Human NKp44⁺IL-22⁺ cells and LTi-like cells constitute a stable RORC⁺ lineage distinct from conventional natural killer cells

We have previously demonstrated that fetal LTi cells are capable of interacting with MSCs and inducing the expression of the adhesion molecules ICAM1 and VCAM1. We also found cells with a similar phenotype to that of the CD127⁺ fetal lymph node cells in postnatal tonsil, but we did not test the LTi activity of those cells in that study (Cupedo et al., 2009). Fig. 1 A shows that both Lin⁻ (CD14⁻CD19⁻CD3⁻TCR-αβ⁻TCR-γδ⁻) CD127⁺CD56⁺ and Lin⁻CD56⁻CD127⁺ LTi-like cells (referred to as CD127⁺ LTi-like cells), isolated from human tonsil, induced ICAM1 and VCAM1 on MSCs, whereas cNK cells had some ICAM-inducing capacity which was much lower than of CD127⁺ LTi-like cells and did not induce VCAM1 expression. The observation that tonsil CD56⁺CD127⁺ cells have functional LTi activity underlines the functional differences between these cells and cNK cells.

It has been suggested that expression of NKp44 identifies a population of RORC⁺IL-22–producing NK cells (Cella et al., 2009). Fig. 1 B shows that Lin⁻CD56⁺CD127⁺ and CD127⁺ LTi-like cells in human tonsils expressed high levels of NKp44 and were positive for CD161, whereas cNK cells expressed NKp44 on only a small subset in the tonsil and were negative for CD161. The CD127⁺ populations expressed CD117, which was not present on cNK cells, but were negative for CD34 (unpublished data). Furthermore, cNK cells expressed higher levels of NKp46 and NKG2D than the CD127⁺ populations. Tonsil LTi cells, but not cNK, also expressed CD25, the IL-2 receptor α chain, whereas the IL-15Rα chain on tonsil CD127⁺ LTi-like cells was expressed at levels equivalent to those on cNK cells (unpublished data). Our data show that the phenotype of CD56⁺CD127⁺ cells is very similar to that of NK22 cells, suggesting that these represent two highly overlapping if not identical cell types.

Recently, we found that fetal CD56⁻ LTi-like cells developed into CD56⁺CD127⁺ cells after 2 wk of culture with IL-15. Although these data suggested a precursor progeny relationship of LTi and CD56⁺CD127⁺ cells, we could not exclude the possibility that the CD56⁺ cells were derived from a small number of highly proliferating CD56⁺ cells contaminating the starting CD56⁻CD127⁺ population. To test whether tonsil CD56⁻ LTi cells also develop into CD56⁺ cells and to exclude the possibility that the appearance of CD56⁺ cells in cultures with IL-15 is the result of outgrowth of contaminating cells, we labeled highly purified tonsil CD56⁻CD127⁺ cells with CFSE and stimulated with IL-2, IL-7, or IL-15. These cytokines were all capable of maintaining LTi viability and proliferation ex vivo, although less proliferation was observed when stimulated with IL-7 alone. In the absence of these cytokines the cells did not survive. After 6 d of culture, the great majority of the LTi cells up-regulated CD56 (Fig. 1 C). Most importantly, CD56 was also up-regulated on CSFE-bright nondividing cells, which proves that the appearance of CD56⁺ cells is not the result of outgrowth. Additional kinetics experiments showed that this up-regulation is underway by day 4 of culture (unpublished data). We cannot completely exclude the fact that cells that did not up-regulate CD56 were more prone to cell death and that, therefore, these cells were underrepresented in our cultures. However, the observations that the cell viability was high (10–15% dead cells) in all conditions and that the cell numbers remained the same or increased during the cultures argue against selective death of CD56⁻ cells. Our data indicate that CD56⁺ cells did not arise from a small contaminating population of rapidly dividing CD56⁺ cells and, thus, this data supports a direct developmental relationship between CD127⁺ LTi-like cells and CD56⁺CD127⁺ cells.

Immature stage 3 NK cells have been shown to develop into cNK cells (Freud et al., 2006). Because considerable phenotypic similarities exist between LTi-like cells and stage 3 iNK cells, we examined whether CD127⁺ LTi-like cells can convert into cNK cells. Using a rigorous sorting strategy, we isolated CD56⁺CD117⁻ cNK cells, Lin⁻CD56⁺CD117⁺CD127⁺ cells, and Lin⁻CD56⁻CD117⁺CD127⁺ LTi-like cells from human tonsils. The use of CD117, in conjunction with CD127 as selecting markers, combined with the rigorous exclusion of Lin⁺ and/or TCR⁺ cells, was crucial for preventing contamination with T cells. These sorted populations were expanded ex vivo using a feeder mixture consisting of irradiated allogeneic PBMCs, JY EBV-transformed lymphoblastoid cells, PHA, and IL-2, which was previously used to expand cNK cells (Phillips et al., 1991). Under these conditions, CD127⁺ LTi-like cells expanded 30–200-fold, and after resting they could be successfully restimulated and further expanded. Importantly, expanded LTi-like cell lines maintained expression of RORC messenger RNA (mRNA), as did CD56⁺CD127⁺ cells, whereas CD56⁺ cNK cells remained negative (Fig. 2 A). At the end of the first feeder cycle, LTi-like cell lines were ∼30–60% CD56⁺ (Fig. 2 B). After subsequent feeder cycles, CD56 expression increased until ∼80–100% of the cells were CD56⁺. After expansion, all populations expressed similar levels of NKp46. CD127 was no longer detectable by flow cytometry; however, LTi-like cell lines continued to express CD127 message (Fig. S1) and to respond to IL-7 (unpublished data). The down-regulation of CD127 is not unexpected, as it is well known that T cells lose CD127 expression after activation and expansion (Xue et al., 2002). Expanded cNK cells expressed CD56, but in contrast to cell lines derived from CD127⁺CD56⁻ and CD127⁺CD56⁺ cells, they lacked CD117.

To exclude potential uncertainties regarding the use of bulk populations, we performed a clonal analysis of CD127⁺ LTi-like cells from two donors. Lin⁻CD117⁺CD127⁺ LTi-like cells were purified and cloned by limiting dilution at one and three cells per well of a 96-well round-bottomed plate. Approximately 8–10% of the wells seeded with one cell per well showed robust expansion and 10–20 clones from each donor were characterized further. All clones expressed high amounts of transcripts of RORC and AHR (Fig. 2 C), a transcription factor which is associated with Th17 cells and IL-22 production (Trifari et al., 2009), and a positive correlation existed between RORC and AHR expression levels (Fig. 2 D). Furthermore, all clones maintained expression of CD117 and NKp46 and most expressed CD56, which is consistent with observations of expanded CD127⁺ LTi-like cell lines (Fig. 2 E). In contrast, parallel cNK cell lines were negative for CD117 and did not express RORC.

LTi cell lines and clones maintained their ability to interact with MSCs and induce the expression of adhesion molecules, whereas cNK cell lines failed to mediate this activity (Fig. S2). Up-regulation of adhesion molecules could be mimicked by addition of exogenous TNF and lymphotoxin, and blocking TNF signaling through addition of TNFRII-Ig inhibited the up-regulation of adhesion molecules on MSCs induced by CD127⁺ LTi-like cell lines (Fig. S2). Collectively, our data demonstrate that LTi-like cells and their CD56⁺ descendants can be expanded ex vivo and retain their phenotypic and functional characteristics. Most importantly, neither the lines nor the clones convert into cNK cells in vitro.

In vitro expanded LTi-like cells have a similar cytokine profile after stimulation as ex vivo cells (Fig. 3 A). Stimulation with PMA/ionomycin was used to reveal the intrinsic capacity of cells to produce cytokines. IL-22 was produced in much higher amounts than IL-17, suggesting that the preferential production of IL-22 that was noted in freshly isolated cells was maintained through in vitro activation and expansion. After expansion, LTi-like cell lines produced amounts of IL-22 that were similar to those of the CD56⁺CD127⁺ cell lines upon activation, which is in contrast to the lower amounts produced by freshly isolated CD127⁺ LTi-like cells (Fig. 3 A; Cupedo et al., 2009). However, in contrast to fresh LTi cells, both CD56⁺CD127⁺ and LTi-like cell lines produced IFN-γ after expansion (Fig. 3 A). The IFN-γ–producing cells within the LTi cell lines may represent differentiated cNK cells that had lost the capacity to produce IL-22. However, the observation that activated IFN-γ–producing LTi cell lines were also positive for IL-22, TNF, and IL-17 (Fig. 3 B) argues against this possibility.

The cytokine production profile of LTi-like cell clones was similar to that of LTi cell lines (Fig. 3 C). Like the LTi lines, the clones acquired the capacity to produce IFN-γ. There was no inverse correlation between IFN-γ production and RORC mRNA in the clones (unpublished data), indicating that IFN-γ production was not associated with a decrease of expression of RORC. Clones derived from donor 1 had higher IL-17 production after stimulation than did clones derived from donor 2. Although most clones produced both IL-17 and IL-22 after stimulation, some clones produced IL-22 but not IL-17, whereas other clones produced IL-17 but not IL-22 (Fig. 3 D). These data indicate that some heterogeneity, with respect to IL-22 and IL-17 production, exists within RORC⁺ LTi cell clones. There was no correlation between CD56 expression and cytokine production profile. Over time, clones largely maintained their phenotype, although with in vitro culture an increase in IL-17 production was sometimes observed (Fig. S3).

It has been suggested that the capacity to produce IL-17 distinguishes LTi cells from NK22 cells (Colonna, 2009). However, our analysis suggests the presence of cells that can produce both IL-17 and IL-22, and even IL-17 only, within the CD127⁺RORC⁺ population which are also NKp44⁺ ex vivo.

We, and others, have documented that freshly isolated LTi-like cells from the tonsil, like their fetal counterparts, are negative for granzyme and perforin (Cella et al., 2009; Cupedo et al., 2009; Sanos et al., 2009). Because CD56⁺ LTi-like cell lines were capable of producing IFN-γ (Fig. 3), we assessed whether they acquired cytotoxic activity. After expansion, LTi-like cell lines and clones expressed perforin, but at much lower levels compared with cNK cell lines, and did not acquire expression of killer cell immunoglobulin-like receptors (KIRs; Fig. 4, A and B; and not depicted). In contrast to LTi-like cell clones, LTi-like cell lines did express some granzyme B but at much lower than cNK cell lines (Fig. 4, A and B). In a cytotoxic assay with K562 cells as targets, cultured LTi-like cells achieved 50% lysis at an effector/target ratio of 19:1, which is ∼13-fold higher than that required for cNK cell lines (Fig. 4 C), suggesting that in vitro–cultured LTi-like cells acquired a moderate cytolytic activity. However, this activity was much lower than that of cNK cells. The clones derived from LTi-like cells did not display any cytotoxicity in this assay. Collectively, our data support the notion that LTi and their CD127⁺CD56⁺ progeny represent a lineage whose developmental pathway is distinct from cNK cells.

It has been reported that circulating CD34⁺CD45RA⁺ cells that are also resident in the lymph node can differentiate into NK cells (Freud et al., 2005). Close inspection of those in vitro–generated NK cells reveals some similarities with LTi-like cells, in that they expressed CD117 and CD161, and part of the cells expressed CD25 (Freud et al., 2005, 2006), raising the possibility that some of those in vitro–generated “NK” cells are in fact LTi-like cells. Indeed, recently the same group reported that the stage 3 iNK cells, which were hypothesized to be downstream of CD34^dimCD45RA⁺ precursor cells, produced IL-22, suggesting that at least part of these iNK are identical to LTi-like cells (Hughes et al., 2009). Given that CD127⁺CD117⁺ cells cannot differentiate into cNK, the cNK precursor within the CD161⁺CD117⁺ iNK population described by Freud et al. (2006) should be negative for CD127. To test this hypothesis, we purified CD127 positive and negative cells from the Lin⁻CD161⁺CD117⁺ population and expanded them in vitro (Fig. 5 A). Fig. 5 B shows that, as expected, CD127⁺CD117⁺ cells maintain the expression of CD117 and had low levels of perforin and granzyme B. In contrast, the expanded CD127⁻CD117⁺ cells lost CD117, were positive for granzyme B and perforin, and had some KIR expression (Fig. 5 B). Both populations expressed NKp46, NKG2D, and CD56. Consistent with the cell surface phenotype, RORC and IL-22 mRNAs were highly expressed in expanded CD117⁺CD127⁺ cells, expressed at low levels in CD117⁺CD127⁻ cells, and not expressed in cNK cell lines (Fig. 5 C). In contrast, IFN-γ mRNA was expressed in all populations but most highly expressed in expanded cNK cells, whereas CD117⁺CD127⁻ cells had an intermediate expression (Fig. 5 C). These data indicate that CD127⁻ cNK precursors are found within the iNK cell population. Our data strongly suggest that iNK cells, as defined by Freud et al. (2006) and Hughes et al. (2009), may contain a mixture of LTi cells and immature cNK cells and do not necessarily contain common NK/LTi cell precursors. If such a precursor exists, it might be present in the CD34⁺CD45RA^± population (Freud et al., 2005, 2006).

The conclusion that cNK and RORC⁺ CD127⁺ LTi-like cells belong to different lineages is consistent with an accompanying paper (see Satoh-Takayama et al. in this issue). These researchers found that development of NKp46⁺RORC⁺ IL-22–producing cells, like that of LTi cells, requires IL-7 and is IL-15 independent. By using cell fate mapping, these researchers also observed that cNK cells never express RORC during their development. Because it is very likely that mouse NKp46⁺RORC⁺ cells, which also express CD127, are the mouse homologue of the human CD56⁺CD127⁺RORC⁺ cells described in this paper, the data of this group are in agreement with our conclusion that human CD127⁺RORC⁺ cells are unable to differentiate into cNK cells. Nonetheless, CD127⁺ LTi-like cells and their CD56⁺ descendants seem to be related to cNK cells because apart from the fact that these cell types share expression of certain markers, development of both lineages in the mouse can be blocked by E2A proteins (Boos et al., 2007) and consequently requires the E protein inhibiting factor Id2. These two lineages may serve comparable functions in the immune response. It is well established that cNK cells are necessary for the initial defense against viruses (Lee et al., 2007). It is possible that LTi-like cells and their RORC⁺ progeny are involved in the innate immune response against extracellular bacteria, much like the IL-22 and IL-17–producing T cells, which also express RORC and are instrumental in the adaptive immune response against bacteria (Aujla et al., 2008; Ishigame et al., 2009; Martin et al., 2009). Indeed, we have observed that LTi cells express toll-like receptors and are capable of responding to TLR stimulation (unpublished data). Non–T cell–derived IL-22 is required for protection from Citrobacter Rodentium infection in the intestine (Zheng et al., 2008), and IL-22–producing RORC⁺ cells contribute to this response (Satoh-Takayama et al., 2008; Cella et al., 2009). The RORC⁺ innate lymphocytes may constitute a heterogeneous group of cells, as indicated by our clonal analysis. In the intestinal lamina propria only IL-22–producing cells may be present, but is very well possible that similar cells producing IL-17 rather than IL-22 are prevalent in other mucosal sites such as the lung. We have also found that LTi cells and CD56⁺RORC⁺ cells are able to produce IL-5 and IL-13 (unpublished data) and RORC⁺ cells might exist in certain tissues that are dedicated to producing these cytokines rather than IL-17 and/or IL-22. Such cells may function in the innate immune response against parasites.

Human tonsils were digested for 30 min in 0.5 mg/ml collagenase IV (Invitrogen) at 37°C and processed to form a single cell suspension. CD3⁺ and CD19⁺ cells were depleted using kits (STEMCELL Technologies Inc.). For flow cytometry sorting, cells were labeled with PE-conjugated anti-CD127 (BD), PE-Cy7–conjugated anti-CD56 (BD), APC-conjugated anti-CD117 (eBioscience), Pacific blue–conjugated anti-CD19 (eBioscience), and FITC-conjugated anti-CD3, TCR-αβ, TCR-γδ, and CD14 (eBioscience) antibodies. In some conditions, PE-Cy5–conjugated anti-CD161 (BD) was also used. Cells were sorted on a FACSAria (BD) to >95% purity. Tonsils were obtained from Bio Options, and blood samples were obtained from the Genentech blood donors program after written informed consent. Ethical approval for the use of this material was obtained from the Western Institutional Review Board.

Human MSCs (Thermo Fisher Scientific) were cultured according to the supplier’s instructions. MSCs were plated in a 48-well plate at a density of 5 × 10⁴ cells per well. LTi-like or cNK cells were added at a 1:1 ratio and co-cultured for 4 d in the presence of 10 ng/ml IL-7. ICAM1 and VCAM1 expression on MSC was determined by flow cytometry. Where indicated, 10 ng/ml TNF or 100 ng/ml LTα1β2 were added at the start of culture. 100 ng/ml TNFRII-Ig was used to block TNF signaling and was added on days 0 and 2.

Flow cytometry–sorted cells were stimulated with irradiated allogeneic PBMCs (5,000 rad) and irradiated JY EBV-transformed B cells (7,500 rad) in the presence of 1 µg/ml PHA (Roche) and 100 U/ml IL-2 (Novartis). IL-2 was replenished every 3 d. LTi-like cell clones were generated using limited cell dilution and expanded using the same method.

Cells were cultured in Yssel’s medium (Gemini Bio-Products; or in house prepared) plus 1% human AB serum (Gemini Bio-Products). Recombinant human IL-7 and IL-15 were used at 10 ng/ml (R&D Systems). For cytokine production, 5 × 10⁵ cells per well were plated in a 200-µl volume in 96-well round-bottom plates.

K562 cells were labeled with a fluorescent dye using the Live/Dead cell–mediated cytotoxicity kit (Invitrogen) and co-cultured with effector cell lines at the ratios indicated for 5 h. Cytotoxicity was determined by propidium iodide staining using flow cytometry.

Total RNA was extracted using either the RNeasy mini or micro kit (QIAGEN) according to the manufacturer’s instructions. Complementary DNA was reverse transcribed using the high-capacity complementary DNA archive kit (Applied Biosystems). Primers and probes were all predesigned Taqman Gene Expression Assays from Applied Biosystems. Real-time PCR was performed on the 7500 Real Time System (Applied Biosystems) machine. The following assays were used: RORC (Hs01076112), AHR (Hs00169233), 18S (Hs99999901), and IL-22 (Hs00220924). All samples were normalized using 18S expression level and expressed in arbitrary units.

Cell surface staining was performed using Alexa Fluor 647–conjugated anti-NKp46 or -NKp44, Pe-Cy7–conjugated anti-CD25, and FITC-conjugated anti-CD62L (all obtained from BD). KIRs were detected using FITC-conjugated anti-KIR2DL2 or -KIR3DL1 (BioLegend). Cells were fixed and permeabilized as described later in this section and stained with PE-conjugated anti–granzyme B and FITC-conjugated anti-perforin (eBioscience). Intracellular staining was performed on ex vivo expanded cell lines and clones stimulated for 6 h with 10 ng/ml PMA and 500 nM ionomycin (EMD) in the presence of 5 µg/ml BFA (BD) for the final 4 h of culture. Cells were stained with LIVE/DEAD (violet) viability assay (Invitrogen) and then fixed in 3% paraformaldehyde in PBS. Cell permeabilization, staining, and subsequent washings were performed using the Fix/perm wash buffer (eBioscience). The following antibodies were used (BD): PECY7-conjugated anti–IFN-γ, and PE-Cy7–conjugated anti-TNF. PE-conjugated anti–IL-17A (eBioscience) and Alexa Fluor 647–conjugated anti–IL-22 (prepared in house) were also used. Data were acquired on an LSRII (BD) and analyzed with FlowJo software (Tree Star, Inc.). Monoclonal mouse anti–human IL-22 antibodies were generated in house by immunizing mice with human IL-22Fc fusion protein. Clone 3F11 was chosen based on its ability to detect intracellular human IL-22 (Trifari et al., 2009). The antibody was then conjugated with Alexa Fluor 647 fluorochrome using a Labeling kit (Invitrogen) according to the manufacturer’s instructions. During analysis, dead cells were eliminated using the live/dead violet dye and doublets were eliminated based on light scattering parameters.

Supernatants were collected, and IL-17A, IFN-γ, and TNF were measured by Bioplex bead-based assays (Bio-Rad Laboratories) and read by a Luminex System (Luminex). IL-22 was measured by ELISA (R&D Systems) according to the manufacturer’s protocol.

Fig. S1 shows CD127 mRNA expression on expanded cell lines. Fig. S2 shows up-regulation of adhesion molecules on MSCs induced by expanded LTi-like cell lines and clones. Fig. S3 follows cytokine production of individual clones over successive expansions in vitro.

We would like to thank Lewis Lanier for critical discussion of this manuscript. We greatly appreciate the excellent work done by Jim Cupp and the members of the Genentech flow cytometry laboratory on cell sorting and by Andres Paler Martinez on Luminex analysis.

T. Cupedo was supported by VENI grant 916.66.018 from the Netherlands Organization for Scientific Research (ZON-MW).

The authors declare that potential financial conflicts of interest may exist in that N.K. Crellin, S. Trifari, C.D. Kaplan, and H. Spits are or were employees of Genentech, a company that develops and markets drugs. The authors have no additional financial interests.