Interleukin-17 (IL-17)–producing T helper cells have been proposed to represent a separate lineage of CD4+ cells, designated Th17 cells, which are regulated by the transcription factor retinoic acid–related orphan receptor γt (RORγt). However, despite advances in understanding murine Th17 differentiation, a systematic assessment of factors that promote the differentiation of naive human T cells to Th17 cells has not been reported. The present study was undertaken to assess the effects on naive human CD4+ T cells of cytokines known to promote murine Th17 cells.
Human naive and memory CD4+ T cells isolated from peripheral blood were activated and cultured with various cytokines. Cytokine production was measured by enzyme-linked immunosorbent assay and flow cytometry. Messenger RNA was measured by quantitative polymerase chain reaction.
In response to anti-CD3/anti-CD28 stimulation alone, human memory T cells rapidly produced IL-17, whereas naive T cells expressed low levels. Transforming growth factor β1 and IL-6 up-regulated RORγt expression but did not induce Th17 differentiation of naive CD4+ T cells. However, IL-23 up-regulated its own receptor and was an important inducer of IL-17 and IL-22.
The present data demonstrate the differential regulation of IL-17 and RORγt expression in human CD4+ T cells compared with murine cells. Optimal conditions for the development of IL-17–producing T cells from murine naive precursors are ineffective in human T cells. Conversely, IL-23 promoted the generation of human Th17 cells but was also a very potent inducer of other proinflammatory cytokines. These findings may have important implications in the pathogenesis of human autoimmunity as compared with mouse models.
Classically, naive CD4+ T cells have been thought to differentiate into 2 possible helper lineages, Th1 or Th2 cells. Th1 cells produce the signature cytokine interferon-γ (IFNγ), a critical factor that promotes cellular immunity. Interleukin-12 (IL-12), acting via the transcription factor STAT-4 in concert with T-box expressed in T cells/T-box 21 (T-BET), is critical for Th1 differentiation. In contrast, Th2 development is initiated by IL-4 signaling with the participation of the transcription factors STAT-6 and GATA-3. The hallmark cytokine secreted by Th2 cells is IL-4, which is crucial for host defense against helminths and the pathogenesis of asthma and allergy. Th1 and Th2 lineage decision appears to be made at a very early stage of T helper cell differentiation, with the respective Th1/Th2 cytokines enforcing their own expression and inhibiting alternative commitment. This occurs by regulation of receptor levels, expression of transcription factors, and epigenetic changes (1–3). Another important aspect of Th1/Th2 counterregulation is interchromosomal interaction between Th1- or Th2-specific cytokine genes (4). As a result of these mechanisms, Th1 and Th2 cells develop into mature effectors with stable phenotypes and important roles in host defense, as documented in a number of murine models.
While the simple dichotomous model of T helper cell differentiation fits well with many models of infection, fitting autoimmune disease into such models has been problematic. CD4+ T cells can also differentiate to become Treg cells, and it is clear that dysregulation of this subset has major consequences with respect to the pathogenesis of autoimmunity (5). An additional complication relates to the discovery of another cytokine, IL-23. IL-23 is a dimer that shares a subunit with IL-12, IL-12p40, and both utilize a receptor subunit designated IL-12 receptor β1 (IL-12Rβ1) (6). The complex biology of IL-12 and IL-23 is relevant to the pathogenesis of autoimmunity, in that gene targeting of IL-12p40 attenuates the development of disease in many models of autoimmunity. Similarly, anti-p40 antibody is efficacious in the treatment of Crohn's disease (7). Although these effects were initially misattributed to interference with the actions of IL-12, using specific deletion of IL-23 (p19−/− mice), it is now recognized that IL-23 (and not IL-12) is the culprit, at least in many of the animal models (8–13).
One way that IL-23 is thought to promote autoimmune disease is through the regulation of IL-17A and IL-17F. IL-17A is a proinflammatory cytokine initially identified in mouse cytotoxic T cells more than 10 years ago. This family now includes 6 members (IL-17 [IL-17A], IL-17B, IL-17C, IL-17D, IL-17E, and IL-17F) (14–17) that share 16–50% amino acid identity and have different tissue expression patterns. IL-17 acts on epithelial cells, endothelial cells, fibroblasts, synoviocytes, and myeloid cells to induce secretion of a variety of mediators, including IL-8, CXCL1, CXCL6, IL-6, granulocyte–macrophage colony-stimulating factor, granulocyte colony-stimulating factor, tumor necrosis factor α (TNFα), and IL-1β. IL-17 family cytokines thus induce cellular infiltration and production of inflammatory cytokines. Dysregulated production of IL-17 is associated with human autoimmune diseases, including multiple sclerosis, inflammatory bowel disease, and psoriasis (13, 18–20). Importantly, studies in a murine experimental arthritis model showed that IL-17 was involved in both the initiation and the progression of the disease. Furthermore, elevated levels of IL-17 were detected in synovial fluid from patients with rheumatoid arthritis, and osteoclast formation was inhibited by anti–IL-17 antibody, suggesting an effect on bone resorption (21–25).
IL-17A was originally reported to be produced by activated CD4+ and CD8+ T cells. More recently, it has been proposed that IL-17–producing CD4+ T cells represent a distinct lineage (Th17), a lineage that does not produce IL-4 or IFNγ (26–28). In fact, these products of Th1 and Th2 cells antagonize the differentiation of Th17 cells. Additionally, cytokines that promote Th17 differentiation are distinct from those that promote Th1 and Th2 differentiation. The current model is that whereas transforming growth factor β1 (TGFβ1) promotes Treg cell differentiation, the combination of TGFβ1 and IL-6 promotes Th17 lineage commitment (29–31). This subset is expanded and maintained by IL-23. Both IL-6 and IL-23 activate the transcription factor STAT-3, which directly binds to the IL-17 promoter to regulate IL-17 expression (32). Conversely, suppressor of cytokine signaling 3 negatively regulates Th17 differentiation by inhibiting STAT-3 phosphorylation (32, 33). In addition to STAT-3, retinoic acid–related orphan receptor γt (RORγt) is an important transcription factor for initiation of Th17 differentiation (34).
The current models of Th17 differentiation are all derived from studies using murine cells. Given the pathogenic relevance of IL-17 family members, it is obviously important to understand how this family of cytokines is controlled in human T cells and to define the conditions under which human naive CD4+ T cells might become Th17 cells. Previously, it was reported that T cell receptor (TCR) crosslinking leads to IL-17 production (35). It has also been reported that IL-23 induces IL-17 expression (26). However, these studies did not separate effects on memory cells versus naive cells. The ability of naive human CD4+ T cells to differentiate into IL-17 producers has not been examined, nor has a comprehensive assessment been undertaken to define the relative importance of various conditions in promoting the differentiation of this putative human lineage. Additionally, the extent to which human T cell subsets fully commit to (or extinguish) IL-17 production has not been addressed. This is important not only with respect to the pathogenesis of human autoimmune disease, but also because human T helper cell differentiation appears to exhibit much more plasticity than murine T helper cell differentiation (36).
In the present report, we show that human memory T cells up-regulate RORγt and IL-17 in response to TCR occupancy. Importantly, in sharp contrast to murine naive CD4+ T cells, in which Th17 lineage commitment is initiated by TGFβ1 and IL-6, human Th17 differentiation does not occur in response to this cytokine combination, even though these cytokines promote RORγt expression. Instead, IL-23 up-regulates its own receptor on T cells and increases RORγt and IL-17 expression. IL-23 also induces other proinflammatory cytokines, including IL-22. These findings may have important implications in the pathogenesis of human autoimmunity as compared with mouse models.
MATERIALS AND METHODS
Peripheral blood mononuclear cells were isolated from peripheral blood of healthy donors (Department of Transfusion Medicine, National Institutes of Health, Bethesda, MD) by Ficoll-Paque Plus (Amersham Pharmacia Biotech, Uppsala, Sweden). Naive or memory CD4+ T cells were further purified using a human naive or memory CD4+ T cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany). The purity of CD45RA+,CD45RO− cells was ∼90%. Naive CD4+ T cells were activated with plate-bound anti-CD3 and soluble anti-CD28 (BD PharMingen, San Diego, CA) and cultured under neutral (Th0), Th1 (IL-12, anti–IL-4), Th2 (IL-4, anti–IL-12, anti-IFNγ), or Th17 (TGFβ1 and IL-6, anti-IFNγ and anti–IL-4) polarizing conditions with IL-2 or stimulated with IL-23 (10 ng/ml; R&D Systems, Minneapolis, MN). On day 7, cells were restimulated with plate-bound anti-CD3 and anti-CD28 and continuously cultured until day 14 under Th0, Th1, Th2, or Th17 polarizing conditions.
Quantitation of cytokine production in cell culture supernatants was determined by enzyme-linked immunosorbent assay (R&D Systems) according to the manufacturer's instructions. Cytokine-producing cells were determined by intracellular staining using phycoerythrin-conjugated anti-IFNγ (BD PharMingen), Alexa Fluor 647–conjugated anti-human IL-17 (eBioscience, San Diego, CA), and Alexa Fluor 488–conjugated anti-human forkhead box P3 (FoxP3; BioLegend, San Diego, CA). Briefly, cells were stimulated with phorbol myristate acetate and ionomycin for 4 hours, and Golgiplug (BD Biosciences, San Jose, CA) was added after 2 hours. Cells were fixed in 4% paraformaldehyde, permeabilized with 0.1% saponin, stained with fluorescent antibodies, and analyzed on a FACSCalibur flow cytometer (BD Biosciences). CellQuest software (BD Biosciences) was used for data acquisition, and Flow Jo software (Tree Star, Ashland, OR) was used for analysis.
Total RNA was isolated with an RNeasy kit (Qiagen, Valencia, CA). Complementary DNA was synthesized with the use of a TaqMan Reverse Transcription kit (Applied Biosystems, Foster City, CA) using random hexamers as primers according to the manufacturer's instructions. Hypoxanthine guanine phosphoribosyltransferase (HPRT) was used as an endogenous control. TaqMan primers and probes for human IL-17A (IL-17), IL-17B, IL-17C, IL-17D, IL-17E, IL-17F, IL-22, IFNγ, IL-4, T-BET, GATA-3, IL-23R, RORγ, and HPRT were purchased from Applied Biosystems, and samples were analyzed using the ABI Prism 7700 Sequence Detection System (Applied Biosystems).
Statistical analyses were performed using Small Stata 9.2 software (StataCorp, College Station, TX). Geometric means were computed separately for each stimulation. The log difference (geometric mean) in IL-17 or IL-22 secretion between the 2 stimulations for each donor was used to conduct a paired t-test and compute a 95% confidence interval.
High levels of IL-17 produced by human memory T helper cells.
We first assessed whether IL-17 production is detected in normal human donors. To this end, we fractionated CD4+ cells into memory (CD45RO+) and naive (CD45RA+) cells. As shown in Figure 1a, freshly isolated memory CD4+ T cells secreted large quantities of IL-17 in response to anti-CD3 and anti-CD28. While there was considerable individual variation in the propensity to produce IL-17, memory CD4+ T cells made significantly more IL-17 than did naive CD4+ T cells (Figure 1b), although modest levels of IL-17 were clearly produced by naive CD4+ T cells (∼15–20% of the levels produced by memory cells). In this setting, plate-bound antibodies and antibody-coated microbeads gave similar results (data not shown).
Human Th17 cells are not generated under optimal conditions for murine Th17 cells.
Since IL-17–producing T cells were readily detectable in human memory CD4+ T cells, we next sought to determine what conditions might favor the generation of such cells from naive CD4+ T cells. Recent studies have shown that TGFβ1 and IL-6, combined with blockade of IL-4 and IFNγ, quickly generate large numbers of IL-17–producing T cells from isolated naive murine CD4+ T cells within a few days (see supplementary Figure 1, available online at www.niams.nih.gov/rtbc/labs_branches/miib/lcbs/publications.htm) (29–31, 37). Therefore, we cultured naive human CD4+ T cells under these conditions (termed “Th17 conditions”) to assess whether this regimen would generate human IL-17–producing T cells.
Surprisingly, the combination of TGFβ1 and IL-6 with anti–IL-4 and anti-IFNγ was no better than anti-CD3 and anti-CD28 alone in terms of IL-17 production (P = 0.67; n = 5) (Figure 2a). This was the case regardless of the concentration of anti-CD3 used, with or without anti-CD28 (see supplementary Figure 2, available online at www.niams.nih.gov/rtbc/labs_branches/miib/lcbs/publications.htm). Kinetic analysis of IL-17 production during 2-week culture further supported this contention (Figure 2b). The effect of TGFβ1 was evidenced by the up-regulation of FoxP3, although, surprisingly, the combination of TGFβ1 and IL-6 enhanced FoxP3 expression, rather than inhibiting expression as is the case in murine cells (31) (Figure 2c). Some protocols for in vitro generation of murine Th17 cells use mononuclear cells along with soluble anti-CD3/anti-CD28 (29) and exogenous TGFβ1 and IL-6. However, this also failed to induce human cells to produce IL-17 (data not shown); in fact, activated monocyte-derived dendritic cells enhanced IFNγ production and inhibited Th17 differentiation (data not shown).
In murine CD4+ T cells, the combination of TGFβ1 and IL-6 with anti–IL-4 and anti-IFNγ not only efficiently generates IL-17–producing T cells, but also extinguishes IFNγ production. This is the basis of the idea that Th17 cells are a distinct functional lineage of mouse CD4+ effector cells (30, 31). Culture of human naive CD4+ T cells with or without exogenous TGFβ1 and IL-6 generated similar numbers of IL-17–producing cells, and notably, half of the IL-17–producing T cells produced both IFNγ and IL-17 (Figure 2d). Therefore, we next polarized naive human CD4+ T cells under Th1 (IL-12 and anti–IL-4) and Th2 (IL-4 and anti–IL-12 and anti-IFNγ) polarizing conditions and assessed their ability to produce IL-17. Surprisingly, IL-17 production was not extinguished under these conditions (see supplementary Figure 3, available online at www.niams.nih.gov/rtbc/labs_branches/miib/lcbs/publications.htm). Despite the absence of IL-17 messenger RNA (mRNA) in cells polarized under Th2 conditions for 7 days, restimulation via TCR occupancy (anti-CD3/anti-CD28 stimulation) permitted considerable IL-17 production, indicating the plasticity of human CD4+ T cells with respect to IL-17 production (Figure 3a). IFNγ and IL-17 dual producers were also observed in freshly isolated human memory CD4+ T cells activated with anti-CD3/anti-CD28, indicating that the incomplete differentiation of human Th17 cells was not simply an in vitro phenomenon (Figure 3b).
IL-23 up-regulation of IL-23R and enhancement of IL-17 production by human CD4+ T cells.
IL-23 is a heterodimeric cytokine that shares a ligand subunit (p40) and a receptor subunit (IL-12Rβ1) with IL-12 (38). IL-23 has been reported to be essential for the development of autoimmunity in mouse models, and current data suggest that this is due to the production of IL-17 (6, 8–10, 26, 39). We found that naive CD4+ T cells expressed low levels of IL-23R mRNA compared with memory CD4+ T cells (Figure 4a). Under Th0, Th1, and Th2 conditions, the expression of this receptor remained low; however, IL-23R mRNA expression was strongly up-regulated in anti-CD3/anti-CD28–activated T cells cultured with IL-23 (Figure 4b). Furthermore, IL-23R was inhibited by “Th17 conditions.”
In mice, IL-23 was initially thought to promote Th17 polarization, but was later viewed as a regulator of Th17 expansion/maintenance (29). Therefore, we next studied the effect of inclusion of IL-23 in cultures on the polarization of naive human precursors compared with other conditions. As shown in Figure 4c, anti-CD3/anti-CD28–induced IL-17 and IL-23 alone had little additional effect at early time points. Addition of IL-23 in Th17 polarizing conditions was also ineffectual. However, after 2 weeks of culture, cells cultured with IL-23 and restimulated with anti-CD3/anti-CD28 produced significantly higher amounts of IL-17 compared with cells cultured under other conditions. Cells cultured under Th1 and Th17 conditions had poor IL-17 production, and addition of IL-23 did not rescue cells cultured under Th17 conditions. IL-23 had relatively little effect on memory (CD45RO+,CD4+) T cells to produce IL-17 (Figure 4d); this contrasted greatly with its effect on IL-22 (see below). Again, IFNγ and IL-17 dual producers were observed under all these conditions (data not shown).
Regulation of RORγt expression in human T cells.
Recently, RORγt was reported to be the key transcription factor that regulates murine IL-17 expression (34). Therefore, we next investigated RORγt regulation in human CD4+ T cells. We found that RORγt mRNA levels were higher in freshly isolated memory CD4+ T cells than in naive CD4+ T cells, but were even more highly inducible in memory cells in response to anti-CD3/anti-CD28 (Figure 5a). Moreover, RORγt mRNA levels were slightly increased by the addition of TGFβ1 and IL-6, and increased even more by the combination of TGFβ1, IL-6, and IL-23 (Figure 5a), despite the fact that IL-17 mRNA levels were not affected (Figure 5b). The effects of cytokines on RORγt mRNA expression were also evident when starting with naive CD4+ T cells, but were more pronounced at 2 weeks of culture, especially after anti-CD3/anti-CD28 restimulation (Figure 5c).
Distinct regulation of IL-17 and IL-22.
IL-22 is a member of the IL-10 family, and its gene is located on chromosome 12q15 between the IFNγ and IL-26 loci (40, 41). Recently, IL-22 has been reported to be produced by murine Th17 cells in vitro and in vivo (12, 42). We sought to determine whether the expression of IL-22 in human CD4+ T cells was regulated in the same manner as that of IL-17A and IL-17F. Again, we found that freshly isolated naive CD4+ T cells did not express detectable levels of IL-22 mRNA. After TCR stimulation, IL-22 production was significantly up-regulated, and the induction was much greater in memory than in naive CD4+ T cells (Figure 6a). Interestingly, IL-23 greatly enhanced IL-22 production, especially in memory CD4+ T cells, contrasting sharply with its minimal effects on IL-17 production (Figure 4d). Addition of IL-23 to cultures of naive CD4+ T cells activated with anti-CD3 and anti-CD28 also resulted in enhanced IL-22 production (Figures 6c and d). In contrast, “Th17 conditions” inhibited IL-22 production (Figure 6b). Conversely, previous studies have shown that IL-22 is produced by Th1 cells (43), and we confirmed the effects of addition of exogenous IL-12 in the experiment shown in Figure 6c. Compared with anti-CD3/anti-CD28 alone, IL-12 enhanced IL-22 production, whereas optimal “Th17 conditions” attenuated IL-22 production (Figures 6b and c). However, IL-23 was clearly a more potent inducer of IL-22 (Figure 6c).
Although the cytokine IL-17 was discovered more than 10 years ago, the existence of a new lineage of T helper cells (Th17 cells) that selectively produce IL-17 was only recently recognized (26–28). More recently, it has been argued that Th17 cells are also major producers of IL-22 (12, 42). The current view of mouse Th17 development is that TGFβ1 and IL-6 are the key cytokines that initiate the differentiation process and that IL-23 plays an essential role in the expansion and maintenance of this lineage. Support for the notion of an IL-23/IL-17 axis has been provided by a number of models of autoimmunity (6, 8–11, 26, 39, 44–46), but despite these compelling data in mice, a systematic analysis of human Th17 differentiation from naive cells has not yet been undertaken.
In the present study, we showed that regulation of IL-17 and IL-22 in human CD4+ T cells is surprisingly different from that in mouse T cells. In the mouse, TCR stimulation with a costimulatory signal alone is not sufficient to induce IL-17, especially in naive CD4+ T cells. In contrast, in human CD4+ T cells, TCR stimulation of memory cells, and to a lesser extent naive cells, is sufficient. Although IL-17–producing T cells are readily detectable among human memory CD4+ T cells, the cytokine cocktail that effectively promotes murine Th17 differentiation (TGFβ1 and IL-6) was completely ineffective in driving Th17 differentiation of naive human CD4+ T cells. Moreover, IL-23, a cytokine that does not effectively induce murine Th17 differentiation, was effective in promoting IL-17 production in human T cells. One aspect of this regulation is the ability of IL-23 to induce its own receptor, which is analogous to other cytokines, such as IL-2, IL-4, and IL-12, which also regulate the expression of their cognate receptors.
While IL-23 is an inducer of IL-17, IL-23 activates both STAT-3 and STAT-4 and is a potent inducer of IFNγ and IL-22 in human CD4+ T cells. In fact, simultaneous IL-17 and IFNγ production was readily evident in freshly isolated, anti-CD3/anti-CD28–activated human memory CD4+ T cells. Furthermore, in conditions that favored Th1 and Th2 differentiation, IL-17 production was not extinguished as is the case in mouse cells. Thus, while IL-23 is a regulator of Th17 differentiation, there appears to be more flexibility and less evidence of “lineage commitment” for selective cytokine secretion in human CD4+ T cells.
Human memory T cells clearly maintain the capacity to promptly generate IL-17; one would expect this locus to remain accessible in memory CD4+ T cells—even under circumstances that favor IFNγ and IL-4 production. Remodeling of the IL-17 locus in memory cells might also allow for efficient production of this cytokine—this possibility and others will need to be examined in the future. Additionally, we consistently observed the coexistence of IFNγ+,IL-17+ and IFNγ−,IL-17+ cells. In this regard, it is notable that IFNγ+,IL-17+ T cells are also evident in the mouse. Whether these 2 populations represent separate subsets of T helper cells or different stages of lineage polarization is also an important topic for future investigation. Although cytokine loci of opposite lineage are typically silenced in fully polarized murine Th1, Th2, and Th17 cells, this is less evident in human CD4+ cells. Investigating the basis of these differences will be an interesting area of further study as we learn more about the transcriptional and epigenetic regulation of cytokine genes.
Stimulation of human T cells with anti-CD3/anti-CD28 and IL-23 was associated with increased expression of the transcription factor RORγt, a key transcription factor controlling mouse Th17 differentiation (34). However, the expression of RORγt and IL-17 are not as tightly linked as they are in mouse cells, suggesting that IL-17 expression is likely regulated by other transcription factors. Kinetic detection of RORγt protein expression in developing Th17 cells may give some insight into the regulation of IL-17 expression, but the field is hampered by a lack of optimal reagents. Clearly, in human CD4+ T cells, TCR occupancy is very important for IL-17 induction. Accordingly, the proximal promoter of the human IL17A gene contains 2 binding sites for nuclear factor of activated T cells (NF-AT), which appear to be important in the regulation of IL-17 (35). However, our data suggest that factors other than RORγt and NF-AT may be important contributors.
The discordant expression of IL-17 and RORγt is also of interest when viewed in the context of the regulation of FoxP3 in human T cells. In murine T cells, Foxp3 expression is tightly controlled and correlates well with the suppressive activity of Treg cells. However, FoxP3 is more widely expressed in activated human T cells, and its expression is not necessarily indicative of a population of cells with suppressive activity (47). Curiously, FoxP3+,IL-17+ cells were present when T cells were stimulated with TGFβ1 and IL-6; the exact functional significance of such FoxP3+,IL-17+ cells remains to be determined.
Additionally, the mechanisms underlying the distinct regulation of IL-17 and IL-22 are not clear. IL-23–dependent IL-22 production is impaired in STAT-3–deficient T cells, and this is likely due to impaired IL-23R expression (data not shown). However, in preliminary experiments, IL-12–dependent IL-22 production was not impaired in STAT-3–deficient mice, presumably because of the importance of STAT-4.
Recent studies have supported the notion of a role of IL-23 in autoimmune responses, and a preponderance of evidence indicates that this cytokine may be more clinically relevant than its close relative IL-12 in causing autoimmunity (8–11, 45). The recent discovery of the relationship between IL-23R single-nucleotide polymorphisms and the prevalence of inflammatory bowel disease highlights the relevance of this cytokine in human autoimmune disease (48). The present data support the importance of an IL-23/IL-17 axis in human CD4+ T cells, consistent with what is seen in murine systems, although IL-23 seems to play a subtly different role. Since IL-23R expression is low in naive T cells, there are likely to be other unidentified cytokines or signals that act in conjunction with IL-23 to regulate IL-17 expression. However, in human CD4+ T cells, IL-23 is also an extremely potent inducer of IL-22, IFNγ, and TNF, cytokines that all participate in autoimmune diseases (12, 49, 50). Therefore, the concept that a major driver of autoimmune diseases can be explained solely by a discrete IL-23/IL-17 connection is likely an oversimplification in humans. The actions of IL-23 are clearly of interest, but this cytokine apparently has very diverse effects on human CD4+ T cells.
Dr. Chen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Chen, Tato, O'Shea.
Acquisition of data. Chen, Muul, Laurence.
Analysis and interpretation of data. Chen, O'Shea.
Manuscript preparation. Chen, Tato, O'Shea.
Statistical analysis. Chen.
We are grateful to Dr. Xavier Valencia for valuable comments and providing reagents. We thank Dr. Michael Ward and Kevin Elias for assistance with statistical analysis. We thank Drs. Tom Nutman, Wendy Watford, and Yuka Kanno for critically reading the manuscript.