An Atypical Population of NK Cells That Spontaneously Secrete IFN-γ and IL-4 Is Present in the Intraepithelial Lymphoid Compartment of the Rat^1 2

The intestine is the largest lymphoid organ in the body, challenged constantly by an enormous quantity and diversity of Ags. The intestinal lymphoid compartment is not only large but also very diverse, and the phenotype and functions of its constituent cell populations are not fully characterized (1). In addition, mucosal lymphoid tissues appear to be regulated by unique mechanisms that govern highly specialized processes such as oral tolerance and controlled (or physiologic) chronic inflammation. Defects in the function of intestinal immune system regulation have been associated with celiac disease (2) and inflammatory bowel diseases (3, 4, 5). In addition, there are specialized lymphoid populations, in particular intraepithelial T lymphocytes (6), that respond to alternative pathways of activation. Coupled with the specialized Ag-presenting capabilities of intestinal epithelial cells (7), these lymphoid populations are thought to participate in immune responses that are under the control of distinct, but as yet poorly characterized, regulatory factors (1).

Using new methodology for the isolation and purification of rat intestinal intraepithelial lymphocytes (IEL),⁴ we previously identified a population of αβ- and γδ-TCR⁻NKR-P1A⁺ intraepithelial NK (IENK) cells (8). These cells were almost completely restricted to the CD4⁻CD8⁻ IEL population, and unlike peripheral NK cells in the rat, they were CD2⁻. We now report the phenotypic and functional characteristics of these IENK cells. The data suggest that the rat IEL compartment harbors a large population of NKR-P1A⁺CD3⁻ cells that function as NK cells, but display an activated phenotype and an unusual cytokine profile that clearly distinguish them from splenic NK cells. Their phenotypic and functional characteristics suggest that these distinctive IENK cells may participate in the regulation of mucosal immunity.

Wistar Furth (WF), Dark Agouti (DA), and Brown Norway (BN) rats of both sexes were obtained from Harlan Sprague Dawley (Indianapolis, IN). Additional WF rats of both sexes were obtained from Charles River Laboratories (Wilmington, MA). All animals were certified to be serologically free of Sendai virus, pneumonia virus of mice, sialodacryoadenitis virus, rat corona virus, Kilham rat virus, H1 (Toolan’s virus), mouse poliovirus (GD7), Reo-3, Mycoplasma pulmonis, lymphocytic choriomeningitis virus, mouse adenovirus, Hantaan virus, and Encephalitozoon cuniculi. All animals were housed in a viral Ab-free facility until used and maintained in accordance with recommendations in the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, National Research Council, National Academy of Sciences, 1996) and the guidelines of the institutional animal care and use committee of the University of Massachusetts Medical School.

For flow microfluorometry, fluorochrome- or biotin-conjugated mAbs directed against NKR-P1A (clone 10/78), CD2 (clone OX-34), CD3 (clone G4.18), CD5 (clone OX-19), CD8α (OX-8), CD25 (IL-2R α-chain, clone OX-39), CD28 (clone JJ319), CD45RC (clone OX-22), CD54 (ICAM-1, clone 1A29), and αβ-TCR (clone R73) were obtained from BD PharMingen (San Diego, CA). The hybridoma secreting the 6A5 rat anti-rat ADP-ribosyltransferase 2b (ART2b)⁵ (IgG1) mAb is maintained in our laboratory (8).

For measurement of IL-4-secreting cells by ELISPOT, paired mAbs (purified OX-81 and biotinylated B11-3) directed against IL-4 were obtained from BD PharMingen. For measurement of IFN-γ by ELISPOT, the capture mAb (clone DB-1) was obtained from BioSource International (Camarillo, CA), and the detection rabbit anti-rat IFN-γ polyclonal Ab was obtained from Torrey-Pines Biolabs (La Jolla, CA) and biotinylated as previously described (9). Isotype control mouse IgG1, mouse IgG2a, mouse IgG3, secondary Ab (anti-rat IgG1), alkaline phosphatase-conjugated streptavidin, and CyChrome-conjugated streptavidin were purchased from BD PharMingen. HRP-conjugated avidin D was purchased from Vector Laboratories (Burlingame, CA). Recombinant rat IL-2, IFN-γ, and IL-4 were purchased from R&D Systems (Minneapolis, MN).

IEL were prepared from rat small intestine using previously described methods with minor modifications (8). Briefly, the entire small intestine was gently flushed with cold RPMI 1640 (Life Technologies, Grand Island, NY) to remove lumenal contents and then incubated in ice-cold RPMI 1640 for 60–120 min to delaminate epithelial cells and IEL from the basement membrane. Detached cells were recovered by flushing the intestine twice with a total of 50 ml of HEPES-buffered HBSS medium containing 1 mM dithioerythritol (Sigma, St. Louis, MO) and 10% Fetalclone I (a defined neonatal bovine serum; HyClone Laboratories, Logan, UT) at 37°C. IEL were then purified by Percoll density gradient centrifugation, and viable cells exhibiting lymphoid morphology were quantified by the method of trypan blue using a hemocytometer.

Thymi, spleens, and mesenteric lymph nodes (MLN) were removed from rats killed in an atmosphere of 100% CO₂. Single-cell suspensions were prepared by gentle extrusion through stainless steel sieves into medium consisting of cold RPMI 1640 supplemented with 5% Fetalclone I as previously described (8, 10). Spleen cell preparations were washed in medium, and erythrocytes were lysed in a hypotonic buffer (0.15 M NH₄Cl, 1.0 mM KHCO₃, and 0.1 mM Na₂EDTA, pH 7.4). For cytotoxicity assays, erythrocyte-depleted splenocytes were suspended in medium and adsorbed against polystyrene plates for 1 h at 37°C, and nonadherent spleen cells were recovered by gentle rinsing of the plates with ice-cold medium.

One-, two-, and three-color flow microfluorometric analyses were performed as previously described (8). Briefly, 1 × 10⁶ viable splenocytes or IEL were suspended in ice-cold FACS medium (isotonic PBS supplemented with 1% Fetalclone I) and reacted with a mixture of 6A5 anti-ART2b mAb and/or biotin-conjugated Abs for 30 min at 4°C. Cells were then washed with FACS medium, reacted with FITC- and PE-conjugated Abs and/or CyChrome-conjugated streptavidin for 30 min at 4°C, washed with FACS medium, and fixed with 1% paraformaldehyde in PBS. FITC-, biotin-, and PE-conjugated isotype control Igs were used for all analyses. Cells were analyzed using a FACScan instrument (BD Biosciences, Mountain View, CA). Lymphoid cells were identified by their forward and side light scatter profiles.

IEL and splenocyte subpopulations were purified by flow microfluorometry. Freshly isolated IEL and splenocytes were reacted either with FITC-conjugated Abs directed against NKR-P1A or with FITC- and PE-conjugated Abs directed against NKR-P1A and CD3, respectively, and sorted by FACScan. As documented in Fig. 1 for one representative sample, the purity of all sorted IEL populations was >95%. The purity of sorted splenic T and NK cell populations was also documented by flow microfluorometry to be >97%. The percentage of contaminating NKR-P1A⁺ CD3⁺ (NK T) cells was consistently very low (<0.1%) in the sorted cell populations in all experiments.

Purified preparations of IEL and splenocyte subpopulations were suspended in isotonic PBS and spun onto glass slides using a Shandon Cytospin 2 (Shandon, Pittsburgh, PA). Cells were reacted with Wright-Giemsa stain and evaluated by light microscopy by a qualified pathologist.

Cytotoxic activity of lymphoid populations was measured using a previously described ⁵¹Cr release microcytotoxicity assay with minor modifications (11). NK-sensitive YAC-1 virus-induced mouse T cell lymphoma (12) and NK-resistant C58NT(D) (13) target cells were obtained from the American Type Culture Collection (Manassas, VA) and maintained in our laboratory in AIM-V medium (Life Technologies) supplemented with 2.2 × 10⁻⁵ M 2-ME. Target cells in growth phase were labeled with ⁵¹Cr as sodium chromate (150 μCi/million cells; New England Nuclear, Boston, MA), and 1.0 × 10^{4 51}Cr-labeled cells were added to each well of a 96-well microtiter plate. Unfractionated IEL, purified subpopulations of IEL, and adherent-cell-depleted splenocytes were added at E:T cell ratios ranging from 75:1 to 3:1, and the plates were incubated for 4 h at 37°C in a humidified atmosphere of 95% air-5% CO₂. All assays were performed in triplicate and averaged.

Total releasable radioactivity (cpm_maximal) was determined by incubating an aliquot of ⁵¹Cr-labeled target cells with 1% Triton X-100. After incubation, cells were pelleted by centrifugation, and 100-μl aliquots of supernatant were transferred to a separate microtiter plate containing 100 μl Optiphase Supermix β-scintillation fluid (Wallac, Gaithersburg, MD) and counted (cpm_test) using a 1450 Microbeta Trilux instrument (Wallac). In all assays reported here, release of ⁵¹Cr in the absence of effector cells (cpm_spontaneous) was <12% and <22% of the total ⁵¹Cr release for YAC-1 and C58NT(D) cells, respectively. Specific cytotoxicity was calculated as a percentage using the raw counts per minute and the formula: specific lysis (%) = ((cpm_test − cpm_spontaneous)/(cpm_maximal − cpm_spontaneous)) × 100%.

Inhibition of peripheral T lymphocyte proliferation by syngeneic WF IEL was measured in a one-way MLC. Responder cells were WF rat MLN cells. Stimulator cells were either syngeneic WF (RT1^u) thymocytes or allogeneic DA (RT1^a) or BN (RT1ⁿ) thymocytes. Stimulator thymocytes were given 2 Gy of γ-radiation (∼0.1 Gy/min) using a ¹³⁷Cs source (Gammacell 40; Atomic Energy of Canada, Ottawa, Canada). Assays were performed in 96-well plates incubated at 37°C in a final volume of 200 μl supplemented AIM-V medium in a humidified atmosphere of 95% air-5% CO₂. In preliminary experiments it was determined that optimal MLN cell proliferation in the absence of IEL was obtained after 5 days of incubation using 4.0 × 10⁵ responder cells and 8.0 × 10⁵ allogeneic stimulator cells. Unfractionated IEL and purified subpopulations of IEL syngeneic to the responder cells were added to cultures at IEL:responder cell ratios ranging between 1:400 and 2.5:1. [³H]Thymidine (1 μCi/well; New England Nuclear) was added to each well for the final 18–24 h of culture, and total incorporation of [³H]thymidine into DNA was determined as previously described (14). Radioactivity was measured using a 1450 Microbeta Trilux instrument (Wallac). All assays were performed in triplicate and averaged. The extent of suppression was calculated as a percentage using the raw counts per minute and the formula: % suppression = (1 − ((cpm_{test allo} − cpm_{test syn})/(cpm_{max allo} − cpm_{max syn})) × 100%, where cpm_{test allo} are counts from the incubation of IEL, responder, and allogeneic stimulator cells; cpm_{test syn} are counts from incubation of IEL, responder, and syngeneic stimulator cells; cpm_{max allo} are counts from the incubation of responder and allogeneic stimulator cells only; and cpm_{max syn} are counts from the incubation of responder and syngeneic stimulator cells.

To determine whether a soluble factor mediated IEL-induced suppression, MLC were performed using 24-well Costar Transwell plates (3 μm pore diameter, Fisher Scientific, Pittsburgh, PA). Responder WF MLN cells were incubated in PBS (5.0 × 10⁷ cells/ml) containing 1 μM CFSE for 15 min at 37°C as previously described (15). After labeling, responder cells were incubated with stimulator thymocytes from syngeneic WF or allogeneic DA or BN rats prepared as described above, with 1.0 × 10⁶ WF responder cells and 3.2 × 10⁶ stimulator cells placed in the lower chamber of each test well. IEL from 5- to 7-wk-old WF rats (2.0 × 10⁵ cells) were added to either the lower chamber or the upper insert chamber of test wells. Cell mixtures were cultured in supplemented AIM-V medium at a final volume of 800 μl/well. Plates were incubated at 37°C for 5 days in a humidified atmosphere of 95% air-5% CO₂. After harvest, the intensity of the intracellular CFSE signal in viable lymphocytes was measured by flow microfluorometry and used as an index of cell proliferation as previously described (16). The absolute number of mitotic events that had occurred per 10⁴ responder cells in the analysis was calculated using the method of Wells et al. (17).

IFN-γ and IL-4 production was first measured using a previously described ELISPOT assay with minor modifications (18, 19). Briefly, 96-well Nunc Maxisorp plates (Fisher Scientific) were incubated overnight at 4°C with coating Abs (1 μg/well) diluted in sterile PBS. Plates were washed once with PBS (4°C) and blocked for 1–2 h with PBS containing 4% BSA (w/v). Plates were then washed once with PBS at 4°C, cells to be assayed were suspended in supplemented AIM-V medium and 100 μl of the cell suspension was added to each well. The optimal number of IEL assayed per well was determined in preliminary experiments and varied from 1 to 9 × 10⁴. The optimal number of splenocytes assayed per well was also determined in preliminary experiments and varied from 1 to 52 × 10⁴. For all ELISPOT analyses, plates were incubated at 37°C in a humidified atmosphere of 95% air-5% CO₂ for 20–24 h in the absence of additional stimulation except as noted. Stimuli included recombinant rat IL-2 (10 or 6000 U/ml) or the combination of PMA (5 ng/ml) and ionomycin (0.4 μg/ml). After incubation, culture medium was decanted, and distilled H₂O containing 0.05% Tween 20 was added to the wells for 5–10 min to lyse the cells. Plates were then washed five times with PBS/0.05% Tween 20 using an Ultrawash Plus plate washer (Dynatech, Burlington, MA). Biotinylated detection Abs (0.5 μg/well diluted with PBS containing 1% BSA) directed against IL-4 and IFN-γ or an irrelevant Ag (trinitrophenol, TNP) were then added to test wells. Plates were agitated on a plate shaker at room temperature for 60–120 min and then washed five times with PBS-0.05% Tween 20.

Next, 100 μl of a 1/1000 dilution of alkaline phosphatase-conjugated streptavidin in PBS containing 1% BSA was added to each well. Plates were again agitated at room temperature for 60–120 min, washed 10 times with PBS-0.05% Tween 20, and then washed once with PBS alone. A solution of 0.6% agarose (w/v, type VII low gelling temperature; Sigma) containing 0.1 M 2-amino-2-methyl-1-propanol (Sigma), 16.2 mM MgCl₂, 0.008% Triton X-405 (v/v; Sigma), and 2.3 mM 5-bromo-4-chloro-3-indolyl phosphate (Sigma) was preheated to 42°C and layered onto the plates, which were then incubated overnight at 37°C in a humidified atmosphere of 95% air-5% CO₂. The number of discrete spots visible in each well was counted using a light microscope at ×2 power. Each cytokine assay was performed using four wells, three of which were assayed using the cytokine-specific developing Ab and one of which was assayed using the irrelevant biotinylated control Ab. For analysis, the number of spots observed in the control well for each quadruplicate assay was subtracted from the number of spots counted in each of the three test wells. The maximum number of spots observed in any control well was three. The adjusted number of spots present in each of the test wells was then averaged. To account for the various numbers of cells added per well, data have been normalized to numbers of spots per 10⁴ cells.

The amount of IFN-γ and IL-4 secreted by IEL was quantified using a modification of the cell-ELISA (20), which is designed to avoid artifact influenced by the consumption of the cytokines that are secreted by cultured cells. For these assays, unstimulated unsorted, sorted, and remixed IEL (1–5 × 10⁴/well) were processed exactly as detailed in the first paragraph of the ELISPOT method described above (through the end of the culture, addition of biotinylated Abs, and washes). For quantification, additional wells were prepared in the same way at the same time, but graded amounts of IFN-γ or IL-4 instead of cells were added to these wells. Threefold dilutions of cytokine were added to wells in amounts that ranged from 1.2 to 900 pg for IFN-γ and from 0.1 to 33.3 pg for IL-4.

After the incubation, addition of detection Abs directed against IL-4 and IFN-γ or TNP and washes, 100 μl of a 1/2500 dilution of HRP-conjugated avidin D in PBS containing 1% BSA was added to each well. Plates were agitated at room temperature for 60–120 min, washed 10 times with PBS-0.05% Tween 20, and then washed once with PBS alone. Next, 100 μl of a solution containing 0.05 M Na₂HPO₄, 0.025 M sodium citrate, 1 mg/ml o-phenylenediamine (Sigma), and 1 μl/ml 30% (v/v) H₂O₂ was added to each test well, and plates were incubated at room temperature for 5–10 min. The reaction was stopped by adding 25 μl of 3 N HCl to each well, and absorbance at 790 nm was measured using a microplate reader (model 3550; Bio-Rad, Hercules, CA). Each IEL population (unsorted, sorted, remixed) was assayed in six wells, of which three were developed using the cytokine-specific Ab and three with the anti-TNP Ab. Each standard was assayed in three wells that were developed using the cytokine-specific Ab.

For wells to which the cytokine alone had been added, a standard curve was generated by linear regression using the software supplied with the plate reader (Microplate Manager PC, version 4.0, Bio-Rad). Curves for both IFN-γ and IL-4 were linear in the region of interest (r > 0.98 for each assay). The amount of cytokine present in each well that had been incubated with IEL was determined by the software. Because the cytokines produced in the course of the ELISPOT were captured and not degraded, the data are presented as picograms of cytokine produced per 10⁴ cells per 24 h. No cytokine was detected in wells that were developed with the TNP-specific detection Ab or in wells to which neither cells nor cytokines had been added.

Parametric data are presented as the arithmetic mean ± SD, except in figures depicting the results of a single experiment performed in triplicate or quadruplicate. In those instances the result is shown as the mean ± SEM. Pairs of means were compared using two-tailed t tests with separate variance estimates and the Bonferroni adjustment for multiple comparisons as required (21, 22).

In a preliminary experiment IEL were isolated from WF rats of different ages, and the percentage of NKR-P1A⁺ cells present was measured by flow cytometry. As shown in Fig. 2, the percentage of NKR-P1A⁺ cells present in the IEL compartment declined with age, decreasing from ∼50% in weanling rats to ∼17% in retired breeder rats >16 wk old. At certain time points, the percentages of NKR-P1A⁺ CD3⁺ or NKR-P1A⁺ αβ-TCR⁺ (NK T) cells present in the IEL compartment were also measured and were consistently <2% (Fig. 2). The absolute number of IENK cells was determined at two time points. We observed that the yield of IEL from 4- to 6-wk-old rats was 5.1 ± 1.9 × 10⁶/rat (n = 7), of which an average of about 2.5 × 10⁶ were IENK cells. The yield of IEL from much larger rats >16 wk old was 27.2 ± 6.4 × 10⁶/rat (n = 11), of which an average of about 5 × 10⁶ were IENK cells. Except as noted, rats 4- to 8-wk old were used in subsequent experiments.

It is known that intraepithelial and splenic NK cells differ with respect to the expression of at least two cell surface Ags. The majority of splenic NK cells express CD2 and CD8α, whereas <10% of IENK cells express these surface Ags (8, 23). To investigate whether intraepithelial and splenic NKR-P1A⁺ CD3⁻ NK cells differ phenotypically in other respects, we measured the expression of a panel of cell surface Ags on both populations. Using 7-wk-old WF rats, we confirmed that the percentage of cells expressing CD2 and CD8α was significantly lower in the IENK than in the splenic NK cell population (Table I). In addition we observed several other differences of substantial magnitude. The percentage of cells expressing CD5, CD45RC, and CD54 was also significantly lower in the IENK than in the splenic NK cell population. Conversely, the percentage of cells expressing CD25 and ART2b was significantly higher in the IENK than in the splenic NK cell population. Only CD28 was expressed on similar percentages of IENK and splenic NK cells. Representative histograms documenting the expression of CD2, CD8α, ART2b, CD25, and CD45RC on IENK and splenic NK cells are shown in Fig. 3.

Because IENK cells differ from splenic NK cells phenotypically, these populations were next purified by flow cytometry as described in Materials and Methods and examined for the presence of the azurophilic cytoplasmic granules that are characteristic of peripheral NK cells (11). Light microscopic analysis revealed that nearly all cells in the purified IENK population had the appearance of large granular lymphocytes with azurophilic granules (Fig. 4,A). These granules were even more prominent than those present in purified splenic NK cells (Fig. 4,B). Azurophilic granules were absent in the majority of intraepithelial (Fig. 4,C) and splenic (Fig. 4 D) T cells.

We next measured specific lysis of NK-sensitive YAC-1 cells (12) and NK-resistant C58NT(D) cells (13) in the presence of unfractionated IEL and monocyte-depleted spleen cells. Both IEL and monocyte-depleted spleen cells killed YAC-1, but not C58NT(D), targets (Fig. 5,A). Flow microfluorometric analysis of the IEL and spleen cell populations in these two experiments revealed that 44–58 and 8–9% of each population, respectively, consisted of NKR-P1A⁺ cells. Reanalyzing the data in Fig. 5,A using E:T cell ratios based on these percentages, we found that the specific killing of YAC-1 cells by NKR-P1A⁺ IENK and splenic NK cells was similar at similar calculated NK to target cell ratios (Fig. 5 B).

To exclude the possibility that cytotoxic T cells may have contributed to YAC-1 killing by unfractionated IEL, the experiment was repeated using purified populations of NKR-P1A⁺ and NKR-P1A⁻ IEL. In this experiment 35% of IEL were NKR-P1A⁺. We observed that NKR-P1A⁺ IEL efficiently killed YAC-1 targets, whereas NKR-P1A⁻ IEL did not (Fig. 5,C). In control experiments we observed that unstained, unsorted IEL and stained, unsorted IEL exhibited similar YAC-1 killing activities. Finally, sorted NKR-P1A⁺ and NKR-P1A⁻ IEL fractions that were remixed at the presort ratio of 35:65 also exhibited YAC-1 killing at IEL to target ratios comparable to those observed using unsorted IEL (Fig. 5 C).

Because splenic NK cells are known to suppress MLC (24, 25), we analyzed NK cells of intraepithelial origin for similar capability. In preliminary experiments, unfractionated IEL isolated from 7- to 15-wk-old WF rats were tested for their ability to suppress a one-way allogeneic MLC. We observed that responder cell proliferation to alloantigen was substantially reduced in the presence of IEL at IEL/responder ratios of ∼1:10 or greater (Fig. 6 A). At such ratios, proliferation was reduced to levels observed in control experiments using syngeneic stimulator cells. In additional control experiments, responder cells did not proliferate in the presence of IEL alone, nor did IEL proliferate in the presence of allogeneic stimulator cells alone (data not shown).

Having demonstrated that IEL can suppress a one-way MLC, we next sought to determine which subpopulation of IEL was responsible for the effect. IEL from WF rats were fractionated into NKR-P1A⁺CD3⁻ IENK and NKR-P1A⁻CD3⁺ intraepithelial T cell (IE-T cell) subpopulations, and their abilities to suppress alloreactive T cell proliferation were compared. As shown in Fig. 6,B, the suppressive activity of sorted IEL appeared to be restricted exclusively to the IENK subpopulation. There was no suppressive activity associated with sorted IE-T cells. The suppressive activity of sorted IENK and IE-T cells that were subsequently remixed was comparable to those of both sorted IENK cells alone (Fig. 6,B) and unsorted IEL (Fig. 6 A).

Having demonstrated that NKR-P1A⁺CD3⁻ (IENK) cells are responsible for IEL-mediated suppression of a one-way MLC, we next sought to determine whether a soluble mediator or cell-to-cell contact was more likely to be responsible for this effect. In this experiment MLC was performed using Transwell methodology, and the proliferation of CFSE-labeled responder cells in contact with unfractionated IEL was compared with the proliferation of responder cells separated from IEL by a permeable membrane. As shown in Fig. 7, retention of fluorescence, indicative of suppression, occurred when responder cells, allogeneic stimulator cells, and IEL were incubated on the same side of the Transwell membrane. Loss of fluorescence, indicative of responder cell proliferation and an increase in the absolute number of mitotic events, occurred when responder cells and allogeneic stimulator cells were separated from IEL by the membrane.

To investigate the possible immunoregulatory role of IENK cells, we measured their secretion of IFN-γ and IL-4 using ELISPOT methodology. In a preliminary experiment IEL were isolated from 5- to 11-wk-old WF rats and incubated overnight in the presence of PMA/ionomycin or IL-2 (10 U/ml). We observed that the frequencies of IFN-γ-secreting IEL under these conditions of stimulation were 25 and 45 spots/10⁴ cells, respectively. However, in three separate assays, the frequencies of cells secreting IFN-γ in the absence of any in vitro stimulating agent were comparable: 33, 43, and 23 spots/10⁴ cells. The results for IL-4 in this same experiment were similar. PMA/ionomycin- and IL-2-stimulated cells generated 22 and 55 spots/10⁴ cells, respectively, and cells cultured in the absence of in vitro stimulation yielded 35, 39, and 46 IL-4 spots/10⁴ cells.

We next tested the hypothesis that IENK cells spontaneously secrete IFN-γ and IL-4 and were the source of these cytokines in our preliminary experiment. Because the secretion of IL-4 has not previously been associated with peripheral NK cells, populations of spleen cells were assayed in parallel with cells of intestinal origin. As shown in Fig. 8, we confirmed by ELISPOT that the frequency of unstimulated IEL that produced either IFN-γ or IL-4 was substantial. In contrast, few unstimulated splenocytes produced IFN-γ, and none appeared to produce IL-4. Addition of up to 6000 U/ml IL-2 to these cultures resulted in only a modest increase in the frequency of cells that secreted IFN-γ and no increase in the number of cells that secreted IL-4.

Analysis of sorted populations of IENK and IE-T cells localized the production of both IFN-γ and IL-4 by IEL to the IENK subpopulation (Fig. 8). Additional assays confirmed that unstimulated splenic NK and T cells did not spontaneously secrete IL-4, and only a few secreted IFN-γ. Addition of 10 U/ml IL-2 to the splenic NK cell cultures had little effect on the number of cells that produced either cytokine. When 6000 U/ml was added to the splenic NK cell cultures, the frequency of cells that produced IFN-γ rose to levels comparable to those observed for unstimulated IENK cells, but the number of cells that produced IL-4 remained low. The frequency of sorted intraepithelial and splenic T cells that secreted IFN-γ was small, and no cells that secreted IL-4 were detectable. Comparable and consistent results were obtained in control assays of stained, unsorted cells and a remixed population of sorted NK and T cells (Fig. 8).

To exclude further any possibility that contaminating IE-T cells may have been responsible for the cytokine production attributed to the IENK population, an additional ELISPOT assay was conducted. In this assay graded numbers of purified IE-T cells were added to a constant number of purified IENK cells, and the number of IFN-γ- and IL-4-producing cells was measured as described above. As shown in Fig. 9, there was no detectable change in the number of cytokine-producing cells as a function of the number of IE-T cells that were added to the IENK cells. In each experimental condition the number of cytokine-positive cells, when expressed as positive cells per 10⁴ IENK cells in the assay, remained constant.

Having quantified the number of IENK cells that produced IFN-γ and IL-4 spontaneously, we next sought to determine the amount of cytokine produced. To circumvent the problem of reuse of cytokines inherent in such determinations, we developed the modified cell-ELISA (20) described in Materials and Methods. As shown in Fig. 10, we again confirmed that unstimulated IEL and IENK cells, but not IE-T cells, produce readily detectable quantities of both IFN-γ and IL-4. IENK cells produced IFN-γ at a rate of ∼150 pg/10⁴ cells/24 h and IL-4 at a rate of ∼1 pg/10⁴ cells/24 h. Results for unsorted and remixed IEL gave consistent results based on the fraction of IENK cells present in these populations.

Using new methodology we have identified a novel population of IENK cells that appear constitutively to secrete IL-4 and IFN-γ. These unusual IENK cells are similar to peripheral NK cells in morphology, in their ability to lyse NK-specific targets, and in their ability to suppress a one-way MLC. However, they are markedly different from peripheral NK cells in two respects. First, IENK cells express high levels of CD25 and ART2, whereas peripheral NK cells do not. Second, unlike splenic NK cells that secrete little or no IL-4 and secrete IFN-γ only in response to stimulation (26), a substantial fraction of rat IENK cells appear to secrete IFN-γ or IL-4 spontaneously.

We (8) and others (23) have previously reported that cells with an NKR-P1A⁺CD3⁻ phenotype are present in the intraepithelial compartment of the rat intestine. Our data now extend these observations by demonstrating that IENK cells appear to represent a new, previously unrecognized subpopulation of rat NK cells. It could, of course, be argued that the IL-4 secretion we observed in rat IENK cell populations resulted from contamination of the IENK cell preparation with either NK T cells or contaminating IE-T cells, both of which are known to secrete IL-4 (6, 27). However, this is unlikely because our assays revealed that NK T and IE-T cells comprised <0.1% of the purified IENK cell preparations used in our ELISPOT assays. This frequency of contaminating cells is insufficient to account for the 1–2% frequency of IL-4-secreting cells we observed in the purified IENK populations. In addition, we observed that the purposeful addition of IE-T cells to the IENK ELISPOT had no effect on the number of cytokine-secreting cells.

It could also be argued that the apparently spontaneous secretion of IL-4 was simply the result of the IENK cells being activated, as are nearly all IELs (6, 28). However, this is also unlikely because our data show clearly that peripheral NK cells, even when activated with high levels of IL-2, secrete little IL-4. The ability of freshly isolated IENK cells to secrete IL-4 suggests that these cells may represent a new subpopulation of NK cells that, in the microenvironment of the IEL compartment, are programmed to secrete IL-4 constitutively. However, it should be pointed out that our data do not formally exclude the possibility that factors other than IL-2 or IL-4 could be present in vivo in the gut, activate IENK cells, and induce them to secrete IL-4. Whether any individual IENK cells secrete both IFN-γ and IL-4 spontaneously also cannot be determined from our data.

Consistent with the view that IENK cells comprise a distinct NK subpopulation is our documentation that they differ phenotypically from splenic NK cells. Expression of CD2, CD8α, CD5, CD45RC, and CD54 is lower in intraepithelial than in splenic NK cell populations. In contrast, expression of CD25 and ART2 is higher. Although the expression of CD25 could simply reflect the constitutively activated state of IENK cells, ART2 expression on peripheral T cells is known to be down-regulated in the setting of activation (29).

Although the cytokine profile of rat IENK cells appears to be unique among cells with the NKR-P1A⁺CD3⁻ phenotype, it is important to note that these cells do share important morphological and functional characteristics with peripheral NK cells. Both are large and exhibit distinctive azurophilic granules (30). We recognize that γδ-TCR⁺ IEL have been reported to have azurophilic granules (31), but this cell subset was excluded from the IENK cell specimens that we analyzed. Our morphological studies also suggest that IENK cells may be slightly larger than those in the splenic population, consistent with their activated state (26). However, this conclusion must be regarded as tentative as we cannot exclude the possibility of artifact induced by the cytospin method used for specimen preparation.

We also observed that, like peripheral NK cells (11, 12, 24, 25), IENK cells exhibit cytotoxic activity and suppress a one-way MLC. The cytotoxicity data, which showed that purified rat NKR-P1A⁺ IEL lyse YAC-1 cells more efficiently than do unsorted or remixed populations, are consistent with studies performed using mouse spleen cells (30). We recognize that NK T cells were present in our cytotoxicity assays, but it is unlikely that they affected the outcome because, unless they are preincubated in the presence of high dose IL-2, NK T cells are ineffective in lysing YAC-1 cells (27). With respect to the suppression of MLCs by rat IENK cells, it should be pointed out that comparable numbers of IE-T cells did not result in suppression, arguing against competition for nutrients in the medium as a mechanism of suppression. NK-mediated suppression of a one-way MLC is thought to occur via killing of allogeneic APC (24, 25), a conclusion supported by our Transwell studies. These studies suggest that cell-cell contact between IENK and stimulator/responder cells may be required for suppression, but it must be recognized that they do not formally exclude the possibility that soluble mediators that are either short-lived or present at low concentrations could be involved.

The constellation of NK-like phenotype and function together with an unusual cytokine profile, i.e., constitutive expression of IL-4 and IFN-γ in the local microenvironment of the gut, suggest that these are specialized cells that may be subserving a gut-specific function. If the IENK cell population is indeed a new subpopulation of intraepithelial cells, how might they influence gut immunity? We propose that IENK cells have previously unrecognized regulatory properties. The gut is known to exhibit unique immune responses that differ from those in the systemic circulation and can influence the response of the peripheral immune system (1).

Two unique immune responses mediated by the gut are regulation of controlled or physiologic inflammation (1), and mediation of oral tolerance upon encounter with exogenous Ags (1, 6, 28, 32). The ability of IENK cells to spontaneously (in essence, constitutively) secrete IL-4 implies an important role vis-à-vis these two unique functions of the gut immune system. IL-4 is known to be a potent regulator of inflammatory processes and biases inflammation toward a nondestructive rather than a proinflammatory destructive process (33, 34, 35). IL-4 has also been demonstrated to protect animals from inflammatory bowel disease (3), and it is required for oral tolerance (36). Interestingly, NKR-P1⁺ cell populations have been implicated in the regulation of a Th1-type immune response in a mouse model of chronic colitis, but it is not known specifically whether NK cells of gut origin subserve that function (37).

With regard to possible immunoregulatory function, it is intriguing to note that ART2 (29) and CD25 are expressed at high levels on rat IENK cells. Expression of ART2 on peripheral rat T cells is a marker of regulatory capability (29, 38), as is expression of CD25 on a population of peripheral mouse T cells that can prevent autoimmune diabetes in NOD mice (39). It is plausible to speculate that ART2⁺CD25⁺ rat IENK cells could play a role in the processes that regulate the physiologic inflammation that is characteristic of the gut (1).

Our data suggest further that defects in IENK cell number or function might lead to unregulated inflammatory processes in the gut that result in disease. Abnormalities in gut immunity may participate in the pathogenesis of autoimmune disorders, including celiac disease (2), inflammatory bowel diseases (3, 4, 5), and autoimmune type 1 diabetes (32). With respect to diabetes, we have documented that the spontaneously diabetic, diabetes-prone BB rat strain of rats is severely deficient in IENK cells (D. J. Todd, unpublished observations). The diabetes-resistant BB rat strain is also deficient in IENK cells compared with the normal WF strain rat used in the present studies. Interestingly, however, when diabetes-resistant BB rat are induced to become diabetic using a standard protocol consisting of anti-ART2 mAb plus polyinosinic-polycytidylic acid (40), levels of IENK cells become nearly undetectable (D. J. Todd, unpublished observations).

There is as yet little information on the abundance and function of IENK cells in species other than the rat. Some IEL in both human and mouse express NK cell surface markers (41, 42, 43), and mouse IEL exhibit NK-like cytotoxic activity (41, 44). However, most IEL in mice (6) and humans (45) are CD3⁺, and spontaneous secretion of IFN-γ (and IL-5) in the mouse has been associated with intraepithelial CD3⁺ T cell populations (46).

In conclusion, we have demonstrated that the rat IEL compartment harbors a large population of NKR-P1A⁺CD3⁻ cells that function as NK cells, but display an activated phenotype and highly unusual cytokine profile that clearly distinguish them from splenic NK cells. Their phenotypic and functional characteristics suggest that these distinctive IENK cells may participate in the regulation of mucosal immunity.

We thank Dr. Bruce Woda for assistance with the interpretation of histology specimens, and Lisa Burzenski, Kelly Lake, and Michael Bates for technical assistance.