Group 2 innate lymphoid cells (ILC-2s) regulate immune responses to pathogens and maintain tissue homeostasis in response to cytokines. Positive regulation of ILC-2s through ICOS has been recently elucidated. We demonstrate here that PD-1 is an important negative regulator of KLRG1+ ILC-2 function in both mice and humans. Increase in KLRG1+ ILC-2 cell numbers was attributed to an intrinsic defect in PD-1 signaling, which resulted in enhanced STAT5 activation. During Nippostrongylus brasiliensis infection, a significant expansion of KLRG1+ ILC-2 subsets occurred in Pdcd1−/− mice and, upon adoptive transfer, Pdcd1−/− KLRG1+ ILC-2s significantly reduced worm burden. Furthermore, blocking PD-1 with an antibody increased KLRG1+ ILC-2 cell number and reduced disease burden. Therefore, PD-1 is required for maintaining the number, and hence function, of KLRG1+ ILC-2s.
Group 2 innate lymphoid cells (ILC-2s) lack antigen-specific receptors and primarily render their function via cytokine signaling (Bartemes et al., 2012; Spits and Cupedo, 2012; Huang et al., 2015). ILC-2s were initially described by several groups and designated as natural helper cells (Koyasu et al., 2010; Moro et al., 2010), nuocytes (Neill et al., 2010; Barlow et al., 2012, 2013), or innate helper 2 cells (Price et al., 2010) that respond to tissue-derived signals including IL-25, IL-33 and thymic stromal lymphopoietin (TSLP). ILC-2s express IL-33 receptor (ST2), IL-25 receptor (IL-17RB), KLRG1 and naturally reside in tissue sites such as the lung, small intestine, skin and adipose tissues. ILC-2s initiate immune responses against parasites (Fallon et al., 2006; Huang et al., 2015), participate in inflammatory processes, such as airway hyperactivity (Chang et al., 2011), allergen induced lung inflammation (Motomura et al., 2014), and allergic atopic dermatitis (AD) in humans (Salimi et al., 2013). ILC-2s also contribute toward lung tissue repair (Monticelli et al., 2011), adipose tissue homeostasis (Brestoff et al., 2015; Lee et al., 2015), and cutaneous wound healing (Yin et al., 2013; Rak et al., 2016). Therefore, elucidating immunoregulatory mechanisms that can modulate ILC-2 cell number and function can identify important checkpoints that can be manipulated for controlling type 2–mediated immune responses.
Recent studies on ILC-2s in airway inflammation have identified a positive regulatory axis driven by ICOS signaling (Maazi et al., 2015; Molofsky et al., 2015; Paclik et al., 2015). Studies on negative co-receptor mediated regulation of ILC-2s has been restricted to the role of KLRG1, which has been previously shown to inhibit ILC-2 effector response (Salimi et al., 2013). Here, we have investigated the role of PD-1 in regulating KLRG1+ ILC-2 subsets and demonstrate the downstream signaling mechanism by which PD-1 regulates KLRG1+ILC-2s. PD-1 is related to the CD28 superfamily and is expressed on activated T cells, B cells, monocytes, and macrophages. It has two binding partners, namely PDL-1 (Dong et al., 1999) and PDL-2 (Latchman et al., 2001; Keir et al., 2008; Fife et al., 2009). Co-stimulation of PD-1 by either of these ligands activate inhibitory signals in T cells which either prevent T cell proliferation or render a regulatory phenotype to the T cells (Fife et al., 2009; Francisco et al., 2009; Amarnath et al., 2010, 2011). These varied immune-tolerant signaling cascades occur through SHP1/2 phosphatases, which are recruited to the ITIM and ITSM cytoplasmic domains of the PD-1 receptor (Okazaki et al., 2001; Parry et al., 2005). The recruited SHP1/2 phosphatases dephosphorylate STATs and/or AKT, thereby dampening T helper cell function (Franceschini et al., 2009; Francisco et al., 2009; Amarnath et al., 2011). In particular PD-1 can specifically inhibit STAT5 signaling in T regulatory cells (Franceschini et al., 2009). It is yet to be clarified if such PD-1–mediated tolerance mechanisms occur in ILC subsets.
Tumors (Wang and Chen, 2011), viruses (Barber et al., 2006; Day et al., 2006; Trautmann et al., 2006), and bacteria (Das et al., 2006; Beswick et al., 2007; Barber et al., 2011) manipulate the PD-1 signaling pathway to evade host immune responses. In particular, clinical trials that use PD-1 blocking antibody have shown phenomenal success in cancer immunotherapy (Topalian et al., 2012; Yaqub, 2015). Parasitic worms also exploit the PD-1 pathway to create an immune-suppressive microenvironment by inducing macrophages with suppressor function (Smith et al., 2004; Terrazas et al., 2005). Hence, PD-1–mediated tolerance mechanisms in adaptive and innate immune cells, with respect to tumors and pathogens, have been extensively studied. However, the cellular mechanism by which PD-1 modulates ILC-2 function during disease pathogenesis is still largely unknown.
In this study, we have explored whether PD-1 regulates ILC-2 cells. We demonstrate that PD-1 is a critical negative regulator of KLRG1+ ILC-2 subsets. Disrupting PD-1 signaling either by genetic deletion or by antibody blockade significantly enhanced KLRG1+ ILC-2 cells in both number and function, thereby efficiently clearing Nippostrongylus brasiliensis worms in mice. In humans, we found that PD-1 is exclusively expressed by ILC-2s (and not ILC-1 or ILC-3) and regulates human ILC-2 function.
Pdcd1−/− mice possess enhanced KLRG1+ ILC-2 subsets
The expression and regulatory function of PD-1 in T cells, B cells, and myeloid cells has been previously characterized (Agata et al., 1996; Okazaki et al., 2002), but its role in ILCs is yet to be defined. Using previously defined gating strategy for ILC-2s (Chang et al., 2011; Monticelli et al., 2011; Halim et al., 2012), we found that Lin−CD45+CD90+CD25+CD127+ KLRG1+ ILC-2s expressed PD-1 in WT mice (Fig. S1, A–C). We next explored the functional consequence of PD-1 expression on KLRG1+ILC-2s by comparing the frequency of these subsets in WT and Pdcd1−/− mice. Initial analysis suggested that WT and Pdcd1−/− KLRG1+ILC-2s showed similar trend in IL-7R and IL-2R surface expression (Fig. S1, D and E). Therefore, we investigated the frequency of ILC-2s in WT and Pdcd1−/− mice. In all organs tested, tissue-resident Lin−CD45+CD90+CD25+CD127+ ILCs were increased and a significant difference was noted in the lungs and skin of Pdcd1−/− mice (Fig. 1, A and B). Next, we evaluated if there was a difference in ILC-2 subsets between WT and Pdcd1−/− mice. A significant increase in KLRG1+ ILC-2s was observed in the lungs of Pdcd1−/− mice (Fig. 1 C). Although there was a small increase in KLRG1+ ILC-2s in the small intestine and skin, a significant statistical difference was not observed (Fig. 1 D). We next evaluated if there was a difference in the cytokine expression within the KLRG1+ ILC-2s. Although both WT and Pdcd1−/− KLRG1+ ILC-2s had similar IL-5 and IL-13 expression by intracellular flow cytometry (Fig. 1, E–G), the cytokine profile of WT KLRG1+ILC-2s that were PD-1− or PD-1+ was altered. In the WT ILC-2s, PD-1− population produced higher amounts of effector cytokines such as IL-13 compared to the PD-1+ subsets (Fig. 1, H and I). These data led us to investigate the kinetics of PD-1 expression on KLRG1+ILC-2s and their potential function in these cells. We found that KLRG1+ILC-2s up-regulated PD-1 expression in response to IL-2 and IL-7 stimulation. PD-1 expression was significantly increased in the presence of IL-33 cytokine, but a similar effect was not noted with IL-25 or TSLP (Fig. S1, F and G). Collectively, these data suggest that PD-1 expression on ILC-2s is regulated by integrative signals from IL-2, IL-7, and IL-33.
PD-1 deficiency does not affect ILC-2 development
We next used a mixed BM chimera experiment to test two aspects of PD-1 function in KLRG1+ILC-2 biology, namely development and function. Mixed BM chimera experiments were performed in which T cell–depleted BM was adoptively transferred into WT B6 recipients after total body irradiation. 8 wk after transplant, ILC-2 frequency and numbers were quantified. Murine recipients that received WT BM had a tendency toward lower numbers of KLRG1+ ILC-2s in the lungs compared with Pdcd1−/− recipients, but this difference did not reach significance (Fig. 2, A–E). A similar trend was noted in the small intestine and mesenteric LNs (Fig. 2, F and G). We next evaluated if ILC-2 cell development in the Pdcd1−/− mice was altered in the presence of WT BM (i.e., in the 1:1 mixed ratio cohort). Our data suggests that ILC-2 cell numbers in the WT and Pdcd1−/− compartments were not significantly altered in the mixed chimera cohorts (1:1 ratio). In addition, these data further identified an intrinsic function for PD-1 in regulating KLRG1+ ILC-2 numbers. In line with the increase in cell numbers, we also found that the IL-13 cytokine production of KLRG1+ ILC-2s was increased in the lungs in the Pdcd1−/− cohorts but did not reach significance in 1:1 mixed chimera cohorts (Fig. 2, H and I). Therefore, lack of PD-1 did not alter KLRG1+ILC-2 cell development or function. To confirm the role of PD-1 in the development and intrinsic regulation of ILC-2s, experiments were repeated in Rag2−/−γc−/− hosts that lack host derived ILCs. These murine recipients were reconstituted with T depleted BM from WT or Pdcd1−/− mice. A significant increase in ILC-2 numbers were noted in the lungs of Pdcd1−/− BM recipients and a trend toward increase in ILC-2 cell numbers were noted in the small intestine (Fig. 2, J and K). No significant changes were noted in the ILC-2 cell numbers in the mixed chimera cohorts, both in the lungs and small intestine, which may indicate inhibitory signals from the WT compartment that can restrict Pdcd1−/− ILC-2 numbers. In summary the mixed chimera experiments demonstrate that PD-1 deficiency does not inhibit ILC-2 cell development.
PD-1 intrinsically inhibits STAT5 signaling in KLRG1+ ILC-2s
Because our data suggested that PD-1 intrinsically regulated ILC-2 numbers, we next tested if PD-1 altered the proliferative potential of KLRG1+ILC-2s. Lung-resident WT and Pdcd1−/− KLRG1+ ILC-2s were stimulated in vitro with various ILC-2 promoting cytokines, of which IL-33 in combination with IL-2 and IL-7 significantly enhanced the proliferation of Pdcd1−/− ILC-2s compared with WT ILC-2s. No significant proliferative difference was noted with other ILC-2 cytokines, i.e., IL-25 and TSLP (Fig. S2, A and B). We next investigated if the in vitro effect was replicated in vivo on IL-33 priming of WT and Pdcd1−/− murine recipients. Toward this goal, WT and Pdcd1−/− mice were treated with rmIL-33 for 3 d as previously described (Maazi et al., 2015), and then numbers of KLRG1+ ILC-2s were determined. Compared with WT mice, Pdcd1−/− mice had significantly increased numbers of KLRG1+ ILC-2 cells (Fig. 3, A and B). No difference in effector cytokine frequency was noted between WT and Pdcd1−/− recipients (Fig. S2 C), but a significant increase in the absolute numbers of IL-13+KLRG1+ cells was observed (Fig. 3 C). We next investigated if in vivo IL-33–stimulated KLRG1+ ILC-2s had enhanced proliferative potential. Our data demonstrated that KLRG1+ILC-2s from Pdcd1−/− mice had significantly enhanced proliferative capacity (Fig. 3, D and E).
On identifying the in vitro and in vivo effect of IL-33, the influence of IL-25 was investigated. IL-25 treatment did not alter the frequency and numbers of KLRG1+ ILC-2s in the Pdcd1−/− mice (Fig. S2, D and E).
The signaling pathways that resulted in an increase in KLRG1+ ILC-2 cells in Pdcd1−/− mice were next explored. Because ILC-2s depend on STAT5 signaling for proliferation and PD-1 inhibits STAT5 signaling in T cells (Franceschini et al., 2009), we explored if a similar mechanism operated in ILC-2s. To measure STAT5 signaling in the KLRG1+ ILC-2 subsets, WT and Pdcd1−/− mice were either treated with PBS or rmIL-33 for 1 d, as previously described (Maazi et al., 2015). KLRG1+ILC-2s were then ex vivo stimulated with IL-2 for 15 min, and then subjected to phospho-STAT5 flow cytometry. STAT5 signaling was significantly enhanced in the Pdcd1−/− KLRG1+ ILC-2 subsets that were treated with either PBS or rmIL-33 (Fig. 3 F). We next investigated if providing PDL-1 ligand had an effect in inhibiting STAT5 activation in WT KLRG1+ ILC-2 subsets. For these experiments, WT and Pdcd1−/− mice were treated with IL-33 for 3 d, and then KLRG1+ ILC-2 cells were stimulated in the presence of plate-coated PDL-1 fc-chimera plus IL-2 or IL-2 alone for 15 min, and then phospho-STAT5 flow cytometry was performed. In the WT mice treated with rmIL-33, sufficient STAT5 signaling was noted, which was significantly decreased in the presence of PDL-1 (Fig. 3, G and H). No difference in p-STAT5 signaling was detected in the Pdcd1−/− cohorts treated with PDL-1. In addition, these data suggest that ILC-2s from Pdcd1−/− mice exhibit robust STAT5 activation (one dose of in vivo IL-33), and this effect is recapitulated in the WT ILC-2s after three doses of in vivo IL-33. The initial boost in STAT5 activation, along with lack of response to PDL-1–mediated inhibitory effects, may contribute toward significant proliferation in the Pdcd1−/− cohorts.
To further explore this mechanism, we sequentially inhibited PD-1 or its downstream signaling components, namely SHP1/2 in ex vivo culture of KLRG1+ILC-2s. IL-33 was administered to WT and Pdcd1−/− mice for 3 d, and then KLRG1+ILC-2s were stimulated for 24 h in the presence of IL-2, IL-7, and IL-33. Certain cohorts were treated with an isotype control antibody (denoted as control in figure), anti–PD-1 antibody, selective SHP1/2 inhibitor (NSC87877), PDL-1 fc, and PDL-1 fc+SHP1/2 inhibitor. After 24-h stimulation, cells from the various cohorts were restimulated for 15 min with IL-2, and then subjected to phospho-STAT5 flow cytometry. We found that blocking PD-1 or its signaling pathway significantly increased p-STAT5 activation in KLRG1+ ILC-2s. This increase in p-STAT5 was inhibited when the PD-1 pathway was activated instead of blocked (by providing the ligand PDL-1fc). Inhibition of p-STAT5 in the cohorts that were treated with PDL-1 fc was significantly reversed upon blocking SHP1/2 (Fig. 3 I). Collectively, these data suggest that altering PD-1 signaling in KLRG1+ILC-2s can modulate STAT5 activation. The impact of these conditions (PD-1 signaling blockade) on nuclear STAT5a/b content in KLRG1+ILC-2s was investigated. A similar experimental approach was used, but the cells were expanded for 72 h instead of 24 h. After stimulation, cells from the different cohorts were isolated and nuclear extracts were prepared and subjected to a nuclear transcription factor assay. We found that blocking PD-1 significantly increased nuclear STAT5 in KLRG1+ILC-2s. An increase was also noted in the presence of SHP1/2 blocker, but did not reach statistical significance. Providing PDL-1 fc in these long-term cultures did not alter STAT5a nuclear content, but blocking SHP1/2 (in the PDL-1 fc group) did increase the content of nuclear STAT5a in KLRG1+ ILC-2s (Fig. 3 J). The nuclear content of STAT5b showed a similar profile to that of STAT5a (Fig. S2 f). Next, we investigated the reason for the lack of PDL-1 fc effect in long-term cultures. We found that IL-33 stimulation for 72 h up-regulated PDL-1 expression on KLRG1+ILC-2s (Fig. S2 g), suggesting that providing PDL-1 fc as a ligand may be redundant under these stimulating conditions. We finally evaluated the proliferation of KLRG1+ ILC-2s under these various culture conditions. Blocking PD-1 or SHP1/2 significantly enhanced the proliferative potential of KLRG1+ ILC-2s, which was significantly inhibited by PDL-1 fc stimulation. Similar to the p-STAT5 assays, blocking SHP1/2 in the presence of PDL-1 fc reversed the proliferative effect (Fig. S2, Ha and I). Therefore, these data suggest that modulating PD-1 signaling in KLRG1+ ILC-2s can alter STAT5 activation status and can also affect proliferation in long-term cultures.
Next, we explored if PD-1 signaling independently (bypassing STAT5) affected proliferation of KLRG1+ ILC-2s. We found that co-injection of both IL-33 and Tofacitinib (JAK/STAT inhibitor) in Pdcd1−/− mice significantly reduced KLRG1+ ILC-2 cell numbers and IL-13 cytokine production in the lungs (Fig. 3, K–M; and Fig. S2, J and K). These data demonstrate that the PD-1 pathway does not operate independently of STAT5 in inducing proliferation of KLRG1+ ILC-2s.
Pdcd1−/− KLRG1+ILC-2s gene expression pattern indicates enhanced effector function
To further explore the molecular mechanism of PD-1 regulation on ILC-2 cells, we performed a microarray assay to compare gene expression patterns of KLRG1+ ILC-2s between WT and Pdcd1−/−. Dozens of differentially expressed genes were identified, with 45 genes up-regulated and 7 down-regulated in Pdcd1−/− (false discovery rate [FDR] ≤ 0.15). Consistently, the expression levels of STAT5 target genes (e.g., slamf1 and csf1) were increased in PD-1–deficient KLRG1+ ILC-2s, whereas cell proliferation or activation-associated genes (e.g., klra9 and lztfl1) were also altered (Fig. 4 A). Gene ontology enrichment analysis of up-regulated genes revealed that the lack of PD-1 affected signaling pathway on ILC-2s, including innate immune responses, macrophage chemotaxis, cell-cell interaction and metabolism (Fig. 4 B). We also performed gene set enrichment analysis against published gene set (Duraiswamy et al., 2011) and the result showed that PD-1 deficiency had similar impact on gene expression patterns between ILC-2s and CD8+ T cells (FDR q-value = 0.036; Fig. 4 C), suggesting that a similar intrinsic regulatory mechanism of PD-1 signal may apply to both innate and adaptive lymphocytes.
KLRG1+ILC-2 cells from Pdcd1−/− mice possess anti-helminth immunity
We next evaluated if PD-1 deficiency modulated ILC-2 function during parasitic helminth infections. To investigate this, WT and Pdcd1−/− mice were infected with N. brasiliensis, and then KLRG1+ ILC-2 subset numbers were evaluated. In the lungs, the frequency and numbers of ILC-2 (GATA3+) or KLRG1+ ILC-2 subsets were similar in both WT and Pdcd1−/− cohorts on day 5 after worm infection (Fig. S3). In the Pdcd1−/− mice, near the site of infection (mesenteric LNs; MLNs), a significant increase in GATA3+ILC-2 frequency (Fig. 5 A) and absolute numbers (Fig. 5 B) was noted. In addition a significant increase in KLRG1+ ILC-2 subset frequency (Fig. 5 C), absolute numbers (Fig. 5 D), and cytokine production (Fig. 5, E–G) was also noted. Next, the function of the Pdcd1−/− KLRG1+ ILC-2 cells independent of adaptive immunity in clearing worm burden was evaluated. For these experiments, Rag2−/−γc-/− mice were infected with N. brasiliensis, followed by adoptive transfer of either WT or Pdcd1−/− KLRG1+ ILC-2 cells. A significant difference in the worm burden (Fig. 5 H) and worm eggs (Fig. 5 I) was noted in recipients that were reconstituted with Pdcd1−/− KLRG1+ ILC-2 subsets.
Blocking PD-1 in Rag1−/− mice enhances ILC-2 function during parasitic helminth infections
Experiments were performed to identify if exogenous blocking of PD-1 using an antibody would have similar effects on KLRG1+ ILC-2 subsets in vivo. Rag1−/−mice were injected with either isotype or αPD-1 antibody at day 0, and then treated with rm-IL-33 from day 1 to 3. In cohorts that were treated with both αPD-1 and rmIL-33 cytokine, a significant increase in KLRG1+ ILC-2 cell frequency numbers was noted (Fig. 6, A and B). To identify if blocking PD-1 can enhance therapeutic potential via ILC-2 cells, we repeated PD-1 blocking experiments in Rag1−/− mice infected with N. brasiliensis. Blocking PD-1 during ongoing helminth infection significantly enhanced ILC-2 numbers (Fig. 6 C) and cytokine production in MLNs (Fig. 6, D and E) and diminished both worm eggs in the feces (Fig. 6 F) and worm burden (Fig. 6 G). These results identify a potential wider application for anti–PD-1 therapeutics in helminth infections.
PD-1 regulates KLRG1+ ILC-2s in humans
We next explored if PD-1 is expressed on ILC subsets in human PBMCs. PD-1 was almost exclusively expressed on ILC-2s as compared with ILC-1 and ILC-3 subsets (Fig. S4, A and B). On further characterization based on IL-7Rα, KLRG1, and GATA3, we found that human KLRG1 subsets also expressed PD-1 at a similar frequency to murine ILC-2s (Fig. 7, A and B). Similar to murine data, PD-1+ subsets secreted significantly lower amounts of IL-5 and IL-13 compared with PD-1− subsets (Fig. 7, C and D). We next evaluated if PD-1 also intrinsically inhibited STAT5 signaling in human ILC-2s. STAT5 signaling was enhanced when PD-1 was blocked on ILC-2s, and this effect was reversed by the JAK/STAT inhibitor Tofacitinib (Fig. 7, E and F). We next investigated if nuclear STAT5a/b content in KLRG1+ILC-2s is altered when PD-1 pathway is manipulated. Similar to murine data, blocking PD-1 resulted in an increase in STAT5a/b nuclear content in human ILC-2s (Fig. S4, C and D). After this, we evaluated the downstream effects of STAT5 activation on ILC-2 proliferation. PD-1+GATA3+ cells were inherently less proliferative as compared with the PD-1−GATA3+ ILC-2s (Fig. 7 G; and Fig. S4, E and F). On inhibiting PD-1 signaling, an increase in proliferation was noted, which was reversed on inhibiting Jak/STAT signaling through Tofacitinib (Fig. 7 G; and Fig. S4, E and F). In summary, PD-1 intrinsically regulates human KLRG1+ ILC-2s by inhibiting STAT5 phosphorylation.
We next investigated if blocking PD-1 by antibody treatment increased human ILC-2 numbers and function in vivo. Initial experiments suggested that similar to our murine data, IL-33 was the most efficient stimuli that induced robust proliferation of PD-1+ ILC-2 cells (Fig. S4 G). NSG recipients were reconstituted with PBMCs and then treated with anti-PD-1 antibody along with rhIL-33 for 3 d. We found that blocking PD-1 significantly enhanced cytokine production in GATA3+KLRG1+ cell subsets (Fig. 7 h). The enhanced cytokine secretion capacity of ILC-2s was inhibited in cohorts that received Tofacitinib in addition to rhIL-33 and αPD-1 antibody treatment. Therefore, in humans PD-1 negatively regulates KLRG1+ ILC-2 cells by modulating the STAT5 activation pathway.
Approximately one-third of the world’s population is infected with one or more types of parasitic helminth worms however treatment options for these ubiquitous infections are limited (Hotez et al., 2008). Type 2 immune responses, which are initiated by ILC-2s, are pivotal in clearing helminth worms in adults, but the regulatory pathways that can modulate the function of ILC-2s are still largely unexplored. Emerging literature on ILC-2 populations in mice and humans (Maazi et al., 2015; Molofsky et al., 2015; Paclik et al., 2015; Seillet et al., 2016; Yu et al., 2016), combined with the data presented here, has identified a pivotal role for co-receptor–mediated regulation of ILC-2s. Whereas immune-activating signals that maintain ILC-2s have been extensively studied, there is paucity in understanding the regulatory pathways that modulate ILC-2 cell number and function in vivo.
This report is the first to demonstrate that IL-33–responsive KLRG1+ ILC-2s are regulated by PD-1 in both mice and humans. When ILC-2s are activated by IL-33 (through the IL-33 receptor ST2), they respond by secreting type 2 cytokines and up-regulate KLRG1. Our data indicates that PD-1 selectively inhibits the proliferation of KLRG1+ ILC-2s via modulating STAT5 function. Here, we propose a new regulatory axis, whereby activated mature KLRG1+ ILC-2s up-regulate PD-1 on exposure to IL-33. PD-1 then inhibits uncontrolled expansion of ILC-2 effector population by blocking proliferation. PD-1, therefore, is a critical negative checkpoint and contributes toward maintaining KLRG1+ ILC-2 cell numbers.
Recent studies indicated that PD-1 is a novel maker for ILC precursors (Yu et al., 2016), but whether PD-1 could regulate the development and function of ILCs, particularly ILC-2s, was unclear. We observed that, first, the frequency of mature KLRG1+ ILC-2s was increased in the Pdcd1−/− mice; second, reconstituted Pdcd1−/− ILC-2 cells were capable of secreting functional cytokines; and third, ILC-2 cell development was not compromised when BM from Pdcd1−/− mice was adoptively transferred into immune-deficient Rag2−/−γc−/− mice. Notably, Seillet et al. (2016) observed that Pdcd1−/− mice possess a similar frequency of ILC-2 precursors to WT mice. In addition, PD-1 has been suggested to be a marker on ILC-2 precursors (Yu et al., 2016). Our study is the first to identify a mechanistic role for PD-1 in mature ILC-2 and propose that although PD-1 is expressed on ILC-2 precursors, lack of PD-1 does not alter ILC-2 development or their cytokine production capacity in mixed chimera experiments.
In keeping with most ILC subsets, ILC-2s are responsive to γc cytokines (Bartemes et al., 2012) and explicitly depend on γc cytokines and STAT5 signaling for proliferation and survival. A specific increase in proliferation was noted in the Pdcd1−/− ILC-2s, which then led us to investigate the interplay between PD-1 and STAT5 signaling in these cells. Notably, STAT5 signaling can be modulated by co-receptors such as ICOS in ILC-2s (Maazi et al., 2015), and PD-1 signaling can inhibit T cell function and proliferation by dampening STAT5 (Francisco et al., 2009). In agreement with these previous studies, a significant increase in STAT5 phosphorylation and proliferation was noted in KLRG1+ ILC-2 numbers in Pdcd1−/− recipients, which was conserved in human ILC-2s. In all these studies (Francisco et al., 2009; Maazi et al., 2015), the PD-1 activating ligand that was primarily responsible for inhibiting STAT5 signaling is PDL-1, and hence we used a similar experimental methodology in the current study. We found that PDL-1 binding to PD-1 specifically down-regulated STAT5 activation in KLRG1+ ILC-2s.
Although the importance of PD-1 signaling in inhibiting STAT5 activation has been previously suggested in T cells (Francisco et al., 2009), two caveats remain. First, no direct mechanistic evidence has linked these pathways; second, it is not clear whether similar mechanisms exist in ILC-2s. Our in vitro and in vivo mechanistic data in murine and human ILC-2s has demonstrated that inhibiting or inducing PD1 signaling pathway can alter STAT5 activation on KLRG1+ILC-2s. Importantly, PD-1 does not operate independently of STAT5 in promoting ILC-2 cell proliferation.
In addition to cellular mechanisms, we sought to verify if Pdcd1−/− KLRG1+ ILC-2s exhibited a gene profile that correlated with enhanced effector function. The microarray data suggested that ILC-2s deficient in PD-1 had increased expression of STAT5 target genes such as slamf1 (Wang et al., 2015), and STAT5 activating genes such as csf1 (Novak et al., 1996). Of note, the gene profile of Pdcd1−/− ILC-2s followed a pattern that is found in effector memory CD8+ T cells. In particular, we found that klra9, which identifies exhaustion in CD8+ T cells (Wherry et al., 2007), was down-regulated in Pdcd1−/− ILC-2s. Gene set enrichment analysis further correlated with the cellular mechanistic assays in highlighting the enhanced effector function of Pdcd1−/− KLRG1+ILC-2s.
The function of Pdcd1−/− KLRG1+ILC-2 cells was found to be superior to their WT counterparts in eliciting host-mediated pathogen responses in in vivo worm expulsion experiments. Given that KLRG1+ ILC-2 subsets are important in the expulsion of worms (Huang et al., 2015), the potency of these Pdcd1−/− KLRG1+ILC-2 cells in Rag2−/−γc−/− identifies a new therapeutic role for PD-1 in inducing robust type 2 responses. Furthermore, the differential numbers of KLRG1+ILC-2 cells in the MLNs of Pdcd1−/− mice during infection suggests the importance of PD-1 as a checkpoint modulator that can regulate ILC-2 cell numbers in the MLN during host–pathogen–mediated immune responses. Of note is the lack of difference in ILC-2 cell numbers in the lungs of infected Pdcd1−/− mice. N. brasiliensis life cycle includes migration through the lungs; however, a clear difference in WT and Pdcd1−/− mice with respect to ILC-2 cell numbers was not observed. Hence, alternate regulatory mechanisms independent of PD-1 may inhibit ILC-2s in different organs during parasitic infections. This differential increase in ILC-2 numbers could be important in translating these findings to helminth immunity.
To confirm the potential therapeutic benefits in host–pathogen–mediated immune responses, PD-1 antibody experiments were performed in Rag1−/− recipients that were infected with N. brasiliensis. To date, the data presented here is the first to show that αPD-1 therapy results in an increase in KLRG1+ ILC-2 numbers, which is reflected in a significant decrease in worm burden. These results suggest that blocking PD-1 can skew immune responses to type 2 phenotype, thereby inducing strong antiparasitic responses. In addition, this observation can be extrapolated to the cancer immunotherapy trials (Topalian et al., 2012; Yaqub, 2015), whereby the primary reason for reduced autoimmune colitis in patients who receive PD-1 therapy can be attributed to increased type 2 responses (Ruyssers et al., 2009; Bashi et al., 2015).
Lastly, the murine N. brasiliensis experiments in Rag−/−γc−/− mice demonstrated that PD-1 intrinsically regulates ILC-2 cell function with minimal contribution from bystander immune cells. In the adoptive transfer experiments, murine recipients reconstituted with Pdcd1−/− ILC-2s had significantly lower worm burden than the WT cohorts. These data demonstrate that an intrinsic defect in PD-1 signaling in ILC-2s contribute to robust antiparasitic helminth responses. Indeed, if PD-1 elicits its function through bystander immune cells, no difference between cohorts that received WT and Pdcd1−/− ILC-2s will be noted.
In summary, we have demonstrated that ILC-2 cells are regulated by PD-1, lack of which enhances KLRG1+ ILC-2 cell numbers and effector function. The cellular signaling mechanism by which KLRG1+ ILC-2 cell numbers are increased in Pdcd1−/− mice is through enhanced STAT5 activation, which results in elevated γc cytokine–mediated cell proliferation and cytokine secretion. Collectively, these data represent an ideal translational route to manipulating ILC-2 cell numbers in vivo. Therefore, PD-1 blockade may be beneficial in situations that require significant ILC-2 cell numbers (helminth immunity) and this effect can be controlled by Tofacitinib to prevent deleterious uncontrolled type 2 inflammatory processes.
Materials and methods
Wild-type C57BL/6 (B6) and B6.Pdcd1−/− littermates were bred and maintained in a specific pathogen–free facility at the National Cancer Institute (NCI; Bethesda, MD). Mice were housed and maintained according to National Institutes of Health guidelines and all experiments were performed in accordance with the NCI Animal Care and Use Approval and institutional guidelines. All mice used in the experiments were age and sex-matched. CD45.1+ C57BL6, NOD-SCID-γc−/− (NSG) and B6. Rag2−/−γc−/− mice were obtained from Jackson Laboratories and maintained at the NCI or at Newcastle University, UK under a home office approved project license. All the experimental procedures performed at Newcastle University incorporated the NC3R guidelines for animal research, and the data were presented according to the ARRIVE guidelines.
Isolation of ILC-2 cells and ex vivo stimulation
Spleens and MLNs were mechanically crushed into single-cell suspensions. Lungs and small intestine were digested in Liberase TL (Roche) and DNase (Roche; Asseman et al., 2003), and ILCs were isolated as previously described. For skin ILCs, a small piece of mouse dorsal skin was removed, minced, and incubated with serum-free media containing Liberase TL and DNase at 37°C. Single-cell suspensions were prepared by mashing the tissue slurry through a 100-µm nylon mesh. Cells were sequentially filtered through 70- and 40-µm nylon filters. To assess production of cytokines, both murine and human ILC-2s were stimulated for 4 h in complete RPMI + 10% FBS with cell stimulation cocktail (PMA, ionomycin, Brefeldin A, and monensin; eBioscience) as per manufacturer’s instruction, followed by intracellular staining. Unstimulated controls were used to determine gating strategy for flow cytometric plots in figures. All human-derived materials were collected under an approved ethics plan (Durham and Tees Valley Research Ethics Committee [12/NE/0121], granted 16th April 2012).
BM mixed chimera experiments
BM cells from B6 or B6.Pdcd1−/− mice were harvested and T cell depleted. Congenic recipients B6.CD45.1+ mice were conditioned with total body irradiation (1,050 cGy) and rescued with 10million T-depleted BM (TDBM) cells from B6.CD45.1+ WT or B6 CD45.2+ Pdcd-1−/−mice. 8 wk later, lungs and small intestine were digested and immune cells were isolated followed by ILC-2 subsets characterization. Experiments were repeated in host B6.Rag2−/−γc−/− mice, where mice were reconstituted with either congenic WT BM (CD45.1) or Pdcd1−/−BM (CD45.2).
The following fluorochrome-conjugated antibodies were used for flow cytometry. Anti-CD3ε (145-2c11), anti-CD8 (53-6.7), anti-CD19 (6D5), anti–Gr-1 (RB6-8C5), anti-KLRG1 (2F1/KLRG1), anti-CD25 (PL61), anti-CD45.2 (104), anti–PD-1 (RMP1-30), anti-CD4 (GK1.5), anti–IL-5 (TRFK5), anti-CD45.1 (A20), anti-CD127 (A7R34), anti-CD25 (PC61), anti-CD5 (53-7.3), anti-CD11c (N418), anti-CD11b (M170), anti-Ter119, anti-CD45R/B220, and anti-F4/80 (BM8) were obtained from BioLegend. Anti–IL-4 (11B11) and anti-Thy1 (53–2.1) were obtained from BD. Anti-KLRG1 (2F1), anti-Ki67 (SoLA15), anti-Gata3 (TWAJ), and anti–IL-13 (eBio13A) were purchased from eBioscience. Anti-ST2 (DJ8) was obtained from MD Bioproducts. Single-cell suspensions were generated from indicated organs, and then incubated with anti-Fc receptor antibody (2.4G2) and stained with Lin PE-Cy5.5 (CD3+, CD4+, CD8+, NK1.1+, Gr1+, CD5+, Ter119+, CD11b+, CD11c+, F4/80+, CD45R/B220+, and CD19+), CD45 APC-Cy7, KLRG1 BV421 or APC-Cy7, ST2 FITC, GATA3 efluor660, and PD-1 BV605, CD127 BV510 or BV711, CD25 Pe-cy7, PDL-1 BV711, Thy1 PE. For intracellular (IC) cytokine measurement, cells were fixed, permeabilized, and stained with IL-13PE, IL-4 BV605, and IL-5 BV421. For human experiments, PBMCs were stained with Lin BV510 (CD19+, CD56+, CD3+, CD20+, CD16+, CD11c+, CD11b+, CD5+, TCRα/β+, and CD14+), PD-1 APC-Cy7 (clone EH12.2H7) or PD-1 Pe-Cy7 (clone:J10S), GATA3 efluor 660, KLRG1 PerCPefluor710 (clone 13F12F2), CD45 FITC (clone 2D1) or CD45 PeCy7 (clone H130), IL13 FITC (clone: JES10-SA2), IL5 BV421, and ST2 APC (R&D systems). ILCs were detected using the following gating strategy. First cells were gated as Lin−CD45+Thy1+, and then ILCs were defined as Lin−CD45+Thy1+ CD25+CD127+. ILC-2s were defined as Lin−CD45+Thy1+ CD25+CD127+ST2+KLRG1+. In human cells, ILC-2s were defined as Lin−CD45+CD127+GATA3+. Cells were analyzed using LSRII flow cytometer with FACSDiva software (BD) and analysis was performed with FlowJo 10.1 software (Tree Star).
In vivo ILC-2 cell stimulation assays
WT or Pdcd1−/− mice were treated with rmIL-33 (R&D Systems; 200 ng/mice) or PBS for 3 d, and then lungs were isolated for ILC-2 characterization and functional phospho STAT-5 phosphorylation assays. In certain experiments, Rag1−/− mice were treated with either αPD-1 antibody (BioXcell; clone RMP1-14; 250 µg/mice/d) or mouse IgG2a isotype control. After antibody treatment, mice were treated with rm-IL-33 or PBS, and then lungs were isolated for ILC-2 characterization. For human into murine experiments, PBMCs were injected into NSG mice (4 million/mouse), followed by i.p injection of rh-IL-2 (500 ng/mouse/d), rh-IL-7 (500 ng/mouse/d), and rh-IL-33 (200 ng/mouse/d). Certain cohorts were treated with mouse isotype control (250 µg/mouse/day) and DMSO Vehicle or with αPD-1 (250 µg/mouse/d) and Tofacitinib (15 mg/kg/d/mouse) for 3 d. Lungs were harvested at day 5 after adoptive transfer, and then human ILC-2 numbers and function were evaluated.
STAT5 phosphorylation assays
For STAT-5 phosphorylation assays, WT and Pdcd1−/−mice were treated with either PBS or rmIL-33 once, and then lungs were isolated the next day. ILC-2s were stimulated with IL-2 (100 IU) for 15 min, and then cells were stained for p-STAT5 and analyzed by flow cytometry. For PD-1 signaling assays, WT and Pdcd1−/− mice were treated with either PBS or rmIL-33 for 3 d, and then lungs were isolated. ILC-2s were stimulated with IL-2 (100 IU) alone or IL-2 plus plate-coated PDL-1-Fc chimera (5 µg/ml) for 15 min, and then stained with p-STAT5 antibody. For PDL-1-Fc chimera coating, PDL-1-Fc was resuspended in PBS (5 µg/ml), and then added to wells in a 24-well plate. The plate was then incubated for 3 h at 37°C and then washed three times with PBS and then used for cell culture. In a second set of experiments, WT ILC-2 cells were treated with rhIL-2 (40 ng/ml), rmIL-7 (40 ng/ml), and rmIL-33 (40 ng/ml). In addition, cohorts were treated with either isotype control antibody (10 µg/ml) or αPD-1 (10 µg/ml), or NSC87877 (SHP1/2 selective inhibitor; 10 µM resuspended in complete mouse media), PDL-1 fc (coated onto plates, and then washed with PBS three times before cell culture), or PDL1 fc plus NSC87877 for 24 h. ILC-2s were then restimulated with IL-2 for 15 min, and then subjected to p-STAT5 flow cytometry. For nuclear analysis of STAT5a/b, ILC-2s were stimulated for 72 h under similar conditions, and then restimulated with IL-2 for 15 min. Nuclear lysates were prepared according to the manufacturer’s instructions (Abcam), and then subjected to transcription factor ELISA-based assay (Abcam). Similar culture methodology was used for detecting human nuclear STAT5 activity in human ILC-2s. For in vivo STAT5 inhibition experiments, WT or Pdcd1−/− mice were treated with rmIL-33 plus DMSO or rmIL-33 plus Tofacitinib (15 mg/kg/day) i.p for 3 d. Lungs were evaluated for KLRG1+ILC-2 cell number and function after treatment.
N. brasiliensis infection
Mice were given subcutaneous injection of 300 third-stage N. brasiliensis larvae (provided by J.F. Urban Jr., US Department of Agriculture) and cell transfer was performed on the same day if required. Mouse feces were collected from each individual mouse on day 7–9 after infection, and eggs were counted. Worm burden in small intestines was measured on day 14.
For cell trace violet proliferation assays, murine ILC-2s and human ILC-2s were stained with cell trace violet in serum-free media as per manufacturer’s instructions. For murine cultures, ILC-2s were stimulated with rhIL-2 (40 ng/ml), rmIL-7 (40 ng/ml), and rmIL-33 (40 ng/ml), or rmIL-25 (40 ng/ml), or TSLP (40 ng/ml), and cultured for 72 h in cell culture media (RPMI, supplemented with 10% FCS, glutamine [2 mM], nonessential amino acids [0.1 mM], 2-mercaptoethanol [50 µM], sodium pyruvate [1 mM], penicillin and streptomycin [100 U/ml]), and then proliferation was measured by flow cytometry. Human ILC-2s were expanded in the presence of rhIL-2 (100 IU), rhIL-7 (40 ng/ml). Certain cohorts were treated with αPD-1 (20 µg/ml) or isotype control (20 µg/ml) alone with either Vehicle (DMSO) or Tofacitinib (0.3 µM). Where indicated, rhIL-33 (40 ng/ml) or rhIL-25 (40 ng/ml), or rhTSLP (40 ng/ml) or rhIL-4 (40 ng/ml), was added to the cultures. Cells were expanded for 5 d, and then proliferation was measured by flow cytometry.
For microarray analysis, flow sorted ILC-2s were pelleted, and then resuspended in TRIzol and shipped to Central Biotechnology Services (CBS; Cardiff University, Cardiff, Wales, UK). Total RNA extraction and microarray experiment was performed by the CBS. In brief, total RNA was isolated, and RNA quality was checked on Agilent Bioanalyzer. All samples used for microarray analysis have high quality score (RIN > 9). RNA was reverse transcribed and amplified, and sense strand cDNA was fragmented and labeled using Affymetrix WT Pico terminal labeling kit. Three replicates (each replicate was from five mice) of each group were hybridized to Affymetrix mouse Gene ST 2.0 GeneChip in Affymetrix hybridization oven at 45°C, 60 rpm for 16 h. Washing and staining were performed on Affymetrix Fluidics Station 450 and scanned on Affymetrix GeneChip scanner 3000. Data were collected using Affymetrix AGCC software. The data discussed in this publication have been deposited in NCBI Gene Expression Omnibus (GEO) and are accessible under GEO accession no. GSE94609. The R/Bioconductor package oligo (Carvalho and Irizarry, 2010) was used for preprocessing and normalization of the microarray data using RMA algorithm (Irizarry et al., 2003). The normalized expression values are log2-scaled, and probes were annotated (MacDonald, 2016). After excluding genes with low variance across arrays (<50% quantile) using genefilter (Bourgon et al., 2010) or without gene annotation, differentially expressed genes were identified as those with at least twofold differences and with FDR < 0.15 using limma (Ritchie et al., 2015). Gene Ontology enrichment analysis for up-regulated genes was conducted using DAVID bioinformatics resources (Huang et al., 2009). The GSEA software (Subramanian et al., 2005) was used for gene set enrichment analysis. The gene expression data were compared against the curated Immunological Signatures gene set (C7.all.v5.2), with the permutation type set as gene_set.
Results are expressed as mean with error bars showing ±SEM unless otherwise indicated. Statistical significance was determined for normally distributed data by using a Student’s t test or a one-way ANOVA, followed by appropriate multiple comparison tests. All analysis were performed using Prism 7.0 (GraphPad Software), and differences were considered significant when P ≤ 0.05.
Online supplemental material
Fig. S1 shows gating strategy for ILC analysis, cytokine receptor expression, and PD-1 expression during homeostasis and cytokine stimulation on ILC-2s. Fig. S2 shows the proliferative effect of ILC-2 driving cytokines on WT and Pdcd1−/− ILC-2s. Effect of in vivo IL-25 administration on ILC-2 cell numbers in WT and Pdcd1−/− ILC-2s, and the importance of PD-1 and STAT5 signaling in WT and Pdcd1−/− ILC-2s. Fig. S3 demonstrates the changes in ILC-2 numbers during helminth infection in WT and Pdcd1−/− mice. Fig. S4 demonstrates the significance of PD-1 in human KLRG1+ ILC-2s.
We thank J.F. Urban Jr. at the US Department of Agriculture for providing N. brasiliensis larvae. We would like to thank Dr. Amanda Redfern, Central Biotechnology Services, Cardiff University, for performing microarray experiments.
This work was supported by the Intramural Research Program, National Institutes of Health and a Newcastle University Research Fellowship, Newcastle University. N.J. Reynolds’s research is supported by the Newcastle NIHR Biomedical Research Centre and the Newcastle MRC/EPSRC Molecular Pathology Node.
N.J. Reynolds’s research group has received grant and consultancy income from Genentech through Newcastle University. The authors declare no additional competing financial interests.
Author contributions: S. Taylor performed experiments and analyzed data. Y. Huang performed N. brasiliensis experiments and wrote the paper. G. Mallett, C. Stathopolou, E.L. Martin, T.C. Felizardo, E.L. Woodward, J. Scott, and M.S. Elias performed experiments. M.-A. Sun analyzed microarray data. N.J. Reynolds critically read the manuscript and helped write the paper. W.E. Paul provided intellectual input and support for N. brasiliensis experiments. D.H. Fowler reviewed and edited the manuscript. S. Amarnath conceptualized, supervised, and planned experiments and wrote the paper.
S. Taylor and Y. Huang contributed equally to this paper.