The basic leucine zipper transcription factor NFIL3 directs the development of a common innate lymphoid cell precursor

1) Since Nfil3-deficient mice appear to have lymph nodes (e.g., Spits and Di Santo, 2011) and Peyer's patches albeit fewer and smaller (present manuscript), it is arguable that LTi cell function is largely preserved during the fetal period in them. Also, NKp46- ILC3 are only mildly reduced in Nfil3-deficient mice (Figure 1). On the other hand, the authors show convincing data for the deficiency of the CHILP compartment, shown to be a common progenitor for helper-like ILCs including LTi cells, in the absence of Nfil3. Is it possible, for example, that Nfil3 is required for the maintenance of ILC progenitors in adult mice but not as strictly required for ILC fate decisions in the fetus? There is also literature suggesting Nfil3-indepedent generation of some ILC subsets; salivary gland NK cells (Cortez et al., 2014), cNK cells after MCMV infection (Firth et al., 2013), as also the mild reduction in NKP cells in Nfil3-deficient mice (Male et al., 2014). The discussion would benefit from acknowledgment and attempted resolution of these issues.

We thank the reviewers for raising this issue and offering insightful suggestions for amplifying our discussion. We have added an additional paragraph to the Discussion section that deals with these points.

2) A fair proportion of the molecular connections between Nfil3 and Tox have been derived from studies with the EL4 lymphoma cell line. While it is appreciated that performing many of these analyses in CLP cells is not feasible, the manuscript needs to acknowledge that EL4 cells are relatively unrelated to any of the cell types being considered and therefore there may be caveats for the interpretations based on data from them.

We agree that it is important to discuss this point and have now added the following sentences to the Results section:

“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.”

3) In Figure 4, it is unclear how the authors excluded contamination of the aLP by CHILP, which might skew the obtained results.

We cannot exclude contamination of the total αLP population by CHILP, and in fact, it is likely that CHILP cells are present in the αLP population. However, because CHILP do not give rise to cNK cells (Klose et al., Cell vol. 157, p. 340 [2014]), our data identify a distinct CXCR6⁺ αLP committed precursor that gives rise to all four ILC lineages including cNK cells (revised Figure 4D). This precursor must therefore lie developmentally upstream of the CHILP. Additionally, in support of our in vitro findings, we now include data (new Figure 4E) showing that the CXCR6⁺ αLP cells also give rise to all four ILC lineages (including cNK cells) in vivo.

4) The authors refer to literature (Male et al., 2014) for the impact of NFIL3 on NKP cells, but that study found only a small reduction of NKP cells in Nfil3-deficient mice. It would therefore be helpful if the authors could present data for NKP cells in Nfil3-deficient mice in their hands.

Figure 1B of Male et al. (2014) actually shows a marked decline in NKP frequencies and absolute numbers in Nfil3^-/- mice. We note that the data are shown in log scale, compressing the visual differences. The percentages of both preNKP and rNKP cells among Lin- bone marrow cells declines by ∼8-10-fold in Nfil3^-/- mice. Likewise, there is an approximately 10-fold decline in the absolute numbers of both preNKP and rNKP cells in Nfil3^-/- mice. Therefore, the Male et al. paper strongly supports the idea that NFIL3 is essential for the differentiation of NKP cells.

5) Figure 2A shows that Nfil3 deficiency results in a 90% reduction in total aLP cells. Nonetheless, T and B cell development is not affected. It would be very useful to show if the CXCR6+ aLP subset specifically and substantially depleted in Nfil3-deficient mice, or if depletion occurs in the CXCR6- aLP cells as well.

We have now included an analysis of CXCR6⁺ αLP cells in wild-type and Nfil3^-/- mice. The percentages of CXCR6⁺ cells among αLP are comparable between wild-type and Nfil3^-/- mice (Figure 4A; Figure 4-figure supplement 1). As a result, the number of CXCR6⁺ αLP cells is markedly diminished in Nfil3^-/- mice (Figure 4B) in parallel with the decrease in total αLP cells (Figure 2A). These data support the idea that CXCR6⁺αLPs require NFIL3 for their development.

Since αLP cells do not give rise to B cells (Figure 3D), it is not surprising that B cell development is not impaired in Nfil3-/- mice despite the marked αLP deficiency. Though αLP cells retain some T cell differentiation potential (Figure 3D and Figure 4C), it is not clear to what extent αLP-derived T cells contribute to the total T cell pool in mice. There is evidence that LSK cells, which are present in normal numbers in Nfil3-/- mice, can enter the circulation and seed the thymus to initiate T cell development (Reviewed in Bhandoola and Sambandam, From stem cell to T cell: one route or many? Nat Rev Immunol. vol.6(2), p. 117-26[2006].) Thus, the marked αLP deficiency in Nfil3^-/- mice is consistent with the relatively normal T cell numbers in these mice.

6) In Figure 5F, the ILC profile from the Nfil3-deficient CD45.2+ hCD2- donor cells (without TOX overexpression) would be an important and interesting control.

The TOX encoding plasmid was delivered into LSK cells by retroviral transduction/infection. It is well-documented that retroviral infection of bone marrow progenitors can by itself impact their ability to differentiate (reviewed in Banerjee et al., Hematopoietic stem cells and retroviral infection. Retrovirology vol. 7, p. 8-17 [2010]). Therefore, comparison of hCD2⁺ and hCD2^- donor cells will not provide interpretable information about the impact of TOX expression on the differentiation potential of the cells. For this reason, we have restricted our analysis to a comparison between Tox-expressing hCD2⁺ donor cells and vector-only hCD2⁺ donor cells.

7) Since the authors later show that only the CXCR6+ subset of aLP have the potential to give rise to all ILC lineages, it would be helpful if they could show the ILC transcription factor profile of these cells. Do they express any Id2? Do they express RORgt, since Possot et al. (2011) showed that 40% of the CXCR6+ aLP in fetal liver were RORgt+?

This is a great suggestion. However, the number of CXCR6⁺ αLP cells in adult mouse bone marrow is exceptionally small, perhaps due to the fact that these cells represent a transitional stage between αLP and CXCR6⁺ α4β7^- cells that lie downstream of αLP (Possot et al., 2011). The small number of cells makes such an analysis exceptionally challenging. As shown in revised Figure 4B, we detect ∼5 CXCR6⁺ αLP cells per 3.3 million Lineage-negative bone marrow cells. As there are, on average, approximately 12 million Lineage-negative bone marrow cells per femur/tibia, this equates to ∼40 CXCR6⁺ αLP cells per femur/tibia. The rarity of these cells makes it extremely difficult to accurately assess transcription factor expression across the population in adult mouse bone marrow by flow cytometry. The analysis in Possot et al. (2011) was performed on fetal liver CXCR6⁺ αLP cells, which account for roughly 0.15% of Linage-negative fetal liver cells and are thus much more abundant than in adult bone marrow. Finally, Possot et al. used quantitative RT-PCR to show that Rorc (encoding RORγt) expression is undetectable in CXCR6⁺ cells from adult bone marrow.

8) The developmental rescue of NFIL3-/- precursors by TOX is of interest, but its quantitative significance would be best interpreted in comparison with the efficiency of rescue of these Nfil3-/- precursors with Nfil3 itself.

This is a great suggestion and we have now included data showing the efficiency of developmental rescue of Nfil3^-/- precursors with NFIL3 in Figure 5F. The results show that developmental rescue is comparable between TOX and NFIL3, with no statistically-significant differences between the two across the various ILC subsets. Although the average rescue efficiency for the ILC2 subset trended higher with NFIL3, the overall difference was not statistically significant.