αLP differentiate into all ILC lineages

It is thought that ILCs differentiate from a common ILC progenitor population (Spits et al., 2013; Tanriver and Diefenbach, 2014). Prior studies have identified progenitor populations that develop into most, but not all, subtypes of known ILC lineages (Hoyler et al., 2012; Constantinides et al., 2014; Klose et al., 2014). The CHILP, identified through Id2 lineage tracing studies, can differentiate into non-NK ILC1, ILC2 and NK1.1⁺ NKp46⁺ ILC3 but not cNK cells (Klose et al., 2014), and the PLZF-dependent ILCP gives rise to all ILCs except cNK cells (Constantinides et al., 2014). These findings accord with the partial ILC deficiencies seen in mice lacking Id2 and Zbtb16 (encoding PLZF) (Boos et al., 2007; Savage et al., 2008). In particular, cNK cell development is not impaired in Zbtb16^−/− mice, while Id2^−/− mice show cNK developmental defects only during NK maturation (Boos et al., 2007). Similarly, ILC2Ps are lineage-specified progenitors of ILC2s with no appreciable potential to differentiate into cNK cells or ILC3 (Hoyler et al., 2012). The broad ILC deficiency (including cNK cells) and impaired ILC precursor development in Nfil3^−/− mice thus suggested that the NFIL3 might be required for the generation of a common ILC progenitor that lies developmentally upstream of the previously identified ILC precursors. We therefore sought to identify NFIL3-dependent precursor populations that differentiate into all ILC lineages.

In contrast to CHILP, ILCP, and ILC2P, αLP cells can differentiate into both cNK cells and ILC3 (Yoshida et al., 2001; Possot et al., 2011) and thus likely represent an earlier stage of ILC development. This idea is supported by expression profiles of key transcription factors known to be involved in ILC development (Figure 2C). Similar to CLP and CHILP, αLPs do not express transcription factors that specify ILC lineages, such as RORγt, GATA3, T-BET and EOMES, suggesting an undifferentiated phenotype. However, in contrast to CHILPs, which uniformly express high levels of ID2, only a small fraction of αLPs are ID2⁺ (Figure 2C). The majority of αLPs express ID2 at levels that are markedly lower than those in CHILPs. Because ID2 is virtually undetectable in CLPs, this suggests that αLPs may represent a transitional stage between CLP and CHILP. This is further supported by the fact that αLPs do not express PLZF while a major fraction of CHILPs express PLZF (Klose et al., 2014), which defines another group of ILC precursors (ILCP) that lack cNK cell differentiation potential (Constantinides et al., 2014).

To determine whether αLP can also give rise to ILC2, we co-cultured purified αLP with bone marrow stromal OP9 cells (OP9-GFP) or OP9 cells expressing the Notch ligand Delta-like 1 (OP9-DL1), which support ILC differentiation in vitro (Holmes and Zúñiga-Pflücker, 2009; Possot et al., 2011; Hoyler et al., 2012). When co-cultured with OP9-DL1 cells in the presence of ILC2-inducing cytokines, αLPs readily developed into ILC2 as the majority of progeny cells expressed ILC2 markers (GATA3⁺ Sca1⁺) (Figure 3A). When OP9-GFP cells (not expressing Notch ligand) were used in this assay, only a small fraction of progeny cells became ILC2 (Figure 3A), confirming that Notch signaling is important for ILC2 differentiation in vitro (Wong et al., 2012; Yang et al., 2013). In agreement with a prior study (Possot et al., 2011), αLP differentiated into ILC3 and RORγt⁻ NK1.1⁺ cells under ILC3-inducing conditions (Figure 3A).

Open in a separate window

Figure 3.

αLPs can differentiate into ILC2 in vitro and in vivo, and can thus give rise to all known ILC lineages.

(A) αLPs can differentiate into ILC2 in vitro. αLPs were purified by FACS and ~25 cells were co-cultured with a bone marrow stromal cell line OP9 (OP9-GFP) or OP9 cells stably expressing the Notch ligand Delta-like 1 (OP9-DL1) for 14 days in the presence of ILC2-inducing (IL-2) or ILC3-inducing (IL-23) cytokines. Cells were then stained and analyzed by flow cytometry. ILC2 cells were identified as CD45⁺ CD3ε⁻ CD19^- GATA3⁺ Sca1⁺, ILC3 as RORγt⁺ NK1.1^-, and ILC1 (including NK and non-NK ILC1) as CD45⁺ CD3ε⁻ CD19^- RORγt⁻ NK1.1⁺. Typical flow plots are shown on the left and combined data are shown on the right. (B and C) αLPs can differentiate into ILC2, ILC3, cNK, and non-NK ILC1 in vivo. αLP cells were purified from wild-type (CD45.1⁺) mice and ~1000 αLP cells were transplanted into sublethally irradiated Rag2^−/−;Il2rg^−/− (CD45.2⁺) mice. ILCs in the small intestine and colon (B) or liver (C) were examined 4–6 weeks later. (D) αLPs failed to differentiate into B cells both in the small intestine and spleen. There were small numbers of T cells in both the small intestine and spleen. Statistical analysis was performed with two-tailed student's t-test. Means ± SEM are shown. N.D., not detected; *, p < 0.05; ***, p < 0.001.

DOI: http://dx.doi.org/10.7554/eLife.04406.009

To assess the potential of αLPs to differentiate into ILC2 in vivo, we transferred ~1000 purified αLPs (CD45.1⁺) into sublethally irradiated Rag2^−/−;Il2rg^−/− mice (CD45.2⁺). After 5 weeks, ILC2 that had differentiated from engrafted αLPs were detected in small intestine and colon of the recipient mice (Figure 3B). We noted that GATA3⁺ ILC2 comprised a small fraction of CD127⁺ ILCs in the small intestine but were the majority in the colon while RORγt⁺ ILC3 showed the reverse tissue distribution pattern (Figure 3B). This suggests that tissue-specific microenvironment influences ILC development or recruitment. Consistent with the previously reported cNK cell differentiation potential of αLPs (Possot et al., 2011), donor cells gave rise to cNK cells in the liver, and also differentiated into non-NK ILC1 (Figure 3C). Differentiation of ILCs from αLPs was not caused by contamination of αLPs with CLPs, as no donor-derived B cells were detected in the spleen and small intestine of recipient mice (Figure 3D). This accords with the loss of B cell differentiation potential in αLPs (Yoshida et al., 2001; Possot et al., 2011). However, there were small numbers of donor-derived T cells, consistent with prior findings that αLP retain some T cell differentiation potential (Possot et al., 2011). Thus, αLPs can give rise to all known ILC lineages in vitro and in vivo. Given the more restricted differentiation potential of CHILP, ILCP, and ILC2P, αLPs are therefore likely to be developmentally upstream of these progenitors, and defective αLP development in Nfil3^−/− mice can thus explain the general ILC deficiency in these mice.

CXCR6⁺ αLP cells differentiate into all ILC lineages but not T- and B-cells

The residual T cell differentiation potential of αLPs suggested that this population includes cells that are not fully committed ILC precursors. Prior studies have shown that when αLP cells acquire CXCR6 expression, they continue to give rise to cNK cells and ILC3 but lose T cell differentiation potential (Possot et al., 2011). CXCR6⁺ cells comprised 3–4% of the αLP population in the bone marrows of adult wild-type and Nfil3^−/− mice (Figure 4A, Figure 4—figure supplement 1). Their absolute numbers were diminished in Nfil3^−/− mice (Figure 4B), in parallel with the decrease in total αLP numbers (Figure 2A). We therefore hypothesized that CXCR6⁺ αLPs might include NFIL3-dependent committed ILC precursors that give rise to all ILC lineages. To assess the developmental potential of CXCR6⁺ αLP cells, we isolated cells by flow cytometry and cultured individual cells with OP9-DL1 feeder cells in the presence of non-polarizing SCF and IL-7 (Figure 4C,D). In contrast to CXCR6^- αLP cells, which retained the ability to differentiate into T cells, CXCR6⁺ αLP cells failed to give rise to T cells in any of the clones examined (Figure 4C).

Open in a separate window

Figure 4.

CXCR6⁺ αLP cells are ILC-committed precursors.

(A) αLP can be divided into two subpopulations based on CXCR6 expression. (B) Enumeration of CXCR6⁺ cells in Nfil3^−/− mice. Bone marrow progenitor cells were enriched by negative selection with Lin⁻ cell counts ranging from 8–12.5 × 10⁶ per mouse. Lin⁻ cells were then stained and analyzed by flow cytometry. The numbers of CXCR6⁺ αLP cells in 3.3 × 10⁶ Lin⁻ cells are plotted. (C) CXCR6⁺ αLP cells lack T cell differentiation potential. CXCR6^- and CXCR6⁺ αLP cells were individually sorted into the wells of a 96-well plate with an irradiated OP9-DL1 feeder cell monolayer. Cells were cultured in the presence of 20 ng/ml SCF and 20 ng/ml IL-7 for 3 weeks. T cells were detected by CD3ε staining. Data are shown as the percentages of CD45⁺ cell-containing wells in which T cells were detected. (D) CXCR6⁺ αLP cells are multipotent precursors to cNK cells, non-NK ILC1, ILC2 and ILC3 in vitro. Individual CXCR6⁺ αLP cells were sorted and cultured as above (240 cells in total, pooled from two independent experiments). ILCs were analyzed by flow cytometry. cNK cells were detected as CD45⁺ CD3ε⁻ CD19^- RORγt⁻ GATA3^- NK1.1⁺ T-BET⁺ EOMES⁺; non-NK ILC1 as CD45⁺ CD3ε⁻ CD19^- RORγt⁻ GATA3^- NK1.1⁺ T-BET⁺ EOMES⁻, ILC2 as CD45⁺ CD3ε⁻ CD19^- GATA3⁺; and ILC3 as CD45⁺ CD3ε⁻ CD19^- RORγt⁺. Each well is presented as a column, with detected ILC lineages highlighted in blue. (E) CXCR6⁺ αLP cells differentiate into cNK cells, non-NK ILC1, ILC2 and ILC3 in vivo. ~1000 CXCR6⁺ αLP cells were purified from 20 CD45.1⁺ mice by FACS sorting and were transplanted into sublethally irradiated Rag2^−/−;Il2rg^−/− (CD45.2⁺) mice. T cells and B cells in the blood and ILCs in the small intestine and liver were examined 4–6 weeks later. Data shown are representative of two independent experiments. Statistical analysis was performed with two-tailed student's t-test. Means ± SEM are shown. **, p < 0.01.

DOI: http://dx.doi.org/10.7554/eLife.04406.010

Figure 4—figure supplement 1.

Open in a separate window

Frequencies of CXCR6⁺ cells in wild-type and Nfil3^−/− αLP cell populations.

Bone marrow progenitors from wild-type and Nfil3^−/− mice were enriched by negative selection and stained to detect CXCR6⁺ αLP cells. The frequencies of CXCR6⁺ cells among αLP cells are plotted. Statistical analysis was performed by two-tailed student's t-test. Means ± SEM are shown. ns, not significant.

DOI: http://dx.doi.org/10.7554/eLife.04406.011

To determine if a single CXCR6⁺ αLP cell could give rise to all ILC lineages, we assessed the clonal differentiation potential of the CXCR6⁺ αLP cells in vitro. While no T cells were detected, mixtures of cNK, non-NK ILC1, ILC2, and ILC3 cells were present in the progeny populations of single CXCR6⁺ αLP cells (Figure 4C,D). Approximately 60% of wells with clonal growth contained multiple ILC lineages. In particular, 43.3% of the wells differentiated from individual CXCR6⁺ αLP cells contained two ILC lineages, and 11.7% three ILC lineages (Figure 4C). Importantly, 2.5% of wells contained all four ILC lineages, demonstrating that individual CXCR6⁺ αLP cells can differentiate into all known ILC lineages. In agreement with the in vitro differentiation data, CXCR6⁺ αLP cells differentiated into cNK, non-NK ILC1, ILC2 and ILC3, but not B cells or T cells, when transferred into sublethally-irradiated Rag2^−/−;Il2rg^−/− mice (Figure 4E). Thus, CXCR6⁺ αLP cells include committed ILC precursors that can differentiate into all major ILC lineages in vitro and in vivo.

NFIL3-dependent ILC development is mediated by Tox

To identify potential mechanisms underlying NFIL3-dependent ILC development, we isolated CLPs from wild-type and Nfil3^−/− mice and surveyed their transcriptomes by Illumina BeadArrays. Nfil3 expression was readily detected in CLPs (Figure 5A), which accords with previous reports (Geiger et al., 2014; Seillet et al., 2014b) and is consistent with the finding that NFIL3 regulates ILC development in a CLP-intrinsic manner (Figure 1C; Seillet et al., 2014a). However, there was no detectable expression in CLPs of other transcription factors that are known to govern ILC development (Figure 5A). These factors include Id2 (Hoyler et al., 2012; Male et al., 2014; Seillet et al., 2014b), Zbtb16 (Constantinides et al., 2014), Eomes (Male et al., 2014; Seillet et al., 2014b), Tcf7 (encoding TCF-1) (Yang et al., 2013), Rora (Wong et al., 2012), Rorc (Eberl and Littman, 2003; Sawa et al., 2010), Gata3 (Hoyler et al., 2012) and Tbx21 (Gordon et al., 2012; Rankin et al., 2013). In contrast, the high mobility group (HMG) transcriptional regulator Tox, which is known to regulate NK and ILC3 development (Aliahmad et al., 2010), was expressed at a detectable level in wild-type CLPs and was down-regulated in Nfil3^−/− CLPs (Figure 5A,B). This suggested that NFIL3 might regulate Tox expression in CLPs.

Open in a separate window

Figure 5.

NFIL3-dependent ILC development is mediated by Tox.

(A and B) Tox expression is lower in Nfil3^−/− CLPs than in wild-type cells. (A) Heatmap comparing expression levels of transcription factors in wild-type and Nfil3^−/− CLPs (left). Factors included Nfil3, Tox and other transcription factors that are known to be involved in ILC development. The absolute signal values and detection p values for each transcription factor in Illumina BeadArrays are also shown (right). Note that only Nfil3 and Tox expression can be reliably detected in wild-type CLP cells. (B) Q-PCR analysis of Tox expression in wild-type and Nfil3^−/− CLPs. (C–E) NFIL3 activates Tox expression by binding directly to the Tox promoter. (C) Tox expression was determined by Q-PCR following shRNA knockdown of NFIL3 (left), and NFIL3 overexpression (right) in EL4 cells. (D) ChIP analysis of EL4 cells using an NFIL3-specific antibody or IgG control. Tox promoter (nt −2105 to −1867) enrichment was calculated as the ratio of the NFIL3-specific antibody pull-down to the IgG control pull-down. The left panel shows results with endogenous NFIL3 levels and the right panel shows results with NFIL3 overexpression. (E) Luciferase reporter assay. A 2.8 kb fragment of the Tox promoter was cloned and fused with the firefly luciferase gene to generate a Tox-luciferase reporter. HEK293T cells were co-transfected with the reporter and an empty vector or an NFIL3-encoding vector. Luciferase activity was normalized to cells transfected with vector-only controls. (F) Restoring Tox expression in Nfil3^−/− progenitors rescues ILC development in vivo. Nfil3^−/− LSK cells (CD45.2⁺) were retrovirally transduced with either an empty vector (MSCV-IRES-hCD2), a TOX-encoding vector (MSCV-Tox-IRES-hCD2), or an NFIL3-encoding vector (MSCV-Nfil3-IRES-hCD2) and then transferred into lethally irradiated wild-type (CD45.1⁺) mice. ILCs were examined 5–6 weeks later. The frequencies of total ILC2 and ILC3 within CD45.2⁺ hCD2⁺ cells are shown. Statistical comparisons between groups were performed with two-tailed student's t-test (B–E), nonparametric one-way ANOVA test with posttests (F). Means ± SEM are shown. ns, not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

DOI: http://dx.doi.org/10.7554/eLife.04406.012

Figure 5—figure supplement 1.

Open in a separate window

Knockdown of Nfil3 by shRNA.

(A) shRNA constructs were screened for Nfil3 knockdown. HEK293T cells were co-transfected with an Nfil3-coding plasmid (CDS only) and shRNA constructs. NFIL3 protein levels were examined by Western blotting 36 hr later. A scrambled shRNA was used as a negative control and ACTIN was used as a loading control during Western blotting. (B) Nfil3 was knocked down by shRNAs, sh39 and sh40, in EL4 cells. HEK293T cells were co-transfected with shRNA constructs, pVSVG and pCMVDR9. Cell culture supernatants were harvested 48 hr later and lentiviral particles were concentrated by ultracentrifugation. EL4 cells were transduced with shRNA-encoding lentivirus by spinoculation in the presence of 4 μg/ml polybrene. Cells stably expressing shRNAs were selected with 8 μg/ml puromycin and live cells (propidium iodide-negative) were purified by FACS. Nfil3 mRNA levels were examined by Q-PCR with Gapdh as an internal control. Statistical analysis was performed with one-way ANOVA with post-tests. Means ± SEM are shown. ****, p < 0.0001.

DOI: http://dx.doi.org/10.7554/eLife.04406.013

Figure 5—figure supplement 2.

Open in a separate window

Experimental design and gating strategy for the Tox rescue experiment.

(A) Schematic illustrating the experimental design. Nfil3^−/− LSK (CD45.2⁺) cells were retrovirally transduced with empty (MSCV-IRES-hCD2), TOX-encoding (MSCV-Tox-IRES-hCD2) or NFIL3-encoding (MSCV-Nfil3-IRES-hCD2) vectors. Cells were then transplanted into lethally irradiated wild-type mice (CD45.1⁺) and ILCs were examined 4–6 weeks later. (B) Gating strategy for examining ILCs in the recipient mice. Live cells were first electronically gated as ZombieGreen-negative and cells transduced by retrovirus were identified as CD45.2⁺ hCD2⁺. ILC2, ILC3, cNK and non-NK ILC1 were gated as Lineage (CD3, CD19, CD5, TCRβ, TCRγδ)⁻ CD127⁺ GATA3⁺, Lin⁻ CD127⁺ RORγt⁺, Lin⁻ NK1.1⁺ T-BET⁺ EOMES⁺ and Lin⁻ NK1.1⁺ T-BET⁺ EOMES⁻, respectively. The results of the experiment are summarized in Figure 5F.

DOI: http://dx.doi.org/10.7554/eLife.04406.014

CLPs are present in small numbers in adult mice (Klose et al., 2014), making it challenging to perform biochemical studies of Tox regulation by NFIL3 using these cells. As an alternative, we found that NFIL3 regulates Tox expression in EL4 cells, a mouse lymphoma cell line (Figure 5C). Knockdown of NFIL3 in EL4 cells with two independent shRNA constructs led to dose-dependent down-regulation of Tox expression (Figure 5C; Figure 5—figure supplement 1). Conversely, overexpression of NFIL3 in EL4 cells increased Tox expression (Figure 5C), indicating that Tox expression is sensitive to NFIL3 levels in EL4 cells in a manner similar to CLPs. A chromatin immunoprecipitation (ChIP) assay with an NFIL3-specific antibody (Yu et al., 2013) demonstrated that NFIL3 directly bound to the Tox promoter (nt −2105 to −1867) and that overexpression of NFIL3 enhanced this binding (Figure 5D). Finally, NFIL3 activated Tox promoter activity as assessed by a luciferase reporter assay (Figure 5E). Thus, NFIL3 activates Tox expression by directly binding to its promoter. EL4 cells are derived from T lymphocytes, a CLP-derived lineage, and thus we cannot exclude the possibility that the regulatory relationship between Nfil3 and Tox differs between T lymphocytes and CLPs. Nevertheless, our studies on CLPs and EL4 cells both support the idea that NFIL3 is an activator of Tox expression.

Because Tox is known to be essential for cNK and ILC3 development (Aliahmad et al., 2010), we postulated that lowered Tox expression leads to the broad ILC deficiency in Nfil3^−/− mice and that restoring Tox expression would rescue ILC development. To test this idea, we cloned Tox coding sequences into a bicistronic vector (MSCV-IRES-hCD2), which allowed expression of the native form of TOX and also marked cells with the cell surface marker hCD2. We then delivered the TOX-encoding plasmid or the empty vector into purified Nfil3^−/− LSK cells (CD45.2⁺) by retroviral transduction (Zheng et al., 2012; Spencer et al., 2014), followed by transfer of these cells into lethally irradiated wild-type mice (CD45.1⁺) (Figure 5—figure supplement 2). Compared to the empty vector control, transduction of the TOX-encoding plasmid led to increased numbers of cNK cells in spleen, non-NK ILC1 cells in liver, and ILC2 and ILC3 in the small intestines of recipient mice (Figure 5F). We observed that rescue of ILC development from Nfil3^−/− LSK cells by Tox was largely comparable to rescue by Nfil3 in the same setting, supporting the idea that Tox acts downstream of Nfil3 in ILC development. Though ILC2 cells developing from Tox-rescued LSK cells were generally fewer than those from Nfil3-rescued LSK cells, the difference between the two groups was not statistically significant. Thus, ILC development is rescued by restoring Tox expression in Nfil3^−/− progenitors, indicating that NFIL3 drives ILC development in part by regulating Tox expression.

Isolation and analysis of intestinal lamina propria lymphocytes

Lamina propria lymphocytes (LPLs) were isolated from the intestine as previously described (Yu et al., 2013). Briefly, intestines were dissected from mice and Peyer's patches were removed. Intestines were cut into small pieces and thoroughly washed with ice-cold PBS. Epithelial cells were removed by incubating intestinal tissues in Hank's buffered salt solution (HBSS) supplemented with EDTA and DTT, followed by extensive washing with PBS. Residual tissues were digested by Collagenase IV (Sigma, St. Louis, Missouri), DNase I (Sigma) and Dispase (BD Biosciences, San Jose, California) for 1 hr at 37°C. Cells were filtered through 100 μm cell strainers and applied onto a 40%:80% Percoll gradient (GE Healthcare, Pittsburgh, Pennsylvania), in which lamina propria lymphocytes were found at the interface of 40% and 80% fractions.

Livers were dissected from mice and cut into small pieces, followed by digestion with Collagenase IV (Sigma), DNase I (Sigma) and Dispase (BD Biosciences) for 1 hr at 37°C. Residual tissues were forced through 100 μm cell strainers. Cells were spun down and applied onto a 40%:80% Percoll gradient as for the LPLs.

Isolated lymphocytes were washed with PBS with 2 mM EDTA and 3% fetal bovine serum (FBS) and F_c receptors were blocked with α-CD16/32 (2.4G2). Cells were then stained with antibodies against cell surface markers including α-CD3ε (500A2), α-CD19 (ebio1D3), α-CD5 (53-7.3), α-TCRβ (H57-597), α-TCRγδ (GL3), α-NK1.1 (PK136), α-Sca1 (D7), α-KLRG1 (2F1), α-NKp46 (29A1.4), α-CD45 (30-F11), α-CD45.1 (A20), α-CD45.2 (104), α-hCD2 (RPA-2.10), and α-CD127 (A7R34). Cells were fixed/permeabilized with eBiosciences (San Diego, California) Mouse Regulatory T Cell Staining Kit #3 per the manufacturer's instructions, and subjected to nuclear staining with α-RORγ (AFKJS-9), α-GATA3 (TWAJ), α-T-BET (4B10) and α-EOMES (Dan11mag). Cells were analyzed with an LSRII (BD Biosciences, San Jose, California) or CyAn ADP (Beckman Coulter, Jersey City, New Jersey) flow cytometer and data were processed with FlowJo software (Tree Star, Ashland, Oregon).

Isolation and analysis of bone marrow progenitors

Femur and tibia were dissected from adult mice and bone marrow cells were released in PBS buffer containing 2 mM EDTA and 3% FBS with a mortar and pestle. Cells were filtered through 70 μm cell strainers and blocked with α-CD16/32 (2.4G2), followed by incubation with biotinylated lineage markers (Lin) antibodies including α-CD3ε (145-2C11), α-B220 (RA3-6B2), α-CD11b (M1/70), α-Gr1 (RB6-8C5), α-Erythroid Cells (TER119), α-CD5 (53-7.3), α-TCRγδ (GL3), and α-NK1.1 (PK136). Cells were then washed and incubated with α-biotin magnetic microbeads (Miltenyi Biotec, San Diego, California). Lineage-negative cells were enriched by an autoMACS sorter with the ‘Depletes’ setting. Surface staining was performed with antibodies including α-biotin (Bio3-18E7), α-CD45 (30-F11), α-cKit (2B8), α-CD127 (A7R34), α-Sca1 (D7), α-Flt3 (A2F10) and α-α₄β₇ integrin (DATK32) and α-CXCR6 (221002). FACS sorting was performed with a FACSAria cell sorter (BD Biosciences) while flow cytometry analysis was carried out with an LSRII (BD Biosciences). In both cases, LSK cells were identified as Lin⁻ Sca1⁺ cKit⁺, CLP as Lin⁻ cKit^low CD127⁺ Sca1^low Flt3⁺ α₄β₇⁻, αLP as Lin⁻ cKit^low CD127⁺ Sca1^low Flt3^- α₄β₇⁺, ILC2P as Lin⁻ CD127⁺ α₄β₇⁺ CD25⁺ Sca1⁺, and CHILPs as Lin⁻ CD127⁺ α₄β₇⁺ CD25^- Flt3^-. Data were processed with FlowJo software (Tree Star).

Cell transfer assay

αLPs were purified from CD45.1⁺ mice by FACS sorting as described above. In order to obtain a large number (~1000) CXCR6⁺ αLP, femur and tibia from 20 CD45.1⁺ mice were pooled together for the cell isolation. Rag2^−/−;Il2rg^−/− recipient mice were sublethally irradiated with a dose of 4.2 Gy on the same day with an XRAD320 irradiator (Precision X-ray, Inc, North Branford, Connecticut). Cells were transplanted into recipient mice by retro-orbital injection. ILCs in recipient mice were examined 4–6 weeks later.

For CLP co–transfer experiments, wild-type and Nfil3^−/− CLP cells were purified by FACS sorting and mixed at a 1:1 ratio before transplantation into sublethally irradiated Rag2^−/−;Il2rg^−/− recipient mice.

For LSK co-transfer, CD90.1⁺ recipient mice were lethally irradiated with two doses of 5 Gy on the same day. FACS-purified LSK cells from wild-type and Nfil3^−/− mice were mixed at a 1:1 ratio and transplanted into recipient mice by retro-orbital injection.

In vitro differentiation assays

αLPs were purified by FACS sorting as described above. For bulk culture, ~25 cells were co-cultured on a monolayer of OP9 cells (OP9-GFP) or OP9 cells stably expressing the Notch ligand Delta-like 1 (OP9-DL1) in αMEM media. To induce ILC2, the culture medium was supplemented with 20 ng/ml Stem Cell Factor (SCF, PeproTech, Rocky Hill, New Jersey), 20 ng/ml IL-7 (BioLegend, San Diego, California) and 20 ng/ml IL-2 (BioLegend). To induce ILC3, 20 ng/ml SCF, IL-7 and IL-23 (BioLegend) were added to the medium. The culture medium was replaced every 3–4 days and, after 14 days, cells were stained and analyzed by flow cytometry. For clonal differentiation, OP9-DL1 cells were irradiated at 1500 rad and seeded at a density of 10,000 cells per well in a 96-well plate. On the following day, CXCR6⁺ αLP cells were individually sorted into the wells and cultured in αMEM media supplemented with 20 ng/ml SCF and IL-7. Cells were analyzed by flow cytometry 3 weeks later. In total, 240 cells from two independent experiments were analyzed.

CLP transcriptome analysis

CLPs from wild-type and Nfil3^−/− mice were purified as described above. Total RNA was isolated with the PicoPure RNA Isolation Kit (Life Technologies, Grand Island, New York). RNA quality and quantity were determined with a Bioanalyzer (Agilent Genomics) with a Pico chip. Samples with RNA integrity numbers (RIN) larger than 8 were subjected to further processing and hybridized to the Mouse WG-6 V2 BeadChips (Illumina, San Diego, California) by the UT Southwestern Microarray Core Facility.

Microarray images were processed and annotated with GenomeStudio (Illumina). Differential gene expression analysis was performed using R together with bioConductor and the Limma package (Smyth et al., 2005; Ritchie et al., 2011). Briefly, signal intensities were first log₂-transformed, followed by background correction and quantile normalization with the NEQC function. Empirical reliabilities of samples were estimated by the arrayWeights function, which gave each sample a weight score accordingly. Samples were then fitted into a weighted linear model by lmFit to detect differentially expressed genes.

Tox rescue experiment

TOX-coding or NFIL3-coding sequences (CDS) were cloned by PCR from total mouse thymus cDNA into the bicistronic retroviral vector MSCV-IRES-hCD2 (a gift from Dr Chandrashekhar Pasare at UT Southwestern) to generate a TOX-encoding plasmid, MSCV-Tox-IRES-hCD2 and an NFIL3-encoding plasmid, MSCV-Nfil3-IRES-hCD2. The MSCV-IRES-hCD2, MSCV-Tox-IRES-hCD2, and MSCV-Nfil3-IRES-hCD2 plasmids were transfected into the Plat-E packaging cell line (Morita et al., 2000) with FugeneHD (Promega, Madison, Wisconsin) to produce retroviral particles. Cell culture supernatant was harvested 48 and 72 hr post transfection. Cell debris was first cleared by spinning at 400×g for 10 min, followed by passage through 0.2 μm sterile filters.

LSK cells were purified from Nfil3^−/− mice by FACS sorting as described above and seeded into round-bottom 96-well plates at a density of 10,000 cells/well in STEMSPAN Serum-Free Expansion Medium (SFEM) (Stemcell Technologies, Vancouver, Canada) (Zheng et al., 2012). During retroviral transduction, cells were mixed with an equal volume of cleared retrovirus-containing cell culture supernatant, supplemented with 2 U/ml Heparin (Sigma), 10 ng/ml mouse Stem Cell Factor (SCF, Peprotech), 20 ng/ml mouse Thrombopoietin (TPO, Peprotech), 10 ng/ml mouse Fibroblast Growth Factor (FGF-1, Life Technologies) and 4 μg/ml polybrene (Sigma). Spinoculation was carried out at 1200×g for 90 min at 32°C to enhance retroviral transduction. 3 hr later, cell media was replaced with fresh STEMSPAN SFEM media supplemented with the above cytokines but without polybrene. Transduction was performed on two consecutive days using retroviral supernatant harvested 48 and 72 hr post transfection, respectively.

On day 3, CD45.1⁺ wild-type recipient mice were lethally irradiated at two doses of 5 Gy as described above. LSK cells were collected from the 96-well plate with Cell Dissociation Buffer (Life Technologies) and washed with sterile PBS. 2000–4000 cells were transferred into recipient mice in 200 μl sterile PBS by retro-orbital injection. ILCs in recipient mice were examined 5–6 weeks later.

shRNA knockdown of NFIL3

Five independent shRNA constructs (sh38-sh42) targeting mouse NFIL3 and a control construct containing scramble sequences (pLKO.1-scramble) were purchased from Sigma. To identify shRNA constructs that could effectively knock down NFIL3, 1 μg of shRNA plasmid and 1 μg of NFIL3-encoding plasmid (Yu et al., 2013) were co-transfected into HEK293T cells in a 6-well plate with FugeneHD (Promega). Cells were harvested 36 hr later, lysed and used for western blotting with anti-NFIL3 antibody.

Two shRNA constructs that could effectively knock down NFIL3 were identified: sh39 and sh40. These constructs as well as the pLKO.1-scramble vector were each co-transfected with the packaging plasmids pCMVDR9 and pVSVG into HEK293T cells. Cell culture supernatants were harvested 48 and 72 hr later and cleared by spinning and filtering as described above. Lentiviral particles were concentrated by ultracentrifugation at 75,000×g for 2 hr and resuspended in RPMI media.

EL4 cells were mixed with lentiviral particles in the presence of 4 μg/ml polybrene, and spinoculated at 1200×g for 90 min at 32°C. 2 days later, EL4 were selected with 8 μg/ml puromycin for 2 weeks. Live cells were sorted with a FACSAria cell sorter as they excluded propidium iodide.

NFIL3 overexpression in EL4 cells

NFIL3 coding sequences were subcloned into MSCV-IRES-hCD2 to generate MSCV-Nfil3-IRES-hCD2. MSCV-IRES-hCD2 and MSCV-Nfil3-IRES-hCD2 were then transfected into Plat-E cells to produce retroviral particles as described above, which were then used to transduce EL4 cells. 3 days after transduction, EL4 cells were stained with anti-hCD2 and hCD2⁺ EL4 cells were purified with a FACSAria cell sorter. Sorted cells were maintained in RPMI media for another 3–4 days, followed by staining and sorting again. The resulting cells were expanded and Nfil3 expression was examined by SYBR green-based real-time PCR.

Chromatin Immunoprecipitation (ChIP)

ChIP experiments were carried out as previously described (Yu et al., 2013). Briefly, EL4 cells or NFIL3-overexpressing EL4 cells were cultured in RPMI medium at ~0.8 × 10⁶ cells/ml. Cells were harvested and fixed with 1% formaldehyde for 10 min in the dark, which was quenched by adding glycine to a final concentration of 0.15 M. Nuclei were released with a Dounce homogenizer (Wheaton, Millville, New Jersey) in Nuclear Isolation Solution containing 10 mM Tris pH 7.4, 5 mM MgCl₂, 25 mM KCl and 250 mM sucrose, and purified by spinning at 1000×g for 10 min over Hypertonic Solution containing 10 mM Tris pH 7.4, 5 mM MgCl₂, 25 mM KCl and 30% (wt/vol) sucrose. Purified nuclei were used for ChIP with the Magna ChIP assay kit (Millipore, Billerica, Massachusetts) per the manufacturer's instructions. The Tox promoter was detected by SYBR green-based real-time PCR with specific primers: Tox-ChIPF6: 5′-GACACTGACAGCAAGGACCA-3′ and Tox-ChIPR6: 5′-CAGGGCTTCATAGCACCGAT-3′, targeting nucleotide −2105 to nucleotide −1867 in the Tox promoter. Enrichment of the Tox promoter was determined by normalizing the level of the Tox promoter in DNA pulled down with an anti-NFIL3 antibody to that pulled down with an IgG control.

Tox-luciferase reporter assay

A 2.3 kb fragment (−2133 to 232) of the Tox promoter was cloned into the pGL3-Basic vector to drive firefly luciferase expression (the Tox-luciferase reporter). HEK293T cells were cultured in a 96-well plate overnight and were co-transfected with the Tox-luciferase reporter and an empty or NFIL3-encoding vector. A pCMV-Renilla-Luciferase reporter was co-transfected into HEK293T cells to serve as an internal control. Luciferase activities were detected using the Dual-Glo Luciferase Assay kit (Promega) and measured with a SpectraMax M5e plate reader (Molecular Devices, Sunnyvale, California). Firefly luciferase activities in each sample were first normalized against Renilla luciferase activities in the same sample and then normalized against that in cells transfected with the empty vector.

We thank Cassie Behrendt Boyd and Tess Leal for assistance with mouse experiments. We thank Dr Nicolai Van Oers and Ashley Hoover for sharing mouse thymus total cDNA, and Dr Hua Wang for assistance with cell irradiation. We also thank Dr Juan Carlos Zúñiga-Pflücker for providing us with the OP9 cell lines, and Dr Paul Rothman for providing the Nfil3^−/− mice. This work was supported by NIH R01 DK070855 (LVH), a Burroughs Wellcome Foundation New Investigators in the Pathogenesis of Infectious Diseases Award (LVH), and the Howard Hughes Medical Institute (LVH).