Nfil3-independent lineage maintenance and antiviral response of natural killer cells

Nfil3 expression levels were evaluated by the ImmGen consortium microarray and confirmed by quantitative RT-PCR in sorted NK cell populations. At rest, NK cells express higher levels of Nfil3 mRNA than CD8⁺ and CD4⁺ T cells (Fig. 1, A and B). Nfil3 expression in NK cells modulates immediately after MCMV infection, suggesting activation-induced regulation of Nfil3 expression, but resting and memory NK cells exhibited comparable levels of Nfil3 (Fig. 1, A and B). We investigated whether elevated Nfil3 transcript (relative to T cell controls) correlated with survival and function in NK cells at various stages before and after viral infection.

Although NK cells are nearly undetectable in Nfil3^−/− mice at steady-state, MCMV infection elicited a profound increase in the overall percentage NK cells, of which the majority (>90%) expressed the activating receptor Ly49H (Fig. 1 C). The expansion of Nfil3^−/− NK cells amounted to an ∼30-fold increase in total cell numbers, in contrast to a roughly threefold increase in WT NK cells (Fig. 1 D). The MCMV-expanded Nfil3^−/− Ly49H⁺ NK cells were detected in both lymphoid and nonlymphoid tissues at day 7 PI (Fig. 1 E). Surface expression of Ly49 family activating receptors, as well as markers of activation and maturation, by WT and Nfil3^−/− NK cells at day 7 PI, indicated that MCMV-expanded Nfil3^−/− NK cells were phenotypically similar to WT NK cells (Fig. 1 F). NK cells from Nfil3^−/− mice expressed comparable levels of activating and inhibitory Ly49 receptors to WT mice, with slightly increased expression of activation markers CD11b and KLRG1, likely a reflection of the greater fold expansion of Nfil3^−/− NK cells compared with WT NK cells (Fig. 1 D). Thus, although Nfil3 is required for NK cell development, the virus-specific NK cells in Nfil3^−/− mice can be driven to expand after MCMV infection.

To address functionality, NK cells from MCMV-infected Nfil3^−/− and WT mice were assessed by ex vivo stimulation using plate-bound antibodies specific for activating NK cell receptors and in vitro killing assays. Ly49H⁺ NK cells from Nfil3^−/− and WT mice produced comparable amounts of IFN-γ, and degranulated similarly (measured by CD107a expression) in response to plate-bound antibodies, or PMA and ionomycin (Fig. 2, A and B). Nfil3^−/− NK cells were capable of mediating in vitro killing of YAC-1 targets similar to WT NK cells (Fig. 2 C). Additionally, MCMV-infected Nfil3^−/− mice were able to reject “missing-self” B2m^−/− splenocytes at day 7 PI (Fig. 2 D), demonstrating the presence of functional Nfil3^−/− NK cells in vivo. Altogether, these findings indicate that NK cells that proliferate in Nfil3^−/− mice in response to MCMV infection represent fully functional effector cells.

Given the selective expansion of Ly49H⁺ NK cells in Nfil3^−/− mice, we investigated whether the MCMV glycoprotein m157 was required for the expansion of Nfil3^−/− NK cells. To test this, we first infected mice with a strain of MCMV in which the m157 gene had been deleted (MCMVΔm157). Nfil3^−/− mice did not demonstrate an increase in the percentage of Ly49H⁺ NK cells after infection with MCMVΔm157, in contrast to infection with parental MCMV, suggesting that expansion of Nfil3^−/− Ly49H⁺ NK cells during MCMV infection depended on recognition of the m157 antigen (Fig. 3 A). To confirm this finding, we engineered a recombinant vesicular stomatitis virus expressing m157 (VSV-m157) and a vaccinia virus expressing m157 (VACV-m157). Nfil3^−/− mice infected with either of these recombinant viruses demonstrated an expansion of the Ly49H⁺ NK cell population that was not observed during infection with control recombinant viruses expressing OVA (Fig. 3, B and C). In all experiments, significant expansion of Ly49H⁺ NK cells was observed in lymphoid and nonlymphoid organs in Nfil3^−/− mice infected with m157-expressing viruses (Fig. 3 D and not depicted). Furthermore, only infection with m157-expressing viruses, but not with OVA-expressing viruses, induced a full maturation phenotype (KLRG1^hi CD62L^lo Ly6C^hi) of Nfil3^−/− NK cells (Fig. 3 E and not depicted). Thus, the presence of a cognate viral ligand is required to promote a robust proliferative program in NK cells lacking Nfil3.

In addition to antigen receptor triggering in T and NK cells, signaling via proinflammatory cytokines such as IL-12 and type I IFNs, generated in response to viral infection, have been shown to play an important role in lymphocyte proliferation and effector function (Sun and Lanier, 2011). To examine the influence of these cytokines during the expansion of Ly49H⁺ NK cells in Nfil3^−/− mice infected with MCMV, Nfil3^−/− mice were injected with blocking antibodies against IL-12, the IFN-α receptor (IFNαR1), or both before infection. By percentage, Ly49H⁺ NK cell expansion was significantly reduced in the spleen and liver of mice treated with blocking antibodies compared with control Nfil3^−/− mice treated with PBS (Fig. 4, A and B). The absolute numbers of Ly49H⁺ NK cells in antibody-treated Nfil3^−/− mice were reduced by >10-fold compared with control animals (Fig. 4, C and D). Reduced numbers were similarly detected in all lymphoid and nonlymphoid tissues of antibody-treated mice (unpublished data). Interestingly, disruption of either IL-12 or type-I IFN signaling results in a defect comparable to ablation of both cytokine pathways, suggesting an absolute requirement for each of these cytokines in mediating the NK cell expansion during MCMV infection.

MCMV infection generates a population of memory Ly49H⁺ NK cells that persists for several months after viral clearance (Sun et al., 2009a). To determine whether Nfil3 influences the formation and persistence of memory NK cells, we measured the number of Ly49H⁺ NK cells in Nfil3^−/− mice at various time points after MCMV infection. Ly49H⁺ NK cells were detected in the blood, liver, and spleen of Nfil3^−/− mice for >60 d after infection (Fig. 5 A and not depicted). Thus, the stages of NK cell expansion, contraction, and memory persistence after MCMV infection could be clearly observed in Nfil3^−/− mice (Fig. 5 B), similar to previous observations in WT mice (Sun et al., 2009a). Interestingly, although optimal numbers of effector and memory Nfil3^−/− NK cells were only observed after infection with MCMV, but not MCMVΔm157, a small population of NK cells appeared in Nfil3^−/− mice infected with MCMVΔm157 at day 14 PI (Fig. 5 B and not depicted), suggesting that ligand-independent signals, such as inflammation, are sufficient to drive a limited degree of NK cell expansion in Nfil3^−/− mice.

To confirm that the persisting NK cells in MCMV-infected Nfil3^−/− mice truly represent long-lived memory cells, and not newly generated cells egressing from the bone marrow at later time points, equal numbers of WT (CD45.1) and Nfil3^−/− (CD45.2) Ly49H⁺ NK cells from mice infected with MCMV (day 7 PI) were adoptively transferred into naive Ly49h^−/− hosts (Ly49h^−/− mice do not express the Ly49H receptor but have normal numbers of NK cells). Donor WT NK cells were derived from NK cells originally transferred to Ly49h^−/− host mice before MCMV infection, to compare NK cell populations that had undergone a similar fold expansion. The absence of CD69 up-regulation on the lymphocytes of recipient mice after adoptive transfer confirmed that virus had not been co-injected along with the NK cells into the naive Ly49h^−/− mice (unpublished data). WT and Nfil3^−/− Ly49H⁺ NK cells were detectable in peripheral blood at 12 h after transfer and were recovered at similar ratios 14 d later (corresponding to 21 d from initial virus exposure; Fig. 5 C), suggesting survival and maintenance of memory NK cell populations occur independent of Nfil3.

Our data demonstrates that NK cell effector and memory cell generation occurs independent of Nfil3. Because Nfil3 has been suggested to act downstream of IL-15 receptor signaling (Gascoyne et al., 2009), we hypothesized that memory NK cells may not have the same requirement for IL-15 as resting NK cells. To test this, resting NK cells (Fig. 6 A) or effector Ly49H⁺ NK cells isolated from MCMV-infected WT donor mice at day 7 PI (Fig. 6 B) were transferred into Il15^−/− or Ly49h^−/− mice. Whereas survival of both the transferred WT NK cell populations was observed in Ly49h^−/− mice, resting and effector NK cells transferred into the Il15^−/− hosts rapidly disappeared (Fig. 6, A and B). Similarly, memory NK cells (purified at day 21 PI) were also found to be dependent on IL-15, as cells transferred into Il15^−/− hosts disappeared within 3 d (Fig. 6 C). Thus, the maintenance of memory NK cells after MCMV infection appears to be an IL-15–dependent but Nfil3-independent process. These results indicate the existence of an Nfil3-independent signaling pathway that facilitates the long-term survival of IL-15–dependent NK cells.

To determine whether mature NK cells continued to require Nfil3 for lineage survival in the absence of infection or inflammation, we generated mice in which Nfil3 could be conditionally deleted in mature resting NK cells. We generated mice in which the Nfil3 gene is flanked by loxP sites (Nfil3^fl/fl; Motomura et al., 2011), a tamoxifen-inducible Cre cassette is driven by the Ubiquitin promoter (Ub^Cre-ERt2; Ruzankina et al., 2007), and the Rosa26 promoter drives a floxed “STOP” with YFP reporter cassette (R26Y; Srinivas et al., 2001). These mice (Nfil3^fl/fl × R26Y × Ub^CreERt2) allow for the inducible deletion of Nfil3 upon tamoxifen treatment, and Cre-recombinase activity monitored by tracking the YFP⁺ population (Fig. 7 A). To address the impact of Nfil3 ablation on the homeostasis of mature NK cells, NK cells from Nfil3^fl/fl × R26Y × Ub^CreERt2 mice (CD45.2) were adoptively transferred into WT recipient mice (CD45.1). Upon tamoxifen treatment, ablation of Nfil3 in both YFP⁺ and YFP⁻ NK cells was confirmed by PCR (Fig. 7 B), suggesting that YFP expression underreports Cre activity and excision of the floxed Nfil3 gene. Interestingly, transferred NK cells were readily detectable 25 d after tamoxifen treatment (Fig. 7 C). Thus, genetic deletion of Nfil3 in mature NK cells had no effect on homeostatic survival and maintenance.

We next investigated whether Nfil3 is required during the developmental stage when NK cell progenitors first acquire their activating receptors (e.g., NKp46). We crossed the recently developed Nkp46^iCre mouse, which expresses Cre-recombinase under control of the Ncr1 (NKp46) gene promoter (Narni-Mancinelli et al., 2011), to the Nfil3^fl/fl mice. Expression of NKp46 is thought to occur immediately after acquisition of the NK lineage-defining markers CD122 and NK1.1 (Narni-Mancinelli et al., 2011). Deletion of Nfil3 was confirmed in NK cells but not other immune cell lineages from Nfil3^fl/fl × Nkp46^iCre mice by PCR analysis of genomic DNA from sorted NK cells and T cells obtained from the same mouse (Fig. 8 A). Surprisingly, Nfil3^fl/fl × Nkp46^iCre mice contained a comparable number of mature NK cells to littermate control mice (Fig. 8 B). Moreover, surface expression of common NK cell maturation markers CD11b and CD27 did not differ significantly between NK cells from Nfil3^fl/fl × Nkp46^iCre mice and littermate controls (Fig. 8 C). We next generated mixed 1:1 WT:Nfil3^fl/fl × Nkp46^iCre bone marrow chimeric mice to evaluate the ability of the Nfil3^fl/fl × Nkp46^iCre NK cells to develop in a competitive environment with WT NK cells. We found that NK cells that have ablated Nfil3 early in development (CD45.2) reconstituted similarly to WT NK cells (CD45.1), suggesting that Nfil3 is exclusively required during the progenitor stages, but not once developing NK cells acquire their first activating receptors (Fig. 8 D).

Lastly, we adoptively transferred the WT (CD45.1) and Nfil3^fl/fl × Nkp46^iCre (CD45.2) NK cells from the bone marrow chimeras into Ly49h^−/− mice to investigate their ability to respond to MCMV infection. Both NK cell populations expanded robustly in response to infection and contracted to form comparable memory cell populations (Fig. 8, E and F). The effector and memory Nfil3^fl/fl × Nkp46^iCre NK cells were phenotypically and functionally similar to their WT counterparts (Fig. 8, G and H). Altogether, these results confirm that Nfil3 is not required for the survival or maintenance of NK cells beyond the earliest stages of development and lineage commitment.

IL-15 and Nfil3 are both factors critical for the development of a mature and functional NK cell pool in the periphery, and mice deficient in Nfil3 or components of the IL-15 signaling machinery lack NK cells at steady-state. However, despite stringent requirement for IL-15 in NK cell development, recent data indicate that viral infection of NK cell–deficient mice (e.g., Il15^−/−, Il15ra^−/−, Rag2^−/− × Il2rb^−/−, and Rag2^−/− × Il2rg^−/−) drives out a substantial population of mature NK cells that are functionally and phenotypically similar to those generated in infected WT mice (Sun et al., 2009b). These findings suggested that ligation of activating receptors (e.g., Ly49H) and signals from proinflammatory cytokines such as IL-12 can bypass the requirement for IL-15 during viral infection (Sun et al., 2009b). Our data indicates that although Nfil3 is indispensable for early stage NK cell development, it is not essential when NK cell maturation and functionality are driven by viral infection. We have established that Ly49H ligation plays a large role in the generation of Nfil3-deficient effector and memory NK cells, as do the proinflammatory cytokines IL-12 and type I IFN. Interestingly, a recent study demonstrated that cross-presenting CD8α⁺ DCs, which are also dependent on Nfil3 for their development (Kashiwada et al., 2011), could be induced to expand in short-term bone marrow reconstitution settings in the absence of Nfil3, and were phenotypically similar to WT DCs (Seillet et al., 2013). Together with our findings in NK cells, these results highlight that under specific conditions, certain lineage-determining transcriptional regulators are dispensable for subsequent lineage maintenance and function.

The precise source of these virally expanded NK cells in Nfil3^−/− mice remains elusive. NK cells differentiate from CLP cells in the bone marrow, maturing through several stages under the influence of specific cytokines (especially IL-15) and transcription factors to result in mature, functional cells (Yokoyama et al., 2004; Martín-Fontecha et al., 2011). Similar to NK cell–deficient mice where IL-15 signals are ablated, the minute number of peripheral NK cells present in Nfil3^−/− mice may be contributing to the robust NK cell numbers after MCMV infection. However, it is also possible that immature NK cells newly expressing Ly49H in the bone marrow are rapidly recruited to peripheral organs during viral infection, even though NK cell development in Nfil3^−/− mice is thought to be arrested at the early NK cell progenitor (NKP) stage (Gascoyne et al., 2009; Kamizono et al., 2009). We have previously demonstrated that adoptive transfer of as few as 10⁴ Ly49H⁺ NK cells into Ly49H-deficient recipient mice, followed by MCMV infection, results in a detectable effector population at day 7 PI (Sun et al., 2009a). Therefore, to determine whether the MCMV-expanded Nfil3^−/− Ly49H⁺ NK cells originated from the small percentage of peripheral NK cells, whole splenocytes from Nfil3^−/− mice were adoptively transferred into Ly49h^−/− recipient mice and infected with MCMV. No effector Ly49H⁺ NK cells could be detected at day 7 PI (unpublished data), suggesting that our whole spleen preparation contained <10⁴ Ly49H⁺ NK cells. Although the origin of the MCMV-driven NK cells remains to be determined, perhaps a combination of the immature bone marrow–derived NK cells and peripheral NK cells contribute to the significant population of Ly49H⁺ NK cells measured in Nfil3^−/− mice at day 7 PI.

Given that NK cells could be induced to expand in mice lacking IL-15 signaling machinery, it is perhaps not surprising that NK cells could similarly expand in Nfil3^−/− mice in response to viral infection. Unexpectedly however, the MCMV-expanded Nfil3^−/− NK cells persisted in mice for a prolonged period after initial infection, even though Nfil3 has been suggested to function downstream of the IL-15 receptor (Gascoyne et al., 2009). Although mature, resting NK cells depend on IL-15 signals for persistence (Kennedy et al., 2000), it was previously unknown whether memory NK cells have the same requirement for IL-15. Our data demonstrates that like resting NK cells, memory NK cells remain fully dependent on IL-15, as memory NK cells transferred into IL-15–deficient hosts disappeared with similar kinetics to both resting and effector NK cells. Together with data demonstrating that tamoxifen-induced deletion of Nfil3 in mature NK cells did not affect NK cell survival, our results predict the existence of an Nfil3-independent prosurvival pathway downstream of IL-15 that permits the maintenance of peripheral NK cells. Because memory CD8⁺ T cells also depend on IL-15 signals for survival (Ma et al., 2006), we tested whether generation of this lymphocyte population depended on Nfil3 in like manner to developing NK cells. After MCMV infection of mixed WT:Nfil3^−/− bone marrow chimeric mice, we observed a comparable percentage of antigen-specific WT and Nfil3^−/− CD8⁺ T cells at memory time points (unpublished data), indicating that this IL-15–dependent, but Nfil3-independent, pathway is not restricted to the NK cell lineage.

Strikingly, the coincident deletion of Nfil3 with NKp46 expression during early development resulted in normal peripheral NK cell numbers, suggesting that the activity of Nfil3 is likely critical only at the early transitional stages from CLP to NKP. Along with the tamoxifen-induced deletion of Nfil3 in mature NK cells, these data together are in contrast to the loss of certain lymphocyte subsets and function during deletion of lineage-determining transcription factors. One of the best characterized examples is regulatory T cell dependence on continuous Foxp3 expression, not only during development but also for lineage maintenance and function (Williams and Rudensky, 2007). Among the transcription factors that govern NK cell development, induced deletion of Eomes in peripheral NK cells has recently been shown to result in de-differentiation of mature NK cells (Gordon et al., 2012), further suggesting that such lineage-defining transcription factors play a continuous role in the maintenance of lymphocyte identity. Thus, future studies will determine why this is not the case for Nfil3, and how Nfil3 initiates programming of the NK cell lineage but then is dispensable for lineage maintenance and function. Furthermore, a greater understanding of the interplay between IL-15 signaling and Nfil3 activity during NK development for the generation of a mature peripheral NK cell pool will inform our efforts to harness this potent effector immune cell for therapeutics and vaccines against pathogen infection and cancer.

WT C57BL/6 (B6) and congenic (CD45.1) mice were obtained from Taconic. Nfil3^−/− (Kashiwada et al., 2010), Nfil3^fl/fl (Motomura et al., 2011), Nkp46^iCre (Narni-Mancinelli et al., 2011), Ub^Cre-ERt2 × R26Y, B2m^−/−, Ly49h^−/−, and Rag2^−/− × Ly49h^−/− mice were bred and maintained at MSKCC. Il15^−/− mice were obtained from The Jackson Laboratory. All mice are on the C57BL/6 background. Bone marrow chimeric mice were generated, as previously described (Sun et al., 2009a). Mice were housed and maintained according to Memorial Sloan-Kettering Cancer Center (MSKCC) guidelines, and all animal experiments were performed in accordance with MSKCC Institutional Animal Care and Use Committee approval and institutional guidelines.

A salivary gland stock of MCMV (Smith strain) was administered to WT mice via i.p. injection at 6 × 10³ plaque forming units (pfu) per mouse. Nfil3^−/− mice were administered 6 × 10² pfu MCMV to induce NK cell expansion without associated spleen and liver pathology observed at higher doses. MCMV-Δm157 (provided by U. Koszinowski, Ludwig-Maximilians-University, Munich, Germany; Bubić et al., 2004) was injected at 4.5 × 10⁴ pfu. Recombinant vesicular stomatitis virus expressing the MCMV m157 protein (VSV-m157) was generated as previously described (Kim et al., 1998) and administered by i.v. injection at a dose of 10⁷ pfu. The recombinant vaccinia virus–OVA (VACV-OVA) was provided by J. Yewdell and J. Bennink (National Institute of Allergy and Infectious Disease, National Institutes of Health, Bethesda, MD). To generate the recombinant vaccinia virus expressing m157 (VACV-m157), a cDNA encoding a fusion protein of the CD8-derived signal sequence, followed by a FLAG-tag and MCMV-m157, was cloned into VACV transfer plasmid pRB21 containing a functional F13L gene, a VACV synthetic early/late promoter, and VACV genomic flanking sequences for homologous recombination and integration into the F13L-deficient VACV vRB12. Identification of recombinant viruses and plaque purification was performed on BSC-1 cells. Recombinant viruses were propagated, purified, and titrated using standard methodology, and 10⁷ pfu were injected i.p.

Mice were administered 8 mg tamoxifen dissolved in 200 µl olive oil by oral gavage on days 0, 1, and 3. In treated Nfil3^fl/fl × R26Y × Ub^CreERt2 mice, effectiveness of tamoxifen treatment was evaluated by monitoring peripheral blood lymphocytes for the appearance of a YFP⁺ population corresponding to the loss of Nfil3, measured by PCR for Nfil3 DNA and qRT-PCR for Nfil3 mRNA.

In vivo blockade of IL-12 and type I IFN signaling during MCMV infection was accomplished by i.p. injection of purified antibodies for IL-12 (clone C17.8) and IFN-α receptor (clone MAR1-5A3) 1 d before MCMV infection. Each mouse was treated with 1 mg of antibody.

NK cells from spleens were enriched using negative selection with magnetic beads (QIAGEN), or by sorting on an Aria II flow cytometer (BD). Purified NK cells were then transferred into Ly49h^−/− or Il15^−/− mice by i.v. injection and monitored by congenic markers and expression of the Ly49H receptor. Single cell suspensions were generated from indicated organs and incubated with the anti-Fc receptor antibody 2.4G2 before staining with indicated mAbs (BioLegend, eBioscience, and BD) for 20 min on ice. Samples were acquired using an LSRII flow cytometer with FACSDiva software (BD). Analysis was performed with the FlowJo 9.6 software package (Tree Star).

Whole splenocytes were obtained from MCMV-infected mice on specified days, and stimulation of NK cells by plate-bound antibodies against activating receptors along with Brefeldin A was performed as described previously (Sun et al., 2012). In brief, 96-well plates were coated with 10 µg/ml of purified antibodies against Ly49H or NKp46. PBS-treated control wells, and wells containing 50 ng/ml PMA and 1 µg/ml ionomycin, or 20 ng/ml IL-12 and 10 ng/ml IL-18 were included. Cells were stained for intracellular IFN-γ and CD107a (LAMP-1).

In vivo cytotoxicity was assessed by i.v. injection of target cells (WT or B2m^−/−) labeled with different concentrations of CFSE at a 1:1 ratio into indicated mice. Spleens were collected at indicated time points and examined for percentage of each target cell population. For in vitro cytotoxicity assays, enriched NK cells were incubated with ⁵¹Cr-labeled YAC-1 target cells in triplicate. Supernatants were harvested and assayed for ⁵¹Cr release after 4 h of incubation at 37°C, and percentage of specific lysis was determined.

CD8⁺ T cells, CD4⁺ T cells, and NK cells were purified (>99%) from total splenocytes using an Aria II sorter (BD), and lysed in Tri-Reagent (Ambion). RNA purification and cDNA synthesis were performed using the RNeasy kit (with DNase I treatment; QIAGEN), and MuLV reverse transcription and Oligo dT₍₁₆₎ primers (Applied Biosystems). iQ Sybr Green SuperMix (Bio-Rad Laboratories) was used for qRT-PCR, and data were normalized to β-actin and expressed as relative target abundance via the ΔΔCT method. Relevant primer sequences are as follows: Nfil3 forward, 5′-ATGCCCAAGAAATCCAGAAA-3′; Nfil3 reverse, 5′-GGGAGAGTGCTTGATGACTG-3′; β-actin forward, 5′-TGCGTGACATCAAAGAGAAG-3′; and β-actin reverse, 5′-CGGATGTCAACGTCACACTT-3′.

Results are expressed as mean ± SE. Data were analyzed using one-way ANOVA, and unpaired Students t test for multiple comparisons. All analyses were performed using Prism 5.0b (GraphPad Software), and differences were considered significant when P ≤ 0.05.

We thank members of the Sun laboratory for technical support and experimental assistance, and members of the MSKCC NK club for insightful comments and helpful discussions. Lewis Lanier, Sasha Rudensky, Ulrich Koszinowski, Bernie Moss, Jonathan Yewdell, and Jack Bennink provided reagents critical to this study. We thank the ImmGen consortium for providing the microarray data used in this study.

M.A. Firth is supported by a fellowship from the Lucille Castori Center for Microbes, Inflammation, and Cancer. S. Madera is supported by a pre-doctoral Cancer Research Institute training grant. A.M. Beaulieu is supported by a NIH T32 award (CA009149). G. Gasteiger is supported by an Irvington Fellowship of the Cancer Research Institute. E. Vivier is supported by the THINK ERC Advanced Grant. J.C. Sun is supported by the Searle Scholars Program, the Cancer Research Institute, and grants from the NIH (AI085034 and AI100874).

The authors declare no financial conflicts of interest.