^high T cells were very low. In all intestinal samples, Vα7.2⁻ CD161^high T cells were 10 times more abundant than Vα7.2⁺ CD161^high T cells (Fig. 4 D).

In contrast with thymic and cord blood cells, Vα7.2⁺ CD161^high T cells express a memory (CD45RO⁺) phenotype in fetal mucosae (Leeansyah et al., 2014). Similarly, in intestinal samples, most Vα7.2⁺ CD161^high T cells expressed CD45RO (mean 91% CD45RO⁺) and high levels of CD69, a marker of tissue-resident memory T cells (Sathaliyawala et al., 2013; Thome et al., 2014; 70–100% being CD69⁺; Fig. 4 E). Conversely, only a fraction of Vα7.2⁻ CD161^high and conventional T cells expressed a memory phenotype, in agreement with recent data in mucosal tissues from pediatric organ donors (Thome et al., 2016).

The very low expansion of Vα7.2⁺ CD161^high T cells after birth or cord blood transplantation may be related to cell intrinsic characteristics or to limited availability of microbial-derived MR1-ligands. We first examined if the expansion of peripheral neonatal Vα7.2⁺ CD161^high T cells was restricted by selective cytokine responsiveness. We found similar prominent expression of IL-18Rα and IL-7Rα in cord blood and adult Vα7.2⁺ CD161^high T cells. Expression of IL-12Rβ2 and IL-23R was much lower but again comparable in cord blood and adult cells (Fig. 5 A). This suggests that the low expansion of naive Vα7.2⁺ CD161^high T cells after birth or after HSCT is not related to hyporesponsiveness to innate or homeostatic cytokines.

The PLZF transcription factor is considered a master regulator of innate-like T cell development and maturation, including NKT and γδ T cells (Kovalovsky et al., 2008; Savage et al., 2008). A gradual acquisition of PLZF expression occurs between stage 2 and stage 3 MAIT cells in the thymus (Koay et al., 2016) and continues during postthymic maturation of MAIT cells in peripheral blood (Cui et al., 2015; Rahimpour et al., 2015). We found that naive cord blood Vα7.2⁺ CD161^high T cells expressed significantly lower levels of PLZF than adult ones (Fig. 5 B), suggesting that final maturation of cord blood Vα7.2⁺ CD161^high T cells requires an early activation signal after birth. Notably, a significant proportion of naive cord blood Vα7.2⁺ CD161^high T cells expressed measurable level of CD25 (Fig. 5 B), and PLZF levels were significantly higher in CD25-positive cells than in CD25-negative cells (Fig. 5 C). This may be related to a low level of stimulation by microbial antigen already available at birth, as also suggested by the higher proportion of cycling cells in cord blood Vα7.2⁺ CD161^high T cells in comparison with adult ones (mean Ki67 expression: 4% vs. 0.65%; Fig. 5 D).

Adult MAIT cells are hypoproliferative in vitro compared with other T cell subsets (Dusseaux et al., 2011; Turtle et al., 2011). To assess the intrinsic proliferative capacity of cord blood Vα7.2⁺ CD161^high T cells, we analyzed their in vitro proliferation in response to mitogen. As indicated by the high proportion of CFSE^low cells, cord blood Vα7.2⁺ CD161^high T cells strongly proliferated after 6-d PHA stimulation, similarly to conventional CD8 T cells. Adult Vα7.2⁺ CD161^high T cells proliferated much less efficiently (Fig. 5 E).

Finally, we investigated the ability of cord blood Vα7.2⁺ CD161^high T cells to respond to short-term stimulation with microbial MR1 ligands. Magnetic bead-enriched CD4⁻ T cells were incubated overnight in the presence of MR1-expressing THP1 cells loaded with paraformaldehyde-fixed E. coli, and IFNγ and Granzyme B (GrB) production was assessed in Vα7.2⁺ CD161^hi cells. E. coli clearly induced intracellular accumulation of IFNγ and GrB in adult Vα7.2⁺ CD161^high T cells, but not in cord blood cells (Fig. 5 F), in line with their naive phenotype. Longer E. coli stimulation (up to 5 d) induced a high mortality of cord blood Vα7.2⁺ CD161^high T cells. Overnight stimulation by IL-12 and IL-18 combination was also unable to stimulate cord blood cells (not depicted). Thus, in contrast to mature MAIT cells in adult blood, cord blood Vα7.2⁺ CD161^high T cells are not able to display immediate effector functions toward bacterially infected cells. These data indicate that, although they have intrinsic ability to proliferate, cord blood Vα7.2⁺ CD161^high T cells need functional maturation and/or expansion after birth to acquire detectable effector activities after microbe-derived antigen recognition.

As the MR1:5-OP-RU tetramer became available during the course of this study (National Institutes of Health [NIH] tetramer core facility), we determined how the Vα7.2⁺ CD161^high and MR1:5-OP-RU tetramer^pos/neg populations overlap. More than 92% of Vα7.2⁺ CD161^high cells were stained with the MR1:5-OP-RU tetramer in adult blood, in contrast with only 2–15% in cord blood (Fig. 6 A), in line with previous results (Eckle et al., 2014; Koay et al., 2016).

We further characterized the α chain of these populations by deep sequencing and found that the Vα7.2⁺ CD161^high tetramer^pos subset in cord blood was highly enriched in the canonical MAIT TCR α chain compared with the tetramer^neg subset (Fig. 6 B). Yet, the canonical MAIT TCR α chains were more frequent in the Vα7.2⁺ CD161^high tetramer^neg than in the Vα7.2⁺ CD161⁻ population. These results fully support our initial analysis of the TCR α chain in cord blood subsets (Fig. 1 A). Altogether, they suggest that the CD161^high phenotype in cord blood Vα7.2⁺ T cells delineates a subset of cells enriched in the canonical MAIT TCR α chain, with only a part reactive to 5-OP-RU presented by MR1, certainly according to the fine specificity imparted by the TCR β chain. It is likely that these few clones then expand to form the adult MAIT pool because almost all Vα7.2⁺ CD161^high cells in adult blood express the canonical MAIT TCR and are stained by the MR1:5-OP-RU tetramer.

We therefore studied peripheral blood samples from 83 healthy infants aged from 1 d to 12 mo and found that the proportion of MR1:5-OP-RU tetramer^pos cells increased very rapidly, representing already more than 50% of Vα7.2⁺ CD161^high cells at 1 mo of life and the vast majority at 6 mo (Fig. 6 C). Conversely, MR1:5-OP-RU tetramer^neg cells tended to disappear (Fig. S2). Moreover at 1 mo, more than 80% of tetramer^pos cells already expressed a mature phenotype, whereas maturation of Vα7.2⁺ CD161^high tetramer^neg cells was delayed (20% at 1 mo, 80% at 7 mo). Reduction in CD8β expression was concomitant to the acquisition of CD45RO. In addition, from 3–4 wk of age, an important proportion of Vα7.2⁺ CD161^high tetramer^pos (but not tetramer^neg) cells expressed CD69, suggesting a recent activation in vivo (Fig. 6 D).

Altogether, these results indicate that expansion and maturation of the tetramer^pos cells within the Vα7.2⁺ CD161^high population occurs in the first days after birth, suggesting the presence of some 5-OP-RU ligand providing an early maturation signal. This signal is likely not quantitatively sufficient to drive significant expansion of peripheral blood MAIT cells, which remain in very low amounts at this stage.

To verify our hypothesis that the fine specificity of the TCR conferred by the β chain must be important for 5-OP-RU recognition and cell expansion, we studied TCR β chain repertoire of adult and cord blood subsets. Adult MAIT cells display oligoclonal expansions with high clonal size and biased TRBV usage (Tilloy et al., 1999; Lepore et al., 2014). Although cord blood MAIT cells display a diverse repertoire as studied with anti-Vβ antibodies (Walker et al., 2012), no extensive molecular characterization of their TRBV repertoire has been performed so far. Using 5′ rapid amplification of cDNA ends (RACE) technique followed by deep sequencing, we characterized the repertoire of FACS-sorted cord blood Vα7.2⁺ CD161^high T cells in comparison to other cord blood CD8 T cell subsets (Vα7.2⁻ CD161^high, Vα7.2⁺ CD161⁻, Vα7.2⁻ CD161⁻) and to the same subsets in adult blood. Importantly, memory (CD45RO⁺) subsets were sorted in adult blood, to allow a meaningful comparison with adult MAIT cells, which are all memory. The rarefaction curves of different subsets, which represent cumulative frequency of individual clonotypes ranked by decreasing frequency, were compared (Fig. 7 A). For cord blood subsets, the curves were straight and overlapping, indicating similar high diversity between the subsets. From cell dilution experiments, we estimated the 5′RACE template exchange efficacy during the reverse transcription step at ∼1–2% (not depicted). Because only 3,000 cells from cord blood were studied, the 30–60 clones making up 80–90% of the reads indicate that all the T cell expressed a different TCR. Identical results were obtained for four cord blood donors (Fig. S3). In contrast, although 10 times as many cells (35,000) were studied, adult MAIT cells encompassed oligoclonal expansions with less than 10 clones making up to 80% of the reads. Albeit less strikingly than in the donor depicted in Fig. 7 A, adult MAIT cells harbored similar oligoclonal expansions in the 13 donors studied (Fig. S3), in line with previous data (Tilloy et al., 1999; Lepore et al., 2014). In many cases, the oligoclonality of MAIT cells was similar to that of the other subsets, certainly because only memory (CD45RO⁺) CD8⁺ T cells were included in our sorting, contrary to a previous study (Lepore et al., 2014), and memory CD8 T cells are known to include oligoclonal expansions (Hingorani et al., 1993).

We then assessed whether cord blood and adult blood Vα7.2⁺ CD161^high T cells displayed similar biases in the TCR β chain repertoire. TRBV and TRBJ biases were conserved between donors, excluding a stochastic process. We observed preferential usage of TRBV 4.3, 6.4, 15, and 20.1 and TRBJ 2.6 and lower usage of TRBV 4.1 and 5.1 and TRBJ 1.2 in adult MAIT cells in comparison with both other memory adult CD8 T cell subsets (not depicted) and cord blood Vα7.2⁺ CD161^high (Fig. S4, summarized in Fig. 7 B, left panel for TRBV, right panel for TRBJ). Importantly, no such TRBV or TRBJ usage bias was observed for cord blood Vα7.2⁺ CD161^high in comparison with other cord blood CD8 subsets (Fig. S5), suggesting that the TRBV bias observed in adult MAIT cells is not related to a pairing problem. However, when comparing cord blood Vα7.2⁺ CD161^high MR1:5-OP-RU tetramer^pos and tetramer^neg subsets (Fig. 7 C), similar to adult MAIT cells, we observed that cord blood MR1:5-OP-RU tetramer^pos population tended to express less TRBV4.1, TRBV5.1, and TRBJ1.2 as well as more TRBV6.4 compared with the MR1:5-OP-RU tetramer^neg population. No similar biases between cord blood MR1:5-OP-RU tetramer^pos and adult MAIT cells were observed for the other segments, but this is probably related to a lack of experimental power because of the very low number of cells studied for the tetramer-positive subset (50 cells). Altogether, these data indicate that the fine specificity of the TCRβ chain is involved in MR1:5-OP-RU complex recognition in accordance with structural data (Eckle et al., 2014).

Here, by combining flow cytometry, deep sequencing, and further MR1:5-OP-RU tetramer staining, we decipher the postnatal development of human MAIT cells and other Vα7.2⁺ and Vα7.2⁻ CD161^high subsets. We longitudinally characterize the maturation and expansion of these populations in infancy and after cord blood transplantation, two clinical settings sharing some similarities for establishment of mature protective immunity. The remarkably superimposable results observed in these two situations allow us to provide a new view of MAIT cell ontogeny in the human.

We characterize three distinct human CD3⁺ CD4⁻ CD161^high populations present at birth: (1) a subset of Vα7.2⁺ cells enriched in the canonical MAIT TCR α chain, among which a fraction is reactive to 5-OP-RU presented by MR1 (possibly as a result of the β chain usage), is the naive counterpart of mature MAIT cells found in adult blood and corresponds to the recently described stage 3 MAIT cells that just exit the thymus (Koay et al., 2016); (2) a subset of Vα7.2⁺ cells not reactive to 5-OP-RU presented by MR1, which predominate at birth but are rapidly outnumbered by MR1:5-OP-RU reactive cells; (3) an abundant subset of Vα7.2⁻ cells, the proportion of which remain stable over time. These three populations are already present at midgestation as observed in the cord blood of high-preterm neonates, in line with their presence in the thymus of aborted second-trimester fetuses (Leeansyah et al., 2014). These Vα7.2⁺ and Vα7.2⁻ CD161^high populations likely share a common prenatal developmental program, as suggested by the strong correlation we observe between their frequencies at birth. Moreover, their percentages are highly similar in both monozygotic or dizygotic twin pairs, suggesting that prenatal environmental factors may also control their development. Although it is highly probable that CD161^high cells expressing the canonical MAIT TCR are selected on MR1 expressed by double-positive thymocytes, the cell type and molecule selecting CD161^high cells that do not express the canonical MAIT TCR remain elusive.

At birth, Vα7.2⁺ CD161^high cells exhibit a naive phenotype and intermediate PLZF levels. Accordingly, they are unable to rapidly produce cytokines or cytotoxic molecules in response to bacterial ligands, in contrast with mature adult MAIT cells. Neither do they respond to stimulation by exogenous IL-12 and IL-18, despite high expression of the receptors for these cytokines. Strikingly, although they represent a minor proportion of Vα7.2⁺ CD161^high cells at birth, MR1:5-OP-RU tetramer^pos cells very rapidly predominate over tetramer^neg cells. By the age of 6 mo, almost all Vα7.2⁺ CD161^high T cells are MR1:5-OP-RU-reactive MAIT cells. Therefore, the term MAIT cells can be confidently assigned to Vα7.2⁺ CD161^high CD4⁻ T cells from the age of 6 mo after birth (or after HSCT), a crucial point for previous or further human studies conducted in the absence of MR1 tetramer identification.

Importantly, this complete shift in the proportion of tetramer^pos MAIT cells within the Vα7.2⁺ CD161^high pool is concomitant with their activation (CD69⁺) and maturation (CD45RO⁺). It is likely that colonization by the commensal microbiota provides a key early activation signal, as supported by the impaired development and maturation of MAIT cells in germ-free mice compared with specific pathogen–free mice (Martin et al., 2009; Koay et al., 2016). Whether MAIT cell maturation in the early postnatal period is correlated with the gut microbiota diversity or function (availability of riboflavin metabolites) is currently under investigation in our laboratory. However, although microbe-mediated expansion of peripheral MAIT cells was demonstrated in mouse infection models (Meierovics et al., 2013; Chen et al., 2017), early- or late-onset neonatal microbial infection is not sufficient to induce a significant expansion in circulating MAIT cells. Indeed, MAIT cells very slowly populate the periphery during infancy, taking ∼5–6 yr to achieve adult levels. This gradual process, observed both in childhood and after cord blood transplantation, is not related to a MAIT-cell intrinsic proliferative defect. Rather, it may reflect specific requirements to allow for continuous proliferation, such as repeated exposure to pathogenic bacteria that provide more than single TCR/MR1:ligand interactions. In mice, administration of synthetic 5-OP-RU alone causes CD69 up-regulation but does not result in MAIT proliferation, whereas the addition of Toll-like receptor agonists causes high levels of activation and proliferation of the MAIT cell pool (Chen et al., 2017). It is possible that some pathogens provide higher amounts of MAIT-activating ligands or an inflammatory/cytokine context allowing larger cell expansion than other pathogens, as shown by the tremendous MAIT expansion in mice during Francisella tularensis infection (Meierovics et al., 2013). Several lines of evidence indicate that MAIT cells can adapt their proliferative and effector responses depending on the microbial and inflammatory signals present (López-Sagaseta et al., 2013; Gold et al., 2014; Slichter et al., 2016; Dias et al., 2017).

Our repertoire analysis of Vα7.2⁺ CD161^high cord blood and adult subsets indicates that the fine specificity of the TCR conferred by the β chain must be important for 5-OP-RU recognition and expansion of Vα7.2⁺ CD161^high T cells expressing the canonical MAIT TCR α chain. These results are consistent with structural data (Eckle et al., 2014). Only the few MR1:5-OP-RU reactive Vα7.2⁺ CD161^high T cells that display a TRAV and TRBV repertoire very similar to adult MAIT cells expand in periphery to form the adult MAIT pool, diluting both the Vα7.2⁻ CD161^hi and Vα7.2⁺ CD161^hi MR1:5-OP-RU tetramer^neg T cells. Collectively, these results suggest that expansion of these few clones already present at birth is driven by repeated interactions with riboflavin-expressing pathogens encountered during the first years of life. Notably, MAIT cell levels exhibit very large interindividual variability (over one log range) in child and adult blood. It is tempting to speculate that microbial infection history contributes to the variable extent of MAIT cell expansion. Only a careful longitudinal analysis of MAIT levels in several populations of various ages in various environments will answer this question.

Although accurate identification of MAIT cells in tissues is limited in the absence of tetramer staining, Vα7.2⁺ CD161^high cells have been found in fetal tissues (Leeansyah et al., 2014), and we detected them in the intestine of 1-mo-old babies. Of note, these cells are tissue-resident memory cells, unable to recirculate, as indicated by the entirely naive phenotype of circulating Vα7.2⁺ CD161^high T cells at birth. Even in the absence of live microbes in the placenta, maternal gestational microbes might be a source of microbial-derived metabolites reaching fetal tissues, as recently demonstrated in a model of transient microbial colonization in pregnant mice (Gomez de Agüero et al., 2016). However, in contrast with their gradual expansion in peripheral blood after birth or cord blood transplantation, we found that Vα7.2⁺ CD161^high cells did not accumulate much in intestinal tissues during childhood. The lack of any detectable Vα7.2⁺ CD161^high cells in the intestine of HSCT recipients, in particular during the course of aGVHD, is quite surprising. Mucosal disruption, microbial translocation, and inflammation contribute to aGVHD pathogenesis and should allow at least a few graft-derived Vα7.2⁺ CD161^high cells to rapidly home to mucosal sites. MAIT cells are highly sensitive to apoptotic stimuli (Gérart et al., 2013). One hypothesis is that microbial translocation triggers activation-induced apoptosis of newly generated MAIT cells within mucosal tissues or that posttransplant CS therapy exacerbates apoptotic sensitivity and prevents MAIT survival in tissues.

In addition to providing a more complete view of the physiological MAIT cell development, our results have important implications for appreciating the impact of MAIT cell loss in other situations. Thus, in line with our observation after cord blood transplantation, MAIT cells remained undetectable up to 2 yr after autologous HSCT in patients with multiple sclerosis (Abrahamsson et al., 2013). MAIT loss was correlated with good clinical response, supporting MAIT involvement in the disease pathology. Whether the benefit is still maintained at a time MAIT cells recover, i.e., 5–6 yr after transplantation, will deserve investigation. Our data may also explain the lack of circulating MAIT recovery in otherwise successfully treated HIV-infected patients (Cosgrove et al., 2013; Leeansyah et al., 2013). Rather than evoking an irreversible MAIT loss in a context of HIV infection, it is more likely that complete MAIT recovery requires a minimum of 5–6 yr after initiation of antiretroviral treatment, as indicated by our preliminary results on long-term treated HIV patients (M. Lambert and S. Caillat-Zucman, personal data).

Finally, our results indicate that the early expansion of MAIT cells after birth dilutes Vα7.2⁺ CD161^high cells not reactive to 5-OP-RU presented by MR1. Part of these MR1:5-OP-RU tetramer^neg express the canonical MAIT TCR α chain and are thus likely selected on MR1 in the thymus. Because a formal demonstration of their MR1 restriction is not feasible in humans, single-cell TCR sequencing followed by TCR reexpression would be necessary to formally characterize the MR1:5-OP-RU tetramer^neg cells that express the canonical TCR α chain. Whether these cells recognize empty MR1 or other MR1 ligand present during fetal life and around birth remains unknown. Nonstimulatory/inhibitory ligands, such as 6F-P and other pterin-based small drug-derived molecules, lead to high levels of MR1 surface expression and can inhibit MAIT cell activation by 5-OP-RU (Keller et al., 2017b). Such MR1 ligands might outcompete microbial ligands until those become quantitatively or qualitatively predominant after birth.

In conclusion, our in-depth analysis of the kinetics of MAIT cell development and its correlation with microbial or nonmicrobial environmental exposures highlights important defects in the newborn, as demonstrated for many other cellular and molecular components of innate immune responses (Kollmann et al., 2017). Whether this may contribute to the increased susceptibility of preterm neonates to severe microbial infections remains an open issue because selective MAIT cell deficiency has not been reported so far in humans. It is tempting to propose that neonatal and even prenatal (i.e., maternal) manipulation of MAIT cell numbers or functions by targeted interventions, such as exposure to probiotics or regulation of metabolic pathways, might optimize newborn defense toward severe microbial infections. The recent success of a synbiotic trial associating Lactobacillus plantarum to fructooligosaccharides to prevent sepsis in rural Indian newborns paves the way for such investigation (Panigrahi et al., 2017).

We thank all the patients and their physicians and the nurse and technician staff from Hôpital Robert Debré who helped with this study. We also thank Adam Woolfe for preliminary analysis of the sequencing data following multiplex PCR.

This work was supported by grants from Agence Nationale de la Recherche (ANR) and Direction Générale de l’Offre de Soins (DGOS; PRTS 14-CE15-0005, acronym NEOMAIT), Institut National du Cancer (INCa; 2012-059), Agence de Biomédecine, Cent pour Sang la vie, IRGHET, and Fondation pour la Recherche Médicale (FRM; ING20150532367). High-throughput sequencing was performed by the ICGex NGS platform of the Institut Curie supported by Agence Nationale de la Recherche grants (ANR-10-EQPX-03, Equipex; and ANR-10-INBS-09-08, France Génomique Consortium) from the “Investissements d’Avenir” program, the Canceropole Ile-de-France, and the SiRIC-Curie program (grant INCa-DGOS-4654).

The authors declare no competing financial interests.

Author contributions: G. Ben Youssef, M. Tourret, M. Salou, L. Ghazarian, S. Mondot, Y. Mburu, M. Lambert, S. Azarnoush, J.-S. Diana, and A.-L. Virlouvet designed research studies, conducted experiments, analyzed data, and/or provided patient samples. V. Houdouin and M. Peuchmaur analyzed data. J.-H. Dalle, O. Lantz, and V. Biran designed research studies, analyzed data, and wrote the manuscript. S. Caillat-Zucman designed research studies, supervised the study, analyzed data, and wrote the manuscript with the help of other coauthors.