Conditional NKT Cell Depletion in Mice Reveals a Negative Feedback Loop That Regulates CTL Cross-Priming

Cytotoxic (CD8⁺) T cells control intracellular pathogens and malignant diseases. Their activation against Ags derived from tumor cells, viruses, or vaccines is facilitated by type 1 conventional dendritic cells (cDC1s), which can cross-present exogenous Ags on MHC class I (1, 2). Immunogenic priming of CTLs (cross-priming) requires cDC1s to express costimulatory molecules and cytokines, that is, to undergo maturation (2, 3). DC maturation can result from the engagement of pattern recognition receptors with their ligands, but pattern recognition receptor ligands are often sparse when Ags need to be cross-presented (3–5). The full DC cross-priming capability also requires CD4⁺ T_H cells that recognize cognate Ag on the same DC (1, 4, 5). Whereas the contribution of different help signals produced by T_H cells is subject to ongoing debate (5) and likely context-dependent (6), the CD40L/CD40 axis has long been known to be crucial for cross-priming (4).

In addition to classical, T_H cell–dependent cross-priming (classical cross-priming [cXP]), invariant NKT cells (iNKT cells) can also induce DC maturation and promote CTL responses by mechanisms involving CD40L (7–13), a process that has been referred to as alternative cross-priming (aXP) (14, 15). iNKT cells are innate-like T cells that harbor an invariant Vα14-Jα18 TCR in mice that recognizes lipid Ags presented by the MHC class Ib molecule CD1d (16). Similar to other innate-like T cells, iNKT cells express the transcription factor promyelocytic leukemia zinc finger (PLZF; encoded by Zbtb16), which controls their effector program (10, 17). They can be activated either by TCR signaling or by cytokines in a TCR-independent fashion and rapidly produce copious amounts of cytokines (10, 17). The function of iNKT cells can be studied by stimulation with lipid Ags, which are also being evaluated for clinical applications (18), or by using constitutively iNKT cell–deficient Jα18^−/− and CD1d^−/− mice (19). Yet, in constitutively iNKT cell–deficient mice, alterations of the gut microbiota, of the thymic development of CTLs (17), and of mucosal-associated invariant T (MAIT) cell homeostasis (20) may interfere with studies on NKT cells. Alternatively, Abs targeting CD1d or the iNKT TCR (21, 22) are used to specifically investigate the role of TCR signaling in iNKT cells in a temporally controlled fashion. However, the latter tools cannot target recently activated cells that internalize their TCR (19, 23, 24) or stop TCR-independent functions. Thus, there is currently no tool to conditionally deplete NKT cells in vivo.

Immunization of mice with a soluble protein Ag together with an iNKT cell lipid Ag (e.g., α-galactosylceramide [αGC]) (7–9, 14, 25), with αGC-pulsed cells (8, 11, 12, 25), or with αGC nanoparticle vaccines (13, 15, 26) induces remarkably strong CTL responses, for instance, against tumor Ags (11, 15). These studies highlight the translational potential of aXP. Such responses are absent in NKT cell–deficient (8, 13, 14, 25) and in splenectomized mice (14), because they rely on splenic iNKT cells providing help to cDC1s (1, 12) and CD40L (7–13). In contrast, IL-12 (8, 14), TNF-α (8), and MHC class II (MHC-II) or CD4⁺ T cells (13, 14) have been shown to be dispensable for αGC-induced CTL responses.

Similar to T_H cells (27, 28), iNKT cells provide help in distinct phases (26). iNKT cells and DCs mutually activate each other within the first ∼6 h after i.v. immunization with an αGC-adjuvanted vaccine (26), triggering CCL17 production by DCs and induction of CCR4 on CTLs (phase 1) (14, 26). At ∼24 h, mature (CD70⁺) cDCs, activated CTLs, and iNKT cells congregate in the splenic white pulp (phase 2) (9, 12, 14, 26). The functional consequences of interactions between iNKT cells, DCs, and CTLs at different time points of aXP have not been studied, mainly because of the lack of options to conditionally manipulate iNKT cells.

In this study, we report two experimental approaches to conditionally deplete NKT cells, even after their activation, and reveal a negative feedback loop in aXP that could not have been detected without a conditional NKT cell depletion technique.

Animal experiments were approved by an Ethics Board of the German state of North Rhine-Westphalia. PLZF^Cre mice (29) were purchased from The Jackson Laboratory. ROSA26^LSL-DTR (inducible diphtheria toxin [DTx] receptor [iDTR]) mice (30) were provided by Ari Waisman. CCL17^eGFP (31), RAG2^−/− (32), Jα18^−/− (33), MHC-II^−/− (34), CD90.1, OT-I × CD45.1 (35), CD40L^−/− (36), and C57BL/6J (wild-type [wt]) mice as well as PLZF^CreROSA26^LSL-DTR mice were bred at the Central Animal Facilities of the Medical Faculty of the University of Bonn and kept in individually ventilated cages under specific pathogen-free conditions.

Irradiated mice (5.5/7.5/9 Gy for RAG2^−/−/Jα18^−/−/wt mice, BioBeam 2000 [MCP-STS]) received 2–5 × 10⁶ bone marrow (BM) cells. For mixed BM chimeras (BMx), BM cells were mixed at a 1:1 ratio of CD34⁺ hematopoietic stem cells. To generate PLZF^iDTR mice, CD19⁺ and DTR⁺ BM cells were depleted magnetically (Miltenyi Biotec) and by i.p. injection of 25 ng/g body weight DTx within 24 h after BM transfer, respectively. Mice were kept on acidified drinking water (pH 2–3) for 2 wk following irradiation.

Naive OT-I cells were isolated by negative CD8⁺ T cell selection (Miltenyi Biotec) in combination with biotinylated anti-CD44 (BioLegend, clone IM7, 6.25 ng/ml) and injected i.v. Cell numbers are indicated in the figure legends.

Mice were injected i.v. with 200 µg of OVA (Merck) plus 0.2 µg of αGC (Enzo Life Sciences) in PBS. Abs against Thy1.1 (Bio X Cell, clone 19E12, 250 µg), CD40L (Bio X Cell, clone MR-1, 250 µg), NK1.1 (Bio X Cell, clone PK136, 100 µg) were injected i.p. AS1517499 (Axon Medchem) was injected i.p. at 2 mg/ml in 20% DMSO/PBS at 5 µl/g body weight once per day.

Splenocytes were pulsed with SIINFEKL (2 μg/ml) and labeled with 0.1 μM CFSE (lo) or were not pulsed and labeled with 1 μM CFSE (hi), after which 2–5 × 10⁶ cells of a 1:1 mix of target and nontarget cells were injected i.v. and relative abundance was quantified 4–5 h later. Specific lysis was calculated as 100% − [100% × (lo/hi)_primed/ Σ(lo/hi)_control/n_control].

Spleens were digested with 1 mg/ml collagenase IV and 100 µg/ml DNase (both Merck) for 25 min or were homogenized. Lungs, livers, and visceral adipose tissue were digested for 40, 15, and 25 min, respectively. Absolute cell number quantification was performed using CaliBRITE APC beads (BD Biosciences).

Cells were stained with titrated amounts of the Abs listed in Table I. Unspecific binding was blocked using anti-CD16/32 (clone 2.4G2, BD Biosciences). iNKT cells and MAIT cells were detected using titrated amounts of CD1d tetramers and MR1 tetramers (37), respectively. CD40L was stained at 37°C for 30 min. Intracellular staining was performed using a Foxp3 transcription factor staining kit (Thermo Fisher Scientific). Dead cells were excluded using Hoechst 33258 (Sigma-Aldrich) or fixable viability dyes (Thermo Fisher Scientific). Data were acquired on LSRFortessa or FACSCanto devices and analyzed with FlowJo (BD Biosciences) and the uniform manifold approximation and projection (38) and PhenoGraph (39) plugins.

Depletion efficiency was calculated based on absolute cell numbers of the indicated subsets in PBS- and DTx-treated chimeras, and the average cell number in PBS-treated chimeras was set to 100%.

Results are expressed as mean ± SEM. Comparisons of two groups were performed using Mann–Whitney U test. Comparisons of multiple groups were performed using ANOVA with a Bonferroni posttest (GraphPad Prism). Statistical significance was defined as follows: *p < 0.05, **p < 0.01, and ***p < 0.001.

We wanted to understand the timing of iNKT cell–mediated help in aXP. Given that the CD40L/CD40 axis has been shown to be required for iNKT cell–dependent CTL cross-priming (7–13) and that CD40L is expressed on activated iNKT cells (Supplemental Fig. 1A) (40), it is commonly assumed that iNKT cells use CD40L to directly activate DCs (10, 11, 13), similar to how CD40L⁺ T_H cells do in cXP, but this has not yet been formally shown. Whereas iNKT cells activate DCs in the first phase of aXP (26), help in cXP is thought to occur at later phases (4, 27). To test the timing of CD40L-mediated help in aXP, we blocked CD40L from 4 h prior to immunization with OVA/αGC (i.e., in phases 1 and 2, Fig. 1A) or from 20 h after immunization, after the peak of the iNKT cell–induced CCL17 production by DCs (Supplemental Fig. 1B). The expansion of Ag-specific OT-I cells was not only reduced after early CD40L blockade, but to a similar extent also after blockade at the later time point (Fig. 1B, Supplemental Fig. 1C, 1D). Although we cannot rule out a contribution within the first 20 h after immunization, that is, in phase 1, this suggested that iNKT cells provide effective CD40L signaling in phase 2 of aXP.

Testing this hypothesis required a method to conditionally deplete NKT cells. As surface markers such as TCR and NK1.1 are internalized upon NKT cell activation (23, 24), we chose a BMx approach, where both CD90.2 Jα18^−/− and congenic CD90.1 Jα18^+/+ BM was transferred into CD90.2 recipient mice (Fig. 2A). In these BMx, all CD90.1 BM-derived immune cells can be depleted by injection of an anti-CD90.1 Ab, whereas the CD90.2⁺ cells derived from Jα18^−/− BM or from the recipient are spared. Anti-CD90.1 removed ∼90% iNKT cells (Fig. 2B, 2C), whereas 50% of conventional T cells (gated as in Supplemental Fig. 2A, Table I) survived (Fig. 2C). Cells lacking CD90 expression, such as B cells, NK cells, DCs, and neutrophils, were not reduced in numbers (Fig. 2C). A recent report suggests that also MAIT cells are absent in the specific line of Jα18^−/− mice that we used (41); hence, those cells may also be absent in the iNKT–cell depleted animals. Yet, given the strong specific activation of iNKT cells by αGC, it seemed unlikely that MAIT cell deficiency would have a major impact in aXP.

iNKT cell depletion 20 h after immunizing the BMx (when phase 2 ensued, Fig. 2D) was efficient (Supplemental Fig. 2B), but unexpectedly it led to a more vigorous expansion of OT-I cells, and especially increased Ag-specific cytotoxicity (Fig. 2E, 2F). Rescue experiments, unfortunately, were not feasible, because iNKT cell transfer is highly inefficient (42, 43), does not deliver the cells to the correct anatomical location, and because activation-matched iNKT cells cannot be sorted due to TCR internalization. Nevertheless, the opposing outcomes after CD40L blockade and iNKT cell depletion challenged the current concept of iNKT cell–mediated stimulation of aXP and suggested a negative feedback loop by which iNKT cells control the expansion of CTLs in phase 2.

To confirm these findings in a different experimental system, and to rule out that the 50% reduction of T cells in the anti-CD90.1–treated BMx (Fig. 2C) confounded our results, we crossed PLZF^Cre mice (29) to ROSA26^LSL-DTR mice (30). PLZF expression during embryonic development (29) necessitated the generation of BM chimeras, hereafter termed PLZF^iDTR mice (Fig. 3A). We used irradiated RAG2^−/− or Jα18^−/− recipients that lack endogenous iNKT cells and eliminated DTR-expressing hematopoietic stem cells by injecting DTx within 24 h after BM transfer (Fig. 3A). DTx administration to reconstituted PLZF^iDTR mice depleted >90% of splenic iNKT cells (Fig. 3B, 3C) and >80% of iNKT cells in blood, liver, lung, and adipose tissue (Fig. 3D), 50% of NK cells (Fig. 3C), and ∼80% of lung MAIT cells (Fig. 3E). Numbers of conventional CD4⁺ and CD8⁺ αβ T cells, B cells, DCs, macrophages, and neutrophils were not altered (Fig. 3C). PLZF expression in Vγ1⁺Vδ6.3⁺ T cells, which constitute ∼15% of splenic γδ T cells (44), was likely responsible for the reduction in γδ T cells in PLZF^iDTR chimeras. Other unconventional T cell subsets, for example, Qa-1–restricted T cells (18), may similarly be targeted in this system. In naive mice, a depleting anti-NK1.1 Ab has a similar targeting profile as DTx has in PLZF^iDTR mice, yet 1) not all iNKT cells express NK1.1 (16), and 2) activated iNKT cells internalize NK1.1, which remains absent for at least several days (23, 24), and therefore anti-NK1.1 treatment did not deplete iNKT cells in our setup (Supplemental Fig. 3A–C). Similar limitations will apply to an Ab targeting the iNKT cell TCR (22), which is presently not commercially available. Lastly, NKT cell depletion in PLZF^iDTR mice does not have a predisposition to target conventional T cells and can thus serve to perform conditional NKT cell loss-of function experiments in the absence of T cell lymphopenia to verify observations made using mixed Jα18^−/−/CD90.1 chimeras. Nevertheless, the reduction of NK cells and other innate-like lymphocytes needs to be kept in mind when interpreting the results.

When we used the PLZF^iDTR system to deplete NKT cells 20 h after immunization (Fig. 4A), we again found expansion (Fig. 4B, Supplemental Fig. 3D–F), effector differentiation (Supplemental Fig. 3G, 3H), and cytotoxicity (Fig. 4C, Supplemental Fig. 3I, 3J) of OT-I cells to be increased, confirming our findings in the Jα18^−/−/CD90.1 BMx (Fig. 2E, 2F). It is unlikely that the concomitant depletion of NK cells, which are usually thought to stimulate CTLs by IFN-γ, but may also eliminate activated CTLs (45), bolstered CTL expansion, because the injection of an anti-NK1.1 Ab to immunized wt mice (Supplemental Fig. 3A) did not impact aXP (Supplemental Fig. 3K). Taken together, both of our NKT cell depletion approaches concordantly demonstrated that NKT cells engage a negative feedback loop in phase 2 of aXP, after their initiating effect during phase 1.

We next followed up on the discrepant effects of CD40L blockade (Fig. 1) and NKT cell depletion (Figs. 2E, 2F, 4B, 4C). When we blocked CD40L signaling in NKT cell–depleted and nondepleted PLZF^iDTR mice, CD40L blockade reduced CTL expansion and cytotoxicity by day 8 (Fig. 4B, 4C, Supplemental Fig. 3D–J) not only in NKT cell–competent mice, but also in the absence of >98% of NKT cells (Supplemental Fig. 3L). Yet, even this small number of residual NKT cells was sufficient to induce aXP (Supplemental Fig. 3M), necessitating a more direct approach to address the role of CD40L on NKT cells. We therefore generated mixed CD40L^−/−/Jα18^−/− BMx in Jα18^−/− host mice in which iNKT cells selectively lack CD40L, whereas in CD40L^−/−/wt control chimeras, CD40L-competent iNKT cells could arise from the wt BM component (Fig. 5A). Indeed, the CTL response was indistinguishable, regardless of whether iNKT cells expressed CD40L (Fig. 5B, 5C). This verified our conclusion that CD40L on iNKT and MAIT cells, which are absent both in depleted PLZF^DTR mice and in Jα18^−/− mice, is not required in aXP.

aXP is effective in the absence of T_H cells (13, 14). Yet, to test whether there is a substantial contribution of T_H cells to CD40L signaling, we inhibited CD40L signaling also in MHC-II^−/− mice (Fig. 6A). CD40L blockade impaired the CTL response at least equally efficiently in MHC-II^−/− mice compared with wt mice (Fig. 6B–D). These data confirmed that CD40L on iNKT cells and T_H cells either redundantly facilitates DC maturation, or that CD40L is provided by other cells. Although CD40L expression by either NKT cells or T_H cells might be sufficient, we also observed induction of surface CD40L on monocytes (Fig. 6E, 6F, Supplemental Fig. 4), which were phenotypically similar to a population that has recently been found to stimulate T cell differentiation by IL-12 production (46). Future studies may elucidate whether these monocytes are involved in aXP.

Finally, we asked whether type 2 cytokines that are produced by NKT cells (10) underlie the NKT cell–mediated feedback inhibition in phase 2. IL-4 and IL-13 signal through STAT6 and have been shown to limit tumor surveillance by CTLs (16). However, pharmacological inhibition of STAT6 during phase 2 (Fig. 7A) did slightly, yet not significantly, increase CTL expansion and had no effect on overall cytotoxicity (Fig. 7B–E). These data suggest that STAT6-inducing cytokines contribute to the negative feedback loop in aXP, but that other factors yet to be identified are more important.

In this study, we present approaches for conditional NKT cell depletion. Using these approaches, we found and explored a negative feedback loop that regulates aXP. Notably, similar to most in vivo cell targeting strategies, neither of our approaches is entirely specific to NKT cells. Yet, given the mostly distinct targeting profiles of either system, the specific activation of iNKT cells that we use, and the consistency between both approaches, we feel confident to attribute the observed effects to iNKT cells. Modifications of our depletion approaches will in principle also allow loss-of-function experiments of other unconventional T cell types (45). These can uncover in vivo functions or reveal biological regulatory loops that are intermingled with processes of opposite function at different time points, as demonstrated here for NKT cell–mediated CTL cross-priming. Furthermore, our PLZF^iDTR chimera approach also allows depletion of the entire unconventional T cell compartment to directly address functions of these cell types as a whole (47).

Our data confirmed that aXP encompasses at least two phases: an early phase of mutual activation of DCs and NKT cells and a late phase in which CD40L-dependent help is provided and the magnitude and nature of the effector response are fine-tuned, reminiscent of cXP (27). Similarly, a small number of NKT cells remaining after depletion, which still outnumber Ag-specific T_H cells in naive animals, were sufficient to induce aXP. This is compatible with our hypothesis of a biphasic mode of action, that is, the loss of inhibition in phase 2 compensating for some loss of stimulatory capacity in phase 1. We highlighted a minor contribution of STAT6 signaling to the fine-tuning of CTL expansion after the first phase of aXP and that this manifests as negative feedback mechanism. Notably, mediators other than STAT6 appear to control not only CTL accumulation, but also cytotoxicity.

Moreover, our approaches allowed insights into the flow of information during aXP. We observed that NKT cells do not require CD40L to induce aXP, as commonly assumed. Instead, CD40L provided by non-NKT cells, such as the monocytic subset we observed or T_H cells, is sufficient, revealing that a more complex cellular interplay than previously anticipated is at work in aXP. Yet, in line with previous results (13, 14), the absence of CD4⁺ T cells did not abrogate aXP and, on the contrary, likely due to the lack of regulatory T cells (48), rather increased the impact of CD40L blockade, ruling out that T_H cells are the sole source of CD40L signaling in aXP. This finding illustrates the need to precisely identify the sources and targets of CD40L signaling (4), which is essential for the clinical application of NKT cell–targeted immunization approaches (18). Finally, NKT cells are assumed to provide help to B cells via CD40L (17), but in the light of our findings, also this assumption might have to be tested formally.

The authors have no financial conflicts of interest.

We thank Daniela Klaus, Moritz Blankart, and Melanie Eichler for excellent technical assistance. We acknowledge support by the Flow Cytometry Core Facility and the central animal facilities of the Medical Faculty of Bonn University. We thank Ari Waisman (University of Mainz) and Andreas Thiel (Charité Berlin) for providing mice and bone marrow, respectively. CD1d and MR1 tetramers were provided by the National Institutes of Health Tetramer Core Facility. The MR1 tetramer technology was developed jointly by Dr. James McCluskey, Dr. Jamie Rossjohn, and Dr. David Fairlie. The material was produced by the National Institutes of Health Tetramer Core Facility as permitted to be distributed by the University of Melbourne.