Research ArticleHuman Immunology

CD4 T Cells with Effector Memory Phenotype and Function Develop in the Sterile Environment of the Fetus

See allHide authors and affiliations

Science Translational Medicine  28 May 2014:
Vol. 6, Issue 238, pp. 238ra72
DOI: 10.1126/scitranslmed.3008748


The T cell compartment is considered to be naïve and dedicated to the development of tolerance during fetal development. We have identified and characterized a population of fetally developed CD4 T cells with an effector memory phenotype (TEM), which are present in cord blood. This population is polyclonal and has phenotypic features similar to those of conventional adult memory T cells, such as CD45RO expression. These cells express low levels of CD25 but are distinct from regulatory T cells because they lack Foxp3 expression. After T cell receptor activation, neonatal TEM cells readily produced tumor necrosis factor–α (TNF-α) and granulocyte-macrophage colony-stimulating factor (GM-CSF). We also detected interferon-γ (IFN-γ)–producing T helper 1 (TH1) cells and interleukin-4 (IL-4)/IL-13–producing TH2-like cells, but not IL-17–producing cells. We used chemokine receptor expression patterns to divide this TEM population into different subsets and identified distinct transcriptional programs using whole-genome microarray analysis. IFN-γ was found in CXCR3+ TEM cells, whereas IL-4 was found in both CXCR3+ TEM cells and CCR4+ TEM cells. CCR6+ TEM cells displayed a genetic signature that corresponded to TH17 cells but failed to produce IL-17A. However, the TH17 function of TEM cells was observed in the presence of IL-1β and IL-23. In summary, in the absence of reported pathology or any major infectious history, T cells with a memory-like phenotype develop in an environment thought to be sterile during fetal development and display a large variety of inflammatory effector functions associated with CD4 TH cells at birth.


Human fetal T cells first appear in the thymus at about 10 weeks of gestation, and their functions and diversity are apparent at about 20 weeks of gestation (1). At this time, fetal T cells begin to colonize the peripheral blood and the secondary lymphoid organs with clearly defined T cell zones. Terminal deoxynucleotidyl transferase contributes to T cell receptor (TCR) diversity and is first expressed at 20 weeks of gestation. Thus, the fetal peripheral CD4 T cell compartment is composed of a naïve T cell (TN) repertoire that is ready to respond to foreign antigens. However, one critical parameter for pregnancy is fetomaternal tolerance, in which Foxp3 regulatory CD4 T (Treg) cells play an important role in both the mother (2) and the fetus (3). Therefore, mature naïve fetal non-Treg cells remain under the control of antigen-specific Treg cells to avoid responses to noninherited maternal antigens (NIMAs) (4, 5). The fetal T cell compartment is thus considered devoid of any memory T (TEM) cells, whereas TEM cells constitute half of the circulating T cells present in adult peripheral blood. Immediately after birth, the safe and presumed sterile environment is abandoned, and neonatal CD4 T cells play a critical role in protection against pathogens; thus, these cells represent a major target for vaccines that could elicit cellular responses and generate antibodies.

CD4 T helper (TH) cells develop along various differentiation pathways and acquire particular functions under the control of the master regulators Tbx21, Gata3, and Rorc, which leads to the development of TH1, TH2, and TH17 effector cells, respectively (6). These functions are mainly mediated by secreted cytokines, primarily interferon-γ (IFN-γ) for TH1, interleukin-4 (IL-4)/IL-13 for TH2, and IL-17 for TH17. A certain degree of plasticity and a large degree of polyfunctionality exist among differentiated CD4 T cells, and both play an important role in immunity (7, 8). Thus, other cytokines can be associated with one of the primary subsets of effector cells described above, or they may, in some cases, generate a specific T cell effector response. This is particularly true for IL-9–producing T cells, which are related to TH2 cells (9), and IL-22–secreting T cells, which are often associated with TH17 cells; however, these cells may also constitute a distinct T cell fraction (10, 11). In addition, T follicular helper (TFH) cells play a critical role in the development of germinal center B cells (12). TH1 cells secreting IFN-γ are critical for the elimination of intracellular pathogens and have been shown to be particularly challenging to induce during early infancy (13, 14). Among these different parameters, the age-dependent maturation of the capacity to produce IL-12 in innate cells appears to contribute to TH1 impairment during early life (15, 16). This has been particularly well demonstrated with cord blood monocyte-derived dendritic cells (DCs), which failed to produce bioactive IL-12 because of epigenetic control of IL-12p35 expression (17). Neonatal plasmacytoid DCs produce type 1 IFN in response to viruses (18) and can thus promote TH1 responses in some circumstances (19). Nevertheless, newborns and very young infants develop a preferential TH2 response to infections and vaccines (13). TH17 cells are often thought of as pathogenic T cells; however, their role in early infancy remains unclear, although their potential contribution to the elimination of extracellular bacteria and fungi renders them important players in immunity (20).

Here, we have identified a population of CD4 T cells in neonatal blood that exhibit an effector memory phenotype and inflammatory properties. We investigated these properties in terms of TH1, TH2, and TH17 cellular function and have thus revisited the issue of developing T cell functions that arise during fetal development, which do not seem to be limited to tolerance.


Identification of the TEM phenotype in the neonatal CD4 compartment

Expression of IL-2Ra and IL-7Ra in CD4 T cells has been established as a potent means to discriminate between TN cells (CD25CD127hi) and Treg cells (CD25hiCD127lo) in human blood (21). Activated CD4 T cells also express CD25, but it occurs at much lower levels in these cells than in Treg cells, and this phenotype can be found in both adult and cord blood (21). Figure 1A shows that under nonpathological conditions, analysis of neonatal CD4 T cells with CD25 and CD127 antibodies can identify TN and Treg cells in addition to a small population of putatively activated T cells (CD25loCD127hi). Because these cells were leaving a tolerant and presumed sterile fetal environment, the presence of activated T cells in neonatal blood was unexpected. We further analyzed this population for the presence of the transcription factor Foxp3, the activation marker CD69, the proliferation marker Ki67, and cell maturation–associated molecules (CD45RO, CCR7, and CD62L) (Fig. 1B). First, a lack of Foxp3 expression distinguishes the CD25loCD127hi neonatal CD4 T cell population from Treg cells. Second, in clear contrast to cord blood TN cells, these cells were activated and underwent proliferation, as confirmed by positive CD69 and Ki67 staining. The expression of these molecules might reflect a continuous flow during which a few cells pass through a transitory activated state before returning to a resting state. In comparison to TN cells, the expression of CD45RO was increased, the expression of CCR7 was decreased, and the shedding of CD62L was detected in most cells comprising the neonatal population of CD25loCD127hi CD4 T cells. Thus, these cells were classified as exhibiting the T cell memory phenotype, which has been previously described in adult blood (22, 23). However, CD45RO staining did not accurately define a population within neonatal T cells, and its expression was slightly lower than in adult memory T cells (fig. S1). Neonatal CD45RO+CD4+ T cells poorly recapitulated the CD25loCD127hi gating on CD4 T cells because most of the cells were CD25-negative (fig. S1). To confirm that these T cells originated from the baby and not from the mother, we performed fluorescence in situ hybridization (FISH) to detect X and Y chromosomes, allowing discrimination of male and female cells (fig. S2). Among the CD25loCD127hi CD4 T cells isolated from male baby cord blood (Fig. 1C), we failed to detect maternal T cells (XX), and therefore, these cells are henceforth referred to as neonatal TEM cells. We repeatedly found neonatal TEM cells in all neonatal blood samples (n = 29), and these cells accounted for 1 to 6% of total neonatal CD4 T cells (Fig. 1D). TEM cells were not recent thymic emigrants, as shown by the lack of CD31 (24) and PTK7 (25) expression (Fig. 1E). To further characterize TEM cells, we compared the TCR repertoire of the different neonatal T cell subsets. Despite the fact that TEM cells expressed CD161 (Fig. 1E), a natural killer T cell marker, only less than 1% of these cells were positive for GalCer/CD1d tetramer staining (Fig. 2A). TCRγδ T cells were marginally found; therefore, 99% of neonatal TEM cells were conventional TCRαβ T cells (Fig. 2A). Expression analysis of TCRV genes covering 70% of the repertoire (26, 27) showed a wide expression of all 10 Vβ genes in TN, TEM, and Treg populations, defining a polyclonal repertoire (Fig. 2, B to D). This was confirmed by the large distribution of CDR3 lengths for four TCRV genes using the Immunoscope technology (Fig. 2, B to D).

Fig. 1. Cord blood contains a population of CD4 TEM cells.

(A and B) Cord blood CD3+CD4+ cells were analyzed for the presence of CD25 and CD127 (A) and for CD45RO, CCR7, CD62L, CD69, Foxp3, and Ki67 (B) with the following gating strategies: CD127hiCD25 (filled gray), CD127loCD25hi (dashed line), and CD127hiCD25lo (bold line). For CD69 staining, the inset corresponds to control immunoglobulin (Ig) staining. (C) X and Y chromosome FISH staining of CD127hiCD25loCD3+CD4+ cells from a male baby. Bar chart indicates the number of nuclei with or without XY staining in a cell preparation. Undefined (Undef.) are nuclei not stained for two chromosomes. (D) Analysis of cord blood (n = 29) for CD4 T cell frequency, grouped as TN cells (CD127hiCD25), Treg cells (CD127loCD25hi), and TEM cells (CD127hiCD25lo). Mean frequency ± SD: TEM = 2.95 ± 1.5%, Treg = 5.3 ± 1.5%, and TN = 91 ± 2%. (E) Cord blood CD3+CD4+ cells were analyzed as in (A) and (B) for the presence of CD161, CD31, and PTK7.

Fig. 2. Neonatal TEM cells are TCRαβ T cells and display a polyclonal TCR repertoire.

(A) Neonatal CD3+ T cells were analyzed for TCRαβ and TCRγδ frequencies and for CD4/CD25/CD127 expression. TEM cells gated as CD4+CD25loCD127hi were also analyzed for αGalCer/CD1d tetramer staining as indicated. (B to D) TN, Treg, and TEM cells were FACS-purified from neonatal blood and subjected to TCRBVb repertoire analysis. The frequencies of different TCRV usage were analyzed among 10 V subfamilies (BV2, BV3, BV4, BV5, BV6a, BV7b, BV19, BV20, BV28, and BV29) and represented with specific colors. The distribution of CDR3 lengths was analyzed for the different TCRBV2, BV4, BV7b, and BV20.

Inflammatory T cells revealed by functional profiling of neonatal TEM

We next sought to determine the functional properties of neonatal TEM cells by assessing TH1, TH2, and TH17 cytokine profiles. We isolated TEM, TN, and Treg cells by fluorescence-activated cell sorting (FACS) (the gating strategy is shown in fig. S1), and we assessed cytokine production after activation. Eight hours after phorbol 12-myristate 13-acetate (PMA)/ionomycin activation, we consistently observed the production of tumor necrosis factor–α (TNF-α) and IFN-γ, but not other cytokines, in both adult and neonatal TEM cells (Fig. 3, A to C, and fig. S3A). After TCR activation with anti-CD3/CD28 antibodies, we observed the production of IFN-γ, very little IL-13, and no IL-17A at 20 hours (Fig. 3C). IFN-γ and IL-13 were produced by TEM cells, but not by TN or Treg cells. In contrast to neonatal cells, adult TEM produced IL-17 (fig. S3B). At 72 hours after TCR activation, we also performed intracellular cytokine staining to observe the production of IFN-γ and IL-13 by neonatal TEM cells (Fig. 3D). IL-17A production could not be detected in any of these experiments. Next, we extended our analysis to a larger panel of cytokines, including TNF-α, GM-CSF (granulocyte-macrophage colony-stimulating factor), IL-4, IL-5, IL-9, IL-17F, IL-22, and IL-21. To ensure accurate detection of cytokines, we chose a late time point, allowing sufficient production of all cytokines. After 72 hours of TCR-induced signaling, we detected TNF-α, IFN-γ, and GM-CSF associated with TH1 cells in the culture supernatant of neonatal TEM cells. TH2 cytokines were also detected, including IL-4, IL-5, and IL-13, but not IL-9 (Fig. 3E). We did not observe any cytokine related to TH17 (IL-17A and IL-17F) or TFH (IL-21) activity and very little IL-22 (TH22). In most samples (n = 13), we consistently observed a hierarchy of cytokine production, with high amounts of IL-13, lower but substantial amounts of IFN-γ, and low amounts of IL-4 (Fig. 3F). Under the same conditions, neonatal TN and Treg cells did not secrete significant amounts of TH1, TH2, or TH17 cytokines compared to TEM cells. These results were consistent with the presence of neonatal TEM cells at birth, based on their ability to produce effector molecules. We can additionally conclude that neonatal TEM cells display a mixed TH1/TH2 profile.

Fig. 3. Cord blood TEM cells display TH1 and TH2 functions.

TN, Treg, and TEM cells were FACS-purified from cord or adult blood as described in Fig. 1. (A to C) Cells from cord blood (A to C) or adult blood (B) were activated with PMA/ionomycin for 8 hours (A and B) and anti-CD3/CD28 for 20 hours, and then the indicated cytokines were analyzed by intracellular staining (A) or enzyme-linked immunosorbent assay (ELISA) (B and C). Results are means ± SD of duplicate or triplicate. (D) Neonatal TEM cells activated with anti-CD3/CD28 were analyzed by intracellular staining for the production of the indicated cytokines after 72 hours. Insets correspond to control Ig staining. (E) Indicated cytokines were tested in culture supernatants from serially diluted activated neonatal TEM cells cultured for 72 hours. The results are expressed as the mean of duplicates and are representative of at least three experiments. (F) Culture supernatants of the indicated activated T cells (30,000 to 100,000) were analyzed for each donor (the same cell number for every donor) to measure IL-4, IL-13, IFN-γ, and IL-17 cytokines by ELISA or Luminex (n = 13 donors). Data are expressed in pg/ml. *P < 0.05, paired t test. Exact P values are provided in table S6.

Neonatal TEM subsets

The functional TEM heterogeneity in terms of cytokine production led us to investigate the complexity of these neonatal T cells by analyzing putative phenotypic T cell subsets. Chemokine receptor expression represents a convenient and powerful means to describe effective functional differentiation (28, 29). CXCR3 is found on the surface of IFN-γ–producing TH1 cells, CRTH2 on the surface of IL-4–producing TH2 cells, and CCR6 on the surface of TH17 cells. We analyzed the expression of CXCR3, CRTH2, CCR6, and CCR4 in adult and neonatal blood CD4 T cells (Fig. 4). In adult peripheral blood, analysis of CD45RACD25lo cells (named TEM in the figure) allows for the discrimination between TH1 cells expressing CXCR3 and TH2 cells expressing CRTH2 (Fig. 4A). In neonatal blood, we identified up to nine combinatorial phenotypes based on the four chemokine receptors analyzed. Along with IFN-γ secretion (Fig. 3), we observed CXCR3+ TEM cells in neonatal blood, and these cells accounted for about 35% of the total TEM population (Fig. 4B). Despite the fact that IL-4, IL-5, and IL-13 were produced, we failed to detect CRTH2 cells (below 0.5% of TEM cells; Fig. 4B). CCR4 is not purely specific for TH2 and was previously found to be associated with CRTH2 cells in adult blood. It was expressed by 80 to 85% of neonatal TEM cells and therefore did not refer to TH2-like cells. Despite the lack of IL-17 secretion, we observed a small frequency of CCR6+ TEM cells. CCR6 is associated with TH17 cells (30), and about 20% of neonatal TEM cells were CCR6+, a third of which were also CXCR3+. In summary, chemokine receptor expression was consistent with the presence of a TH1 population among neonatal TEM cells but did not contribute to the phenotypic discrimination of other TH functions.

Fig. 4. Chemokine receptor expression analysis defines a large variety of neonatal CD4 T cells.

(A) Adult and neonatal T cells were gated as TN, Treg, or TEM cells, as described in Fig. 1, and analyzed for CD25, CD45RA, and chemokine receptor expression. (B) Population frequency analysis of neonatal TEM cells for the indicated phenotypes (mean of six donors ± SD).

Molecular profiling of neonatal TEM subsets

The low frequency of TEM subsets appeared to be a limiting factor for a clear functional description of neonatal TEM cells via these assays. This led us to further analyze TEM subsets at the molecular level by performing a transcriptional analysis of the main TEM subsets. Neonatal TEM cells were sorted by sequential gating on the basis of the presence of chemokine receptors and included all TEM cells with the exception of the rare CRTH2 population (gating strategy and purity of the subpopulations are detailed in Fig. 5A, fig. S1, and table S1). First, CXCR3+ TEM cells were purified as CRTH2CXCR3+ (about 35% of TEM cells) and CCR6+ TEM cells were purified as CRTH2CXCR3CCR6+ (about 10 to 15% of TEM cells). Then, CCR4+ TEM cells, corresponding to CRTH2CXCR3CCR6CCR4+ (about 45 to 50% of TEM cells), and CCR TEM cells, which are negative for all chemokine receptors analyzed (about 5% of TEM cells), were purified. Neonatal TN cells isolated as CD45RA+ T cells were used as a control. Whole-genome analysis using Agilent chips was performed on each of the four neonatal TEM populations and on TN cells isolated from three male donors (Figs. 5 and 6). Principal components analysis (PCA) showed a good clustering pattern of each population for all three donors (Fig. 5B). All TEM subsets were different from neonatal TN cells and from each other, indicating that they are indeed distinct subsets. CXCR3+ TEM cells and CCR6+ TEM cells were closely clustered on the basis of their transcriptional patterns. CXCR3+ TEM cells expressed moderate levels of TBX21 but very high levels of IFNG gene (Fig. 5C), and can thus be defined as a TH1 subset; however, CCR6+ TEM cells expressed high levels of RORC, but failed to express IL17A. TH2-associated GATA3 was similarly found in all TEM cells, and IL4 was expressed in CXCR3+ TEM cells, CCR4+ TEM cells, and CCR TEM cells, but not in CCR6+ TEM cells. We also compared the neonatal TEM cells with published gene signatures for TH1 (80 genes), TH2 (335 genes), TH17 (177 genes), TFH (165 genes), and Treg (234 genes), as detailed in table S2. For each TEM subset, we calculated the sum of intensities of all selected genes relative to the TN from the same donor to take into account both the genes and their expression normalized to their TN counterpart. Figure 5D confirms that CXCR3+ TEM cells and CCR6+ TEM cells are closer to TH1 and TH17 signatures, respectively, than to TH2, TFH, or Treg. However, CCR4+ TEM cells were broadly associated with many TH signatures, including TH17 and Treg, but not related specifically to a TH2 signature. As shown in Fig. 6, unsupervised statistical hierarchical clustering of more than 3500 gene probes highlighted the extent to which CCR6+, CXCR3+, and CCR4+ cells activated various gene sets after TCR engagement. These gene clusters were distinct from one subset to another; however, all TEM cells down-regulated genes that were found to be up-regulated in activated TN cells (Fig. 6 and table S3). CCR TEM cells up-regulated several genes independently of those found in other TEM and TN cells, and were similarly strongly associated with all TH signatures (Fig. 5D). This potentially indicates that CCR TEM cells are still capable of acquiring many possible fates, and they may represent a flexible intermediary or transitory phenotype between TN cells and other TEM cells. Together, the functional study documents TH1 and TH2 functions, whereas microarray analysis of four phenotypes of TEM cells based on chemokine receptor expression shows distinct subsets, among which CXCR3+ TEM cells and CCR6+ TEM cells show a close affiliation to TH1 and TH17, respectively.

Fig. 5. Chemokine receptor expression pattern defines molecularly different cord blood TEM subsets.

(A) FACS analysis of neonatal TEM cells purified as CRTH2CXCR3+ (CXCR3+), CRTH2CXCR3CCR6+ (CCR6+), CRTH2CXCR3CCR6CCR4+ (CCR4+), and CRTH2CXCR3CCR6CCR4 (CCR). Each graph corresponds to the analysis of the indicated chemokine receptor for the different populations. (B and C) After activation for 24 hours with anti-CD3/CD28, whole-genome microarray analysis was performed on the indicated purified TEM subsets (n = 3 donors). (B) PCA analysis of the indicated T cell subsets performed on microarray data. (C) The expression of the indicated transcription factors and cytokine genes related to TH1, TH2, and TH17 subsets is shown and expressed as the mean ± SD of fold change of expression compared to TN CD4 T cells for the corresponding donor. *P < 0.05, one-way analysis of variance (ANOVA). Exact P values are provided in table S6. (D) Microarray data for each neonatal subset were analyzed and compared for the expression of a selected gene set corresponding to TH1, TH2, TH17, TFH, and Treg subsets. Radar plot shows the mean ratio (n = 3) of the sum of intensities of indicated TEM for each plotted set of genes normalized to their respective TN. The color code for the populations applies to the whole figure.

Fig. 6. Microarray analysis of cord blood T cells can discriminate between TEM subsets.

Heatmap display of hierarchical clustering analysis of microarray data of TN CD4 T cells and CXCR3+, CCR6+, CCR4+, and CCR TEM cells from three donors, as described in Fig. 5, generated with Qlucore Omics (3500 probes; P = 0.035).

Assessment of TH17 potential for CCR6+ TEM cells

The CCR6+ TEM cells are poorly defined, and the low frequency of these cells limits the possibility of performing a functional analysis of isolated cells. The strong expression of RORC led us to specifically extract a number of TH17-associated genes (20, 31, 32) from the microarray analysis (Fig. 7A). The biased analysis of TH17-associated genes in CCR6+ TEM cells in comparison with TN cells shows an increased expression of transcription factors such as RORA, MAF, PRDM1, and BATF in addition to more recently described genes such as FAS, ETV6, POU2AF1, and TSC22D3 (32). Nevertheless, in agreement with the lack of IL-17A secretion, IL-17A transcripts were not detected, and IL-17F and IL-22 were not consistently detected. With regard to TH17 cell surface receptor transcripts, the chemokine receptors CCR4, CCR6, and CCR8 were expressed. We also found cytokine receptors for IL-1 (IL1R1) and IL-23 (IL23R and IL12RB1), which are important for TH17 differentiation, expansion, and maintenance (33) in CCR6+ TEM cells. We therefore determined their capacity to perform TH17 functions. In the absence of IL-6 and TGF-β (transforming growth factor–β), which are the key mediators of TH17 differentiation, there was no induction of TH17 cells from TN cells in the presence of IL-1β and IL-23 (Fig. 7B). However, under similar conditions, neonatal TEM cells (but not Treg cells) developed a TH17 response, characterized by IL-17A and IL-17F secretion. IL-22 was also produced by TEM cells, but this occurred in much lower amounts than in TN cells. In addition, IL-22 induction in TEM cells was independent of IL-1β and IL-23, indicating that it was not associated with the TH17 response. We failed to detect a TEM response to NIMA using the mother’s antigen-presenting cells (APCs) (fig. S4). In conclusion, at birth, the neonatal T cell compartment is not entirely naïve and contains CD4 T cells with an effector memory phenotype, which display a large variety of inflammatory effector functions similar to those observed in adult blood.

Fig. 7. Neonatal TEM cells can develop into TH17 cells.

(A) Expression of TH17-associated genes in neonatal TEM cells from the microarray analysis of three donors. Data are expressed as fold change over TN cells. (B) Neonatal TN, Treg, and TEM cells were stimulated with anti-CD3/CD28 in the presence or absence of IL-1β and IL-23 for 4 to 6 days. Supernatants were tested for IL-17A or IL-17F and IL-22. The results are representative of two experiments. *P < 0.05, paired t test. NS, nonsignificant. Exact P values are provided in table S6.


Here, we identified a population of CD4 T cells with a memory phenotype in the cord blood of healthy neonates. Neonatal TEM cells were characterized on the basis of their phenotype as CD25loCD127hi and accounted for 1 to 3% of the CD4 T cell compartment in all cord blood samples tested (n > 50). Severe combined immunodeficient patients have often shown a propensity to allow maternal T cell chimeras (34), and this phenomenon has also been observed for various immune cell populations in nonpathological situations (35). However, in our study, X and Y chromosomal FISH staining demonstrated that the TEM cells mainly originated from the baby. It has been thought that the T cell compartment in neonatal blood only contains TN and Treg cells. Recently, however, human Treg cells have been shown to be heterogeneous and to include a small fraction of cells expressing inflammatory/effector cytokines (36, 37). Despite the expression of CD25, neonatal TEM cells did not express Foxp3, as assessed by FACS intracellular staining, and thus are distinct from Treg cells. Upon activation, these cells behaved as inflammatory cells with a mixed TH1- and TH2-like function, because they secreted a panel of cytokines, including TNF-α, GM-CSF, IFN-γ, IL-4, and IL-13, and were able to secrete IL-17A and IL-17F when stimulated with IL-1 and IL-23.

There are some limitations to our study, in particular linked to the fact that we did not describe any antigen specificity, raising an issue about the origin and the role of these T cells in the fetus. Another limitation concerns the possibility that we could not totally rule out the possibility that a small fraction of TEM (less than 5%) could be of maternal origin.

We used chemokine receptors associated with effector functions (29) to separate neonatal TEM cells into cell subsets for further analysis. Only a very small fraction (<0.5%) of neonatal TEM cells expressed CRTH2, which is a robust marker for TH2 cells in adult blood (38, 39). In contrast, large amounts of IL-13 and IL-4 were produced. CCR4 could not be used as a surrogate marker for a TH2 memory population, because it was found on almost all neonatal TEM cells, and CCR4+ TEM cells did not show any clear evidence for a TH2 signature. IL-4 transcripts were detected in CCR4+CXCR3CCR6 TEM cells, CXCR3+ TEM cells, and CCR TEM cells, but not in CCR6+ TEM cells. This result seems to indicate that there is a general trend of neonatal TEM cells toward TH2 cytokine production. Such a preference has been previously reported in murine naïve fetal and neonatal CD4 T cells (40) and is the result of epigenetic regulation of these T cells at the IL-4 locus (41). Unlike human T cells, murine T cells only start to egress from the thymus after birth. Therefore, the extent to which this applies to humans remains to be determined. Because cord blood T cells represent a common source of naïve cells to study the differentiation of the various TH cells, this suggests that naïve neonatal T cells remain unbiased and still open to various TH cell fates.

As expected, IFN-γ transcripts were only up-regulated in neonatal CXCR3+ TEM cells. This result is in agreement with CXCR3 expression on TH1 memory cells in adults (29). These TH1 cells account for 30% of neonatal TEM cells. Therefore, our results highlight the capacity of neonatal immune cells to develop inflammatory TH1 cells (and also pro-TH17 cells) in the noninflammatory and presumably sterile fetal environment. It is notable that inflammatory mediators (which are low–molecular weight molecules) from the mother can easily cross the placenta and thus can contribute to the type of fetal T cell responses. Whatever their origin, the detection of these neonatal TEM cells that produce IFN-γ illustrates the capacity for TH1 development during early life. Congenital cytomegalovirus infection was previously shown to be associated with the fetal development of mature T cell responses, including IFN-γ–producing T cells, a process that involved CD4 and CD8 αβ T cells (42, 43) and the γδ T cell compartment (44). Here, we have shown that this also occurs in the absence of severe infection and without any pathophysiological consequences.

One of the particular features of neonatal TEM cells is the expression of a TH17 gene signature by CCR6+ T cells, a characteristic that is associated with an inability to directly release the IL-17 cytokine family. Although they secrete TH1/TH2 cytokines, neonatal TEM cells express CD161 (KLRB1), which is associated with TH17 precursors. Cosmi et al. (45) have previously described pro-TH17 cells that are specifically found in the cord blood. These CD161+CD4+ T cells express a TH17 gene signature (RORC, IL23R, and CCR6) and are unable to produce cytokines; however, after 1 to 2 weeks of culture in the presence of IL-1β and IL-23, they become able to secrete IL-17. These TH17 precursors account for 1 to 2% of CD4 T cells in the cord blood (45), whereas the neonatal CCR6+ TEM cells we identified in the present study represent less than 0.3% of CD4 T cells. In addition, neonatal TEM cells are CCR7 and CD45RO+, indicating that these cells poorly overlap with TH17 precursors in phenotype and function. Thus, the current study highlights the plasticity of neonatal TEM cells and opens perspectives to explore the detailed mechanisms of neonatal immune regulation.

The traditional view of memory T cells has evolved since the original linear scheme in which naïve cells become effector cells and then are converted to central/effector memory cells (23). Thus, the CD45RO+RA phenotype cannot be considered the only memory T cell phenotype. Song et al. (46) provided evidence for an early memory pool of CD45RA+ROCD62L+ T cells, which are able to secrete IFN-γ and IL-4, depending on the expression of CXCR3 and CCR4, respectively. Similarly, pro-TH17 cells are atypical when their CD45RA+RO phenotype is taken into consideration (45, 47). More recently, a particular population of T cells known as memory stem cells has been described for both CD4 and CD8 compartments (4850). These T cells exhibit stem cell properties and further challenge the canonical phenotype and function of memory cells, because they exhibit a CD45RA+ROCD27+CD127+CCR7+ naïve-like phenotype and are associated with the production of IL-2, TNF-α, and IFN-γ upon CD3/CD28 stimulation.

Although the CD45RO/RA balance is no longer a canonical feature of memory cells, the neonatal TEM cells we have described here appear to be the classical type, with their CD3+CD4+CD45RO+RACD62LCCR7 phenotype. Earlier investigation using 20- to 38-week gestation cord blood described a 1 to 2% population of CD45RO+RA lymphocytes, with a frequency that did not correlate with gestational age (51). How and why these cells develop remains to be investigated. One can imagine that neonatal TEM cells develop in response to transplacentally acquired maternal antigens; however, we failed to detect a TEM response to NIMA using the mother’s APCs (fig. S4). Fetal T cell exposure to foreign antigens in utero could also occur during mild/asymptomatic infections or vaccination of the mother during pregnancy. Such antigen transfer has been proposed in the context of the influenza vaccination. Detection of influenza-specific T cells using tetramer staining of cord blood cells has been reported (52). However, these results should be considered with caution because of the extremely low number of T cells that were detected by FACS in these experiments. Therefore, this interesting study still awaits additional supportive reports for the same or another vaccination. The antigen specificity of neonatal TEM cells remains to be determined, including their potential self-reactivity. Tubo et al. (53) recently showed that various TH subsets can be generated from the very same TN, resulting in a diverse pattern of effector functions from a single original clone. Our analysis of TCRVb repertoire for the neonatal T cell populations shows the polyclonal nature of TEM cells. Other hypotheses could also be proposed. Similar to naturally occurring thymic-derived Treg (nTreg) cells, natural TH17 cells directly exiting the thymus have been reported (54). Neonatal TEM cells would also represent a pool of polyfunctional “natural” memory cells that could arise directly from the thymus. Such cells could thus be quickly mobilized during early life in the absence of any preexisting memory stimulus. Neonatal TEM cells express CD69 and are Ki67-positive, as if they have been recently activated; however, the lack of CD31 and PTK7 expression rules out the recent thymic emigrant nature of neonatal TEM cells. These cells could also correspond to a pool of memory cells undergoing proliferation in response to endogenous signals. Alternatively, recent characterization of nonpathogenic commensal bacteria in the placenta may represent a source of microbial signal for immune maturation (55). Immunohistochemistry studies reported CD45RO+CCR5+ T cells within the fetal intestine, another possible place for immune maturation (56).

Our final point concerns the qualitative impact of these neonatal TEM cells on the immune response to microorganisms and to vaccination. T cell responses develop via cross-reactivity from an existing repertoire, and therefore, frequency, function, and specificity of both naïve or memory T cells may influence the outcome (57, 58). TH2 responses develop in response to many vaccines (13); however, neonatal Bacille Calmette-Guérin vaccination can lead to TH1 responses (59, 60). The potential cross-reactivity of neonatal TEM cells and the influence of these cells on the ongoing immune response are not yet defined; thus, further investigation should include the evaluation of the CXCR3+ TEM cells versus other subsets in response to mycobacteria. Future experiments should also aim to decipher the origin and the impact of these neonatal inflammatory effector memory T cells on the response to vaccines and infections. The possible role of these neonatal TEM in graft-versus-host disease after the use of mismatched umbilical cord blood as a source of stem cells in transplantation to treat hematologic disorders (61) could also be investigated.


Study design

The study was based on an original observation of a specific phenotype of memory T cells in cord blood. All experiments were then designed and performed on cord blood samples collected from delivery of healthy pregnant mothers without any reported infectious disease and used in an unsupervised manner.

Sample size was not defined in advance. Neonatal TEM analysis was performed on 29 donors for the phenotypic analysis. Thirteen of these donors were also used for functional analysis, and three others for microarrays. For functional analysis, all samples were included, complete with the three T cell populations (TN, TEM, and Treg). All data were included to take into account dispersion and heterogeneity of the responses. For phenotypic analysis, samples were selected according to blood volume, allowing sufficient TEM purification. Outliers were not excluded. For microarrays, only male donors were used for PCA analysis and were chosen to avoid sex interference within PCA of the three donors.


Buffy coats were obtained from adult donors by the Etablissements Français du Sang (Paris, France). Heparinized cord blood samples were collected between 2010 and 2013 from healthy full-term neonates at the Maternity Port Royal and Bichat (Paris, France). Inclusion criteria for the mother consisted of a normal pregnancy and negative HIV, hepatitis B virus, and hepatitis C virus tests. Additional exclusion criteria eliminated donors that experienced preeclampsia, fever, acute infection, diabetes, and/or chronic disease. Written consent was obtained from the mothers. This study was carried out with the approval of the Ethics Committee of the Institut Pasteur and the French regional committee (Comité de Protection des Personnes Ile-de-France IV) in agreement with the principles of the Declaration of Helsinki.

Cell purification

Cord blood mononuclear cells (CBMCs) or peripheral blood mononuclear cells (PBMCs) were separated from cord blood or adult buffy coats by gradient centrifugation with a Lymphoprep kit (Axis-Shield). T cell fractions were positively enriched from CBMC or PBMC preps with anti-CD4 magnetic beads using an AutoMACS (Miltenyi Biotec). In some cases, T cells were negatively enriched for CD45RO cells with a T cell memory kit (Miltenyi Biotec). To sort the subset of T cells present in the T cell–enriched cord blood, samples were sorted on the basis of surface markers to obtain the following fraction: CD3+CD4+CD25CD127hiCD45RA+ (TN cells), CD3+CD4+CD25hiCD127lo (Treg cells), and CD3+CD4+CD25loCD127hi (TEM cells). Cells sorted by AutoMACS on a FACSAria III cell sorter (BD) were routinely >90% pure for TEM cells and 97 to 99% pure for Treg and TN cells.

Culture medium and reagents

Complete medium consisted of RPMI 1640 supplemented with 10% fetal calf serum (ICN Biomedicals Inc.), 5 × 10−5 M 2-mercaptoethanol (Sigma), and antibiotics (Gibco BRL). TGF-β was purchased from R&D Systems. Cytokines IL-1β, IL-6, and IL-23 were purchased from PeproTech. The antibodies used in this study are listed in table S4.

Cell stimulation and cytokine/chemokine detection

Cells were generally cultured in complete medium or X-VIVO 20 and stimulated in a volume of 125 μl for 48 hours. IFN-γ, IL-4, IL-13, and IL-17 were measured by ELISA kits from eBioscience. For large panels, cytokines were detected with multiplex Luminex kits (Affymetrix).

In vitro T cell differentiation

T cells (105, 5 × 104, or serially diluted) were restimulated by plate-bound anti-CD3 (OKT3) and soluble anti-CD28. Two to 6 days later, supernatants were collected and tested for cytokines. T cells were restimulated with PMA/ionomycin (BD Biosciences) and subjected to intracellular staining. To detect intracellular cytokines, cells were fixed, permeabilized, and stained with antibodies against IL-4, IL-13, IL-17, IL-22, IFN-γ, and TNF-α. Cells were detected on a FACS LSRFortessa (BD) or CyAn (Coulter) machine and analyzed with FlowJo software (Tree Star).

Microarray analysis

Negatively enriched cord blood TEM cells were FACS-sorted as detailed in fig. S1 and table S1, and 5000 cells were stimulated for 24 hours with anti-CD3/CD28. After 24 hours, cells were lysed and pellets were frozen at −80°C and subsequently used for RNA isolation. The four TEM subsets and the TN fraction were isolated from three independent donors, all of whom were male. The gene expression profiles were measured by Miltenyi Biotec with an Agilent DNA chip. We used the Agilent 60-mer Whole Human Genome Oligo Microarray containing about 59,000 known gene and candidate gene probes. Data sets were analyzed with IPA (Ingenuity) and Qlucore Omics software.

Fluorescent in situ hybridization

Fluorescent in situ hybridization (FISH) was performed to detect the X and Y chromosomes (MetaSystems) in the nuclei of TEM cells according to the manufacturer’s protocol. The sorted cells were fixed in freshly prepared fixative (3:1 methanol/acetic acid) for 30 min at 4°C. The DNA probe mixture contained repetitive sequences specific for the chromosome X (green) or Y (orange) centromeric region. Slides were mounted with antifade Vectashield (Vector Laboratories) containing 4′,6-diamidino-2-phenylindole to counterstain the nuclei. Microscopic analysis was performed using a Zeiss Axioplan 2 imaging microscope and AxioVision software.

Gene set analysis

We extracted gene lists from microarray data sets previously published for human adult T cell subsets (table S2). Only genes that were up- or down-regulated twofold were selected. Gene sets for TH1 and TH2 cells were from in vitro–differentiated CD4 T cells from Aijö et al. (62). TH17 genes were defined by selecting genes with TH17/TH0 ratios >2 or TH17/TH0 ratios <0.5 from Ciofani et al. (31). Genes for TFH were from ratios of CXCR5+ TFH gene expression over TEM (>2 for up-regulated and >0.5 for down-regulated) (63). The gene list for Treg was based on ratios of Treg versus Teff (effector T cells) (64). The level of transcriptional expression of all tagged TH genes was determined by transcriptome analysis of the three different donors for a neonatal TEM subset normalized to its corresponding TN. For each gene, the average of the three donors was calculated, and a sign was assigned depending on whether it was up-regulated (plus sign) or down-regulated (minus sign). Finally, the sum of all these signed averages was used to draw the radar plot of Fig. 5D.

T cell repertoire analysis

Total RNA was prepared from cord blood T cell subsets with the Total RNA Miniprep kit (Sigma-Aldrich), and complementary DNA was synthesized with SuperScript II Reverse Transcriptase (Invitrogen). The different BV germline genes can be clustered in 24 families according to their level of homology (IMGT nomenclature, Ten BV families corresponding to BV2, BV3, BV4, BV5, BV6a, BV7b, BV19, BV20, BV28, and BV29 and representing around 70% of a normal adult T cell repertoire were selected for the analysis. For quantitative repertoire, polymerase chain reactions (PCRs) were carried out by combining a reverse primer and a specific fluorophore-labeled probe for the constant region (MGB-TaqMan probe), with 1 of 10 primers covering the different BV chains (table S5). Real-time PCRs were subsequently carried out in a 25-μl reaction mixture with a final concentration of 400 nM of each oligonucleotide primer, 200 nM of the fluorogenic probe, and 1× FastStart Universal Probe Master Mix (Roche). Thermal cycling conditions comprised FastStart Taq DNA Polymerase activation at 95°C for 10 min, then 40 cycles of denaturation at 95°C for 15 s, annealing, and extension at 60°C for 1 min. For all of these reactions, real-time quantitative PCR was then performed on an ABI 7300 system (Applied Biosystems). For Immunoscope profiles, products were then subjected to run-off reactions with nested fluorescent primers Vic-ACACAGCGACCTCGGGT, Pet-ACACAGCGACCTCGGGT, and Fam-CCTTTTGGGTGTGGGAGA-MGB specific for the constant region for three cycles. The fluorescent products were separated and analyzed using an ABI Prism 3730 DNA analyzer. The size and intensity of each band were analyzed with Immunoscope software (65), which has been adapted to the capillary sequencer. Fluorescence intensities were plotted in arbitrary units on the y axis, and CDR3 lengths (in amino acids) on the x axis.

Statistical analysis

Unpaired or paired t tests were used to compare two groups of data. Data are presented as means ± SD. P values of <0.05 were considered statistically significant. One-way ANOVA was used in Fig. 5C. Summary of statistical calculations is shown in table S6.


Fig. S1. Phenotype of neonatal T cells and purification and gating strategy for isolation of TEM cells.

Fig. S2. FISH staining.

Fig. S3. Functional response of adult T cells.

Fig. S4. Coculture of maternal APCs with cord blood T cells.

Table S1. Purity of CD3+CD4+ T cell populations sorted for microarray analysis.

Table S2. List of genes for CD4 TH subset population.

Table S3. List of 150 genes allowing discrimination of all TEM populations.

Table S4. List of antibodies used for FACS.

Table S5. List of TCRV primers.

Table S6. Details of statistical analysis.


  1. Acknowledgments: We thank J. M. Treluyer (CIC 0901) for the cord blood sample collections. Funding: This work was supported by an Agence Nationale de la Recherche (ANR) grant (ANR 09-MIEN-017) and by the Fondation pour la Recherche Médicale (grant no. DEQ20120323719). This study also received funding from the French Government’s Investissement d’Avenir program, Laboratoire d’Excellence “Integrative Biology of Emerging Infectious Diseases” (grant no. ANR-10-LABX-62-IBEID). X.Z. and S.L. were supported by ANR and by the European Commission FP7 ADITEC program (HEALTH-F4-2011-280873). X.Z. was also partially supported by the Shanghai Rising-Star Program (grant no. 12QA1403600). B.M. was supported by the Pasteur-Paris University (PPU) International PhD program and by the Institut Carnot. D.Z. was supported by DIM Malinf et region IdF. We acknowledge the Center for Human Immunology at the Institut Pasteur for support in conducting these studies. Author contributions: X.Z. and B.M. performed the research. S.L., E.D., A.L., and D.Z. performed the experiments and analyzed the data. A.V. did bioinformatics analysis of the transcriptomics data. O.L., G.R., E.A., C.L.R., and C.L. contributed to experimental design. R.L.-M. designed and supervised the project. X.Z., B.M., S.L., C.L., and R.L.-M. wrote and/or critically revised the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Microarray data sets are available at
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article